HP Forums

Full Version: A weird HP-67 card reader problem
You're currently viewing a stripped down version of our content. View the full version with proper formatting.

I have in my possession not one, but TWO HP-67 processor boards that share a common problem: attempting to read from a magnetic card fails under normal conditions, but it SUCCEEDS when the supply voltage is reduced to approx. 3.3-3.5V (at which point the calculator still operates, but the low power LED is already faintly lit.)

I have checked the obvious: I have tried the processor boards with "known good" readers and sense amplifier boards, and I have also REPLACED the CRC, to no avail. I have also tested the internal power supply voltages, and when I detected an abnormal ripple in the -12VDC supply on one of the boards, I replaced a capacitor to correct it. None of this solved my problem: both boards fail to read cards at a normal supply voltage.

To make things even weirder: both CPU boards can WRITE cards just fine; furthermore, the two boards are of different vintage, one having the 1818-0550/1818-0551 ROMs and two transistors in place of the 1818-0226/1818-0228 ROMs and the metal canned module on the other. Yet, their behavior is the same.

If anybody has encountered this problem and has some suggestions, that'd be much appreciated!

Viktor

You're right -- it's weird.... Firstly, do the output voltages of the switching converter change when you reduce the battery voltage? You can check them at pins 1 and 2 of the ACT?
How do they compare with the voltages in a working HP67? You've found one defective capacitor that caused ripple on the Vgg line on one board -- what about the other electrolytics on the board?
Ripple on the Vss (+6.25V) line, which is used by the card reader sense amp, will cause all sorts of problems.
Does the card move more slowly when the PSU voltage is reduced (or can the motor driver in the card reader sense amp cope)? If it does, maybe the ACT is clocking too slowly to read the cards at normal
speed. Check the frequency of the Phi_1 or Phi_2 clock signals.
I will now have to think some more....

Viktor:

PLEASE BE CAREFUL!!!

I had a 67 with this _very_same_problem_ and I was NOT careful, and ended up frying the calc's brains.

I never did solve this problem, except to have it overwhelmed by larger, more significant ones. Another reader of this forum now has my old calc, and after some major organ donations and transplants, I do believe it is back to this stage, with no obvious cause or fix in sight.

Good luck, and please do be careful!


Dan M.

Does your 67's have a pot for adjusting the motor speed. Perhaps the motor is passing the cards too quickly through the reader. You can readjust the pot or replace the resistor that sets the voltage across the card reader.

I should have mentioned that I also tried adjusting the motor speed, and it didn't make a difference. (My previous experience also shows that a healthy HP-67 is very tolerant with respect to the motor speed, even when the transport is slightly uneven. In this case, the transport of the card is smooth.)

And yes, Vss is also correct, within tolerance with or without the reduced supply voltage (the difference is only a few hundredths of Volts) and there's no appreciable ripple. Phi_1 and Phi_2 are also correct.

The only difference I see with an oscilloscope is a slight difference in the oscillator frequency in the DC-DC converter when I reduce the supply voltage, but my understanding is that it's entirely normal behavior and besides, that frequency does not appear anywhere outside the converter anyway.


Viktor

Well, I am usually rather careful, actually... so how, exactly, did you manage to fry your HP-67?


Viktor

I am not normally very careful, so now I try harder to be.

The death of this HP67 was done somewhere in my stupid act of manually making a poor connection (battery pack - calculator) in order to get the "low battery" light on so the card reader would work... really careless and foolish and I paid the price, as did the poor calculator.

On the good side, I got the 67 fairly inexpensively, and it was sold to me as "broken," in that the display was flashing and the thing wasn't working. It turns out that its problem was actually the wall wart power supply, and since it was in a security cradle without the key, nobody thought to open the thing up and try a different power supply. (Well, actually, somebody had thought of it but learned that a screwdriver will not pry an HP-67 out of a security cradle without destroying it!) Live and learn, I suppose! BTW, I found it much easier to open the security cradle by "picking" the lock with a paperclip than by using forcible methods. And a locksmith is able to inexpensively make a key to fit the cylinder.

Live and learn! And be careful!

Dan

Hello Viktor

Have you checked for a good ground connection between the reader and mother board. I have found that Ohm's law is normally the problem in situations that involve working at lower voltages. This manifests itself as a voltage drop when the motor turns on.

