"Horror Room" Symptoms of Bad CIs

10 posts / 0 new
Last post
Macintosh_nik's picture
Online
Last seen: 1 hour 36 min ago
Joined: Jan 8 2021 - 05:18
Posts: 69
"Horror Room" Symptoms of Bad CIs

Hello everyone, I think I'm ripe for writing my first column of my own on this wonderful forum.  Immediately I apologize if the text is a little clumsy, I do not speak English, I use Google translator.  The idea is this: upload here a photo (or even better video) depicting the symptoms of bad microcircuits.  Not too experienced A1 amateur builders like me would be able to identify a "troublemaker" from these photos or videos and replace him on their own without having to purchase expensive equipment.  I am sure that every A1 builder has such situations during debugging, and the presence of a "room of horrors" - the name was invented by Uncle Bernie, will greatly simplify this task.

Video

 I'll start with the Signetics 74123 chip, board location B3.  I had this "grandfather from the 70s", when I pressed reset, it did not stop displaying /////////, with a short hold, @ appeared, but as soon as I let it go, everything continued again.  The selection method was used to find the problem and replace the "troublemaker".

Offline
Last seen: 6 days 4 hours ago
Joined: Apr 1 2020 - 16:46
Posts: 157
Actually, my name for this was "Chamber of Horrors",

... but when it goes though Google translate twice, from English to Russian to English, it's amazing how good the translation is.

 

I really appreciate and welcome macintosh_nik for opening this thread and I hope I will soon find some time to post screenshots of my "chamber of horrors" exhibits.

 

Allow me a comment on the 74123: the symptom of '//////////////' appearing on the screen is typical for a marginally working DRAM. The oneshot in the 74123 responsible for the /CAS timing should produce a 480ns wide pulse. But I have found that, more often than not, the 27K resistor at the 74123 will not produce 480ns, but a much longer time, which is bad, and the DRAM will not work. I have found that 74123 from different manufacturers or different lots have great variations in the pulse width they produce, even for the same resistor and capacitor value. Those from my current tube of 74123 all need 20KOhm resistors to make the 480ns, and the variations between ICs from this same tube are very small.

 

So it is not necessarily a "bad" 74123 which macintosh_nik has found, it just did produce the wrong timing. Which may be one of the reasons why digital designers back in the  1970s learned to hate these oneshots. If you want to get any reproducible timing out of them, you need to use a trim pot as the timing resistor, but be careful, if you just use a trimpot, you can adjust it down to near zero, and the 74123 may get damaged by that. Its datasheet specifies a minimum resistor of 5KOhm. So if you experiment with a trim pot in the Apple-1 to replace the (almost always wrong) 27K resistor, use a 5K Ohm fixed resistor in series with a 20K trimpot.

 

You can then adjust the trimpot even without having an oscilloscope. Set the resistance between pins 15 and 16 of the 74123 to 22K (power off of course, and the 74123 not in the socket). Then enter a DRAM test program (if possible, if not, try other settings of the resistance, such as 18K, 20K, 24K, 26K).

 

Let the DRAM test program run for a while, hopefully with no errors yet, and then very slowly adjust the trim pot until the first DRAM errors appear.  Turn power off, remove the 74123, measure and note the resistance between pins 15 and 16. Set resistance again to the value where the DRAM did work. Put the 74123 back and repeat the experiment, but this time turn the trim pot into the other direction until DRAM errors are seen.  Power down, remove 74123, measure resistance. Use the average value between the two values where the error "cliffs" are. Adjust the trim pot until the resistance between pins 15 and 16 is at that optimum value. Re-run the DRAM test. You should not see errors anymore for days / weeks / months. Then you are sure all is OK and you can replace the 5Kfixed/20KTrimpot with a fixed resistor of the same optimum value. If errors still do occur, you may have a bad DRAM chip, or your Apple-1 may need my reliability mods added, seen here in post #4:

 

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

 

Entering the DRAM test program several times is tedious, especially if you have no keyboard yet ;-)

 

So I started to provide PROMs with a hidden burn-in firmware page with my IC kits. More on this in a separate post. My notebook battery is running too low now. Stay tuned !

 

 

 

 

 

Macintosh_nik's picture
Online
Last seen: 1 hour 36 min ago
Joined: Jan 8 2021 - 05:18
Posts: 69
Hi Uncle Bernie!

Thanks for the detailed comment!  I already read about a trimmer resistor instead of 27kOhm in your earlier post, but somehow I did not connect it with my "bad" 74123. Now everything becomes clear.  In the near future I will definitely undertake this.  I will write about the results, good luck in all your endeavors!

