Help needed with strange terminal issue (maverick builder)

17 posts / 0 new
Last post
Offline
Last seen: 1 month 3 weeks ago
Joined: Feb 9 2022 - 08:53
Posts: 18
Help needed with strange terminal issue (maverick builder)

Hi,

First of all I would like to apologies for posting in the wrong thread. I started in the ""Horror Room" Symptoms of Bad CIs" thread but I couldn't proove it was a faulty IC somewhere so I created this new thread.

For those who starts reading this thread form there here's my problem. When D4/D5 sockets are fully populated with shift registers (I'm using the TO-99 package MM1404AH and they seems to be working) my terminal section starts to delete all the characters from screen resulting in a very fast screen scrolling up. If I put a 0.1uF capacitor between D4a pin 4 (-5V) and D5 pin 8 (GND) screen stops scrolling up but the clear screen won't clean the cursor shift register. I also have a ram issue, when all 16 sockets are populated the high nibble of every byte is corrupted (in reading it outputs "D" or "F" only).

If only bank X is populated (0000-1000) ram works but it loose data.

 

I also had a spacing issue on screen that I was able to solve today, the 0.001uF was broken and replacing it solved that problem. I also changed the 27k resistor with a 22k resistor and I made myself sure to have a 47pF capacitor near the 74123 (B3). The value of this capacitor is exactly 47.5pF (read with my tester) but still no luck. I asked a friend to help me measuring this capacitor in case my tester is not good and the capacitor has a different value. The electrolitic capacitors I'm using are 2 x 2200uF by Vishay and a 4700uF capacitor in place of the 5300uF one. I tested ripples with my oscilloscope and they seems ok. As dram chips I'm using 16 x TMS4116 by Texas Instruments. I suspect some problems with my -5V line because it seems to activate the fake cursor killer circuit, anyway I've done a test and since I'm using an smd replacement for the LM323k I lowered the output voltage from 5.00V to 4.81V and shift registers seems to be more stable. Do you think this could be caused by a too weak trasformer or do I "solved" something by lowering the voltages?

 

Offline
Last seen: 16 hours 58 min ago
Joined: Apr 1 2020 - 16:46
Posts: 842
Here is a general plan on how to proceed, "Maverick" builder:

Here are a few hints (thanks for opening this new thread, BTW, helps to keep this forum more orderly).

 

First, check that all regulated voltages are in spec, meaning +/- 5% which you can expect from quality regulators. I know about the pain with the LM323K as I myself was affected by the usurious prices. What happened is that the major manufacturers of the LM323K did a "last call" around year 2014 and then ceased production. Now only TI is left and these LM323K cost  a whopping $68.96 a piece, see here:

 

https://www.mouser.com/ProductDetail/Texas-Instruments/LM323K-STEEL-NOPB?qs=R0RaSr%2FznUSEfDillJatbA%3D%3D

 

... which is not viable for Apple-1 builders. When I ran out of my 'last call' LM323K from ST Microelectronics, I looked for alternatives and found some. But as so often, the IC broker forced me to buy ALL his stock. I had to write a check over some thousand dollars. Now I can do a 'bath' in fine Fairchild UA323K like Uncle Scrooge McDuck does in his gold coins. Hundreds are left over. But still, better than buying the currently made TI ones at monopoly / usurious prices.

 

The thing with the reduced 5V voltage helping with Tom98's problem is rooted in MOS transistor physics. Subtract the threshold voltage of those MOSFETs and the remainder squared is proportional  to the current they can draw. So a little reduction in power supply voltage can have profound effects. But this is not the right remedy ! It can only point to the culprit(s) !

 

One remedy is to improve the power supply bypass capacitor network. This means adding the 'reliability mods' as seen here:

 

https://www.applefritter.com/content/part-path-towards-rock-solid-apple-1-builds

 

The "photo" in my post #4 in that thread actually is a high res scan that was stitched together. So to see all the details, download the photo (in most browsers, right click on photo, choose "save as ...") and then click on the downloaded picture and magnify it. I guarantee you can see each and every detail. Down to the things you don't want to see ...