Have you monitored the current as you move the voltage above 3.5V. Many times, a damaged CMOS component will "turn-on" and act like a lazy ZENER regulator. You should detect this with a high resolution milliameter.

A friend of mine recently had a similar problem with an HP-65. With normal battery voltage the card reader would not feed cards properly because the motor driver in the sense amp chip (or the power supply) was going into current limit. It would work with lower voltage packs.

The solution was to solder a 1/2 or 1 ohm resistor (1/4 watt) in series with the reader motor.

Another thing to try is to replace the capacitor that is across the reader motor. I use a 5-10 uF tantalum cap. Make sure you get the polarity right. If this cap is flakey you can get noise feeding back into the sense amp chip that is dependent upon the supply voltage.

Well, I usually use either a regulated power supply or some series diodes to reduce the voltage to the desired level :-)

As for the HP security cradle, I have no experience with them, but I found it fairly easy to pick the lock on TI PC-100 printer cradles. If HP used a lock of similar quality, then I imagine picking the lock cannot be that difficult.


Viktor

Good suggestions, much appreciated. Unfortunately, I believe all connections are good, both ground and supply power.

As for the current, at 3.16V I measured 172mA (calculator on, 0.00 displayed); at 4.02V I measured 160mA. I.e., there's no sign of abnormal behavior at the higher voltage, but these two current readings appear to be a bit on the high side. I'll compare it against the readings on a known good machine. I did not (yet) measure the current during a card read.


Viktor

Another set of excellent suggestions, thanks! Sadly, they don't seem to solve my problem.

In my case, the card reader is feeding cards through fine at normal voltage (remember, it WRITES cards just fine) it just displays Error on a read attempt.

As for the capacitor, I measured and found no ripple on the motor contacts, so I believe the capacitor is good. Besides, the bad processor boards continue to malfunction even when used in conjunction with a "known good" reader.


Viktor

Yes, it's quite normal for the PSU oscillator frequency to vary with battery voltage.
The PSU circuit in just about all LED HP calculators uses a single transistor as an osicillator (with the little transformer on the board). A second transistor is used as a regulator -- the base is connected to the Vss line through a zener, the emitter is grounded, the collector goes to the base of the
oscillator transistor. When Vss rises too high, the zener turns the transistor on, reducing the bias on the oscillator. A simple circuit that works well.
The outputs from the PSU (in the HP67) are Vss (obtained by rectifying the pulses on one pin of the transformer and smoothing the result) and Vgg (obtained by a voltage doubler circuit off another pin on the transfomer).
Vss is normally around +6.25V, Vgg around -12V.
It is also quite normal for the machine to draw more current if the battery voltage drops -- the oscillator then has to 'work harder' to keep the output voltages correct.
I am assuming you know the fault is on the logic board -- that is to say that you've checked the boards on several known-good reader PCBs, and they fail on all readers. I think that pretty much eliminates the motor speed, the capacitor across the motor, the sense amp, etc.
I really would check all the electrolytics on the logic board, since they're the most likely thing to fail. Is it possible that the current spike when the motor starts up is putting a glitch on the Vbatt line (don't tell me you have a regulated PSU -- it will do nothing for this). There is
a capacitor across the battery lines on the logic board -- maybe that's gone low in value or high in ESR. Check it. I don't see how it can be one of the logic chips since the voltages across these are the same no matter what battery voltage you apply, so
the chips can't know the battery voltage has changed. The only chips with connections to Vbatt are the card reader sense amp and the display cathode driver, and neither are on the logic PCB.

If you havn't actually tried the resistor trick, I would highly recommned giving it a try. Before the '65 started stalling out, it gave the same symptoms (no read, would write).

The fact that you have two totally different CPU's giving the same problem is HIGHLY suspicious. Read but no write problems are almost always related to the analog circuitry on the reader card. All the signals to/from the reader circutry from the CPU card are digital in nature. My feeling is there is something right on the edge of working and some power supply or noise related problem is at the heart of it. If you swapped the "bad" CPUs between a few machines I bet something would start working somewhere.

The HP sense amp chip is quite a tricky piece of silicon. It has very low level signal amplifiers on the same chip as high current/noise motor driver circuitry. Keeping the two isolated from each other is a chore even today, much more so 25 years ago. I've found that weird read problems are almost always related to this.

