WARNING: Counterfeit DS0025 from China (oh, no surprise !)

27 posts / 0 new
Last post
Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
WARNING: Counterfeit DS0025 from China (oh, no surprise !)

Hello Apple-1 builders worldwide !

 

Despite I sell Apple-1 IC kits with most of the passive parts added, too, I also appreciate the maverick builders who go out on their own quest for parts.

So they get scammed by IC counterfeiters, because some of the  ICs used in the Apple-1 are rare and obsolete and did not sell for 40+ years and now suddenly there is a demand for those hard-to-find parts. The Chinese are particularly skilled in grinding off IC part numbers and putting another part number on.

 

Here is the latest scam that was brought to my attention:

 

 

These are counterfeit DS0025 which were sourced in China at a ridicolously low price. Keep in mind that genuine DS0025 now cost $40-$60 each at IC brokers. So it's tempting to try to get a few from China for a tad above $1 each (last time I looked they wanted $2-$3 - greed ! - it seems that everytime somebody buys one, they jack up the price ... this rule applies to all Apple-1 ICs unfortunately).

 

Here are some clues how to spot the counterfeits:

 

1. A real DS0025 has a 250 ohm pulldown resistor from its input pins #2 and #4 to the negative supply pin #3. The actual value varies with process tolerances, I have measured 210 to 270 Ohms on various lots of real DS0025 I encountered. The counterfeits have no such input pulldown resistor, but they do have reverse biased PN junctions from the input pins to the to supply pins (#3 and #6, the latter is the positive supply). This means there is an industry standard ESD clamp in these inputs which is typical for a CMOS part. But with power applied, the inputs draw no DC current (nA perhaps). This means these inputs lead to MOSFET gates. So it's proven the counterfeits were made with  a CMOS or BiCMOS process technology, and not with a plain vanilla bipolar process as the DS0025.

 

2. A real DS0025 made by National Semiconductors (to the best of my knowledge, correct me if I'm wrong) NEVER had laser engraved type numbers because the semiconductor industry had no such laser engraving machines at the time the DS0025 was discontinued and its production stopped.  The counterfeits have laser engraved type numbers, the modern National Semiconductor logo that was not used for real DS0025, and they have a fantasy production lot / date code. Also, if you take an Acetone soaked Q-Tip to the counterfeits, black paint comes off (as seen in the photo) exposing a scratched surface witnessing that the original part number was ground off.  

 

These counterfeit DS0025 have been made from some other type of IC where the original part number was ground off, black paint applied to make the surface smooth, and then laser engraved with a new type number. Standard practive for Chinese IC counterfeiters seen on many, many more counterfeit parts coming from China.

 

So what are these magical mystery parts which pretend to be real DS0025 ? They don't work in the Apple-1 but seem to have the same footprint in terms of inputs, outputs and not connects. If you put them into an Apple-1, all the 2504 type shift registers get hot but other than that you get the same symptoms as with having no PHI3, PHI4 clocks. While I was taking measurements with an oscilloscope and the 2504 (in my case, AM2804 substitutes) removed from the sockets to protect them from a heat stroke, I inadvertently touched the input pins with my bare fingers (the IC was in a riser adapter to get better access for the scope probes) and - voila ! - the thing came to life and produced the required PHI3, PHI4 clocks on its outputs, pin #5 and pin #7. Wow ! Part of the mystery was solved: whatever this mystery part is, it's at least an industy standard  dual, inverting, MOSFET driver, and not just some poor opamp or 555 timer fished out of the electronic junk recycling stream.

 

How come ? Well, see my point 2 above. These mystery ICs have CMOS inputs which on the Apple-1 are being driven by 10 nF DC blocking capacitors, so the standard TTL signal swing between 0V and +5V gets transferred down to the -12V power supply rail (the 250 Ohm pulldown resistors in the real DS0025 complete this level shifting circuit). On the counterfeits there are no such pulldown resistors and so the operating point for the CMOS input stage was wrong, and the IC could not do its job. When I touched the riser with my bare fingers, the skin resistance coupled all pins in that side of the IC to the negative supply (pin #3, conveniently located between input pins #2 and #4, pin #1 being not connected internally, saving a bond wire). This charged the floating input nodes to an operating point near the negative supply rail and the input stages would start working. Of course, the charge would leak away after a short while and the clocks would stop - and being an inverting driver, they would get stuck at -12V. The two phase dynamic logic in the 2504 does not like the two clocks being active at the same time, and the dynamic gates get shootthrough current, which heats the chips up. Fortunately, PMOS is too weak to cause currents / heat that would destroy these ICs all to quickly.

 

At this point I knew that I was close to solve the mystery. I added some pulldown resistors at the inputs and put the 2504 back into the sockets, knowing from the oscilloscope that the PHI3, PHI4 clocks would be good. Bingo ! The Apple-1 came to life ! Wozmon started up but then crashed (this was a lab rat machine without my famous reliability mods). Aha ! A worse situation than with the DS0026 (no typo) which is a faster version of the DS0025. With good bypass capacitors, the Apple-1 works OK when using a DS0025, but is close to the cliff of unreliability. This is because of the inadequate power supply bypassing of the original design, especially on the -5V rail. If the PHI3, PHI4 clock edges get too fast, the 2504 shift registers pump a lot of spikes into the -5V rail and the DRAMs don't like that, causing random program crashes. With the DS0025 it almost works, most of the time. With the faster DS0026 the Apple-1 gets pushed over the cliff and it won't work. Hence, my reliability mods (you can find them here on Applefritter). This is also the reason why I keep a few lab rats around without any reliability mods, to be able to demonstrate this effect.

 

With the mystery part, the counterfeit DS0025, this effect is much,  much worse than even with the DS0026. This is the final clue: we are dealing with a modern, very powerful CMOS or BiCMOS based MOSFET gate driver which can pump an incredible amount of current to and from that MOSFET gate. I did a datasheet search and found that there are numerous such modern MOSFET gate drivers around, from many manufacturers, and they can pump up to 4 Amperes (!) at their outputs. Their footprint follows the industry standard for such 8-pin MOSFET gate drivers, which - surprise, surprise  ! - was set back in the late 1960s / early 1970s with the venerable granddaddy of all MOSFET gate drivers, the DS0025.

 

So the mystery is almost solved. I still don't know the exact type number or manufacturer of the powerful modern MOSFET gate drivers the Chinese counterfeiters have turned into fake DS0025. I don't care because they won't work in the Apple-1 anyways. The circuit and the PCB layout would need to be modified to make these modern parts work in the Apple-1 without dire effects. But anyways, you have been warned ! Case closed as far as I am concerned  !

 

For sake of completeness, here is a list of known counterfeit ICs from China which target unsuspecting Apple-1 builders:

 

1. fake DS0025 which are restamped modern MOSFET gate drivers. They don't work in the Apple-1 but may work in some other DS0025/DS0026 sockets.

 

2. fake 2504/1404 shift registers which are restamped 555 timers or 741 footprint opamps. In the Apple-1 they only get hot and melt your sockets.

 

3. fake 2513N character generators which are restamped RO-3-2513 character generators made by General Instrument (later, named Microchip Technology) ... these were designed to be drop-in compatible substitutes for the 2513N so at least they do work in the Apple-1. One was recently spotted on an original Apple-1 which was featured here on Applefritter. Oh, and I don't wonder why they make these fakes: real Signetics 2513N now cost $150-$200 each at IC brokers and no better prices can be had.

 

4. fake 6502 which are laser engraved with the Rockwell logo ... look at Ebay to see what I mean. These typically are not even restamped NMOS parts but more modern and easier to find CMOS versions of the 6502 - which extra instructions and some subtle differences in CPU cycle counts in some of the NMOS style instructions, too. As CMOS technologies are much faster than NMOS they produce stronger current spikes on the power supply lines which in the Apple-1 are too weak anyways.

 

5. fake 6520 PIA, also laser engraved with the Rockwell logo. See point 4. At least these seem to work in the Apple-1, I was told by some builders. As with the fake 6502, the giveaway is the laser engraving. Rockwell Semiconductors NEVER made NMOS 6502 with laser engraving. Same thing as with the DS0025 - they did not have laser engraving machines for ICs at the time. Everything was done with a transfer printing process using ink.

 

6. fake LM323K regulators. There is a thread here on Applefritter about those. These typically are restamped generic BJT (bipolar transistors) in TO-3 packages and they won't work as regulators of course. It's no wonder that the Chinese counterfeiters make them because of the usurious moon prices asked for them on places like Digikey and Mouser (about $60 each). This pain came about when ST Microelectronics stopped making them some years ago. Since LM323K are needed as spare parts to keep a lot of vintage equipment alive (Arcade video games, pinball machines, etc., etc.) the LM323K from the "last call" production run were quickly used up, and the pain is large enough that some people even make little TO-3 shaped PCBs with a modern switchmode power supply on them, which can drop into a LM323K place. Not good for authentic looking Apple-1, though.

 

7. fake PROMs. These are not really counterfeit, but used (programmed) and hence, worthless. The Chinese sell them as "used" (who would buy them ?) but some fraudulent sellers claim these are virgin / blanks. So buyer beware. Oh, I actually had one case where a builder had bought a few "used" PROMs and plugged them into the A1, A2 sockets, hoping the Apple-1 would work. Alas, there was no Wozmon inside ... funny attempt !

 

So this is the currently known list of fakes and frauds related to Apple-1 ICs. Buyer beware ! You have been warned ! And, of course, there are quality issues with some genuine Signetics parts of the 2519N and 2504V/1404 types. There must be some bad - possibly "reject"  lots sitting at IC brokers who peddle them at moon prices. It's no fun to buy 7 of these 2504V for $8 each and then find 3 of them are bad. It's even worse if you buy a 2519N from that usurious Ebay seller for $65 each plus shipping and then 20-25% of them are bad or die soon.  Oh, and then there are the character generator sellers who sell character generators having the wrong character set inside. This is why I had to stop selling my kits for six weeks. An otherwise reputable IC broker sold me RO-3-2513 with the wrong CGR-005 character set despite prior closing the deal I had a lot of email correspondence with them to make sure they are RO-3-2513 CGR-001 (which have the same character set as the Signetics 2513N CM2140) and again, same thing with the 2513N, there are numerous IC brokers out there who list 2513N but the CM number is wrong and there is the wrong character set inside. And they still want $150-$200 for them. Madness !

 

All these perils of course can be avoided if you buy one of my famous Apple-1 IC kits. I screen all the ICs for being genuine and each IC set I sell is burned-in in a real Apple-1 (well, a clone, of course) running 24/7 for four weeks. I have 6 burn-in rigs. If only one IC dies during burn-in, and is replaced, a new 4 week cycle starts on that rig. So you always get an IC set with my kits which are proven in a real Apple-1 and are known to work. There also is a lot of know-how in the selection of the passive components I add in the kits. These kits are your best way to get instant build success with your Apple-1 project. And the best part is, you can buy these kits directly from me, here on Applefritter, just send me a PM. You get a much better deal than on Ebay ... these Ebay fees eat a lot of money that is better spent on more parts !

 

Comments invited !

 

Macintosh_nik's picture
Offline
Last seen: 10 hours 8 min ago
Joined: Jan 8 2021 - 05:18
Posts: 427
Hi Uncle Bernie!

Very useful post, I also want to add from me that Mostek 4027n-1 from China also do not work in Apple-1, only get hot. 

P.S. Glad to hear you're back with your famous Apple-1 build kits!
Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
About the fake MOSTEK DRAMs from China of post #2 !

Mostek was founded in 1969 and existed until 1985, when it merged into ST Microelectronics, which was the end of "Mostek" branded IC products.

 

Back in 1985 nobody in the semiconductor industry used laser engraving for the lettering on the ICs, because this type of machine was not invented yet. Cost effective semiconductor based high power lasers did not exist ... and all other lasers like YAG lasers used for trimming of hybrid circuits and some ICs, which theoretically could be used for engraving, were far too expensive and maintenance intensive to be economically viable for lettering of IC packages in some of the second or third world countries where the IC encapsulation happened. 

 

So, in 1985 end of the Mostek brand name and no laser engraving machines for lettering ICs. The ICs shown in your photo look laser engraved (most likely, try with Acetone / nail polish remover to see of the black paint comes off, too) and the 0138 "date code" implies they were made in year 2001, week 38.

 

Absolutely impossible for real Mostek parts !

 

Uncle Bernies Tip:  NEVER buy ICs for the Apple-1 which have laser engraving !

 

(All vintage ICs of the time period we are interested in used transfer print lettering using ink !)

 

As Apple-1 builders, we are lucky that ALL of the vintage ICs NEVER had laser engraving so it's easy to spot counterfeits of our ICs. The general electronics industry of today however is badly hit by Chinese counterfeits of modern ICs, which normally do use laser engraving, so this easy detection of counterfeits does not work for more modern ICs. The Chinese fake literally everything, not only ICs, and it has gotten so bad that nobody wants to by ICs anymore which come out of China - or even were close to China. This hurts a lot of legit businesses based in China who have real ICs they want to sell. Unless the Chinese government cracks down on the counterfeiters (------------------) the Chinese electronics components business is toast, because sooner or later nobody wants to buy anything from them anymore.

 

Comments invited !

 

Off-topic and inappropriate parenthetical removed. - Tom

Macintosh_nik's picture
Offline
Last seen: 10 hours 8 min ago
Joined: Jan 8 2021 - 05:18
Posts: 427
Hi Uncle Bernie!

Well, I would not say that it is so bad, in China you can buy working chips for the Apple-1 even with laser markings. Of course they will not be vintage chips, trick or treat, I don't know. I wouldn't recommend to buy chips for your first assembly from China, but if you already have working vinyl boards to check the goods and you need a second (spare) set of chips, if you are ready to wait (long shipping) and if you are not a "hunter" for correct datacodes then China is a good enough option. As you can see there are a lot of "ifs"...

The photo below is what I bought at different times in China and it all works in the Apple-1.

Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
About Apple-1 ICs with laser markings.

In post #4, macintosh_nik wrote:

 

"Well, I would not say that it is so bad, in China you can buy working chips for the Apple-1 even with laser markings. Of course they will not be vintage chips, trick or treat, I don't know".

 

Uncle Bernie's comment:

 

This is correct insofar as more recently made ICs (after year 2000 or so) typically have laser markings from their legit manufacturers. IIRC a few manufacturers started to use laser marking in the late 1990s and the rest followed sooner or later. Nowadays, with a few exceptions, all ICs produced are laser marked. The ink transfer print process was messy and required skills to set up the machine, and time for the ink to dry.

 

But MOS Technology never had laser marked ICs, so your MOS 6502 from the first photo  is definitely a fake, even if it works.  The TI SN7432N in the second photo is laser marked and very likely legit. TI still makes TTL ICs (SN74xxx, not only the LS etc.) which I found out when I searched 7400 which ran out. If you buy 1000, you can buy directly from TI, and they are reasonably priced, a tad above $0.50, and not the usurious price asked by others, I have seen them for $1 and more, which for four NAND gates is a moon price. Alas, I don't need 1000 of those 7400. The SN74154N  in your first photo looks ink marked to me.

 

So how can we spot counterfeits ? 