 

After all these mods are in you know you have a good, healthy Apple-1 motherboard, which shall work even without my specially selected IC sets from my kits. (As long as your ICs are not bad ... most TTL are good out of the tube, though).

 

There is another issue, fake DS0025 from China. I wrote about them in another thread but here is the abridged version: the original MH0025 from National Semiconductors spawned the DS0025 which then lead to an improved and faster version, the DS0026, all still bipolar technology, and these spawned gazillions of CMOS based 'MOSFET drivers' used in the industry throughout, up to today, more than half a century later. All with the same socket footprint as the late 1960s MH0025.  Crafty Chinese IC counterfeiters grind off the type numbers of these modern and super strong CMOS parts and laser engrave them with "DS0025" because they know that real DS0025 cost $50+ each at IC brokers, if you can find them.

 

Now here is a secret. If you have an Apple-1 that works perfectly fine with a real DS0025, and you plug in a DS0026, it is highly likely the DRAM will have occasional errors (more often under a DRAM test program, of course). Put in Uncle Bernie's reliability mods as per the above link and it will work fine even with a DS0026. How can I be so sure ? Well, my first Apple-1 build failed because I did use a DS0026. The DS0025 was unobtainium for me back then. I could not start selling my early kits before I had sorted out the problems, because all I had was DS0026. This is the dirty little secret why the 'reliability mods'  were published in Fall 2020 - when I started to sell these kits. I had no other option - only much later I was able to find a few DS0025 which have been in my kits since then. But I learned another lesson about lousy PCB layout and lack of bypass capacitors of adequate performance ! 

 

Now, let that secret sink in. If you don't have a real DS0025, your Apple-1 will not work, ever, period. Unless you put my 'reliability mods' in. Only then it will work. If you got any Chinese fake, and some are re-stamped DS0026, it simply won't work unless you put all my 'reliability mods'  in.

 

But if you are unlucky and have a fake DS0025 based on a modern CMOS MOSFET driver which can source >3 Amperes of gate charging current within nanoseconds of being activated, it will never, ever, work even when all the 'reliability mods' of the world are put in. These modern monsters need special multilayer PCB layout techniques to work at all !

 

There is a way to make these brutal and much too fast CMOS drivers work in an Apple-1 - I have engineered a solution when I got my first Chinese fakes of DS0025 - but before that, please put all the reliability mods in, including the six 390 Ohm damping resistors.

 

Then, check for the proper timing of the 74123 oneshots. You mentioned you have an oscilloscope. Set trigger to positive edge, timebase to 100ns/div and measure the positive pulse width at pin #13 of the 74123. It should be 480ns wide, as in the Apple-1 schematic (in the a1man.pdf found online). I calculated this number myself and this time Woz did not screw up - he is spot on, and he took the "fake" PHI2 timing in his Apple-1 into account. So we know he did it right. The trouble is, the timing window of the original 1 MHz 6502 and the original MK4096 1st generation DRAMs  is only 70ns wide ... +/- 35 ns around the 480ns center. It is all but impossible to meet this with unselected components. Woz must have known this because the 47 pF capacitor in the originals was a mica capacitor ... back in the day the most expensive and most stable capacitors available. They were sold in 2% tolerance bands. Mica is a natural mineral which lived in the Earth for 100's of millions of years, so, no, it will never change its dielectric properties in a human lifetime. The art of making mica capacitors is long lost. Humanity is tumbling into idiocracy. Even simple items like bolts can't be made to spec anymore. When the manufacturing was moved to China by greedy, parasitic, psychopathic globalist billionaires (see the "Elysium" movie to learn about their traits and what they plan to to to you and your family), the know-how was lost. The last "high strength" bolt I broke using a torque wrench failed at half the specified tension. Both the bolt and the torque wrench "Made in China". Who is the culprit ? Me buying "Made in China" ? So far "Idiocracy". Don't you worry where the bolts fastening the jet engines to the wings in the airliners are made ? See, this is why I fly my own plane.

