Challenging Woodstock Repair



#9

I've got 3 Woodstocks (a 22, a 25 and a 27) all with the same problem. Their RAM chips (RAM/ROM in the case of the 27) don't work until they've heated up a bit.

The problem manifests itself as follows: In the case of the 22 and 25, the CPU chip works fine (i.e., you can do any command that just uses the stack) but the if you do any command that accesses the external memory chip you either read a zero or a crazy number, a write will do nothing. On the 27 the whole machine is locked up until it's warm, because the RAM and ROM are both on the same external memory chip.

Holding the circuit board near a light bulb for a couple of minutes will produce enough warmth to fix the chip temporarily -- but you need to power reset it to get it going. Leaving the machines on for about 15 minutes in whatever state they come up in works too - with the same power reset needed. The real test, however, is to heat the memory chip directly with a point heat source (soldering iron for example) applied directly to an unused pin (there are lots of them on the RAM/ROM chips) and within a second or so, you can reset the calculator to normal. In all cases the calculator performs normally (including off/on cycles) until the chip cools back down to room temperature.

I've tried bypass caps and assorted other obvious remedies with no luck. So, can any of you PhD EE's out there tell me what's going on inside these chips and how can the problem be corrected without my having to move to Death Valley?


#10

Hey, are there old cathode ray tubes used in those units ?? Then you really got an oldie!

I have an old radio (from '51) which also needs to warm up for about 20 minutes before it works.

Menno


#11

I know this is not exactly a helpful tangent we're on, but I also had an old walkman (no, no vacuum tubes in it) that, if left unused for a long time, needed like 20 minutes before its sound came on slowly.


Viktor

#12

Just guessing: There are very slow chemical and electrochemical processes in an integrated circuit (oxidization, electromigration, and others) that may affect the different layers of the IC. I don't know any detail about them, but have read, for instance, that aluminum IC features tend to develop sort of microcracks due to chemical issues and also to the voltage applied, heat, etc. Keep in mind that such layers are vapor-deposited on a chip during manufacturing to serve as contact point to the outside.

Of course, this kind of processes may or may not have to do with the problem described, but 25 years may be enough time for solid-state devices to show that "solid" is a relative concept. These RAM chips are some of the first digital IC products of such years, I understand that current manufacturing technologies allow for more reliable chips on the long term.

If there is anyone here related to chip manufacturing and reliability (perhaps a Chemical PhD rather than a EE), I'm sure his/her opinion can offer some interesting light about this.

"Practical" suggestion: Set a "heat pipe" from the big resistor on the side of the PC board, connect good rechargeables inside, and run your Woodstocks from external power. I remember the warmth from the side of my (missed) HP 25... :-)


#13

It may be that some chip (with age or some other damage) has lost the ability to switch fast enough to keep up with the clock in the calculator.

The optins are to slightly increase the supply voltage (this tends to allow devices to switch faster) or to reduce the clock speed.

I would tend to suggest the former option because it's less dangerous.

If you can find the part of the calculator that generates the clock, you should be able to resuce the clock speed by increacing the value of a capacitor. This is relatively easy because you can place another one in parallel with the one that's there.

I have no ides where to find the capacitor, what value you should use (but I'd suggest using the same as the original - which will double the capacitance and halve the speed) or whether this could help. But it's an option.

#14

I have an HP-27 with exactly the same problem. You tried to help me with this problem earlier by having me replace the filter caps, but it didn't work. I have found that the best way to make my calculator work is to put it on the charger for a while, turn it on, and leave it on for about 1/2 hour.

BTW: Thanks for trying to help earlier.


#15

I've had some success in fixing the 27. Thanks in part to Steve's suggestion of trying to slow down the clock.

First, there are two clock signals on the 27, one is approximately 180Khz the other is about 1.6KHz. Both seem to be internally generated on the chips and brought out to the rest of the circuit. None of the capacitors are used in RC oscillators other than in the power supply inverter (which runs at about 93KHz, BTW).

What I found was that when the RAM/ROM worked, the 1.6KHz line was really 1.6KHz. But when it was not working it was exactly twice this frequency (i.e., 3.2Khz)! I isolated the clock line and found that the CPU chip is the one that generates this clock and outputs it on pin 20.

For some reason that I don't understand, depending on the load on this pin it runs at 1.6KHz or 3.2Khz. I found that if I resistively couple this clock line to the rest of the circuit I can get it to run reliably at 1.6Khz.

