Teardown/repair : Time Electronics 9823 10ppm/year AC-DC calibrator

[TiN]

Let's dig into something rare. Not Keithley, not Fluke, but Time Electronics 9823 straight from England!



This beast was bought on eBay (seller still have lesser version, 30ppm/year 9822).

There is hardly any information on the web about this old beast, but from what is available, it's specifications are well competitive to monsters like Fluke 5500A/Datron 4808! I could not resist, you know.

Thing powers up, provide output but it's barely functional and have issues. E.g. if I set output 10.00000V, I get 9.845V. Adjustment to 10.1V making voltage jump and than quickly go back 9.845V. Sounds like digital/DAC issue to me. So, let's take it apart...



Construction is using aluminum frame with many boards. Most of parts numbers are dated 1988, some 1990. Everything is DIP and boards are 2-layer. Should be rather easy to recover schematics, yay.



Take a look on boards:
I let you guess which one is what.





















Some cleaning would help...



Too bad DAC and reference are potted in large blocks . Seems it will remain a mystery what reference is used in this machine.

Anyone had experience with this calibrator? Any manuals/service info?

[Vgkid]

Thanks for the teardown, certainly an interesting beast.
Shame about the potting...

[IanJ]

Hi all,

Well I recognize the Euro card 19" rack kit..........sold by Farnell in the UK, we bought tons of them. Made by Vero I think. The green plastic guides were brittle and the little location protrusions at the ends would snap off rendering them useless. The dark green ones are original, looks like the lighter ones are a newer design.

I notice that Time Electronics elected to use a ribbon cable backplane rather than one long PCB.

Ian.

[quarks]

Hello TiN,
how do you find all this juwels and even more interesting, how do you all things you do in only 24h per days?
Thanks again for sharing.
Bye
quarks

[TiN]

how do you find all this juwels and even more interesting, how do you all things you do in only 24h per days?

You call that doing? Man, I don't want even making by backlog list, believe me. Some of the gear sitting in the piles for years(!). Blame all to precision-nuttery.
And since it's funded by personal cash, simple "go buy F5720A" ain't happening, leaving room only for obscure and wierd gear like this one .

For example, right now I need solve dilemma which one to work first, 9823 or 2002. 2002 is much easier, but still will take 20-40 hours, while 9823 can be 5 hour fix, or 500 hour one. I'm going to buy manual for it, hoping for schematics/service info. And that's ignoring any other projects I have to date..

DiligentMinds.com
To my regret, I don't have access to X-Ray.

IanJ
I got impression that outer rack kit is something standard. Thanks for details and info!

[manganin]

it's specifications are well competitive to monsters like Fluke 5500A / Datron 4808!

Datron 4808: not even close in real life. Fluke 5500A: probably comparable, except the AC frequency upper limit, which can be a problem when calibrating some multimeters following the manufacturers original procedure. (The "high" frequency AC is where most of the weight and cost of the high end calibrators comes from.)

The good side is that the Times instrument is very simple and straightforward design. Seen what you have done before, should be easy for you to get it working.

Basically a good instrument for calibrating up to 6.5 digit multimeters, but most often seen in calibrating handhelds.

The problems are usually high voltage/current AC (follow the burn marks), connector (especially the ribbon) or relay related.

Too bad DAC and reference are potted in large blocks

I am not familiar with that specific module, but the company has always potted the key circuits in their equipment. Usually there are a few standard ICs and some passives inside. Protecting their (repair) business rather than hiding any innovations.

Any manuals/service info?

The company still exists. They can probably send/sell the User Manual, but getting service documents has been very difficult. (My guess would be that their current calibrator models are still very similar inside.)

[TiN]

I got user manual and calibration manual from Time (linked in article).
For service info, I'm attempting to get paper copy, will see how that goes.

Will be interesting to get this box fixed and perhaps pimp it out for better performance, replacing reference with LTZ-module and building better DAC?

[ManateeMafia]

It definitely looks more serviceable than my Datron 4700. You don't even need an extender board.

[TiN]

Alright, thanks to fellow volt-nuts I got few pieces of schematics for REF and DAC boards. Full paper manual I paid for should be on it's way on monday too.