 

Where Woz did screw up with the DRAM timing (or did he) was to use a carbon composition resistor as the timing resistor for the 480ns pulse. This won't work until hand selected and measured after soldering. In my kits I provide a pre-tinned carbon composition resistor hoping that it would not move again when soldered in. And a metal film substitute which is guaranteed not to change under soldering. I also furnish faster 6502B to widen the timing window.

 

The other trouble you have observed with missing characters may have been the other oneshot of the 74123. It is generally quite non-critical in its timing, anything above 2us works, but around 2.9us, 3.9us, 4.9us ... there is a "sour spot" where the 6520 PIA gets metastable and the handshake logic fails, causing dropped or duplicated characters. The 6820 PIA did not have that issue as far as I can tell from my experiments trying to provoke the fault. I'm sure that the 6820 has it, too, but I think they did something to mitigate the metastability (trick circuits exist) and this was not done in the 6520. Ironically, I was unable to find proper setup and hold time specifications for the CB1 pin of the 6520 PIA in any datasheet I have. Seems like an oversight. Like in politics, the things they don't tell you are the important ones.

 

Measuring the 2nd oneshot of the 74123 is a bit tricky because it's only fired when a character is being output. In the Apple-1 manual, there is a small machine language program example which you can use to produce a constant character stream. But you need a fully functional DRAM to enter it with the Wozmon. In my kits, the A1, A2 PROMs have a "diagnostic page" which runs even with no DRAM being there, so you can do the oneshot measurement anyways. Set the oscilloscope trigger to 'negative edge' and check the 74123 pin #12 to have 3.5us wide "low" pulse width. Anything goes unless it's close to the x.9 us point or lower than 2 us.

 

So far my comments and hints for today. I always appreciate those daring 'maverick' Apple-1 builders ! Because their struggles prove that my IC kits indeed have "hidden value". Or shall we call it "surplus value" according to Karl Marx ? Or "secret sauce" as it is called in the industry ? This is  a very elusive thing ... hard to prove it is even there ... but believe me that none of the builders of my kits goes through the same "pain and suffering" episodes as the 'maverick' builders do. Which is why we need those mavericks ... to tell their story ... so others can learn and give me their hard-earned money for one of my kits. Get them as long as they are available ! Send PM for better price than on Ebay.

 

- Uncle Bernie

Offline
Last seen: 2 months 1 week ago
Joined: Apr 25 2019 - 11:52
Posts: 12
Help needed with strange terminal issue (maverick builder)

Toms98,

you mentioned in your post about using TMS 4116 drams. 

Those are 16kx1 bit chips but the apple 1 uses 4K x 1 dram chips like mk4096 or equivalent.  

pinouts similar but pin 13 is cs# on 4096 and A6 on 4116. different beasts.  

 

 

Offline
Last seen: 1 month 3 weeks ago
Joined: Feb 9 2022 - 08:53
Posts: 18
This explain the ram problem

Thanks a lot UncleBernie! I'll startd doing all the mod and tests you suggested. I'm quite sure of the source of my DS0025 and also of my PIA (I used a 65c21s); I found a problem with 74123 but now seems solved, just need to scope the 480ns signal.This explain the ram problem when all 16 chips are in socket, thanks stanleyruppert!

Offline
Last seen: 2 months 1 week ago
Joined: Apr 25 2019 - 11:52
Posts: 12
Help needed with strange terminal issue (maverick builder)

One additional suggestion/ caveat. You mentioned using a smd switch mode replacement for the lm323. I've found a few vintage boards ( with marginal / sensitive designs) are really impacted by the significant high frequency noise on the 5v output of some of the smd sm lm323 replacements. Spent hours chasing transients and glitches on one recent trainer ( Heathkit) I repaired, that was completely stable after I put back a real to-3 regulator. On other more tolerant systems they seem to work just fine though. You might check the hf output noise on the particular smd regulator you have. 