Macintosh_nik's picture
Online
Last seen: 1 hour 36 min ago
Joined: Jan 8 2021 - 05:18
Posts: 69
Signetics 2504

Hello everyone, we continue the "Horror Room" column, today another exhibit from the 70s is Signetics 2504. There are 7 such IC's on the board, different symptoms appear in each place, so there will be a lot of photos.  As you know, a working Apple 1 appears with a grid when turned on (24 lines of 40 characters each, @ symbols are blinking).

If the bad IC is set to D4B, then the _ will be W (@ symbols are blinking).

D4A, then the _ will be [

D5B, then the _ will be ]

D5A, then the _ will be ^

D14B, then the _ will be ?

D14A, then the _ will be O

Finally, if the bad 2504 is set to C11B, the grid will look fine, but ClearScreen will not work (no single blinking @ symbol appears in the upper right corner).

I hope this information will be useful.

 

 

 

 

Macintosh_nik's picture
Online
Last seen: 1 hour 36 min ago
Joined: Jan 8 2021 - 05:18
Posts: 69
Signetics 2519

Hello all! Another exhibit of the horror room, the most difficult to find Signetics 2519. It's very sad when you're looking for the right chip for a long time (I needed one with vertical markings), you find it, buy it, wait over a month until it's delivered and it turns out not to work. 

 

So, the whole screen is a grid of static question marks, clearscreen doesn't work. Shit happens...
Offline
Last seen: 6 days 4 hours ago
Joined: Apr 1 2020 - 16:46
Posts: 157
Uncle Bernie's horror show of bad 2519 ICs ...

Hi Macintosh_nik:

 

Thanks for posting this, but I have had my unfair share of misbehaving 2519, too,  actually, a little pile of them ""worth" several hundred dollars (and people still complain my 100% tested and burned-in ICs kits are too expensive. Idiots. So far I have not made one dime of profit out of these kits, and if I would ever get out of the sea of red ink, I would upgrade the kits to eat these profits away again ... I loathe the taxman that much.)

 

Here are a few more symptoms of bad 2519 (all were outputs of my burn-in firmware in PROM running):

 

 

And here is another one:

 

 

It seems that there is a pattern sensitivity in these "bad" 2519 so they are not 100% dead.

 

But the really bad news is this: with the exception of maybe 2-3 of them which were bad right out of the tube, all the other bad ones appeared to work at the beginning, but then they died during the burn-in. Most died over the first night of burn-in, so I found the mess after breakfast, and could replace them without affecting the schedule, but a few, maybe 3-4, died in the first week.

 

The takeaway from this is that I highly recommend any builder using parts sourced from other vendors to do a full burn-in over weeks to find those ICs which die prematurely so they could ask for a replacement in due time. It's not smart to put a newly built Apple-1 away after running it only briefly.

 

DRAMs are the second worst offenders, and they also like to die during burn-in. But other than these 2519 I saw DRAMs dying in each week of burn-in, so I keep a set of already burned-in ones ready to replace those without disturbing my one-IC-kit-per-week rhythm.

 

In my next post I will show you the true magnitude of these horrors !

 

 

 

 

 

 

Offline
Last seen: 6 days 4 hours ago
Joined: Apr 1 2020 - 16:46
Posts: 157
UNCLE BERNIES GRAVEYARD OF DEAD APPLE-1 ICs

 

Here is my "graveyard" which is located on top of the monitor of my "developer's" Apple-1:

 

 

And these are only those ICs who were bad out of the tube. I use the same table to "harvest" and replace the ICs from my burn-in rigs, this is why the bad ones end up there. The burn-in rigs have been moved from my lab upstairs to in front of the open fireplace downstairs, because there they can sit on fireproof bricks, just in case. There is another, smaller pile of ICs there which died during the burn-in, but sorry, no photo for them.

 

Some ICs have gotten dots with nail polish left to me by my ex wife (she took everything else of value). The number of dots allow me to discern them, they all have interesting ill effects, and maybe I find some time later to document these effects in photos.

 

If you look closely you can spot about 9 pcs of bad (but very expensive) 2519, about a dozen bad DRAMs, some 50 bad MM1404, and a few bad PIAs and bad TTLs, too.

 

And now the shocker: all this mayhem and all these dead ICs come from the production of just 25 of my 100% tested and burned-in IC kits !

 

You may now appreciate how much pain, suffering, desperation and monetary losses I save Apple-1 builders from by offering these 100% tested and burned-in IC kits.

 

If I would just sell IC kits (or single ICs) to Apple-1 builders with not testing and not burning-in these ICs to weed the bad ones out,  I'm certain all hell would break loose and I would have lots of unhappy customers who want their money back or threaten me with lawyers. This, Ladies and Gentlemen, may be the root cause why everyone, every single one, vendor of complete Apple-1 kits in the past has given up, except maybe that magical Unicorn, but they ring up very high prices for their "complete" Apple-1 kits (still with no PROMs ?)  and hide them very well on their website, so it's hard to order them from there.

 

I have come to the conclusion that selling untested ICs to Apple-1 builders is not a viable business, because it will cause frustrations and is a phone call generator. One of the companies I worked for as a chip designer had the rule: "Don't design a phone call generator." --- Because if you do, these phone calls will steal all the time you should spend on designing follow up products.

 

With the Apple-1, the two Steves in their garage had created such a phone call generator and worse, they had sold enough of them to have the phone ringing all the time. Which kept Woz - the only person at Apple who could answer these questions and help these disgruntled and/or desperate Apple-1 owners  - from working on the Apple-II. This is why the two Steves (?) decided to "buy back" the Apple-1 to destroy them. This was done in form of a voucher so a customer who turned in his Apple-1 for destruction would get a new Apple-II at a discounted price. And I bet that Steve Jobs even back then was too much of a savvy businessman to have given too much of a discount - meaning that they for sure still made a profit on these discounted Apple-II. But this is only my conjecture. The rest of this paragraph is how the "official" Apple story is being told, but rendered with my own words.

 

Before I did publish my reliability mods here in Applefritter the situation even was probably much worse:

 

I estimate that, prior to these mods, that  from each 3 Apple-1 clones ever built:

 

1/3 did not work at all.

1/3 did work very unreliably and did crash often.

1/3 worked somewhat OK and could run programs for prolonged periods of time.

 

It all depends on the particular combination of ICs , tolerances of timing components, and the type of bypass capacitors. It's a hit-and-miss. But by using a DS0025 clock driver from the original BOM, you could improve your chances towards getting your build into the last class, which worked somewhat OK. But DS0025 are very hard to find, and if you can find them, they are heinously expensive: Rochester Electronics, a chip broker, wants close to $50 each. The DS0026 which is plug-in compatible but has faster rise and fall times (which causes huge current spikes within the 2504/1404 shift registers, via various capacitances of the MOSFETs therein)  is much more abundant and cheaper, but I think no Apple-1 build using a DS0026 will work robustly unless my reliability mods were put in. With my mods, even an Apple-1 using a DS0026 works perfectly robust.

 

Here is the thread with these mods. The picture showing them is in post #4:

 

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

 

So far today's contribution of mine to the "CHAMBER OF HORRORS" thread here on Applefritter. The title chosen by Macintosh_nik is what the translation software english->russian->english made out of my original idea, sorry, it's too late to fix it. A "chamber of horrors" or "cabinet of horrors" is a secluded room (i.e. in a University teaching medical doctors) in which they show terrible mostrosities, tumors, etc., all preserved in Formaldehyde. But most industrial companies had such "chamber of horrors", too, full of specimen of parts which had failed in the field, and caused angry (or mutilated or dead) customers, to train young engineers to avoid such bad designs. All this is gone now, due to the bean counters, and due to Globalisation, and so bad designs are coming back to haunt the customers (and the manufacturers).

 

A little personal sidenote about "bad designs" and why nowadays all fields of industry suck more or less:

 

 I'm glad to be retired so I can't design horrible ICs anymore which come back to haunt me. Actually, this never happened to me ... all the 40+ years ... my own designs always were based on paranoia against possible process flaws and they were 100% robust ... but occasionally I had to fix bad designs of others (most of those designers had left the company in question before their SHTF) and, once, my brilliant analysis of a field failure and my proposed fix drew the wrath of a narcisstic psychopathic superior, which cost me my career at that company. Because he himself was the "complete moron" or "cunning saboteur" who had meddled with that circuit design and he had forced  these designers to put his "pet" clamp circuit in, which then later caused field failures (a bandgap reference would not start under certain rare power cycling / ramp up conditions, and then got stuck there, no further power cycles would help, a very nasty effect, you had to wait for 10-30 minutes powered off until the charge on an internal capacitor leaked away enough to allow for another startup attempt, the high quality of today's wafers contributing to the slow decay of that charge).

This director is another example of a lousy engineer who has turned to the "management" career ladder because he knew he's a bad designer, but then, of course, he used his position of power to incompetently meddle with designs again, each circuit block done under his rule had to have his special signature "mark of the beast" in it, even if it was a superflous addition. And, of course, he was not fired even after an ISO9001 auditor exposed the utter incompetence and failure of that psychopath even  in management tasks ... he consequently was promoted, IIRC to the level of vice president. I wished they had made him CTO instead, such that this obnoxious vulture of a semiconductor outfit would go under even quicker ! 

(no, it's not TI, but another very, very big outfit which due to lack of in-house innovation continues to take over and strip mine smaller competitors, just to install their own sh*tty management methods and their Orwellian spyware on all the computers. And stock options only for their management vermin, not for us chip designers who do the actual productive work and the  innovation. Suckers ! I'm glad that I'm out of there.)