Reference board:



There is 14-bit AD DAC , output of which goes to DAC PCBA (net LS 14 bits, perhaps meaning least significant 14 bits?).
Then return from DAC module comes back to REF PCB (MS Bits return, most significant?) making 20 bits output?

Schematics:



Elements location:



DAC PCBA

Block diagram:



Calibration trimmer locations:



Schematics of DAC module.



Now need to understand how this thing works to troubleshoot.

REF signal measurement against BUFFERED 0V show 10.42260V, pretty stable. But output from calibrator with 10V set shows stable 9.83417 VDC and it does not change if I change setting to 10.01 for example.
I'll make a jig to test all those optocoupler (ILDN14). There are many of them everywhere, perhaps some are dead?..

[Jay_Diddy_B]

Hi,

looking at the circuit for the 9800 D/A Board, I would guess that the module contains JFET switches or something like that.

Regards,

Jay_Diddy_B

[TiN]

Tested all ILD74 optocouplers. They all seem to be fine, but superslow, at 8kHz already barely useable, with ~11-28us of rise time.
Going to test AD7534 DAC on REF board now and check some local powers.

[TiN]

Scope capture of optocoupler output on CS/WR signal to REF DAC:



I'm not sure if this horrid rise time normal or not. Will see tomorrow, when I get digikey box with some new ILD74 couplers..

Disassembled more of front panel side to get access for high-current power amplifier/TA:



Front panel connections and cabling:



Main binding posts are gold-plated, very much likely made by copper. Shielded by steel cage.



High-current output shunt formed by four brass wires?



PA with all cables attached:



It's bit PITA to measure boards, as everything need to be plugged, seems like I need eurocard extender. Anyone have extra?

[SeanB]

Not surprising with a 100k pullup resistor, but you are going to be limited by CTR ratio if you reduce the value too much to get a better rise time. You might be able to replace the 100k pull ups with 10k and get a better rise time, but that might fail if you get some with the minimum 12% transfer ratio.

http://pdf.datasheetcatalog.com/datasheet/siemens/ILD74.pdf

Probably would be a good idea to look for a pin compatible one with better transfer ratio and which has a faster turn on time. 3us is not exactly fast, and turn off is really slow there, there must be more modern devices in the same package ( or in a package that can be made to fit the DIL socket using a small board) that are both faster and higher transfer ratio, so you can drop the pull up resistor to 1k for most of the data lines.

[Kleinstein]

The circuit once worked with these slow optocouplers. So there is no need to make them much faster, though a little faster might help. For this just a slightly reduced resistors (e.g. 68 K instead of 100 K) and maybe selecting the couplers to exclude the very poor ones should be good enough. With a 100 K (and in some places even 470 K) resistor such a slow rise time could be normal - it's also rare that old opto-couplers  get slower. The normal type of ageing is a lower CTR, thus not fully reaching the low level and at the same time getting slightly faster as they don't reach saturation any more. 

Even the poor looking signal shown from the scope should still give valid logic levels.

Besides the optos there is also the possibility for caps getting old. Access to testpoints might be a little hard without a riser board, but still possible. Some of the testpoint even have pins for connectors.

R22 on the DCI board looks really interesting - have not seen a resistor in TO3 before.

[Vgkid]

R22 on the DCI board looks really interesting - have not seen a resistor in TO3 before.

Itis a 4 wire power resistor in that case. I have 2 in my CS-152 calibrator. Vishay vhp series.


I'm tempted to replace the other 4 resistors, with some that are closer to being in spec/ less drifty with temps.

[TiN]

Yep, that's hermetic 4-terminal Vishay foil resistor.

What bothers me in optocouplers that they are specified for 3us both rise and fall times. To test I had breadboard bodge with datasheet conditions, they show still slow ~25us rise times, while fall is ok.
Anyway, I just got DK box with set of new ILD74's (Vishay made), so will test those to see if it's better. I know logic levels would be still valid, but if some signals go in wrong time, than register would not latch correct data.

I don't think it's the cap issues here, as - there are not many of them elsewhere expect PSU/PA, and second - output is not adjusting unless I go in big steps in settings, but it should change if I just increase setting e.g. +1mV at 10V range. So far my eval shows that something wrong with DAC generation on REF board, so only MSB 6-bit DAC ladder on DAC PCB working.