Offline
Last seen: 1 month 3 weeks ago
Joined: Feb 9 2022 - 08:53
Posts: 18
I have good news and bad news

I have good news and bad news today.

The good news is that the terminal section is now working thanks to the mods suggested by UncleBernie. Actually I used only 3 x 0.1uF capacitors, two in the D4/D5 locations and one in C11B. This solved the problem caused by these TO-99 shift registers. Thanks UncleBernie! Without your help this debugging preocess would have been way harder!

The bad news is that ram still don't work. I still haven't put the resistors suggested by UncleBernie because I'm not sure that TMS4114 dram can work without some kind of modification. I put a 22k resistor (measured with a tester) and a 47pF capacitor (measured, exactly 47.7pF) in the 74123 location. I still have to scope the signal now (I haven't got the time today, I'll do this tomorrow).Did someone here that used 4114 dram could help me? Do I have to buy new dram chips?

Offline
Last seen: 16 hours 58 min ago
Joined: Apr 1 2020 - 16:46
Posts: 842
About the damping resistors and wrong DRAMs

In post #6, Toms98 wrote:

 

"I still haven't put the resistors suggested by UncleBernie because I'm not sure that TMS4114 dram can work without some kind of modification."

 

Uncle Bernie comments:

 

The six damping resistors mitigate the ringing on the multiplexed address bus which is caused by using 74S257 Schottky TTL drivers together with the PCB traces being too long for that. This problem is well understood and has been measured and confirmed by many builders who have fast enough oscilloscopes. Some DRAMs are more sensitive than others to ringing on the power supplies and/or ringing in the multiplexed address bus (6 lines). With the Intersil MK4027N-3 date codes from early 1980s I provide with my current kits, almost nobody needs extra bypass capacitors but the six damping resistors are needed more often than not, otherwise the diagnostics page will throw the occasional DRAM error.

 

About the use of 4116 type 16kx1 DRAMs in the Apple-1: I tried to modify one of my lab rats (I have 14 Apple-1 in total) for 20K RAM, by keeping the upper 4k bank as is and replacing the lower bank in Row B with those 16k x 1 types. There are several problems with that. One is that the 16k x 1 DRAMs need an extra address mux and an extra refresh counter bit. But if you simply  keep their A6 input L or H, they should behave as 4k x 1 DRAMs and no extra multiplexers and refresh counts should be needed. Note that this is just an educated guess of mine, I did not try this, as I put in all the extra stuff to really get 16k Bytes.

 

But there is trouble which is a fundamental problem. 1st generation DRAMs (4096 type) and 2nd generation DRAMs (4027 type) had the feature that they kept the read data on their outputs live for a minimum of 10 microseconds, even if /CAS and /RAS were deasserted (driven "H"). The output would keep that state (the read data) until the next falling edge of /CAS. Only at the falling edge of /CAS the DRAM would decide if it is selected by /CS (pin #13) or not and change the output drivers to 'Z' if not selected. But until this point, the previously read data from the previous cycle was available at Dout. With the 3rd generation DRAMs, pin #13  was changed to A6 to provide two more address bits, and /CS went away. Now they had to change the way to control the Dout and they did that by disabling the tristate driver after the /CAS rising edge. From this point on, the read data on the PCB traces starts to crumble away. It is only held by parasitic capacitance there.  Alas, in the Apple-1, the crumbling starts before the 6502 cycle is complete.

 

This "read data hold" feature offered by the early DRAMs was lost at that point in DRAM history. Some call it "progress". It came back much later under the name "EDO" or "Extended Data Out". Which was hailed and sold as a great invention. Idiots.