What I did was cut the trace from pin 20 to the rest of the circuit and inserted a 20K resistor. The calculator now runs perfectly regardless of the ambient temperature (well, within reason at any rate).

Thanks to all of you for your input on this and I'll keep you posted on my 22 and 25 repairs. Victor (Rollinger), if I find that the same problem exists on the 22 and 25 I'll know that it's likely to be your problem too and tell you exactly how to fix it.

-Katie


#16

Congratulations for a very good work and for posting the detailed solution here !!
I am glad that my guess about ICs with internal "aging" was wrong, it would have been almost impossible to repair.

#17

I've now managed to repair the broken 22 and 25 that I have in addition to the 27. In both the 22 and the 25 I found that the problem was noisy clock/data signals. The noise quiets down after the calculators have been on for a few minutes and they start working fine (it's sort of interesting to watch this pretty abrupt shift on a 'scope).

On the 22 I found that the 175Khz-clock signal was coming through the unused pins. And did the following fix: I simply connected one of the unused pins on the RAM chip to ground. (At least some of the unused pins appear to be connected to the substrate of the chip.) In particular, I connected a jumper between pin 7 (ground) and pin 10.

On the 25 this repair did not work, although it too had clock noise coming through on the same unused pins. However, I found that the data signal line was quite noisy too. The fix here was to install a 20K resistor between the data line and ground. The data line is pin 11 of the CPU chip and there is a convenient ground just across the chip on pin 12.

In the process of all this work, I found that my earlier assumption about clock generation was not completely correct. There is a sort of external RC circuit governing the speed of the on-ship (CPU) oscillator. It's a 10 ohm resistor shunted across a rather strange diode-looking device (a bilateral trigger diode?). The diode-like device has around 300pf capacitance, which makes no sense given the resonant frequency of around 700Khz and the 10 ohm shunt resistance, but that's what it measures! Anyway, you can slow down the clock with a parallel capacitance (300 pf will cut the frequency in half).
The CPU divided the 700Khz oscillator by 4 and generates the 175Khz clock and by something around 100 to generate the lower frequency clock. (BTW, slowing down the clock was not a fix for either the 22 or the 25.)

The lower frequency clock seems to be the main RAM/ROM access clock (very, very low given that these chips are bit serial address and data circuits) the higher frequency clock seems to be needed for the display driver chips. Why it also connects to the RAM/ROM chips is a mystery to me, unless these are dynamic RAM chips and need that clock for internal refresh purposes.

Anyway, I hope that some of this will help in your repairs. One of these days, I suppose that I should get around to drawing a schematic of the Woodstocks now that I've at least got a good idea of many of the pin functions.


#18

About that RC circuit in Woodstocks... it's not an RC circuit.

It is an LC circuit with a 330 pF capacitor (yes, it's a capacitor, even if it looks diode-like) and a 140 uH inductor. Or something close anyway; these values are from the HP-97 Service Manual (available on the MoHPC CD set right here) but I found identical values in other Woodstocks I repaired.

I also found that you cannot adjust the frequency of this LC circuit arbitrarily. If you try to make it too low by adding a larger capacitor, for instance, the oscillation simply stops.


Viktor


#19

Viktor, which makes perfect sense and explains why I couldn't decode the color-coding bands on the inductor. One of these days I've got to get an inductance (RLC) meter so that I can figure out what these parts are. I agree totally with your statement about the limited range of oscillation in this LC circuit. I was able to get it down to around 300Khz (w/ parallel capacitance) and up to around 820Khz (w/ series capacitance) but that was it -- it quits outsides of that range. BTW, do you know if static or dynamic RAM is used in the Woodstocks?


#20

Those frequency values are consistent with my experience. I should also add that near the limits, the waveform gets increasingly distorted, and the calculator no longer operates reliably.

As for the RAM, I don't _know_ for sure, but I think it's a fair bet that they're static. They're obviously static in the C-MOS version, and even if they were dynamic in the N-MOS variety, that'd be of little concern to us, since any refresh circuit would necessarily have to be on the chip itself; the calculator does not provide any refresh mechanism. So for all practical intents and purposes, the RAMs are static.


Viktor

#21

I do not think I could add anything of value to your already wonderful work on Woodstocks, but let me just mention two details, obtained mostly from HP Journals of those years:

The issue about the HP 65 states the memories were recirculating shift registers. May some of this influence the IC design or operation of later models? I agree with Viktor: the continuous memory models mandates static circuitry.