I've indeed swapped the bad CPU PCBs around and they stay bad, with "known good" readers and reader PCBs (known good because they work fine with good CPU PCBs.)

In other words, I've isolated all variables external to the CPU PCB by cycling through several sense amplifier boards, card readers, even display/keyboard PCBs (on the odd chance that the problem might be due to a misbehaving display driver chip for instance), and the problem remains, and it's consistent. And yes, I too am dumbfounded by the fact that TWO CPU PCBs exhibit the same behavior, and they're not even of the same vintage!

I've also cycled through three different CRCs, all three of which working fine in good CPU PCBs. What remains, then, is:

1. The DC-DC converter (all supply voltages appear good on my scope)
2. The processor
3. The ROM chips
4. That extra support chip (is it an ACT? I never had an HP-67 service manual, so I'm only guessing on the basis of the HP-97 service manual) marked 1818-0268.

Maybe it's time to start swapping these chips from a known good CPU board, but right now all my good CPU boards are in perfectly good, working calculators, and I hate to mess with working calculators!


Viktor

Thanks for your valuable advice! I saw no signs that the capacitor across Vbatt would be the culprit, but I decided to try changing it anyway. Well, guess what... to my considerable surprise it seems to have solved the problem. I am going to try and see if this solves the issue on the second CPU PCB as well.

What is the nominal value of this capacitor? I tried to measure it, and my meter says it's around 50uF, which seems way too high. The highest valued tantalum I had in my toolbox is 15uF, and that's what I used as a replacement, and that seems to work fine.

Well, this was educational to say the least. Thanks again to everyone who tried to help!


Viktor

Yes, both boards work fine now that I replaced this particular capacitor. Wow.


Viktor

At last I guessed right :-). Mind you we'd virtually elminated everything else.
About the capacitor value. 50uF could be about right -- my notes say 47uF. And 22uF for the Vss smoothing cap, 2.2uF for the 2 caps in the Vgg voltage doubler. Since 15uF works in place of the original 47uF, and that the 47uF measured 50uF on your meter, it would
appear that the fault was not that the capacitance got too low, rather that the ESR got too high. This is a common failure mode for old electrolytic capacitors (of all types)

Although you've solved the problem, I might as well say what I think the chips on the HP67 logic board are :
The 22 pin chip is the ACT (Arithmetic, Control, and Timing). It's the processor. It's similar to the one in the 97, but it also has the keyboard column lines connected to it (no PIK in a 67...).
The 28 pin chip, as you know, is the CRC (Card Reader Controller). It's the only chip (AFAIK) that's also used in the 97.
The 8 pin chips are memories. I believe all of them are combined ROM and RAM, but I can't see any easy way to prove that.
The 18 pin chip is one of the ROMs (ROM0 I think) and the display decoder. It provides the segment drive lines to the display interface.
On the other boards, you know about the motor controller/sense amplifier on the card reader PCB. On the keyboard/display PCB, the 20 pin chip is the display cathode (digit) driver, and is HP custom.
The 16 pin chip is the only off-the-shelf part in the machine -- it's a 7-transistor array for driving the display segments and IIRC it's an RCA CA3082.
Oh yes, the transistors. The 4 on the keyboard/display PCB are the remaining segment/anode driver (there are 8 segments total -- the 7 in the '8' pattern and the decimal point)
and some 'logic' for the - signs (the cathode driver can drive 14 digits, there are 15 actual digits on the HP67, but in 2 of them only the - sign segment is used).
On the logic board, 2 of the transistors are for the PSU, the other 2 form a current mirror to set the display current.
No, I don;t have an HP67 service manual either, but it's not that hard to figure this out...

22-pin chip is the ACT... yes, that's right of course. Neglecting for a moment what the acronym ACT stands for, I temporarily labored under the assumption that we have a processor _and_ an ACT...

Incidentally, this same chip I believe is also used in the HP-25, which does in fact have separate RAM chips, i.e., the RAM is not integrated onto the processor, so I think that pretty much confirms your guess: one or more of the ROM chips is probably a combined ROM/RAM chip. I'd suspect it's the 18-pin chip, because all the 8-pin chips have numbers in the same series as the numbers of the ROMs in the HP-97... of course, numbers mean nothing, and besides, the HP-97 Service Manual could be lying about the function of those chips, too.

