HP 34970a repair advice

[Antonio90]

Hello everyone.

Returning after a long hiatus from posting. I've kept reading and watching EEVblog videos, but work has been rather unforviging this year.

I've recently adquired an used HP 34970a, with the 20 channel MUX and the DAC modules. It is working nicely so far, except for the usual 202 and 204 errors from the dead battery. Thankfully, no leakage in the PCB AFAIK. I don't really need a DAQ, but I wanted a 6.5 digit DMM, and this was the only device close to that which did fit my budget.

I have no means to calibrate a device like this, not even close. Apart from a few 0.02 voltage reference ICs and 4.5 digit multimeters, if the cal constants disappear, the device is most likely toast for me.

So I started looking around for a i2c programmer, and found this https://whatever.sdfa3.org/hp-34970a-data-acquisition-unit-part-2.html and this https://btbm.ch/8-pin-i2c-memory-programmer-with-arduino-uno/ links. Bought a few FM24C04-B and tested writing and reading them successfully. So, today, I dumped the contents of the external modules, starting with the DAC, as it is the least useful part for me. I tried to do it in circuit, as I'm still a beginner soldering, and don't have a hot air station or a desoldering gun. The results are attached as .TXT files.

After the (apparent) success, I proceeded to open the instrument and remove the DMM module to dump its F-RAM. So far I was assuming every FRAM had the address "0" on their A1 and A2 pins, and had them tied to GND. That was a mistake, as the one in the DMM has it's A1 pin (or was it A2?) tied to VCC (EDIT: It was A1, its right in the service manual). So I shorted VCC to GND and almost killed my arduino. Thankfully it shut down, and the instrument seems to be working well, despite the fail.

Now, when I opened the case just before the FRAM snafu and after having the unit plugged in standby, I noticed the dmm shield was rather hot, also, the plug-in modules and the transformer were warm. I don't have a mains power meter, but I measured the standby current into the unit and it was north of 40mA. That's 9 Watts, and seems a bit too high for a standby current, isn't it? I don't really know though, and haven't been able to find the standby power specified in the service manual or anywhere else for that matter. The transformer hums too while the unit is on standby, but I can't hear it over the fan while powered on, as it's quite low. I have it on 240 volts right now, but just measured the line voltage and it's more like 225-227V. It heats too on the 220V mode.
I attach a photo of the transformer, which seems to have some kind of burnt gunk in a thin line over the coils.

Now, I have to dump the main unit F-RAM, replace the battery (I might put a socket for a CR2032), and track down the idle power draw, if it's a problem. Maybe replace the fan too, but the unit wasn't used very much and it seems to be in good shape.

So, a few questions:

- Could someone take a look at the contents of the F-RAMs to see if they look right?
- I don't have a hot air station, is it a good idea to use low melting point solder with indium to replace the tiny SOIC8 FRAMs without overheating the PCB?
- Any ideas on the idle power draw? Whether it is normal or not, and any tip you could have to start troubleshooting.

Thank you very much.

Regards, Antonio.

[Kleinstein]

The no load current of a classical transformer has quite some phase shift. So the 9 VA measurend are considerable less than 9 W.  For a transformer of this size and low load 9 VA sound reasonable. So there may not be anthing wrong with this.

Why would you want to change the FRAM if the battery is empty ?  Normally FRAM is quite reliable and no real need to change.

For just a SOIC8 part, there is no real need for hot air. A conventional iron with a suitable tip (e.g. wide enough to heat up all 4 pin of one side at a time) is good enough. If one can get a small quantity a lower melting point solde could be a good idea, but no absolutely needed. Leaded solder is alread OK.

[Antonio90]

The no load current of a classical transformer has quite some phase shift. So the 9 VA measurend are considerable less than 9 W.  For a transformer of this size and low load 9 VA sound reasonable. So there may not be anthing wrong with this.

Why would you want to change the FRAM if the battery is empty ?  Normally FRAM is quite reliable and no real need to change.

For just a SOIC8 part, there is no real need for hot air. A conventional iron with a suitable tip (e.g. wide enough to heat up all 4 pin of one side at a time) is good enough. If one can get a small quantity a lower melting point solde could be a good idea, but no absolutely needed. Leaded solder is alread OK.

Oh well, I guess taking 9VA as 9W is really silly. Thank you very much for the correction.

These FRAMs are not so reliable, and they are written every time a relay actuation happens. There are a lot of HP 34970a with dead FRAMs which translates to 74X errors and lost of cal, as they also store the calibration constants.
According to Free Electron on this forum they always die. They changed them in later revisions, but I think it is a good idea to replace them with new ones that meet their reliability specs.

The battery just powers the RTC and keeps the stored readings, which I don't really need right now, but it's long past its time and could leak all over the place.


Thank you!

[gamalot]

Don't worry about the standby current, all three of my 34970A consume more than 60mA when turned off.

[Antonio90]

Thanks! It only worried me because the DMM shield was significantly hot. Not too hot to keep the finger on it though.

[Kleinstein]

I would expect the writing after relay action to only effect the MUX cards, not the main DVM part. The mux cards should not have much cal data - if at all some offsets, and this part of calibration does not need very special gear - just some shorts.
It would be a kind of fail if the FRAM to count relay actions has less endurance than the relays itself.

If you have a update of the FRAM contend to write it to a new one, this should be good enough. One could still replace the FRAM if it really dies. I see no reason to take the risk already now. AT could be OK to get a new chip, as long as they are still easily available.

[Dr. Frank]

Hello,
if the FRAM is still ok (no calibration error message), then please leave it as it is.. especially if you are not familiar with SMD soldering.
A calibration is probably also not necessary, because the 34970A and its parent DMM, the 34401A, are very stable over many years.
If you are able to read the FRAM in situ, go ahead, to save its content for a possible later failure.
The original supplier was Seiko, please search the eevblog forum, which was known for data losses.
So I also ordered FRAMs from another supplier, but will change only, when the old one fails (20 years old).
The other FRAMs on the scanner cards also do not pose a problem, because they do not suffer from wear-out.

The battery only supplies the clock, and several SRAMs, which contain setup and logging data. That has nothing to do with calibration data.
Therefore, only replace the lithium cell. Better use the original type as a battery holder is difficult to mount

It's normal that the 34970A draws exactly 9W power in off mode, because most of the circuits are running when the DMM is switched off, please have a look in the schematics.. that has the advantage that no warmup time is needed. In power on mode, my unit draws about 13W.
(The explanation of this other guy about the transformer with no load is simply bullshit).   

The thermal design of this instrument is very poor. One of the voltage regulators is getting quite warm, and it is sitting directly under the analogue PCB, heating the whole metal shield of the DMM PCB, and also all analogue components, making the DMM less stable than the 34401A (which is very similar) .

The air stream of the fan is  blocked by the power plug, so that's not very efficient in cooling, and this blocking creates noise. W/o the plug, the fan is silent, so no use to replace it.

Frank 

[Antonio90]

Hello,
if the FRAM is still ok (no calibration error message), then please leave it as it is.. especially if you are not familiar with SMD soldering.
A calibration is probably also not necessary, because the 34970A and its parent DMM, the 34401A, are very stable over many years.
If you are able to read the FRAM in situ, go ahead, to save its content for a possible later failure.
The original supplier was Seiko, please search the eevblog forum, which was known for data losses.
So I also ordered FRAMs from another supplier, but will change only, when the old one fails (20 years old).

The battery only supplies the clock, and several SRAMs, which contain setup and logging data. That has nothing to do with calibration data.
Therefore, only replace the lithium cell.

It's normal that the 34970A draws exactly 9W power in off mode, because most of the circuits are running when the DMM is switched off, please have a look in the schematics.. that has the advantage that no warmup time is needed. In power on mode, my unit draws about 13W.
(The explanation of this other guy about the transformer with no load is simply bullshit).   

The thermal design of this instrument is very poor. One of the voltage regulators is getting quite warm, and it is sitting directly under the analogue PCB, heating the whole metal shield of the DMM PCB, and also all analogue components, making the DMM less stable than the 34401A (which is very similar) .

The air stream of the fan is  blocked by the power plug, so that's not very efficient in cooling, and this blocking creates noise. W/o the plug, the fan is silent, so no use to replace it.

Frank 

This is awesome info. Thank you very much. I have read the posts in the forum and a couple of blogs on the matter, that's why I wanted to backup the cal data from the DMM. The rest of the backups were a test to see if my setup is working, as I read that the FRAMs can be reinitialized (with the DMM unit unplugged) and don't have critical data bar the cal constants (only on the DMM module).

Thank you for the info on the power consumption and the voltage regulator dissipating heat. I knew most supply rails were on when in standby as it's in the service manual, but I should really have had a better look at the schematics. I have little time for the hobby, and the hurry to make the most out of the time I can spare often gets the best of me. The "on" current draw was a little bit more than 60mA, so everything seems to be alright in that respect.

I have a few Cypress semiconductor FM24C04-B 4Kb FRAMs, which are a drop-in direct replacement AFAIK, but will certainly follow your recommendation of leaving it alone if it's working, especially provided that my soldering skills are not up to the task. I also read about the lithium battery function and replacement procedure, and will replace it ASAP.

Thank you very much.

[razvan784]

Regarding the battery, I chose to eliminate it completely in my instrument (which was bought as defective and needed cleaning and repair).
The instrument will not boot without a battery because the voltage supervisor IC needs to see a voltage that resembles a battery. I fixed the issue by replacing the battery with a 13k resistor that I had on hand. It forms a voltage divider with a resistor and diode that are already installed between +5V and the battery positive, ensuring the supervisor IC is happy.

[Antonio90]

The resistor solution seems interesting, and I might do it later on if I find I don't need any of the features backed up by the battery. But right now I put a CR2302 holder and a new battery, and it's working alright. There seems to be a couple of plated holes with the right spacing for the job and connected to the original battery pins, so it was a rather simple affair.

I thought about buying a similar battery to the one that was installed, but I just made an order to Mouser, with all the bits and pieces I could think of to reach the free shipping, but forgot the battery, and I'm not paying 25 bucks for it.

Unfortunately, I didn't manage to dump the FRAM contents, it takes about 5 seconds to start reading, then freezes after one page and a half. I'll look on the schematics to see what could be the problem, but I'm not keen on trying too hard. Otherwise, I'll just replace it when it dies and reinitialize it. Or ask around for a known working image.



Thank you all for the help.

[Dr. Frank]

I definitely recommend to replace the battery!
At first, otherwise you'll always get annoying error messages when powering on.
The mode and state selection is very clumsy, due to the serial menu, and due to the lack of mode buttons.
I think, w/o a battery you even have to enable the DMM each time [shift utility => ..... DMM enabled / disabled].

The storage of the last chosen state of the instrument is a very comfortable feature, especially compared to the 34401A, where you always have to call the menu several times, just to select 6 1/2 digit, Slow and R_in > 10¹⁰ Ohm.

W/o battery, the use as a DAQ is practically not feasible, as you need correct Date and Time for data logging.

The Li cell ( or a similar one) is also offered by other distributors.

Concerning the FRAM, please identify which supplier was used.. maybe you have a younger unit than mine, which already has Cypress or RAMTRON inside, which are not known for failures, like Seiko.
Also, please identify the firmware version, [shift utility => ...... REV XX-01-02].
Initially my unit had old Rev 08, which had a lot of  bugs, giving severe errors in the Ohm mode for some ranges.
I burnt a new OTP with latest FW 13, and now everything works great.

Frank

[Antonio90]

I definitely recommend to replace the battery!
At first, otherwise you'll always get annoying error messages when powering on.
The mode and state selection is very clumsy, due to the serial menu, and due to the lack of mode buttons.
I think, w/o a battery you even have to enable the DMM each time [shift utility => ..... DMM enabled / disabled].

The storage of the last chosen state of the instrument is a very comfortable feature, especially compared to the 34401A, where you always have to call the menu several times, just to select 6 1/2 digit, Slow and R_in > 10¹⁰ Ohm.

W/o battery, the use as a DAQ is practically not feasible, as you need correct Date and Time for data logging.

Hello Frank,

Well, you are right, it doesn't make sense to skip the battery replacement, being an easy repair as it is. I don't need a full featured DAQ, but just as a logging multimeter, the date and time can be crucial data.

Concerning the FRAM, please identify which supplier was used.. maybe you have a younger unit than mine, which already has Cypress or RAMTRON inside, which are not known for failures, like Seiko.
Also, please identify the firmware version, [shift utility => ...... REV XX-01-02].
Initially my unit had old Rev 08, which had a lot of  bugs, giving severe errors in the Ohm mode for some ranges.
I burnt a new OTP with latest FW 13, and now everything works great.

Frank

Mine is also REV 08, i'm afraid, and most date codes point to 1999, so it's 20-22 years old. My FRAMS read the usual: RIC 24C04-S, but the letters are red, or brown or something (I'm a bit colorblind) instead of the usual white ones I've seen in previous posts, see photo attached. I really don't know where to find who's the manufacturer.

Also, excuse my ignorance, but what is OTP? Also, I haven't been able to find any procedure to update the firmware in the service manual. Is there any resource with a procedure to do it?

EDIT: I have found the service notes for the firmware problems in revs 1-9. The usefulness of the search feature is awesome 
It seems that the firmware is not publicly available though, and you need to buy the ROM with it preinstalled?

Thank you very much for your help, really appreciate it.

Antonio.

[Dr. Frank]

Hello Antonio,
no problem, that's the purpose of this forum, I guess.

OTP = One Time Programmable (Device), in this case it's an EPROM, Erasable Programmable Read Only Memory, w/o a glass window for erasing, i.e. a PROM.

The square IC in the socket in your picture is the PROM of the 34901 scanner card, U101.
The main PCB A1 contains two OTPs, U305, which is a one time programmable µC (I/O controller, 80C51 derivative, I guess).
The other one, U401, is the PROM you need to replace, look for its sticker with 'Rev. 8'.

I append a photo of the PROM in question, it was V 9, on the left, and the PROM type from ATMEL which I used to burn the V13 inside. A friendly volt-nuts colleague had sent me his newer instrument for downloading the content of its firmware.

That is the only way to update the firmware, as this design from about 1995 or so had no FLASH memory yet.

My instrument from 2000 also has this 'RIC' designator on the FRAMs, so I also assume it is also affected.
There is one FRAM on the main PCB, and one on the DMM PCB, A4, which contains the calibration constants .. maybe the data loss only occurred for a certain batch from Seiko / RICOH only.. I can't find any further details in the forum, or in the Service Note of hp.

Frank 

[Antonio90]

Hello Antonio,
no problem, that's the purpose of this forum, I guess.

OTP = One Time Programmable (Device), in this case it's an EPROM, Erasable Programmable Read Only Memory, w/o a glass windows for erasing, i.e. a PROM.

The square IC in the socket in your picture is the PROM of the 34901 scanner card, U101.
The main PCB A1 contains two OTPs, U305, which is a one time programmable µC.
The other one, U401, is the PROM you need to replace, look for its sticker with 'Rev. 8'.

I append a photo of the PROM in question, it was V 9, on the left, and the PROM type from ATMEL which I used to burn the V13 inside. A friendly volt-nuts colleague had sent me his newer instrument for downloading the content of its firmware.

That is the only way to update the firmware, as this design from about 1995 or so had no FLASH memory yet.

My instrument from 2000 also has this 'RIC' designator on the FRAMs, so I also assume it is also affected.
There is one FRAM on the main PCB, and one on the DMM PCB, A4, which contains the calibration constants .. maybe the data loss only occurred for a certain batch from Seiko only.. I can't find any further details in the forum, or in the Service Note of hp.

Frank 

I see, thank you. The DMM FRAM doesn't worry me too much now, since I already backed it up, saved it in the computer, and programmed a new FRAM for replacement when the need arises.

Thanks again, I'll look around to see if I can find it, but at least it should be a pretty decent Voltage and Temperature logging meter.

[Dr. Frank]

...

Thanks again, I'll look around to see if I can find it, but at least it should be a pretty decent Voltage and Temperature logging meter.

That's right, it's really nice and versatile, especially with this scanner functionality.
If you want to measure temperatures, maybe you also need the appropriate precision NTCs for all the HPAK instruments? That's 2k25, 5k, or 10k, which are either no more available from KS, or only for an exorbitant price.

Could you please describe, how you downloaded the content of both FRAMs, i.e. which hardware you used, and which programmer?

Thanks - Frank

[Antonio90]

That's right, it's really nice and versatile, especially with this scanner functionality.
If you want to measure temperatures, maybe you also need the appropriate precision NTCs for all the HPAK instruments? That's 2k25, 5k, or 10k, which are either no more available from KS, or only for an exorbitant price.

Could you please describe, how you downloaded the content of both FRAMs, i.e. which hardware you used, and which programmer?

Thanks - Frank

Yeah! It was pretty straightforward, really. I used an arduino Mega, following this post https://btbm.ch/programming-cypress/, and this library https://github.com/battosai30/FRAM (made ad hoc for the I2C Cypress Fram), and a Pomona SOIC8 clip (see attached).

I connected the VDD and VCC pins of the FRAM to the GND and 5v pins of the Arduino Mega, and the SDA and SCL to it's respective pins. If someone is going to do it with a different arduino, it is necessary to pull up SDA and SCL to VCC, the Mega has them already up. Also, the usb power is not enough to drive the FRAM in circuit.


At first I was trying to connect every pin but, really, you just need the VDD, VCC and the Data and clock lines, the circuit and power lines are already there.