 

On the 3rd generation DRAMs are some other subtle timing changes involving /RAS and /CAS and this is the reason why most don't work in the Apple-1. I tried numerous 4116 type 3rd generation DRAMs and none worked (despite I had the added address mux and the added refresh bit). But one particular one did work: the MM5290N-4 made by National Semiconductors. I also used a very foul trick to keep the read data alive for a bit longer, added capacitors:

 

 

This is a very ugly kludge which however was quick to do and it made the DRAM pass all the test programs with no error.

 

It's not an acceptable solution, however.  The timing needs to be fixed elsewhere, in the upstream logic. But alas, my logic analyzer has been in parts for more than two years, and I had no time to put it back together again, after I repaired all the problems it had on several subassemblies. If I can ever find all its guts I might be able to put it together later this year and then use it to get to the bottom of the DRAM problem. Without knowing exactly what is going on in the circuit I don't want to waste any of my precious RQLT on tinkering around any further. This is one of the very few cases where an oscilloscope is not good enough to measure everything in a proper way. A logic analyzer is needed because it allows for more complex triggering on events and then captures what has happened before and after the trigger event. Analog scopes can't do this because of their analog delay line - you can just see the trigger event but not much before that. Oh, and I despise digital oscilloscopes, being an analog IC designer. They do have their merits but unless you get the super high end ones you can't see many type of events which are easily seen on an analog scope, and these cheaper digital bastards may trick you into thinking all is good while it ain't. (Know the limits and pitfalls of your instruments). Furthermore, the digital scopes typically cannot be repaired anymore, so don't waste too much of your money on them.  It's difficult enough to keep old analog scopes made in the 1980s by Tektronix or HP in a good functional condition. With digital scopes it's hopeless because normally you can't even get service manuals with all schematics, and they are full of custom ICs you can't find anywhere other than in another scope of the same type.

 

OK, this said, my best recommendation for Toms98 is to get some 4027 type DRAMs (speed grade is irrelevant for the Apple-1, all 2nd gen DRAMs are fast enough). At some point in the future the woes with the 4116 type DRAMs will hopefully be sorted out and then we can have reliable 20 kByte or 32 kByte Apple-1 still using "honest" 1977 era DRAMs. I need that for my Disk-1 project as with only 8k RAM you can't run a DOS in any useful way (yeah you can run the DOS alone, but not much else). So I will get there eventually. But don't hold your breath. It's not the most important project I have. I did not try any of the published 20k and 32k mods from back in the day, because after I studied their schematics I came to the conclusion they may have even worse timing issues than the approach I adopted.

 

- Uncle Bernie

 

 

Offline
Last seen: 1 month 3 weeks ago
Joined: Feb 9 2022 - 08:53
Posts: 18
TMS-4116 ram can work (but I have to figure out how exactly)

I have interesting news about the TMS-4116 dram.

Meanwhile I'm looking for 16x4027 ram chips  as UncleBernie suggested I decided to experiment with the ram I have. I realized that the loss of data I'm experimentig it's due to the lack of refresh. In fact when I access the ram at 0000-0FFF the !CS line goes low, but that's connected to pin 13 which - as was pointed out - is line A6 in the 4116, so the refresh is performed with that line pulled high. Without changing the ram configuration I now tryied to access E000-EFFF locations (sicne I wired W to E this ram is accessible because !CAS can be pulled low and bank X !CS is now high) and here I had luck! Despite some problems probably caused by the lack of mods suggested by UncleBernie I was able to safely store data in bytes from over E042. I was even able to run the 11 bytes test program adapted to run at that location! I think the 4116 ram can be used to replace the 4027 by adding a cas enable with a pseudo-nand made with two diodes and a pull-down resistor. I'll draw a schematic of what I have in mind and on Monday I hopefully will be able to work on this.

The problem I still have are that ram from E000 to E040 is filled with D0 and can't be write. D041 and D042 sometimes are corrupted, I don't know what can cause this but I suspect it can be caused by some IC. I forgot to mension that I'm using 74LS367AN in lieu of 8T97 and 74F257 in lieu of 74S257. Can these ICs works as long as you know?

Offline
Last seen: 16 hours 58 min ago
Joined: Apr 1 2020 - 16:46
Posts: 842
On 4116 and 74LS367 in the Apple-1

In post #8, Toms98 wrote:

 

"I forgot to mension that I'm using 74LS367AN in lieu of 8T97 and 74F257 in lieu of 74S257. Can these ICs works as long as you know ?"

 

Uncle Bernie comments:

 

Wow, this is the true 'maverick' spirit ! You can be proud that you choose to beat  this  path into the jungle. How daring ! (Oh, and I'm not snarky ... I mean it ! There are very few of us who, when faced with a quirky design, choose up the ante even more by adding further deviations from "The Path" and further complications.)

 

One thing first. I never claimed that the 4116 does not work in the Apple-1. It almost works in my 20 kByte Apple-1 mod, but it never worked well enough to run a full DRAM test with no errors. There are subtle timing problems which need to be addressed, and I am sure that once these are resolved, any 4116 (or uPD416) will work just fine. At the current state of my work, only the MM5290N, under the scrutiny of my DRAM test programs, does run for hours or days without any error. And only after these kludge capacitors were added. There are simpler ways to extend the /CAS active time than that, but I wanted to prove the (few) errors seen (before the caps were added) were caused by crumbling Dout data, and nothing else. It was just the cheapest trick to test this hypothesis.

 

The 74LS367 runs just fine in my "all LSTTL" Apple-1 and so I see no reason why it should cause you some problems. It not the same thing as the 8T97, though. IIRC it's a bit weaker. But in the Apple-1 that does not matter. If you attach any extension board to the 44-pin edge connector (not not be confused with the "slot") and this one has no buffers on all the address and data lines, the Apple-1 won't work anyways.

 

The 74F257 is not a proper substitute for the 74S257. I've experimented a lot with various substitutes (such as 74HCT257 and 74AS... etc) but none worked all too well. I did not try the 74F257, though, because I have none of those. I would not recommend to look for higher speed logic families. High speed logic has fast edges and the fast edges cause reflections / ringing if the trace length exceeds a critical threshold. I have seen super fast CMOS self-destruct because the reflection coming back from a long trace caused negative undershoot on the driver which triggered latch-up, and the next thing is the IC starts to smoke and melt into its socket. As a general rule, never, ever, go faster than necessary to accomplish the mission. Applies to digital logic the same as to driving cars. And I know what I'm talking about. I drove 160 mph (or more) on public roads (lawfully) when I was younger and more stupid. Today, I would not even try. Had a close call all too often. Only super quick reflexes and nerves of steel - which at my advanced age I don't have anymore - saved me (and the offenders who pulled over into the fast lane).  And in my professional realm I saw all too many ambitious IC projects fail because the designers had not slowed down the output drivers. With modern CMOS you have to add lots to extra circuitry to make the edges of these drivers slower. Manager class vermin (their bonus depends on being in time and on budget) don't understand the implications of transmission line effects and fast edges so they don't allow for this extra design work to slow things down. Idiots. Their parasitic pattern of crime is to push impossible schedules, deliver the tape out on time, and then when the IC can't be sold, blame the designers. And they collected the bonus for being "on time" and never get hit with a clawback when a project fails. Instead, they get promoted to "vice president" where they can clown around, get even more money, but can't sabotage projects anymore, or at least not that easily. But I digress. Designing IC could be such a great job without that manager vermin.

 

The takeaway here for the general readership is a follows: when you work in electronics, always choose the  slowest  logic family that can do the job, and stay away from the fast stuff as much as you can. Already back in the early 1990s CMOS got so fast that an IC which toggled its own output drivers  could collapse its own power supply to a point where internal flipflops would flip erratically. And about the CMOS ICs which self destruct when the "shockwave" they made on a trace comes back to hit them, I already told you. Proper, competent design by engineers who understand transmission line effects can prevent such disaster. Alas, the Apple-1 layout from 1976 is not fit for that. There are no impedance controlled microstrip transmission lines anywhere.

 

Plug in these speed demons at your own peril !

 

- Uncle Bernie

Online
Last seen: 29 min 27 sec ago
Joined: Jul 5 2018 - 09:44
Posts: 2536
You are correct Uncle Bernie,

You are correct Uncle Bernie, in that the 74LS367 is weaker than the 8T97.  I'll take your word that it does not matter in an Apple-1.  It does matter in an Apple ][ or ][+.  If you substiture 74LS367s in those it will work fine with no, or maybe one card in the system...  But add a couple cards to the slots and you will start to get weird issues and glitches.  If I remember right, 74H367 did work better as a substitute, but at least as recently as a few years ago good 8T97s were still available, and that's what I'd definitely run in any ][ or ][+.

 

Offline
Last seen: 16 hours 58 min ago
Joined: Apr 1 2020 - 16:46
Posts: 842
About the 74LS367 in the Apple II (off topic)

In post #10, softwarejanitor wrote:

 

"It does matter in an Apple ][ or ][+."

 

Uncle Bernie agrees:

 

Apple II is  a different animal. Those 8 slots when fully loaded can be a stressful load to any bus drivers. I once made a killing getting an early Apple II motherboard for a song ... did not work ... the culprit was the 74LS367 some previous owner of this motherboard had substituted for the 8T97.

 

The issue with the Apple-1 is different. There are no eight slots. And if you attach anything to the 44-pin edge connector you will lenghten the traces above the length threshold where the reflections will make it unreliable, even when the six damping resistors of my 'reliability mods' are in. This may be the reason why back in the day there never was an expansion board for the Apple-1.

 

It can be done if the "expansion board" has full bus transceivers near the edge connector for everything, adresses, data bus, control signals, everything. But  none of the known ones ever sorted out all the problems. I think that Woz never tried to use the 44-pin connector for anything close to an "expansion board" although there are some photos suggesting there was once a hand-wired plug-in which - most likely - had some early 512 bytes EPROMs to hold the BASIC.  This conjecture is based on the observation that on these "BYTE SHOP" promotional pictures the upper DRAM bank on the Apple-1 motherboard was unpopulated. But the "BYTE SHOP" stuff seen on the screen most likely was from a for/next loop in BASIC. 

 

It has been a stupid idea to load a BASIC interpreter into heinously expensive RAM from back in the day. So Woz "invented"  (or adopted ?) the ROM cartridge which later was seen in the Atari VCS. Steve Jobs worked at Atari. Hmmm. The bad part is that Apple (the "Apple Computer Company") never followed through with providing the Apple-1 BASIC in EPROM. Users had to buy the heinously expensive DRAM. I found no cheaper offer than $19.98 in any of the BYTE Magazines of the time, per 4k x 1 DRAM IC. Eight were needed to the tune of $159.84 plus shipping which back in the day was a lot of money --- for only 22 times that you could get a nice new American made car with a V8 engine (just to show you the perspective).

 

- Uncle Bernie

 

Online
Last seen: 7 min 46 sec ago
Joined: May 4 2021 - 06:35
Posts: 153
External RAM expansion

Hi all,

 
UncleBernie wrote:

"It can be done if the "expansion board" has full bus transceivers near the edge connector for everything, adresses, data bus, control signals, everything. But  none of the known ones ever sorted out all the problems."

 

 

While designing/testing the Juke-Box Board, back in 2020, I came through many of the mentioned problems.

As a result, my Juke-Box board (https://p-l4b.github.io/jukebox/) and my microSD Storage Card (https://p-l4b.github.io/sdcard/) have full bus SMD transceivers.

 

Their built-in RAM expansion increases total capacity up to 32 kB, with the ability to replace onboard RAM as well, if and when needed.

