Advantest R6581 Service Manual

[TheDefpom]

I was thinking that there must be an ASCII data stream somewhere to grab the measurement data from, rather than trying to decode the screen data, probably the same for the mode data too I expect, I mean the data is viewable over GPIB so it must be in a raw format somewhere other than the screen.

[Mickle T.]

The front panel interface is not a serial or parallel interface in the usual sense. In fact, it is a shared memory area that can be accessed asynchronously by both the main assembly processor and the front panel processor.

As far as I know, the first 47 bytes of external memory on the H8 MCU side are a serial buffer of character codes for the 1st and 2nd lines of the display. The next 18 bytes are the annunciators flags (0x00 - off, 0x01 - on).
But besides this, the same memory contains a keyboard buffer, a matrix of pressed key flags, an interrupt messaging mailbox, and a large number of flags, the purpose of which I do not know (most likely they are related to scrolling, blinking, etc.).
The H8 MCU firmware contains a character generator with the standard ASCII character set, however, I found that characters with codes 0xE0...0xE8 and 0x7F are used in debug mode to smoothly fill the display with vertical and horizontal lines.

[Mickle T.]

I decided to take a chance with my working multimeter and test the translator with the first monochrome OLED display that came to hand (Winstar WEH002004 in graphic mode).
Apparently I’ll still have to look for another R6581D/T, or wait until the display on my multimeter completely goes out

[picburner]

I have an R6581T also with a practically invisible display. I was thinking of doing a similar job to what Mikle is doing using a Futaba GP1294AI 256x48 VFD, basically it is a COG that incorporates SPI input logic and HV drivers and is currently sold at a very affordable price.
The dimensions of the display are similar to the pre-existing VFD but, to do a job well done (and to fit in the limited space), I would have to redo the entire original PCB.
I've already done similar work for a Wavetek 1271, but in that case I really had no other choice.
In this case, however, I'm a little hesitant: is it really worth doing all this work?

[Mickle T.]

I think the GP1294AI is the perfect solution. If the indicator is suitable in thickness, it can be connected with a thin flexible cable to an external mezzanine board on which the microcontroller and high-voltage DC-DC can be placed. A complete redesign of the entire PCB carries the risk of damaging the H8/322 microcontroller, the firmware of which is not yet available.

[branadic]

Thanks to MickleT., who deserves all credits, we have a wonderful service manual. I was wondering, why noone has reported INL improvement just yet? Maybe the math is too much of a hurdle?
Hence why I've put together this simple spreadsheet, hoping that more people will benefit from MickleT.'s wonderful work and report their success here.

I've tested the spreadsheet myself against our 3458A and an ADR1399 upgraded AN3200, taken readings with H0 ... H15 reset, calculated the new constants, copied them into the device using Python and rerun the INL test. Attached are my results.
Obviously summer time is not the best time to do this, even if the lab is AC stabilized. But the windows are facing south and temperatures outside were quite high, so there is quite some temperature change during the measurements. However, even with that drawback INL turned out <±0.1 ppm.

I hope that motivates people to get their R6581(D,T) INL down too.

-branadic-

[leighcorrigall]

I'm relocating at the moment. No lab to perform measurements/adjustments. I have 2x 3458As that I can use for a linearity adjustment

I will reply again when I have time to review the manual and the scripts. I agree, the first time I read the manual, it was difficult to understand the heaviside functions.

[Mickle T.]

Announcement of the next update.

[branadic]

Wow, great progress, I'm in for one

-branadic-

[TheDefpom]

Indeed that looks very good, I picked up a couple of different displays, I haven’t got around to trying to do anything with them yet though. I may not need to if I can buy these from MickleT, or just hook up a pro mini and dump some code into it.

Good to see the larger display can simply be plugged into the other units connector and it just works.

[gamalot]

I've also thought about DIY replacement display, but it seems difficult to find one with the right aspect ratio.

[bsw_m]

I found this display. It fits in place of the original pretty well.

[chekhov]

One more alternative could be this display:
https://www.buydisplay.com/bar-type
-4-58-inch-320x960-ips-tft-lcd-display-spi-rgb-interface

It has the same dimensions as the one above, but less colors and maybe a bit lower contrast.
I have both of them but haven't compared yet, wanted to revive my failed Keithley 200X display replacement project.

[bastl_r]

Hi

The link does not work for me with the line break.
Here without now.

https://www.buydisplay.com/bar-type-4-58-inch-320x960-ips-tft-lcd-display-spi-rgb-interface

[gamalot]

Thanks guys, I just ordered two 4.58" LCDs to see if they will fit. 

[dietert1]

Thanks for the hints. I found those bar displays at digikey, 1528-5805-ND. They don't have display memory, so i ordered STM32H730 MCUs that can drive the display with their LTDC peripheral. They have enough RAM to hold a 1-byte per pixel memory, so 256 colors with a programmable color lookup table (palette).

Regards, Dieter

[Mickle T.]