The 16-pin chip is a 7-transistor array? I didn't know that. Mind you, that wouldn't stop me from using a transistor to substitute for a broken segment driver... in fact, I've done that before, although not in HP machines. (I have, however, replaced fried 4-transistor arrays in an HP-91 with discrete transistors. Worked fine.)

I love these old machines. For one thing, they are actually repairable, unlike those present day machines that require dynamite for disassembly, only so that you can then find a superfine PCB with a single blob of an SMT chip in the middle...

Thanks again for all the help!


Viktor

Where exactly is this capacitor? Are there schematics available somewhere. You post seems to suggest there are.

Yes, the HP67 is electronically very close to the Woodstock series (and the Topcat for that matter) -- they all use the ACT processor.
I beleive there are several types of ACT (the one in the topcat series doesn't have the keyboard inputs (or at least those pins aren't wired) and there are serveral
type numbers used for the ACT. So you probably can't use chips from one machine to repair another :-(. But at least the pinout is similar (to the extent that if a pin is used it will always have the same signal on it, no matter what the machine is) so it's possible to
use information known about one machine to repair others.
I am almost certain that the 18 pin chip in the 67 is ROM and display decoder only -- no RAM. I don't think it connects to the Data line. The HP97 manual is misleading here -- in that machine, most of the 8 pin chips are
combined ROM/RAM, but there's one odd one (different pinout) that's ROM only. Or at least that's what the signals imply to me.
Yes, I like these older machines too -- they're a lot easier to work on than the modern stuff. Did you know that there are even testpoints in some
HP calculators that carry all the important signals. You can virtually determine which chip has failed in a 35 or 45 using these testpoints only...

Tony,

The reason I mentioned the HP-25 is that the chip identifier there is _exactly_ the same as that used in the HP-67: that is, I've seen 1820-1523 in both machines. Furthermore, I think I also saw this ACT in at least one HP-21, but I'm not 100% sure (I don't always make notes.)

Yes, I knew about the test points, but in the absence of a service manual, I find them to be of somewhat limited use, since I am relying on (educated) guesses, not certain knowledge. (For instance, if I see a signal that looks suspicious, in the absence of a service manual, I might not know what the correct signal should look like.) Which is why I hope that one of these days, someone might be able to contribute more service manuals (say, at least the HP-67 service manual and a Woodstock service manual, and perhaps the HP-65 as well) to a future Museum CD. In the meantime, I find the service manuals already present immensely valuable, because even when they're not for the exact model you're working on, there are many similarities.


Viktor

Mike,

I don't know where you can find the schematics for the HP-67 (if you do, I sure would like a copy!) but schmematics are available for the HP-97 on the Museum CD set, and you can deduce a lot about the HP-67 by observing the physical circuit and comparing it with the HP-97 schematics.

The capacitor in question is near the top of the processor board (i.e., on the side opposite to the row of connector holes), next to one of the transistors. If you check where it's connected to, you'll find one of its pins wired to ground, the other wired to Vbatt. This is the only large capacitor between the battery supply voltage and ground.


Viktor

I didn't know that the 67 and 25 use the same ACT (although it would be reasonable) -- the only Woodstock I've worked on seriously is the 21
(and some versions of that use a different ACT). FWIW, the chips in the 21 are the ACT (22 pins), ROM/anode driver (18 pins) and cathode driver (20 pins). Many of the signals are labelled in the PCB etch on some machines as well :-)
As regards testpoints, the HP67 and HP21 (at least) have 3 testpoints. On the HP21 they're the 3 unmarked holes at the front edge of the PCB. On the HP67 they're 3 of the pins on the connector between the keyboard/display PCB and the logic PCB (they have no traces going to them on the keyboard PCB). They are the Phi_1 clock, Sync and ISA (ROM address/instruction line) in some order. Those signals are
sufficiently different that it's easy to work out which is which.
On the HP35 amd 45 there are 12 testpoints along one edge of the PCB. From the display end of the machine they are : Vss (+6.25V); Gnd; Phi_1; Phi_2 (CPU clock); Vgg (-12V); IS (ROM data); IA (ROM address); Sync (cycle synchonisation); WS (Word Select); Init (reset); Carry; Data (seems to be a bit-serial version of the X register).
Viktor, if you e-mail me privately I can give you a little more info.