In any case, I don't have a way to check the integrity of the files without replacing the FRAMS, but a few notes:

The DAC module gave me quite a few problems, but it also was the first I tried to read. It is the least important to me by far too.

The 20 channel MUX was straightforward: the readings were consistent with every try. The little program written by the owner of the blog linked has an option to compare the readings against a predefined table, and it worked fine.

The DMM board: I hate it. Removing the shield without bumping anything is hard, and extreme care is required. The readings of the FRAM were also consistent, and the only problem is one of the pins is pulled up to VCC, so you have to change the address of the I2C device.

As for the mainframe one, each reading yields a different table. I don't know if that's how it works, maybe it also stores the number of times it has been read, but there's one to three values that change every time I tried to read or compare it.

So, I think it's worth to try and read the DMM FRAM, being the most critical. Just be prepared to take quite a bit of time getting the board and ribbon cable out of the shield.


Never heard of those thermistors, is it a calibration kit or something for HPAK? The temperature measurements are mainly because I want to fiddle with a JVR, I hope the relative readings, if consistent, are more important than absolute accuracy.

Best regards,

Antonio.

[Kleinstein]

Reading the I2C chips in circuit could be tricky:
The power to the FRAM chip on board would also power some of the circuit - even the CPU may come alive.  If one has trouble, I would consider forcing the CPU in reset state in some way, so you could not get 2 masters on the I2C bus. At least chances are the CPU would not do extra FRAM access if not really needed, so likely a few reads at the start and than no more.
There should be already be I2C pull up resistors on the board, that would help - the AVR internal pull up's are rather weak and thus only for a low capacitance circuit (like 1 chip close to the µC).

[gamalot]

Hi Antonio,

I just saw your message, I think the position where you installed the battery holder is correct.

However, I found out that I made a mistake! After I installed the CR2302 battery, I can’t remember how long it took, it swelled up like it was pregnant. I believe it was caused by charging when the instrument was plugged into the mains.

My solution is to replace the 14.7K resistor R183 with a 510K resistor. So far it seems to work well.

I thought I had shared this information a long time ago, but after checking I found out that I didn't. I'm really sorry about this.

[Antonio90]

Hi Antonio,

I just saw your message, I think the position where you installed the battery holder is correct.

However, I found out that I made a mistake! After I installed the CR2302 battery, I can’t remember how long it took, it swelled up like it was pregnant. I believe it was caused by charging when the instrument was plugged into the mains.

My solution is to replace the 14.7K resistor R183 with a 510K resistor. So far it seems to work well.

I thought I had shared this information a long time ago, but after checking I found out that I didn't. I'm really sorry about this.

Thanks! No problem at all. I'll try the same mod then. It's a good excuse to get some SMD resistors and other parts, as I only have through hole parts.
I might try and buy an equivalent battery to the one that was already installed.
And BTW, wouldn't it be a reasonable idea to install a supercapacitor instead of a battery? It's true that the instrument should be unplugged when not in use for long periods and it would drain the charge, but maybe is one less thing to worry about while being enough to keep the useful functionality when needed.

Reading the I2C chips in circuit could be tricky:
The power to the FRAM chip on board would also power some of the circuit - even the CPU may come alive.  If one has trouble, I would consider forcing the CPU in reset state in some way, so you could not get 2 masters on the I2C bus. At least chances are the CPU would not do extra FRAM access if not really needed, so likely a few reads at the start and than no more.
There should be already be I2C pull up resistors on the board, that would help - the AVR internal pull up's are rather weak and thus only for a low capacitance circuit (like 1 chip close to the µC).

There has to be something like that happening in the circuit. Also, I could not get any aknowledge from my read attempts while the DMM board was connected, but I got complete readings (just not really reliable) when the board was disconnected.

I might look in the schematics for a reset pin on the uC, but I'm a bit afraid to damage something. I'm not really familiar with digital circuitry anyway.

Thank you very much!

Antonio.

[Kleinstein]

Supercapacitors have a considerably lower capacitance compared to the old NiCd/ NiMH batteries. The original battery may last 6 months and the super cap maybe only a few days or in old design with higher power drawn maybe just hours. It takes a really low power design to get months of back from a super cap.  Than a LI primary cell could also run for many years.

Another possible way to get the cal constants would be a bus sniffer, while the DMM starts. So let the DMM start and read the cal data and just listen and record. I remember someone did this one a different meter (AFAIR not standard I2C) - still quite some work, but likely feasable if there is not much more than the CAL ram on the I2C bus.