 

The answer is in the photo of my post #1 above. When the counterfeiters grind off the original markings, the surface of the IC is not smooth, it has grind marks, and no amount of polishing can turn it into the same kind of surface the real ICs have. So the counterfeiters apply a layer of black paint and then do the laser engraving which removes the paint where the letter are. If you test the ICs with an Acetone soaked Q-Tip (nail polish remover is the same) then the black paint comes off, and this is the giveaway that the IC is counterfeit.

 

Note that there are vintage ICs which do have black paint from the legit, original manufacturer, but these cases are few, and exceedingly rare with laser marked ones.

The reason is that from time to time, mistakes did happen and  ICs were marked with the wrong code or the ink transfer print process went wrong (the rubber piece which transfers the ink wears out and if not replaced in time, the print will be bad). Then they used the same black paint process to get a new surface to undo the mistake and when that ink was dry, they ran these ICs through the ink transfer printing process again, this time with the right markings. 

 

I encountered a few ICs where the legit manufacturer did this black paint / remarking job. In one case it was DRAMs whose speed grade was downgraded to a lower speed, obviously they had mixed up a few tubes from different speed bins coming from the tester. In another case it was military ICs in CERDIP which had an US Department of Defense part number (without the "secret catalog" you can't identify which 74xxx it is) and then were remarked with the civilian 74xxx  code to be able to sell them as commercial / civilian ICs. So legit ICs with black paint do exist, too.

 

But these Chinese counterfeits are a different animal. There may not even be the right die (for laymen: the silicon chip) inside. They may be the same kind of function, but if so, possibly made on a different process technology, such as CMOS instead of NMOS, or BiCMOS instead of bipolar (seen in the case of the counterfeit DS0025 of post #1). This is bad because the speed and the requirements for the power supply bypassing rise dramatically and the Apple-1 original PCB layout can't sustain that. Logic levels for CMOS also are not TTL compatible unless it's the 74HCT family. 

 

Some counterfeits are outright frauds, which will never work, such as the Signetics 555 timers re-stamped to pretend to be Signetics 2504 shift registers. They did that because the grey mold compound used by Signetics in the late 1960s and early 1970s was Signetics proprietary and no other manufacturer had that. But these fake 2504 at least had no black paint job (would be a giveaway due to the grey package) and they had a real (and well done) ink transfer paint job, so they really looked like the real deal. But of course, being 555 timers, in the 2504 sockets they only would get hot and melt the socket !

 

Just yesterday somebody contacted me on Ebay and complained about fake AM1404APC he bought on alibaba (the one with the 40 thieves, mind you !).

 

The photo he sent was too poor quality but it looks like the typical Chinese counterfeiter black paint / laser marking job. These counterfeits have the following lettering:

 

1st line: AM1404

2nd Line: the AMD logo, followed by APC

3rd line: 09ZBEH   or   0928BEW,   to hard to read.

 

So be warned ! This brings to our attention that they already have started to counterfeit the 2nd tier / substitute ICs for the Apple-1 after nobody believes in their counterfeit "original" Signetics ICs anymore. Except for a few fools, maybe. But the word is out and every serious Apple-1 builder reads the forums and the posts on the FB Apple-1 group and knows about the fake "Signetics" ICs from China.

 

The Apple-1 IC situation is getting worse by the day ! Not only do we have the counterfeiters, but also the usurers who sit on huge stockpiles of the original Signetics parts. It's not that they are gone / sold out, noooo, there are plenty, enough to build at least 10 x the number of Apple-1 clones that exist right now, literally thousands of Apple-1 clones could be built from this parts stock, putting it to good use, but as far as I am concerned I don't pay $150 ... $200 for a Signetics 2513N character generator, and I don't pay $330 for a set of seven Signetics 2504V, and I don't pay $65 plus shipping for Signetics 2519N date codes from 1977 from which 20-25% are bad out of the tube or die during burn-in. I can show you the piles of dead 2519N from that seller which I have, but I bought them from him years ago before he became greedy with his prices. Oh, I don't pay $60 for LM323K (current prices at Mouser, Digikey) or $40 for DS0025N. At these moon prices my famous IC kits would cost me $730 for the beforementioned parts alone, plus all the other parts not mentioned, and this simply is not commercially viable.

 

I do not support usury. I boycott these usurers, and I hope you all do the same. Shall they sit on their moon priced ICs until the moon falls from the sky. Their heirs will then throw these stockpiles of genuine Signetics ICs into the recycling bin, for whatever little amount of gold there is in the bond wires.  It's the same sort of idiots who stockpile rare vintage great wines in their cellar and never open a bottle because they fear that the contents is spoiled.

 

But enough of that for today !

 

Just don't buy your parts from counterfeiters nor from usurers, and hold your money tight. Maybe the usurers come to their senses when they find out they don't sell these parts anymore, and the moon prices will come down to earth, and be affordable for Apple-1 builders again, and since these parts are Signetics originals, this would also stop the Chinese counterfeiters dead in their tracks. All problems in the world could be solved with just a little bit of common sense. But noooo, all we get is madness and insanity.

 

Comments invited ! 

 

Offline
Last seen: 7 hours 38 min ago
Joined: Feb 27 2021 - 18:59
Posts: 498
In another case it was

In another case it was military ICs in CERDIP which had an US Department of Defense part number (without the "secret catalog" you can't identify which 74xxx it is)

You mean 54xxx? The 5400-series predates the 7400-series and is for military/aerospace.

Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
Im post #6, robespierre wrote

Im post #6, robespierre wrote:

 

"You mean 54xxx? The 5400-series predates the 7400-series and is for military/aerospace."

 

Uncle Bernie answers:

 

No, this was a different animal. Made by National Semiconductors with date code 7619 (correct for the Apple-1 by coincidence) these 74160 had a weird number I was told by the seller it's either a Department of Defense number, or a special number for the  US Air Force.  I'm not to familiar with these so I must believe the seller. I sold a few with my recent kits until I was able to find "civilian" 74160 marked as such.

 

It's getting more and more difficult to even get ordinary  TTLs. There seem to be a few whales who suck up all the inventory of all Apple-1 parts they can find. I had all ICs needed for 100 kits in stock, 2 years ago, with a few exceptions and these exceptions caused me a lot of headache to find them. At least I am now in a position that I can deliver the 50 kits I promised for this year. After kit #100 it's probably over. Not that I would hate to help builders worldwide to get instant building success on first power up, to the contrary, it's really nice to get all the good feedback, praise and thanks, but when the ICs I have run out, then even I can't conjure up new ones out of thin air like the Chinese counterfeiters do ;-) 

 

You are right about the 54xx series. These are the same as the 74xx of the same "xx" number, as in the beginning (1964) these ICs were so heinously expensive that only military and aerospace could afford them. They came in hermetic packages (typically CERDIP, but I've seen flatpack versions used in satellite electronics, too) and they are guaranteed to work over the full military temperature range.  The SN74xx introduced in 1966 by Texas Instruments was a clever marketing ploy to make them attractive for the consumer and industrial marketplace. Much of the cost savings came from the plastic packages. In 1972 they were already cheap enough so hobbyists could build the first digital clocks (meant are those for private people ... some government organisations had digital clocks even back in the 1940, but based on vacuum tubes, and a rack full of those).

 

I think with the 74xxx TTLs getting scarce, the 54xx may be a yet untapped alternative. Surplus Sales of Nebraska has a lot of different 54xx ICs in stock, but last time I looked, they did not have all types needed in the Apple-1. 74LSxxx is much more abundant but when I tried to build an LSTTL based Apple-1 I ran into numerous problems around the 1k x 1 PMOS shift registers. None of them worked except the MM1404 from National Semiconductors. I did not get to the bottom of it why a LSTTL  based Apple-1 may act up like this, because once it worked, I was satisfied. So I only trust TTLs in the Apple-1 for now. In some places we can use 74LS without causing problems, but this is not a complete solution for a fully LSTTL based Apple-1.

 

I wonder if the international Apple-1 builder crowd would accept 54xx substitutes.