If that does not help, I'll have to make some test jig to check DAC and opamps, but that will be likely at weekend.
Today will be Kei2002 repair time, as now all parts are here I expecting that one go easy.

[amspire]

If the optocouplers are just controlling the 6 bit DAC, I cannot see any reason it needs speed.

That potted part of the DAC I think just contains the switches for the 6 bit DAC. Probably discrete MOSFET switches. I assume the 6 bit DAC provides the 6 most significant bits of a 20 bit DAC and the AD7534 provides the lower 14 bits.

It is not as if you have to change the 20 bit DAC output voltage 1000 times a second. 25uS is super fast compared to the 20 bit D/A settling time.

Not sure why the DAC switches are potted but it could be a number of reasons, including thermal reasons or to make it hard for other companies to copy their switch circuit since it is the secret to the 20 bit DAC. The R-2R bit is easy, but getting the switches right is for a 20 bit DAC is extremely hard. That module is Time's magic sauce to make the calibrator work to specification.

The switch resistance has to be constant to within 10 mOhms over temperature, but you cannot just stick in a big MOSFET because you also need off-current leakages of about 0.1nA or less. The leakage current increases exponentially with temperature, so they probably designed for near 1pA OFF leakage at 20 deg C. They are very tough specs to meet. They may be using a small geometry switch with a fairly high ON-resistance such as 0.1 ohms or even 1 ohms, and adding a temperature compensation network to each switch MOSFET to make the switch resistance constant to 10 mOhms over temperature. When the R-2R is calibrated, it will allow for the switch resistance. As long as the switch ON resistance is constant to 10mOhms and the OFF leakage is less then 0.1nA over temperature, the switches will have a no effect on the DAC accuracy.

Richard

[TiN]

Got new DAC from UTsource, did not help the issue though.

To ensure it's this section issue, I removed DAC's output opamp OP27 which converts DAC's current output into voltage, and connected my HP 3245A source instead to supply stable voltage instead.
And it worked as expected, adjusting source's output changes voltage accordingly, and there is no issue to get exact output voltage.

For example 10V:



I ran it for 10 minutes, voltage is nice and stable, hovers around 1ppm, even with everything open and hanging around.

[TiN]

I think I've gotcha.. Here's LA capture on DAC's digital inputs...



And here's from datasheet:



See anything wrong?

[ManateeMafia]

IC5?

Never mind. Looks like they are supposed to be LO.

[TiN]

Nope, i don't think they are supposed to be LO, as we have three write operations (CS goes down three times).
So it should be A[01] to load first 6 bits (which is 0x14 in LA above), then A[10] to load least 8 bits (which is 0x96 in LA above) and last one A[11] to load all 14 bits into DAC (data dont matter here, so 0x00 in capture).

But now instead DAC programmed to 0x1414 then to 0x9696 and then directly to 0x0000, Which is ain't right.
I guess I'll have to get service manual to trace where those A1/A0 go on processor board.

This brings my attention to processor board, where to PIA Motorola 68A21P are located next to CPU. I thought - ok, they are same, and not contain ROM inside, so what happen if I swap them in sockets.
Aha, now A0/A1/CS toggling like crazy but Data bus at DAC stay dead zero. Perhaps one of PIA dead?

[ManateeMafia]

You're right. I missed the footnote for both address lines being LO at the same time.

[TiN]

I think this photo is self-explaining on the progress



Here's also DAC signaling on jig from RPi connected as controller for test earlier:



Btw, later DAC I've got from UTsource was either fake or broken. Marking was looking legit though, I'll take macro shots later on of it.
So this is still original AD7534 in the calibrator now.

I hooked output to 3458A for overnight log, will see if it stays nice and stable, even though MFC is in bits on the table without any covers.

[TiN]

Almost finished assembly, everything seem to be working fine. Here some shots with 10 VDC, kOhm and VAC.



As usual, more pics and details in my repair worklog.

Next step - testing and calibration.

[ManateeMafia]

... Next step - testing and calibration.

No need for cal, sticker says it is not required.

Great job!