On the question of the 1/100 clock signal: since the serial bus used a line for instructions and addresses, each cycle of this line had some idle time, followed by 10 instruction bits, a second idle phase, then 56 data bits, a third idle phase, and then the cycle starts again. Is it possible that the 1/100 clock was a SYNC signal and that the full ISA cycle was equal to 100 clock (bits) periods? On a BCD machine, a 100 divider "fits".

I apologize in advance for the vagueness of these comments, I am writing this from memory, recalling some experiments and measurements made 20 years ago...

#22

Tried it on a 27 that was taking close to 1 hour to come up on ac and then a while longer to work correctly, assuming I got the right location, and nothing, too bad, was a really neat job but it will be easy to reverse. Tried with a 22K ohm and then as that did not work put a 15K ohm in parallel with that for approx. 10K, nothing. I don't want to see how long it takes to come up, was hoping inmmediate results according to the fix. I'm assuming the pin 20 you mention is the third pin from the left on the top pin row of the bottom chip, then when looking at the board with the chips down, is the third pin from the right on the top row of that chip. The trace that was cut runs to a via. If I have the wrong spot, please let me know, Katie or others. As some of you know, I sold my scope recently, and can't look at the signal, but did verify the resistor implant by meter.
Regards, Frank


#23

Frank, You've got the right pin and it sounds like you did the repair as I described it. That *did* fix my 27, but not my 22 or 25. I posted another message about those repairs and have reason to believe that they may be even more likely to be your problem. (The reason being that noise problems seem to disappear more quickly with heat than the double-clocking problem. In fact, I think that the double clocking problem was due to a noise problem itself.) Both of those repairs are easier than the one you just tried and they are compatible with each other, so you can do both of them at the same time without caring about which one was actually needed.

The 22 and 25 have an extra ROM chip on the board, but the circuitry is identical to the 27 in the region of the repairs.

BTW, The fastest test that I use to check if a Woodstock (22,25,27,29) is functional; is to simply press the SIGMA+ button several times. This tests several RAM locations and a bunch of ROM too in addition to the basic functions. I press it several times, to check for stuck higher order bits (which I have seen). What do other people use as Woodstock self-tests?


#24

OK, thanks Katie, I will give those fixes a try as well

#25

This may sound obvious, but I'm just covering all bases.
You've checked the battery voltage? (Should be better than
2.5 volts with no load.)

Sometimes when a nicad is going, it will appear as almost
zero volts until it warms up for awhile. (I guess the
crystals that are forming in the cells are shorting things
out, and the heat might melt the crystals.) Since the
battery acts as a filter capacitor, this would probably
cause the behaviour that you see.

The other thing is the electrolytics. They can dry out.

I've often wondered too, if ROM's would eventually lose
their links like EPROMs. I don't think so as it's a more
permanent process I think. But EPROMs will eventually bleed
out and drop bits. I'd hate to think that ROMs might do
this too (and eventually leave all us collectors with
paperweights!).

Cheers.


Possibly Related Threads...
Thread Author Replies Views Last Post
  Usefulness of Woodstock Cs without their C Matt Agajanian 5 605 12-10-2013, 02:05 PM
Last Post: Geoff Quickfall
  Woodstock, Classic & Spices for iPhone Matt Agajanian 8 744 07-10-2013, 06:10 PM
Last Post: BShoring
  A Woodstock 67 Matt Agajanian 4 454 05-20-2012, 10:57 PM
Last Post: Mike Morrow
  Disassemble Battery Pack 82028B (woodstock) Jacques Karensky 0 257 05-19-2012, 03:39 AM
Last Post: Jacques Karensky
  Woodstock/Spice Keyboard Cleaning Revisited Matt Agajanian 0 253 05-12-2012, 03:43 PM
Last Post: Matt Agajanian
  Just a challenging hobby Matt Agajanian 0 161 05-07-2012, 04:35 PM
Last Post: Matt Agajanian
  Woodstock parts compatiability? Russell Dobson 7 619 03-28-2012, 05:35 AM
Last Post: Luiz C. Vieira (Brazil)
  Cleaning keys on a Woodstock (HP-27) Geir Isene 2 345 10-30-2011, 03:53 PM
Last Post: Geir Isene
  Woodstock LED question Mark Henderson 9 683 07-30-2011, 02:22 PM
Last Post: Mark Henderson
  Woodstock plastic yellowing db (martinez, ca.) 1 240 06-26-2011, 07:31 PM
Last Post: Ben Moss

Forum Jump: