HP-67 Reads cards but does not write

[Trevor Gale]

I have searched the forums about this but not found a similar thread, hence my question here.

I have amongst my HP calculators which I still use, an HP-67 which I went to use 2 or 3 weeks ago and found that it would not save a program I wrote on it. It simply returns the LED display "Error" after passing a card through the reader/writer slot.

It passes the card perfectly well, but has not actually written the program or anything on it - I checked this on my HP-97, as well as passing the card again through the '67 in 'run' mode to read it, which only returns the "Error" display again.

This HP-67 will read a card almost every time perfectly well - I have tried cards from an original HP application pack and also a card containing a program which I wrote and then saved on the HP-97, both of these are read perfectly by the HP-67.

There is no question of any "gummy wheel" problem in this HP-67, the battery is a new one correctly charged (checked in my electronics workshop); also I have tried this with many various cards (some unused, new, which I subsequently checked on my HP-97 and HP-41's), and I was not so stupid as to have the corner "write protect" cut away! All other functionality of this HP-67 works just fine.

Does anyone have any constructive ideas?

[teenix]

When a non-protected card is sent through the reader, the write protect switch operates before the head switch. For write protected card, the opposite happens. If the switches are not properly adjusted, then the head switch may operate first telling the CRC chip that the card is write protected and an error will be displayed.

If you have a multimeter and the circuit of the HP-67, then you can check the switch sequence by inserting a card into the slot with the calculator turned off. This takes a little force, so you can also remove the motor and drive shaft to enable the card to pass over the switches easier. When a switch closes, it is shorted to ground, or negative side of the battery.

When looking down at the calculator with the back cover and CPU board removed, with the connecting pins at the bottom, pin #1 is on the left. GND is pins 1 and 2. HDS is pin 14 and the WPS is pin 17.

The HP-97 service manual has procedures for adjusting the switches and is the same for the 67.

cheers

Tony

[Trevor Gale]

Thank you Tony,

I shall take the HP-67 back into my workshop this evening or tomorrow and examine those switch offset settings. I have a very well-equipped electronics workshop (also with mostly HP / Agilent test & measurement equipment!) and I have performed several repairs on my HP-41 card readers in the past, so I am fairly familiar with the procedure to dismantle the things, etc..

I have never come across this particular problem before, hence my original query - I had not thought of the pressure-switches, so I must be getting old!

I will tell of my results back here when I have had the time to examine the HP-67.

[John Garza (3665)]

And before you reach for a screwdriver, it's handy to check if it really is the switch with the Kemper correction card...

If it works, then you know it's the leaf switch.

-J

[Trevor Gale]

(02-10-2024 01:31 PM)teenix Wrote:  When a non-protected card is sent through the reader, the write protect switch operates before the head switch. For write protected card, the opposite happens. If the switches are not properly adjusted, then the head switch may operate first telling the CRC chip that the card is write protected and an error will be displayed.

If you have a multimeter and the circuit of the HP-67, then you can check the switch sequence by inserting a card into the slot with the calculator turned off. This takes a little force, so you can also remove the motor and drive shaft to enable the card to pass over the switches easier. When a switch closes, it is shorted to ground, or negative side of the battery.

Hi Tony,
The WPS 'finger' was indeed not making contact with the gold pad on the reverse side of the keyboard board. I traced the contact pad to pin 14 of the processor card which leads to pin 3 of the microprocessor chip.

I turned the WPS finger screw 1/4 turn clockwise, watching carefully with a card inserted and non-inserted. On re-assembly with all the self-tapping (black) screws back where they came from and checked again with a multimeter between 'ground' (-ve) and pin 3 and pin 5 (HDS) of the microprocessor chip and the contact are made as they should be.

With a fully-charged battery (no charger connected to the calculator of course) I inserted a card with a known working program on it - it was brought through the reader no problem, but now it gives an error on reading. I tried this with both a write-protected card and with an un-protected card, both with the same program. I checked the cards afterwards in my HP-97 and they read perfectly.

There is a speed-control preset potentiometer on the CDC card, and I set this 1/8 turn to the left and then 1/4 turn to the right, then back to the original position - this made no difference to the result but did change the speed of the card pass-through, as expected. I also checked the eccentric spindle for the feed gear and the card is pulled smoothly through the reader - although now there is sometimes a short extra motor-on time after the card has been passed.

As I mentioned, there is no question of a 'gummy wheel' or any gummy waste left lying around anywhere in the reader.

I have repaired quite a number of HP-41 card readers successfully, and also one of my HP-65's successfully which out of curiosity I wrote and read a program card using a card from the same batch of cards I have been using, also perfectly fine!

Would you have any further ideas please?

[John Garza (3665)]

While you're in there, you should replace the tantalums. One failure mode is the ability to write but not read.

-J

[teenix]

With the switch leaf contacts, they should all be above the circuit board by the same distance. It is easy to try and adjust them and they become uneven. I'm not sure if the CRC chip is upset by this and causes R/W errors.

The short motor on time is due to the software. This usually happens when an error occurs, and the motor can speed up and run for a short time just as the card has passed through. As this seems to have occurred after you tried adjusting the switches, I'm wondering if this is due to the way the switches are adjusted, maybe it comes back to CRC timing.

The capacitor that seems to give card problems is marked on the diagram. If you decide to change it, then as John kindly mentioned, you may as well do them all.

The R/W head might need a cleaning also.

A couple of assembly tips,

When reassembling, make sure the keyboard assembly sits properly in the calculator before tightening the 6 screws. Sometimes the black plastic frame is not sitting properly (can rock back and forward a bit) and warps when tightened. This may affect the card reader.

Check the wiring at the top of the keyboard. These can fray if the disassembled calculator is moved too much and may limit battery current. My Classic Notes, page 44.

Lastly, be careful with the Power and Prgm/Run switches. The golden contacts can fall out causing a short on reassembly and when the battery is inserted can destroy the display Anode or Cathode drivers. I found this out the hard way.

cheers

Tony

[Trevor Gale]

(02-13-2024 09:47 PM)teenix Wrote:  With the switch leaf contacts, they should all be above the circuit board by the same distance. It is easy to try and adjust them and they become uneven. I'm not sure if the CRC chip is upset by this and causes R/W errors.

The short motor on time is due to the software. This usually happens when an error occurs, and the motor can speed up and run for a short time just as the card has passed through. As this seems to have occurred after you tried adjusting the switches, I'm wondering if this is due to the way the switches are adjusted, maybe it comes back to CRC timing.

The capacitor that seems to give card problems is marked on the diagram. If you decide to change it, then as John kindly mentioned, you may as well do them all.

The R/W head might need a cleaning also.

A couple of assembly tips,

When reassembling, make sure the keyboard assembly sits properly in the calculator before tightening the 6 screws. Sometimes the black plastic frame is not sitting properly (can rock back and forward a bit) and warps when tightened. This may affect the card reader.

Check the wiring at the top of the keyboard. These can fray if the disassembled calculator is moved too much and may limit battery current. My Classic Notes, page 44.

Lastly, be careful with the Power and Prgm/Run switches. The golden contacts can fall out causing a short on reassembly and when the battery is inserted can destroy the display Anode or Cathode drivers. I found this out the hard way.

cheers

Tony

Thanks again, Tony, for spending the time with all my queries.

I checked the tantalums with an ESR meter I have (also measures ESR at high frequencies so it's not just a cheap 'Eastern' box!) and they were okay so I put them back in circuit (I have to re-order a batch of components to refill my workshop 'spares'). I also re-checked the 'goldfinger' (sorry for the pun, I couldn't resist!) contacts for height and operation and they are okay.

I re-adjusted the eccentric spindle just a little again, and now this HP-67 *WRITES* fine, every time, but it will not read!!! It will not even read the card that it has just written - but it did write the card perfectly (on 2 different cards I tried) as they are perfecly readable with my HP-97 machine, both cards.

It is a kind of puzzle I have not encountered in many years of experience with the HP-67, HP-41's that I have - I'll admit that I have not yet put an oscilloscope on the CRC chip to check data waveforms but I somehow doubt that that would give me a good clue as to what is happening... after all, if it writing okay then it should be able to read the card especially one that has just been written by the same head in the same machine under the same conditions... Any more ideas?

Kind greeetings - Trevor.

[teenix]

Maybe try changing the 47uF tantalum power supply cap anyway. It has a known issue with card reads.

HP stated that some sense amps run at marginal specifications and can cause read errors, (probably more so now that they have aged) so anything that improves the conditions that the sense chip operates in will help. The caps on the sense board may also need attention.

cheers

Tony

[John Garza (3665)]

Also, the 6.8uF tantalum at the top of the bridge board (near the sense amp chip) could cause this problem.

-J

[Trevor Gale]

My apologies for the 'long' delay in responding, however I have the 'flu or a bad cold... and waited for a (delayed) deivery of some extra workshop spares.

I replaced the two 22uF tantalum capacitors and the 6.8uF tantalum on the CRC card, and also replaced the 47uF and the 22uF capacitors in the mainboard supply.
I have double-checked the head connection wiring, checked for normal winding resistance, and for safety I have checked all solder joints on the CRC board and that it sits properly on the pin-points of the mainboard. Batteries (I have more than the one set) fully charged.

The result is that it began to write occasionally (i.e. readable on my HP-97 and an HP-67 from a collague) but still not read any cards; after a few passes it now no longer writes successfully (perhaps 1 in 20 times it will write correctly)... this is indeed a very frustrating fault to follow up.

I am not thinking of looking again at the motor working, clutch status, and anything else that comes to mind; having also checked how the card 'feels' though the path I detected no obstacle of change which was not normal...

[AndiGer]

IIRC sometimes it was reported the card reader doesn‘t work with battery fully charged but worked when low battery indicator began to light.
Can‘t remember if that issue was finally solved …
Andi

[Trevor Gale]

One further (surprising) advance: after trying a few things with my mates' HP-67 as well, it turns out that the entire pack of cards (new) that I have been using are far, far less reliable than another new pack I have just tried.

So now, my HP-67 that originally read but did not write, is now back to writing perfectly well all the time but will not read - not even cards that have been check-read by my mates' HP-67 and my own HP-97.

I am really at a loss, the first time I have ever been at a loss wuth these card readers...

[AndiGer]

You wrote you rotated the eccentric cam and after that it wrote but didn‘t read.
Have you tried re-adjusting it?

[Trevor Gale]

I have tried adjusting that eccentric pin, and I can feel / hear / see it makes a little difference, but apart from a series of settings where it refuses to write at all, the best position is where it will consistently write but still will not read...

[teenix]

It might be time to get the oscilloscope into action just to see the read signals going to the CRC chip on the RA RB pins.

There are some scope trace pictures in my Classic Notes starting on page 35.

cheers

Tony

[Trevor Gale]

Thank you again Tony, and yes, I also thought it was time to clear the bench and get this HP-67 connected to the oscilloscope. I made a number of other measurements as well.

Firstly, though - I see a difference in the circuit on my card-reader interface board between that in the MoHPC notes and what I have on both my and my friends' HP-67 - I have attached the circuit diagram of mine to this post (if I have done things correctly).
   
I did several tests with the calculator ready to read a (properly written with a short program) card each time whilst 'scoping the Hx head inputs and the corresponding Rx data outputs; the result was always the same - channel A supplied no RA data while channel B supplied RB data as expected and as shown in your excellent notes document.

'Scoping the head input of *both* channels gave me similar waveforms (+/- 200mV) to your example shown, this tells me that the heads appear to be functioning correctly.

However, what I assume to be the "bias levels" for the amplifier-comparators in the chip seem to be different: for the (working) channel B I find a bias on chip pin 9 of +1.41V which would be in agreement with the waveform crossovers you illustrate. However on channel A I find on chip pin 5 a level of +1.32V which is too low, and generates a constant "high" logic output on RA during a card read attempt.

An obvious test for the channel amplifier was to "persuade" the bias level for channel A to be higher, this was done using a 560K-ohm resistor to the nominal 6V line from pin 19 of the chip to pin 5 - which brought that level to near +1.43V, sufficient to bring it into the threshold region. Having done that, trying to read the card did generate an RA logic signal similar to channel B but with incorrect data which still returns an error (I could not expect to set this "bias" to the exact level required, and in any case doing this kind of modification will change the hysteresis points away from what they are intended to be).

So it is still an enigma, more so by the day...

[teenix]

I measured the sense pins 5 and 9 and got 1.45V on each.

If you have a storage CRO then you may be able to capture an image of the data being read from the faulty channel and compare that to the same data from a known good reader. You would only need maybe a 50mS sample so it's clearly visible for comparison. HP mentions +/- 5% motor variation and +/- 5% CPU clock variations is allowed so slightly altered bit width should be ok within reason.

As I understand it, the sense input is a 200 gain amp, some filtering, a comparator and latch to provide the logic interface. The resistor may not be doing the intended function as it is probably altering the input to an op amp rather than change the comparator trip point.

It is possible that the analog circuitry for that channel inside the sense chip is failing and cannot be fixed. As mentioned, HP said some sense chips that were manufactured were operating with marginal specifications, so if anything is failing inside, then it may just cease to function. In this case the only repair was replacement.

Have you tried replacing the motor capacitor?

cheers

Tony

[Trevor Gale]

Right! I do have both a digital storage 'scope and an analog storage 'scope (the analog one is a Tektronix!) - but before I coupled that up I thought that I would further inspect the interface board for any reason behing the differing bias levels for the two amplifier channels - after all, to change one of these levels only required a very high resistance... I wanted to double-check things like leakage resistance of the new capacitors and such.

I had replaced these tantalum electrolytics on the board of course, but here's the rub - as I normally do with new work these days, I use lead-free rosin-cored solder (in various guages, in this case very fine). Any rosin that is left over after the cleaning e.g. with isopropyl alcohol can present a high resistance - and with one of the capacitor connections being close to the CRC chip, this turned out to be the case.

This chip is soldered into the board with very little space between the bottom of its package and the board itself - so my short-brushing to clean away the rosin did not reach under the chip. I cut a small strip of paper card, soaked it in isopropyl alcohol, and managed to slip this in and out underneath the chip, rubbing away the left-over rosin. I could see that this was working by looking at the discolouration of the card when I extracted it from underneath the chip. I performed this with a second piece of card, and a third, at which point I detected no further discolouration on the card.

I re-situated the board in position, re-adjusted the eccentric spindle to what I thought was the last good working position (I had made a small mark on the side of the screw head), and placed a battery set to hold in position and turned on the calculator. That worked of course (I would be surprised if it had not);

Keeping the battery in as before, I measured the bias voltages - these had changed to 1.42V on BOTH channels, therefore I was perhaps on the right trail!

To check that it was still at least writing correctly I then wrote in a very short program and saved it to an empty card. This read perfectly well on my HP-97.

I then wrote a slightly different program with a different start label ("D" instead of "A") on the HP-97 and wrote that to a new card. I placed that card into my HP-67 and to my amazement (and glee!) it read in perfectly.

With this having been achieved, I took some of my earlier-written and more complex, longer programming already saved to cards and read many of these into the HP-67 with no problems - I checked the operation of the programs just to be sure - and I also re-wrote new empty cards with some of these programs (many using both sides of a card) in both directions between the HP-97 and this HP-67 - with again no problems with either function in either direction.

I am overjoyed to have found a solution to this problem, but at the same time I thought I would examing the possible results of using this lead-free solder in other high-impedance situations and compare it so some of the older Pb-Sn leaded solder I still have. I found that the residue which can be left by the rosin-cored lead-free solder can be an order of magnitude less in ohms than the flux residue left by lead-tin solder.

In my other occasional hobby interest, restoring vintage radios using valves (tubes in U.S. language) I also use lead-tin solder when making joints or replacing components, since it 'wets' the component leads and solder tags far easier and cleaner than does any lead-free silver-containing solder.

I suggest that as and when we have to repair boards or replace capacitors in our HP calculators then we do so using lead-based solder - if we can get some since it is banned from sale in some countries. The solder we would use and the calculators we have are not toys and any miniature amount of lead we will use is not going to be chewed at by young children.

In any case I would like to take the opportunity to thank everyone who made suggestions and/or ideas to assist me, especially Tony who clearly has a good in-depth knowledge of our classic and other HP calculator models. I wish you all a fine weekend.

[teenix]

Hi Trevor,

Your persistence may have provided some insight as to the repair of other card reader boards. In that they might simply need a good wash to restore them to a working state. Interesting.

I have found that cleaning flux residue with isopropyl alcohol can sometimes leave a white powder like substance, especially so if you clean with denatured alcohol. I use a small ultrasonic cleaner and usually give a warm detergent wash and rinse afterwards to help clean up any residue, although this doesn't guarantee cleanliness under some parts.

Nice work.

cheers

Tony