Advantest R6246 "Repair" and Teardown

[42Khz]

Hi everyone,

Quite a while ago I got my hands on a "for parts" Advantest SMU, the R6246. The case had a large dent. I suspect the unit has been dropped (it weighs 33kg!), or something heavy was dropped on it.
The ebay listing indicated that display showed garbage after turning on. For this reason, I was betting on the display driver having some issue and the unit being fine otherwise.
In case anyone knows how much these units cost new, please let me know. I'm curious how good of a bargain I got.

First, the specifications. Since they are quite extensive, I'll only mention the highlights here.
Channel A:

• Banana connectors;
• Can source voltage from 620mV (10μV source resolution, 1μV measure resolution) to 62V (1mV, 100μV) full-scale;
• Can source current from 62μA (1nA, 100pA) up to 20A (1mA, 100μA);
• Maximum output power is 200W (+/-20V up to 10A).

Channel B:

• Triax connectors;
• Can source voltage from 620mV (10μV, 1μV) to 220V (10mV, 1mV);
• Can source current from 6nA (100fA, 10fA) up to 2A (100μA, 10μA);
• Maximum output power is 24.8W (+/-40V up to 0.62A).

A nice feature of this unit that I think is worth mentioning is its ability to measure either its output current or output voltage, regardless of the source setting. This is unlike my Keithley 237, which can only measure what it's not sourcing. The datasheet also claims it to be stable into large load capacitors on the higher-current settings (1mF from 600mA upward, I didn't test this), which I believe is fairly unique.

Since it was wired for 90-110V while I live in a 240V country and there were no schematics available, I decided to fully tear the unit down first. I took photographs of all the analog stuff in the process, to be able to check everything for damage and possibly reverse-engineer everything. Though I haven't gotten around to the latter.
There was no nice mains potential setting switch but the unit can be wired for 240V, so I reconfigured the transformer connections. I used seg-el's notes for another Advantest unit as a guide (https://www.eevblog.com/forum/repair/advantest-r6245-dual-channel-smu/msg3456330/#msg3456330). If I can find my own notes on the R6246 I'll post them here.
I would be amiss not to plug that thread in its entirety (https://www.eevblog.com/forum/repair/advantest-r6245-dual-channel-smu/). There, user leighcorrigall documented his own teardown and more of a sister to the R6246: an R6245 with two low-current measurement channels.

After the initial teardown and rewiring the transformer, the unit booted up. However, the display would start drawing random garbage characters and the user interface became unresponsive.
By pressing some buttons during boot I managed to get into the menu where the unit would not freeze up. At this point I was confident the display was not at fault. I dreaded the possibility of some unobtanium processor being kaput, or some EEPROM corrupted.
Fortunately, after many additional teardowns and debug sessions, I ended up finding a grand total of three bent connector pins in several of the unit's PCB cards. After I bent the last pin back in place and put the unit back together, it seemed to behave perfectly well and I have not found any obvious issues since. That was an easy fix! (which still took me several days of tracing puzzling symptoms)

To start off, here are photographs of the unit close-up, and with channel A sourcing 10V into a 10kΩ resistor with an HP 3456A keeping an eye on the voltage.

[42Khz]

The most interesting parts to me were the signal conditioning hardware for the two output channels. Each channel consists of what I suspect is a controller board (which is the same design for both channel A and B) and a current and voltage ranging board. For each channel, these boards are held together with some brackets which slot into the chassis.

Unfortunately I don't have great photographs of the entire unit, this snap of the channel cards will have to do for now. You can recognize the two channels by the brackets on the bottom right. The block on the top left is the output power stage, which I'll get to. The block on the bottom left has a whole bunch of regulators and big caps on it. Beside the channel cards there is a card containing the DACs for both channels, and a card containing some digital control stuff including some EEPROMs.