A military Apple-1 ! Yikes !

 

Comments invited !

Macintosh_nik's picture
Offline
Last seen: 10 hours 8 min ago
Joined: Jan 8 2021 - 05:18
Posts: 427
Hi Uncle Bernie!

It looks like I have a couple of these factory repainted Fairchild 74157's, the top is black and the bottom has the usual color of chips in a ceramic package. The paint is very even, no smudges and the font is the same as the original Fairchild font. I bought it in the USA, not in China, so I don`t think, that someone would re-label a chip, which is not very rare. Although everything can be. And about the gray chips, you said that only Signetics could have that color, but I have several Fairchild chips in the same gray package.

Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
About the ICs in post #8:

In post #8, macintosh_nik wrote:

 

"It looks like I have a couple of these factory repainted Fairchild 74157's."

"And about the gray chips, you said that only Signetics could have that color, but I have several Fairchild chips in the same gray package."

 

Uncle Bernie replies:

 

The 74157 look genuine to me, it's the same look as those 74160 which were re-painted and re-marked military parts I mentioned in my previous post.

 

Oh, and your discovery of the grey plastic Fairchild packages proves that memory of  old people is fading. Of course you are right. Now when I see your photo I remember that in the early 1970s I had some computer scrap PCB boards with these grey Fairchild 9Nxx ICs from which we kids tried to desolder them, which was difficult, we first sawed them in strips with single IC rows and then also made a cut between the pin rows. Then, with a fine blade, a cut between all the pins was made. After that, the PCB pieces only had one pin and could easily be desoldered from the ICs. This is how poor we were ... we had to get our ICs from scrap PCBs. Interesting how the human brain works ... give it one clue, a picture, or a sound, and all the memory comes back , after half a century !

 

So Fairchild obviously did use a similar encapsulation process as Signetics did, at least in the early 1970s,  your "youngest" IC is from week 20 in 1973.  The industry switched to the black mold compound around that time. But Signetics kept their grey mold compound for many years after. So grey mold compound Signetics ICs are more often found ! I don't know what drove the switch of mold compounds, but it could be that it has to do with patents and their expiry, or the grey mold compound turned out to be inferior in terms of reliability for the ICs. This is speculation of mine, of course. In my later professional life I worked for semiconductor companies and designed ICs and I can tell you that IC encapsulation processes and materials are very tricky and the chemistry of the mold compound mix is (or was) a well guarded secret. Since a few years there is a switch to so-called "green mold compound" but the color is not green, it's still black, the "green" means "environmentally friendly", and these "green mold compounds" lead to inferior quality of the ICs and to lesser reliability and lifetime. "Go green, get junk products that don't last". The lead free soldering process also is full of problems, the solder joints tend to crack under thermal cycles and tin whiskers can grow which may make shorts. So all lead-free soldered products in my opinion are junk and I try to avoid to buy them. There is a reason why military electronics, and avionics, and many medical electronics (i.e. implants, think pacemakers) are exempt from the lead free solder regulations. The consumer grade trash that civilians have to buy (there is no other choice) uses the unreliable and lifetime shortening lead free solder. So after a few short years the product fails and has to be thrown into the trash because it can't be repaired. Where is this "environmentally friendly", much more electronics junk in the landfills ? And all the rare materials and all the energy that went into the production wasted ?

 

Like everything "green" / "environmentally friendly" it's a big scam to rip off the customer and to ruin the competitors who do not own the patents for the "green" product. All these evil corporations need to do is to bribe corrupt politicians (who are incompetent morons in these technical matters) to nod off the new "environment protection law" the lawyers of the corporation wrote to leverage the patents the corporation owns against their competitors. This is the nature of the game.

 

And try to run a car factory with electricity from windmills and solar farms. Good luck with that. When this insanity continues we will end up with a pre-industrial society with no mass produced products, everything will be hand made, and to be able to afford a good sword with a good blade you would either need to be an aristocrat exploiting thousands of serfs or work your whole life to be able to buy it.

 

This of course is just another example how my line of thought wanders from green mold coumpounds to how "they" wreck our industrialized civilisation until we are back to the dark ages (literally, no electrical lights anymore for you, serf !) But if I prevented you, dear reader, from buying into this "green" scam, it was worth your time to read. If you can, only buy products that last more than one decade, and can be repaired, and avoid all that modern, "green" junk that does not last.

Offline
Last seen: 7 hours 38 min ago
Joined: Feb 27 2021 - 18:59
Posts: 498
For a long time the lead-free

For a long time the lead-free solder was quite bad but a large factor was the same patents/trade secrets issues you referred to for mold compounds. SN99C and SN100C were locked up by patents and the most accessible alloy was (still is) SAC305, which makes dull brittle joints. I always prefer to use lead alloys, especially Sn62, but there are more choices now for ROHS solders.

macnoyd's picture
Offline
Last seen: 13 hours 44 min ago
Joined: Oct 15 2012 - 08:59
Posts: 839
Regarding body-coated parts ...

Having worked for a Corporation that created IC printing equipment and inks, be aware that IC's and other electronic components have been ink coated and laser printed since the early 1980's.  The coating was primarily added as a median layer to allow laser prints.  You will find silver, blue, black and clear coated parts with scencil type prints on them from the early 1980's until the late 80's.  The reason for the stencil type was primarily because vector mechanisms weren't developed in quantity yet.  Took nearly a decade before that came about.  Laser prints using stencils allowed for quick change, but when vector printing came about, stencils quickly disappeared.  Vector printing remains a standard today, though dot matrix is also quite common.  Check out the violet printing on the bottom of your bevridge cans... that is either clear coat that chemically changes color with laser, OR it's an electrostaticly charged liquid ink particle. Both are common.  As of late, beer bottles are laser date-coded.  The composition of the bottle had to be changed to allow for a more visible print. (that's another story)

The traditionl ink you see on old IC's is primarily a UV cured ink compound that uses an iron-oxide dye.  It was often applied using an etched plate with the IC image masked within and transferred using a silicone pad and UV lamp cure.

Today it's mostly lasers as our Chineese counterparts often use.

Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
About early laser marking attempts ...

In post #11, macnoyd wrote:

 

"be aware that IC's and other electronic components have been ink coated and laser printed since the early 1980's."

 

Uncle Bernie adds and clarifies:

 

True, but it's a fact that in the 1980s, laser markings of ICs were exceedingly rare and only used in special cases. The  big and very expensive CPUs for PCs or large proprietary VLSIs were the trailblazers for laser marking, as some had individual serial numbers laser marked in, which is impossible with the transfer print process. These ICs did cost some $100 each (or even thousands, the sky is the limit) and so laser marking to be able to identify and trace each individual IC made sense. Mind you that even back then there were IC counterfeiters, and one of the common counterfeit trick was to re-label the ICs for a higher speed grade, which they weren't. This way the counterfeiter could rake in the price differential between the standard speed and the rarer and much more expensive high speed grade versions. Back in the day these counterfeiters had no laser marking machines (prohibitively expensive and very rare) so having laser markings by their original manufacturer served as a sort of counterfeit protection for expensive ICs. For 20 cent TTLs not viable (yet).

 

The cheaper and smaller DIL packaged ICs of the 1980s typically were ink marked. No IC from the 1970s or earlier was laser marked. As macnoyd wrote, the first laser marking machines for ICs came up in the early 1980s. They did not exist in the 1970s. The (very big) semiconductor company I worked for in the 1990s did not use laser marking until the mid 1990s and only when SMD IC packages  became more common, they started to use laser marking on those. Overall, the whole semiconductor industry followed the same pattern. IC collectors can see that pattern. After year 2000, almost every electronic component was laser marked.

 

For the Apple-1 we are primarily interested in the rare Signetics 2513, 2519 and 2504 PMOS ICs from the 1970s, and the early 4k x 1 DRAMs, of the latter the 1st generation DRAMs used in the original Apple-1, the MK4096, being exceedingly rare and also a possible target for the counterfeiters of today  These ICs all are from the 1970's, at least I've never seen any with 1980s date codes, because in the end of the 1970s these ICs already were obsolete and out of production.

 

It eludes my why anyone would counterfeit the 2nd generation DRAMs (4027 type) or the DS0025, but as we can see in this thread, the counterfeiters even do that.

 

So despite of macnoyd's assertion that laser marking machines did exist in the 1980s, you can still heed my advice to avoid laser marked ICs for your Apple-1 build, and you will be able to dodge most, if not all counterfeits, which are offered today for unsuspecting buyers. Mind though that the counterfeit Signetics 2504V which were reported to me two years ago were ink marked,  on a grey plastic DIL-8 package, but of course also came out of China.

 

So to be absolutely safe against Chinese counterfeit ICs, do no buy anything that comes out of China, was ever near China, or was ever touched by a Chinese seller. This of course is exactly how the counterfeiters hurt the legit Chinese businesses who have fished out genuine ICs from the electronics scrap / recycling stream. This applies to socket or solder pulls. But lots and lots of overaged and unwanted new-old-stock ICs in their original factory tubes also went to China after RoHS regulations caused the IC brokers to throw them into the recycling stream. So the Chinese do sit on a treasure trove of vintage ICs they got for pennies or for free. It's a pity that the risk of counterfeits is so high that we can't buy them anymore.

 

As far as I am concerned I would not buy any IC from China at any price, even for pennies a piece. Sorry you sh@t into your own nest, dear Chinese counterfeiters !

 

One final word of warning about IC brokers: there are two types, the "stocking brokers" and the "non stocking brokers". The first ones have the ICs they offer in their own stock, in the best case since they bought them as surplus from the electronics industry, before that industry was dismantled and moved to ... China. The key point here is that these ICs never left the country (i.e. USA) so they are no counterfeits, ever. The "non stocking brokers" do list ICs they saw listed elsewhere and when you commit to buy, they "bring them in" ... from China !?  This is where the risk lurks.  You had learned your lesson with the Chinese counterfeiters and now buy from U.S. IC brokers and they bring in the same Chinese counterfeit crap. At least you can make them give you your money back, in most cases. But it's always a disappointment, a hassle, and loss of priceless RQLT. So they should not only give you the money back, but also pay you, say, $600 per hour as fair compensation for the RQLT lost to investigate the counterfeits, do the phone calls, and initiate the returns. But they don't. So whenever you buy ICs and then get counterfeits instead, you lose, even if you get your money back. So buyer beware ! 

macnoyd's picture
Offline
Last seen: 13 hours 44 min ago
Joined: Oct 15 2012 - 08:59
Posts: 839
Agree with UncleBernie ...

I agree with UncleBernie regarding the specialized chips, but with one exception...

IBM was using our earliest Laser printer  back in 1983.  Also Martin-Marietta as well. But as UncleBernie indicatess, both were creating high cost military processors with unique part numbers in a fairly low quantity. (low being in the 1000's to 10k range because they were for specific purposes)

The one exception to that was Fairchild.  They adopted  laser printing in their TTL and CMOS IC's earlier than anyone else (to my knowledge) including Motorola, TI, Signetics and others.  The only reason I know that was because the folks who gained Fairchild's laser print business was our direct competitor, Lumonics.  They had a superior laser product from what we had at the time and made me wish I was working for them! :-D

But UncleBernie is spot on regarding IC's for the Apple 1.  Stay away from laser printed chips.  Shouldn't be hard to do at the moment.  But as time marches forward, so does the cost, so don't wait.

Macintosh_nik's picture
Offline
Last seen: 10 hours 8 min ago
Joined: Jan 8 2021 - 05:18
Posts: 427
Hi Uncle Bernie!

Yesterday I received a small batch of 74157 from China with laser markings, working, but one died after 10 minutes while testing.

Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
In post #14, macintosh_nik wrote:

 

"Yesterday I received a small batch of 74157 from China with laser markings, working, but one died after 10 minutes while testing."

 

Uncle Bernie's opinion:

 

Unless you can rub some black paint off using an Acetone (or nail polish remover) drenched Q-Tip, this particular ones probably are not counterfeit.

Why should anyone counterfeit humble TTLs which are worth almost nothing (a few dimes each) anyways ? Even for the Chinese, not worth the effort.

 

So most likely, you have just encountered the awesome "quality" of indigenous "Made in China" semiconductors, which do not have the best reputation.

 

The Chinese have bought the dismantled, long obsolete wafer fabs from the West, which were contaminated anyways (think "pig pen", but not in the literal sense, the problem is more subtle), and then they rebuilt them in China, after a long travel at sea, and note that the salt in sea water / sea spray contains ... sodium. And sodium ion contamination is the worst thing you can have in your wafer fab. It typically comes from the skin particles the cleanroom personell is shedding and so the contamination grows over time and first wrecks MOS, and then later, bipolar, too. There are procedures to mitigate the problem but this is part of the "secret sauce". Short, it's not easy to keep the cleanroom clean, and then all hell breaks loose.

 

The mainland Chinese are particularly inept to make semiconductors, which is evidenced by the fact that lots and lots of Chinese semiconductor companies did fail and went out of business. As simple as semiconductors look, successfully making them requires a lot of know-how not found in books and not easily stolen by industrial spies. The cleanroom personell knows not much, they just move wafers around from tool to tool, and press buttons but the actual "recipe" is hidden from sight, it's loaded into the tool via LAN. In the old times they had to actually dial in the settings of the tools, but this went away when the tools got "smart". So even if the Chinese would put a spy into a cleanroom, he could not steal enough secrets to run the same process in a Chinese copycat wafer fab.

 

It eludes me why the island Chinese ("Taiwan") are different and  can produce world class semiconductors (think TSMC). The culture of cheating and cutting corners should be the same. But maybe the Taiwanese have found procedures to avoid that in their wafer fabs, and so their semiconductors come out good.

 

If you want good TTLs, go buy then from Texas Instruments, they still make them, and have tens of thousands in stock. Of course when buying directly from TI (which is possible) you have to buy thousands of each type, like a distributor does. It has been my experience that distributors are very reluctant to order or restock TTL in DIL plastic packages, which the electronic industry does not use anymore, everything is SMD now. Actually finding the real 1970's and 1980's TTLs for the Apple-1 is getting difficult. The industry phased plain vanilla TTL out when LSTTL came along, because LSTTL was cleverly designed to be drop-in compatible with TTL, and it saved a lot of money for the power supplies and the cooling fans.

 

Alas, the drop-in rule (use LSTTL in lieu of TTL) does not apply to the quirky Apple-1. I got one example working after a lot of experimenting and head scratching (all 74xxx being LSTTL, no exception, except for thr 8T97). It seems that only the MM1404 from National Semiconductors do work in such a machine. None of the 1404 style shift registers from any other manufacturer did work. Maybe it's the input clamping Schottky diodes of the LSTTL. I did not have the time or interest to investigate that further once the LSTTL machine came to life. But I think that in the Apple-1, many TTLs could be replaced by LSTTL, except for a few connected to the outputs of the PMOS 1404.

 

Somebody should investigate this and make a substitution list. LSTTL are much more easy to find than TTL and they are cheaper to buy. No need anymore to buy Chinese made 74xxx of dubious quality !

 

Offline
Last seen: 3 weeks 6 days ago
Joined: Jul 5 2018 - 09:44
Posts: 2587
I don't claim to be an expert

I don't claim to be an expert but back in the day for example for the 8T97 chips I found in the Apple ][ I usually couldn't substitute a 74LS367 unless the machine pretty much had no cards in it, except maybe a Disk ][ Controller Card.  A normally loaded machine would fail in strange ways.  However I had better luck with other chips like the 74H367 or 74F367.  Those may not be any easier to find these days though.  It is also often easier to find C or HC variants of many chips, but compatibility in different applications varies widely.

 

Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
8T97 substitutes and a warning against 74C and 74HC in Apple-1

In post #16, softwarejanitor wrote:

 

"It is also often easier to find C or HC variants of many chips, but compatibility in different applications varies widely."

 

True about them being easier to find, but be aware that 74C or 74HC families have CMOS logic levels, which at 5V supply voltage are *not* TTL compatible. Go for the 74HCT family instead, which have been specifically designed to have TTL compatible input logic levels, and their outputs swing rail to rail, unlike TTL, giving much better noise immunity to the TTLs (or other logic) downstream. However, all CMOS logic families except for the very early metal gate CD4xxx series switch much faster than TTL and need  more power supply bypass capacitors than the Apple-1 has. The faster 74ACT family is even more demanding for the power and ground grid.

 

So far I've not succeeded in populating an Apple-1 with CMOS logic ICs. But I was able to make one example work with a CMOS 6502.

 

With the 8T97 you are right - the 74LS367 is weaker and the Apple II is demanding if many daughter cards are plugged in. But in the Apple-1 it works fine.

Offline
Last seen: 3 weeks 6 days ago
Joined: Jul 5 2018 - 09:44
Posts: 2587
UncleBernie wrote:In post #16
UncleBernie wrote:

In post #16, softwarejanitor wrote:

 

"It is also often easier to find C or HC variants of many chips, but compatibility in different applications varies widely."

 

True about them being easier to find, but be aware that 74C or 74HC families have CMOS logic levels, which at 5V supply voltage are *not* TTL compatible. Go for the 74HCT family instead, wh

 

You're absolutely right that the HCT parts are a better choice for compatibility, especially with older designs.  I've actually had C and HC parts work where LS originally were, but as I said, it varies a lot whether that is true.  I also know basically nothing about Apple 1, only Apple ][ and various peripheral cards and other accessories.  Many of these are newer and more permissive.

 

 

Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
How to spot counterfeit 6502 !

Hi fans,

 

it has been brought to my attention that Chinese sellers still sell counterfeit 6502 on Ebay. These typically are blatantly laser engraved with a huuuge Rockwell logo. Alas, Rockwell never used laser engraving on the originals, these were all rubber transfer printed.

 

Now, on Ebay it's usually easy to claw back your money via Ebay buyer protection. But you need to prove your point. Here is a tip:

 

The original NMOS 6502 used a different ESD protection structure at their inputs, while all CMOS 6502 have the industry standard dual diode ESD protection structure: one diode from the pin to ground and one diode from the pin to VCC. The latter diode never is present in the NMOS version. This is how you can test for the diode:

 

Digital Multimeter set to "DIODE" measurement. Red (+) probe to pin #2 of the suspect 6502. Black (-, COM) probe tip to pin #8 of the suspect 6502. The 6502 of course NOT in the socket. If it's a genuine NMOS 6502 you can see "overrange". If it's a counterfeit one (re-labeled CMOS version) you can see a diode forward voltage of 600mV ... 700 mV. This is the proof it's a counterfeit part, if it was sold as a genuine NMOS 6502.

 

Armed with this knowledge, you can expose the counterfeits. Be aware that the above rules do not apply if the "6502" is a total, fully fraudulent fake, where some other 40 pin IC was re-stamped as "6502". Some of the Chinese vendors are so cunning that they mix re-stamped CMOS 6502 with re-stamped junk. This proves that even the CMOS 6502 are getting rare in China.

 

Do not fall prey to these swindlers and criminals. If you want a CMOS 6502 you can buy perfectly fine and genuine WDC65C02S6TPG-14 originating from Bill Mensch / Western Design Center. I have one and if you bend up / disconnect on the PCB the pins #1, #5 and #36 and connect IC pin #36 to IC pin #8, then it will work in a standard NMOS6502 socket ... you may need to add a high performance power supply bypass capacitor, though. These CMOS ICs are so fast that they brutally kick into the power and ground rails, and vintage computers typically were not designed with that effect in mind.

Offline
Last seen: 21 hours 24 sec ago
Joined: Jul 31 2021 - 08:38
Posts: 21
Hi,Just looked into my

Hi,

Just looked into my leftovers box and saw a few 6502s from a very early Far East order, long before I discovered this great forum here with all the helpful tips on how to identify fakes.

Not that I'm surprised by anything about the "Chinese Refinement Activities", but why should one relabel a Rockwell (?) R6502 to a MOS 6502, if it's only sold for 69 cents anyway?

 

 

 

The "diode test" shows no CMOS CPU and it runs without any problems in my A1. 

By the way, in the box I also found some MOSTEK DRAMs with date codes '99 and '01 (Uncle Bernie, watch your coffee... ;-)

 

 

I wish you all a good start into a Happy New Year!

-Pete

 

 

Offline
Last seen: 4 hours 27 min ago
Joined: May 4 2021 - 06:35
Posts: 158
What about a 6502 with Datecode 4102...?

Interesting topic, I'd like to add my recent findings.

Some time ago I came across a counterfeit 6502, clearly sandblasted, with a datecode 4102 added somehow.
I promised myself to do some analysis on it, which I report here hoping it will be useful to someone.
 
TL;DR is extremely likely that my unit is actually a NMOS 6502B, or a 6502 that can be rated at 3 MHz.
 
A few more details:
The unit works perfectly, both on my Full-size Replicas and on Replica-1, so I'm pretty confident that it's actually a 6502 anyway.
UncleBernie "diode presence test" between pin 2 and pin 8 tells NO DIODE.
I then decided to do some measurements on the phi1 and phi2 clock signals since they are generated internally starting from the phi0 master clock.
In fact, I told myself that any speed difference should be detectable there, particularly in the Rise Time of those signals, which should then be an acceptable proxy for the real thing.
 
I used my Briel Computer Replica-1, which generates a very steep and clean ~1 MHz clock signal (see picture 0_REFERENCE_CLK), whose Rise Time is 6 ns, just to avoid giving the microprocessor an insufficiently steep signal to work with.
Phi 1 is the yellow channel, Phi 2 is the purple channel.
 
 
The "unknown" 6502 presents (picture 1_UNKNOWN):
- phi1 Rise Time = ~84 ns, phi2 Rise Time = ~105 ns
 
I then did some comparisons with some 6502s I had in the drawer, and this is what came up:
 
Genuine MOS 6502 (image 2_MOS_6502):
- phi1 Rise Time = ~102 ns, phi2 Rise Time = ~133 ns
 
Synertek P6502B (image 3_SYNERTEK_P6502B):
- phi1 Rise Time = ~89 ns, phi2 Rise Time = ~103 ns
 
As a further comparison, I also measured a Rockwell 65C02, rated for 14 MHz and in CMOS technology, which obviously has very small Rise Times (image 4_ROCKWELL_65C02): ~11 ns for phi1 Rise Time and ~25 ns for phi2.
 
Comparing the various measurements, it becomes clear that the 6502 under investigation is significantly faster than genuine 6502 and has nearly identical characteristics to the Synertek P6502B.
Since the 6502B is rated for 3 MHz I can assume that the unknown chip is itself a NMOS 6502B manufactured by either MOS or Synertek.
 
 
Side note: the measured Rise Times are a bit higher than those on the datasheets.
This is because of the measurement methodology: the oscilloscope measures the time from 10% to 90% of the voltage value, the datasheets usually report the voltage values as "0" and "1" and the times are measured from there. For example, in the MOS 65xx manual, values of 0.4 V and 2.4/1.5 V are given for clocks and 0.8 V / 2V for other TTL signals.
Also, other variables such as impedance/load capacity etc. were not measured.
For this reason, my measurements should be taken as a comparison under the same working conditions and not as absolute values.
 
Anyway, if you have any other tests in mind let me know!
 
Claudio - P-LAB
 
 
 
 
 
 
 
 
 
 
 
 
 
Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
Not all NMOS "6502" are real 6502. Japanese ripoffs ?

In post #21, p-lab wrote:

 

"UncleBernie "diode presence test" between pin 2 and pin 8 tells NO DIODE."

 

Uncle Bernie comments:

 

Looking for PN diodes between inputs and VDD (the positive supply pin) on unknown ICs is a good way to discern between NMOS and CMOS ICs, but it's not absolutely failsafe against cases where a CMOS IC for some reason does not have that diode. One such reason is to have 5V compatible inputs on a CMOS IC which is powered from a 3.3V supply voltage. In this case, they can't use the typical two diode CMOS input protection structure. But most CMOS input protection structures used throughout  the industry which do not need this "over the top" capability have that diode. You should see a forward voltage of 600...700mV at room temperature on that diode. The red probe of the multimeter (in "diode" setting) must be on the input pin, of course, for the "P" of the suspected diode. And the black probe must be on the VDD pin (for the "N" of the suspected diode). 

 

It eludes me why these Chinese counterfeiters go through the effort to re-label 6502-ish ICs to "6502" and then ask for a pittance. However, I have a suspicion. When the Japanese semiconductor industry started to make knock-offs of the original 6502 without wanting to get sued for patent infringement (the sole patent protecting the NMOS 6502 is U.S.-Pat. 3,991,307), so the crafty Japanese just left out the decimal mode. They did not need it for their game consoles anyways. Said patent protects the on-the-fly decimal correction logic of the 6502 ALU. So with no decimal mode, no correction logic needed, and the patent is dodged. I think said patent also was the reason why Western Design Center of Bill Mensch implemented a different way to do the decimal correction - his 65C02 processors needed one cycle more to do that. It's the same method the old 8080 processor did use, except in the 8080, it's an extra instruction ("DAA") which you use after a binary add or subtract of decimal (BCD) numbers. Again, the '307 patent is dodged.

 

The Japanese made gazillions of these crippled "6502" ripoffs having no decimal mode, and this may have been the opportunity for these Chinese counterfeiters to have an abundant source of material to turn into fake "6502".

 

This, however, is just a conjecture of mine. I'm just trying to find a logical explanation for the findings of p-lab. I think it would make no sense to re-label real NMOS 6502 from any manufacturer to again turn them into "6502".  These Chinese counterfeiters are scoundrels, but they are not stupid, and would not expend efforts if not needed.

 

If you want real NMOS 6502, you can buy them at Anchor Electronics in Santa Clara / CA / USA. These are from Atari, "house marked" with "CO14377" and they are genuine 6502B rated for 3 MHz clock speed. There is no "6502" on them, however. Price is reasonable.

 

- Uncle Bernie

Offline
Last seen: 4 hours 27 min ago
Joined: May 4 2021 - 06:35
Posts: 158
Testing the Decimal Mode

I quickly scribbled down a little program, which I don't even know if it is an exhaustive test:

 

LOC  CODE       LABEL    INSTRUCTION

0000 F8                  SED

0001 A9 05               LDA #$05

0003 18                  CLC

0004 69 05               ADC #$05

0006 20 DC FF            JSR $FFDC

 

 

Basically, the Decimal Mode is triggered, and the sum of 5+5 is performed. The call to the PRBYTE routine shows the contents of the Accumulator:

- if "10" appears the Decimal Mode worked

- if "0A" appears the Decimal Mode did not work

 

Apparently my unit has Decimal Mode implemented (see image below).

 

If the computer is reset (the CLD instruction - CLear Decimal - is the first to be executed by the WOZ monitor in case of a RESET) and the program is executed from location 1 (skipping the SED instruction - SEt Decimal) the 6502 of course performs the operation in binary mode.

 

I don't know what to say, I agree with UncleBernie about the apparent uselessness of the operation, but maybe I was just "lucky" not to get a unit lacking Decimal Mode taken from the huge bin of sandblasted 6502s...

 

If anyone has other ideas for further (or better) tests, I am happy to help...

 

Claudio - P-LAB

 

 

PS - test quickly conducted on Briel Computers' Replica-1, sorry for poor image quality coming from the small monitor..

 

 

Offline
Last seen: 21 hours 24 sec ago
Joined: Jul 31 2021 - 08:38
Posts: 21
I tried your little test

I tried your little test program and it seems that in this/my strangely relabeled 6502 the decimal mode has actually been implemented as well.

 

I guess I got lucky there too...

 

Best, Pete

Offline
Last seen: 13 hours 30 min ago
Joined: Apr 1 2020 - 16:46
Posts: 885
About the decimal mode test of post #23

This is a useful test to find out if it's one of these Japanese 6502-ish knockoffs without the decimal mode.

 

Well done !

 

But it's not enough to tell if the "6502" is a version which implements the decimal mode by adding an extra CPU cycle for ADC and SBC when the decimal mode is turned on. You would need to program a loop and use an oscilloscope to measure the CPU cycles needed. But it can be done.

 

I agree with p-lab's conjecture in post  #23 that the Chinese counterfeiters might have a pile of 6502-ish ICs with the original markings ground off, and then then add the laser engraving to make "uniform" fake "6502", so  if you buy more than one, you will get a mixed bag, but all look alike. (Try this with spare parts for airliners - counterfeit brake discs of Chinese origin have been found in some airliners already, but the article did not say if they had a "yellow tag". As the old adage says: "paper is patient". Yellow tags can also be forged).

 

Armin told me that he had bought some hundreds of "6502" out of China for his "wave-soldered Apple-1 project" which looked all alike but were - of course - laser engraved / relabeled they indeed were a "mixed bag",  some did work in the Apple-1, some didn't, so he had to send the whole lot back for a refund, which he got. He then ruefully sourced his 6502 and 6520 from me . . . 

 

As a side note on CMOS 6502 in the Apple-1: I succeeded to run a WDC  W65C02S6TPG-14 with a Y2011 date code in an Apple-1 (pins #1, #5 and #36 bent up, #36 stub with a wire to pin #8) on an Apple-1 and 'regular' CMOS 6502 might work sometimes, too, but these CMOS parts internally switch so fast that they bounce VCC/GND badly, and this exacerbates the issues the Apple-1 has in terms of ringing power, ground, and signal lines. It's a hit-and-miss and certainly is not a recipe for reliable operation. I found that reducing the power supply voltage of the CMOS 6502 to 3.3V can greatly reduce the stress on VCC and GND rails and leads to a better reliability and less program crashes. Added bypass capacitors also help. So in the end we are not forced to use NMOS 6502 for the Apple-1 forever.

 

There is another way to tell of the "6502" in your Apple-1 is a real NMOS one, by running my 'Codebreaker' game I have published here on Applefritter (make sure you get the latest version with the cheating bug fixed). It will complain when it runs on a CMOS 6502 and advertise my IC kits. But you still can play it . . .  it's a nice game !

 

- Uncle Bernie

Offline
Last seen: 4 hours 27 min ago
Joined: May 4 2021 - 06:35
Posts: 158
Cycle analysis of ADC on 6502/65C02

Thank you! :-)

 

To answer the question on the clock cycles I had to use my gizmo (I knew it would come in handy one day...) to dump buses and control lines:

 

https://p-l4b.github.io/A1-STEPPER/

(video: https://www.youtube.com/watch?v=UwX5D4P9iEI)

Here are the results of comparative tests conducted on the "unknown" 6502, on a genuine 6502 MOS, and on a Rockwell 65C02, all of which I have used before.