The first post has been updated.
A new topic about replacing the display has been created: https://www.eevblog.com/forum/repair/advantest-r6581-vfd-replacement/msg5634443/

[jaromir]

Hello guys.
Thanks to Mickle T and branadic I was able to straighten out the linearity of my Advantest R6581T. Attached is INL chart against known good HP3458A.
This chart contains three curves:
1, Blue is original INL of the untouched instrument.
2, Red curve is INL of the instrument when H0..H14 were zeroed and H15 was equal to 1.0
3, Green curve is INL after H0..H15 were calculated - from separate voltage measurement sweeps at the point of interest - using the recipe of Mickle T and checked against the spreadsheet provided by branadic, entered into the instrument and saved into EEPROM.

Attached is a file containing set of original H coefficients, too. It looks to be identical to those values reported by Mickle T.

[leighcorrigall]

Hi everyone,

I turned on this instrument recently as I have not had a lot of time in the last couple of years to tinker with it.

Interestingly, I noticed that a Fluke 732A measurement would change by several ppm over continuous operation. If I bumped the unit or moved it, the measured input value would sometimes jump significantly.

On review, it appears that the Vref board connector to the mainboard is not very good! If I poked the Vref board, the measured input would easily jump 10 ppm.

I removed the connector and hardwired the Vref board directly to the mainboard with bare copper wire and regular solder. This minor modification made a massive difference on the overall performance.

You have been warned. I hope this small contribution helps the community.

---

My next task is to clean the board in preparation for an artifact calibration

Setup:
Fluke 5440B
Fluke 5450A (possibly an ESI 925 for additional resolution)
HP 3458A

[dietert1]

I can confirm that the reference module connector isn't pure copper but includes brass springs. Was replaced by soldered copper wires in our three instruments. Maybe i already mentioned it before.

Regards, Dieter

[Mickle T.]

Another way to troubleshoot and improve the performance of the R6581 DMM is by adjusting the control pseudocode tables, including the ACAL routines. For example, this could significantly improve the reproducibility of artifactual calibration of problematic ranges such as 1 A, 10 nA, 1000 MΩ, and some others (at the expense of process time, of course).
Since the service manual will no longer be updated, it might be worth considering a technical manual for this multimeter in the future, including a detailed analysis of the algorithms of key routines and data structures in the firmware.

[Hydron]

What sort of changes are you suggesting? Increasing the number of averaged readings for these steps?

I plan to do some modifications to the unit I recently got, including:
- tweaking the temperature coefficient
- adding a bucking transformer to reduce input voltage (my mains supply is stable but very high at 250V)
- fixing the reversed capacitor, relay drive, input protection (not sure what resistors to use here, or if I should try something else like depletion mode FETs)
- patching the "display off" message in the ROM to be all spaces to prevent VFD wear
- do a linearity calibration once I have access to something to compare it to

Will document the new mods (rom patch and transformer) here, though they are nothing too exciting.

Many thanks to Mickle T. for your work on this meter - I don't think I'd have a 8.5 digit meter otherwise!

[Mickle T.]

What sort of changes are you suggesting? Increasing the number of averaged readings for these steps?

Absolutely right, the simplest way is to increase the number of averaging iterations. I tried it on the 1 A range, and instability, as expected, decreases according to the square root law.
The second, more complex method is to optimize the analog switch code order and delay timings.

[Ismsanmar]

- patching the "display off" message in the ROM to be all spaces to prevent VFD wear

Mickle already uploaded a modified firmware with that feature a couple of months ago: https://www.eevblog.com/forum/metrology/advantest-(adcmt)-r6581-dmm-firmware/msg5776843/#msg5776843

The only indicator he left active is the "SMPL" indicator.

- input protection (not sure what resistors to use here, or if I should try something else like depletion mode FETs)

If you want to do it like Advantest did on their R6581 models, they did it by installing 4 resistor towers in positions R008 to R011, each with 4 22kohm resistors in parallel, and a 1kohm resistor in series with a 1.5nf 3kv ceramic disc capacitor on relay K008.

[Hydron]

Mickle already uploaded a modified firmware with that feature a couple of months ago: https://www.eevblog.com/forum/metrology/advantest-(adcmt)-r6581-dmm-firmware/msg5776843/#msg5776843

The only indicator he left active is the "SMPL" indicator.

- input protection (not sure what resistors to use here, or if I should try something else like depletion mode FETs)

If you want to do it like Advantest did on their R6581 models, they did it by installing 4 resistor towers in positions R008 to R011, each with 4 22kohm resistors in parallel, and a 1kohm resistor in series with a 1.5nf 3kv ceramic disc capacitor on relay K008.

Thanks for the FW link, didn't realise it was already done! In the end I just dumped and patched my A01 firmware (no AC mode in the T model so I don't want the other changes), see attached file. A HN27C4096ACC-10 from eBay (used, albeit sold as new) did the job, and I can erase it again to add any new tweaks that Mickle T. comes up with as it's windowed (if it were a one-off mod I would probably just buy an OTP part from a reputable source). Note that it seems 100ns access time is quick enough (I think the original OTP part is quite a bit faster).

Thanks also for the protection details from the non D/T models - I recall discussions about potential better ways to do this (was it to reduce noise?), though I'll probably just go with what Advantest did. I'm assuming it's safe to use without the mod as long as there's not range switching or transients in the HV range?