 

The global problem, as I see it, is that such effects and mechanisms and "power play" by managers is now everywhere, pervades every company, and the larger the company is, the worse it gets. Back in the day when brilliant engineers seeing such rot and management incompetence at their employer could just walk out, find venture capital, and start their own company where everything was done right, no manager vermin infesting the place yet. Some of the finest and most successful companies came out of that effect, and they are legendary now, but most did not survive as an independent company, and  got swallowed by these much larger vultures mentioned above. How can that happen to a company making stellar products ? The reason is very easy: once a company exceeds a certain size, more managers need to be drawn in to run it. And these come from the outside, often leaving behind a lot of stink at the companies they came from. These new hires more often than not turn into "management vermin" again infesting the place, but the founders in the meanwhile, over the decades,  got too rich and too old to care about that infestation, the same type of infestation they once had walked out from. So, like termites eating your house, rot sets in within the company, and at some point the company is weakened enough to be a takeover candidate for the big ones. Which, of course, only will make matters worse. But they will strip-mine all the assets and all the patents and the technology. And they will drive out all the brilliant but too expensive engineers ... if these can find new employers, or are wealthy enough to retire.

 

This is how companies rise and fall and get eaten by the vultures. The problem today is much worse, because most hi-tech is too expensive nowadays for startups. For instance, in the semiconductor industry of the early 1980s, you could set up a wafer fab using the newest equipment for somewhere between 10-25 Million USD of capital invested. And then you had a very competitive process technology. Nowadays, you need 1-2 Billion USD to do the same. In the 1980s, a single IC designer could design a new IC. More complex ICs, like a microprocessor, needed small teams of maybe half a dozen designers. Nowadays, to really design the 100's of millions of transistors that go into any leading edge IC, you need hundreds if not thousands of IC designers to get the work done in a useful period of time. Hope you can see the problem ! This is why I sold all my semiconductor stocks I had accumulated over the decades from company stock options. The semiconductor industry is now mature, much like the car industry, or the aircraft industry, or the food industry, and the good days are over, and the products suck.

 

 

 

Offline
Last seen: 9 hours 9 min ago
Joined: Apr 9 2021 - 04:31
Posts: 26
You should send your

You should send your graveyard to one of the chinese fake chip refurbish processors, they could make some nice sets for none working showcase boards.

Seriously some people might be interested in such a set, but they should be deeple marked on the bottom as none working chips.

Offline
Last seen: 6 days 4 hours ago
Joined: Apr 1 2020 - 16:46
Posts: 157
Another bad MM1404 shift register ....

... this one is very peculiar. It almost works but there seems to be a pattern sensitivity, so some characters are wrong:

 

 

The above is the output of the hidden burn-in test page in my A1,A2 PROMs featured here:

 

https://www.applefritter.com/content/uncle-bernies-burn-proms-and-gimmick-switch

 

The takeaway from this is that you really need to watch this screen output closely and not just be happy if you see a marching pattern that seems to be almost right. These pattern sensitive errors normally are not easy to spot (this one is easy !) and they also might happen somewhat randomly, sometimes they manifest (i.e. warm ICs) and sometimes they may not appear (cool ICs, such as after startup).

 

All these 1Kx4 dynamic shift registers are typical mee-too products of the very early 1970s and the companies who were making them wanted to make a quick buck and still sell them cheap.  I don't know which semiconductor outfit made them first, but is certainly wasn't a huge design effort. Their inner circuit is so simple and regular that if you have a modern CAD workstation running Cadence and you can call up library cells for the I/O pads, you probably could do the complete circuit design, layout and simulation of such a chip in one day. In the early 1970s it was not so quick as they had no circuit simulator, no library cells, and no powerful layout tools. Still, they already had the first digitizers and could tell the CAD tool to put one cell down many times in the layout once the hand-drawn cell was digitized.  So, even back then it probably was the quickest and still useful chip you could design. Of course it was PMOS. They had no NMOS process yet, which explains the weird supply voltages. I really wonder where the "huge" market for these stupid shift registers could have been. Being bit serial devices they make very slow RAMs if used as such. Think of a semiconductor "magnetic drum memory" which early 1950s computers had. Some early electronic desktop calculators also used serial memory of a few hundred bits, but this typically was  an acoustic delay line made out of a special wire bent into a spiral. Other than the Apple-1 and the early TV typewriter published by Don Lancaster in the September 1973 issue of "Radio Electronics" magazine I don't know of any applications for these serial shift registers. Maybe that's the reason why we can still buy them today ... all the unsold ones went to the chip brokers.

Offline
Last seen: 6 days 4 hours ago
Joined: Apr 1 2020 - 16:46
Posts: 157
And another type of Symptom:

this is caused by a 7404 (half of which are for the clock generator) and the clock is faulty, mainfesting itself in this garbled screen output:

 

 

This happend on the third day of burn-in. (And another IC bites the dust ...)

Log in or register to post comments