Note: to facilitate the capture I replaced the screen printout of the sum with a JMP at the start of the program, to create an infinite loop.

 

Long story short: the 6502 "unknown" doesn't appear to be a 65C02 in disguise.

 

From the dumps below, we can see that both the "unknown" 6502 and the genuine 6502 always perform ADC in the same number of cycles (11 cycles).

The 65C02, on the other hand, takes an extra cycle in Decimal mode (so 12 in total), as indicated in the documentation and pointed out by UncleBernie.

(source: https://downloads.reactivemicro.com/Electronics/CPU/65C02%20Datasheet.pdf )

 

For full interpretation of the dumps, in relation to the SYNC signal used to trigger the Op-Code and the pre-fetch function of the 6502, I have given some examples in the manual (all Stepper project is totally free, under Creative Commons CC-BY 4.0).

 

In this specific case, in a nutshell, the extra cycle is already pre-fetching the JMP Op-Code ($4C) but in reality the 65C02 is still finishing the previous ADC instruction (in fact the $4C Op-Code is not decoded as such because the SYNC signal has not been raised yet --> it will be lifted next cycle).

 

Regarding the use of 65C02 on Replicas: the 65C02 unit tested here works perfectly on both my Replicas (a Newton and a wave soldered by Armin).

It's rated for 4 MHz only, I'll try the WDC's 14 MHz one when I'll have the chance to place the next order...

At this address, in addition to the information already provided by Uncle Bernie, you will find the complete comparison and other useful information.

https://www.westerndesigncenter.com/wdc/AN-002_W65C02S_Replacements.php

 

Happy New Year!

 

Claudio - P-LAB

 

PS: @UncleBernie: Yes, Codebreaker "complains" when a 65C02 is in use.. (see picture below) ;-)

 

DUMPS:

 

"UNKNOWN" IN DECIMAL MODE 

0000000000000000  11111000  $0000 $F8 r  SED 11                                  

0000000000000001  10101001  $0001 $A9 r   A9 11                                  

0000000000000001  10101001  $0001 $A9 r  LDA 11                                  

0000000000000010  00000101  $0002 $05 r   05 11                                  

0000000000000011  00011000  $0003 $18 r  CLC 11                                  

0000000000000100  01101001  $0004 $69 r   69 11                                  

0000000000000100  01101001  $0004 $69 r  ADC 11                                  

0000000000000101  00000101  $0005 $05 r   05 11                                  

0000000000000110  01001100  $0006 $4C r  JMP 11                                  

0000000000000111  00000000  $0007 $00 r   00 11                                  

0000000000001000  00000000  $0008 $00 r   00 11

 

"UNKNOWN" IN BINARY MODE 

0000000000000000  11011000  $0000 $D8 r  CLD 11                                  

0000000000000001  10101001  $0001 $A9 r   A9 11                                  

0000000000000001  10101001  $0001 $A9 r  LDA 11                                  

0000000000000010  00000101  $0002 $05 r   05 11                                  

0000000000000011  00011000  $0003 $18 r  CLC 11                                  

0000000000000100  01101001  $0004 $69 r   69 11                                  

0000000000000100  01101001  $0004 $69 r  ADC 11                                  

0000000000000101  00000101  $0005 $05 r   05 11                                  

0000000000000110  01001100  $0006 $4C r  JMP 11                                  

0000000000000111  00000000  $0007 $00 r   00 11                                  

0000000000001000  00000000  $0008 $00 r   00 11

 

-----------------------------------------------

 

GENUINE MOS 6502 IN DECIMAL MODE

 

0000000000000000  11111000  $0000 $F8 r  SED 11                                  

0000000000000001  10101001  $0001 $A9 r   A9 11                                  

0000000000000001  10101001  $0001 $A9 r  LDA 11                                  

0000000000000010  00000101  $0002 $05 r   05 11                                  

0000000000000011  00011000  $0003 $18 r  CLC 11

0000000000000100  01101001  $0004 $69 r   69 11                                  

0000000000000100  01101001  $0004 $69 r  ADC 11                                  

0000000000000101  00000101  $0005 $05 r   05 11                                  

0000000000000110  01001100  $0006 $4C r  JMP 11                                  

0000000000000111  00000000  $0007 $00 r   00 11                                  

0000000000001000  00000000  $0008 $00 r   00 11

 

GENUINE MOS 6502 IN BINARY MODE 

0000000000000000  11011000  $0000 $D8 r  CLD 11                                  

0000000000000001  10101001  $0001 $A9 r   A9 11                                  

0000000000000001  10101001  $0001 $A9 r  LDA 11                                  

0000000000000010  00000101  $0002 $05 r   05 11                                  

0000000000000011  00011000  $0003 $18 r  CLC 11                                  

0000000000000100  01101001  $0004 $69 r   69 11                                  

0000000000000100  01101001  $0004 $69 r  ADC 11                                  

0000000000000101  00000101  $0005 $05 r   05 11                                  

0000000000000110  01001100  $0006 $4C r  JMP 11                                  

0000000000000111  00000000  $0007 $00 r   00 11                                  

0000000000001000  00000000  $0008 $00 r   00 11

 

-----------------------------------------------

 

ROCKWELL R65C02P4-14 IN DECIMAL MODE 

0000000000000000  11111000  $0000 $F8 r  SED 11                                  

0000000000000001  10101001  $0001 $A9 r   A9 11                                  

0000000000000001  10101001  $0001 $A9 r  LDA 11                                  

0000000000000010  00000101  $0002 $05 r   05 11                                  

0000000000000011  00011000  $0003 $18 r  CLC 11                                  

0000000000000100  01101001  $0004 $69 r   69 11                                  

0000000000000100  01101001  $0004 $69 r  ADC 11                                  

0000000000000101  00000101  $0005 $05 r   05 11                                  

0000000000000110  01001100  $0006 $4C r   4C 11    <-- EXTRA CYCLE                              

0000000000000110  01001100  $0006 $4C r  JMP 11                                  

0000000000000111  00000000  $0007 $00 r   00 11                                  

0000000000001000  00000000  $0008 $00 r   00 11

 

ROCKWELL R65C02P4-14 IN BINARY MODE 

0000000000000000  11011000  $0000 $D8 r  CLD 11                                  

0000000000000001  10101001  $0001 $A9 r   A9 11                                  

0000000000000001  10101001  $0001 $A9 r  LDA 11                                  

0000000000000010  00000101  $0002 $05 r   05 11                                  

0000000000000011  00011000  $0003 $18 r  CLC 11                                  

0000000000000100  01101001  $0004 $69 r   69 11                                  

0000000000000100  01101001  $0004 $69 r  ADC 11                                  

0000000000000101  00000101  $0005 $05 r   05 11                                  

0000000000000110  01001100  $0006 $4C r  JMP 11                                  

0000000000000111  00000000  $0007 $00 r   00 11                                  

0000000000001000  00000000  $0008 $00 r   00 11

 

-----------------------------------------------

 

 

 

 

 

 

 

 

Offline
Last seen: 4 hours 27 min ago
Joined: May 4 2021 - 06:35
Posts: 158
Cycle analysis of ADC on 6502/65C02: clarification

A necessary clarification (my bad, it was late in the night):

 

>From the dumps below, we can see that both the "unknown" 6502 and the genuine 6502 always perform ADC in the same number of cycles (11 cycles).

>The 65C02, on the other hand, takes an extra cycle in Decimal mode (so 12 in total), as indicated in the documentation and pointed out by UncleBernie.

 

ADC operation itself is performed in 2 cycles on genuine NMOS 6502, "unknown" unit and CMOS 65C02 in Binary mode.

It is completed in 3 cycles on CMOS 65C02 in Decimal mode.

11 cycles and 12 cycles stated above refer to the clock cycles taken to complete the whole program.

 

Claudio - P-LAB

 

 

Log in or register to post comments