With a single expansion board all RAM tests have never detected any problems, even after hours of operation, and my Replicas do NOT have any reliability-mod in place.

 

I have also conducted extensive tests of my boards with (very long) passive bus extenders but the results have not been conclusive, although mostly positive.

I also know of someone using an active bus extender successfully with my boards.

 

Since, to my knowledge, there is no a single "de-facto" popular bus extender model I always advise against using these devices in conjunction with my boards.

 

@Toms98: if it's RAM you're looking for, you might consider using one of these boards and not taking a risk with the 4116s.

 

BR - Ciao,

Claudio - P-LAB

 

 

 

 

 

Offline
Last seen: 1 month 3 weeks ago
Joined: Feb 9 2022 - 08:53
Posts: 18
About RAM solutions

Thanks Claudio, I'm definetly intrested in your solution! I saw your card on  a video and it would be a very nice add to my build. Anyway I would like firstly to fully debug my build, I also noticed that when I issue a write commend to the ACI board it corrupts all the ram. I'm finishing the reliability mods and in a couple of weeks I should recive the proper 74S257. Also I keep an eye on the proper 4027 dram.

Online
Last seen: 7 min 46 sec ago
Joined: May 4 2021 - 06:35
Posts: 153
sdcard

Happy to hear that!

No rush, take your time and feel free to PM/email me anytime.

 

Enjoy your build! :-)

 

Ciao,

Claudio

Offline
Last seen: 1 month 3 weeks ago
Joined: Feb 9 2022 - 08:53
Posts: 18
Working! Thanks for your help!

I can finally say that my board is working! I adapted the 4116 ram and is working!Apparently 74F257 works as a replacement for 74S257. Anyway I bought a tube of 74S257 and now I'm using the right chips on my board.

The only problem I had was that my tape recorder burned the ACI LM311, but I had two so I replaced it immediatly. I think I'll use an mp3 player from now on.Thanks everyone for your help!
Offline
Last seen: 16 hours 58 min ago
Joined: Apr 1 2020 - 16:46
Posts: 842
An intact cassette recorder can't kill the LM311 ...

In post #15, Toms98 wrote:

 

"my tape recorder burned the ACI LM311"

 

Uncle Bernie's comment:

 

This is impossible. No audio source plugged into the TAPE IN can fry the LM311. That LM311 must have been dead or morbid all the time. Here is the reasoning: there is a 100nF coupling capacitor between the LM311 and the TAPE IN. Impedance seen by this on the LM311 side is 5 kOhms. 100nF at 60 Hz is 26.5 kOhm impedance. This is a voltage divider by 6.3. No way that any voltage seen on the earphone output of an intact cassette recorder can fry the LM311. The sole exception I can see is that this cassette recorder is a decrepit cassette recorder from hell which due to an internal defect puts the line voltage on the earphone output. I once had a vacuum tube based tape recorder (with the big spools) which my teenage self had pulled out of a dumpster which had this very problem. Everything plugged into it was "hot". As the "hot" wire on the line cord ! Other than that it worked fine - it just tried to electrocute you. The problem was traced back to a bad transformer which I rewound and the problem was gone. Similar problems have been seen with electric guitar amps. The strings being "hot" can lead to funny guitar riffs ...

 

Other than that, congratulations for your successful build !

 

74F257 for 74S257 works of course but the ringing issues on the multiplexed address lines will be a little bit worse. However they are so bad anyways that a little bit worse doesn't matter. The added six 390 Ohm damping resistors from the 'reliability mods' will take care of this ringing in any case, 74S257 or 74F257, or 74AS257.

 

- Uncle Bernie

Online
Last seen: 7 min 46 sec ago
Joined: May 4 2021 - 06:35
Posts: 153
Well done!!! Will you share

Well done!!! Will you share your solutions for RAM and 2513? Others might be interested...!

Cheers!

:-)

Log in or register to post comments