I did take extensive photographs of one of the channel controller boards (identical between the channels) from both the front and back. This should enable someone sufficiently motivated to reverse-engineer most of the control loop (excluding power stage and ranging). If someone wants the original, higher-quality photographs I can share them. I compressed them a bit for the forum.

[42Khz]

Let's continue with a look at channel A's (up to 62V, up to 20A) signal conditioning board. This contains (almost) all the current sensing resistors and I suspect also the voltage ranging.

We see some very large relays as well as some thick cabling. The screwed-down terminals next to the MRT-3044S relays are wired to the power stage through thick cables.
Moving to the other side of the board, we find some precision resistors and some JFET switches (2N4393). Towards the back of the board is a nice BB opamp in a can.

[42Khz]

Now it's time for channel B, this is where femtoamps live. There is a big shield over the sensitive current-ranging stuff, and also one over the output connectors.
Removing the can, we see a whole bunch of teflon cups, tons of air-wiring and a bunch of reed relays, among other cool stuff.

[42Khz]

Next up: the DAC board. This board contains the same circuitry twice (note the 015 and 016 towards the board corners in the photographs), with each one isolated from the other. Each copy contains a pair of analog devices AD7846 16-bit two-stage DACs along with some precision opamps from Analog Devices (OP27, OP07) and Linear Technology (LT1008) and an LT1021 buried-zener reference. There is also some digital stuff, and what seems to be a small bodge board with a NAIS AOV214 PhotoMOS relay.

The closeup of the DACs shows, among other things, IDAC and VDAC testpoints.

[42Khz]

Finally we get to the power stage. It truly is a beautiful piece of Japanese origami. I dared not to fully tear it apart, especially because I was quite sure it was working correctly.

A side view shows a bunch of Toshiba (2SK1530, 2SJ201) and Hitachi (2SK1303) transistors, along with what seems to be two pairs of diodes (biasing?), some drivers and a pair of Isabellenhütte precision resistors. The latter must be the high-range current sense resistors for channel A (which is why I wrote that most current sensing stuff is on the signal conditioning board). Also visible are some very thick board-interconnecting wires.

Unfortunately I don't have any photographs of the "inside" of the power stage. There is a whole bunch of power-looking stuff inside the "wind tunnel".
Another side shows big fat tracks along with some thick connectors, this must be part of channel A.
For dessert we have a look at the opposite side. I suspect the circuitry here is related to channel B.

[42Khz]

I promised mostly pictures of analog boards, but I do have one of the digital board that sits beside the analog channel boards.
The halves of this board, like the DAC board, seem to be mostly identical. There are some EEPROMs and a whole bunch of digital chips with red markings on them. I have no idea why they felt a need to do that, are they perhaps selected in some way?

[42Khz]

For anyone interested in rewiring their unit, here are some before and after photographs of the transformers and their connections.
I don't recall the details, but I'll update this post if and when I uncover my old notes. Basically I measured the transformer taps to figure out how they were connected and labeled the wires as B for the Big channel A transformer and S for the Smaller channel B transformer. One detail I do remember is the fuse that seems to be embedded in the transformers, this is something to watch out for.

[42Khz]

Unfortunately, although everything seems to be working properly under manual control (although the memory check does seem to skip the first 64kbyte), I cannot seem to get measurement results read out over GPIB.
I can control the instrument just fine, reset it, set the outputs and turn the channels on. But when I try to read back measurement data the unit sends me its identification along with its EEPROM revision numbers. If I keep reading after that I get pure garbage. Thus I have no nice measurement results/IV curves to show (yet).

Does anyone have an idea what might be going wrong? I tried changing the delimiter to CR+LF as indicated in the code snippet below (unfortunately it's in Japanese, there is no English manual), as well as only CR or LF.
I'm using an ebay-bought (so most likely counterfeit) Keysight 82357B USB to GPIB adapter with MATLAB. I used MATLAB's visadev and sent the commands shown in the code snippets.
I will double check, but I'm quite sure I managed to read from my HP 3456A just fine in the past with the same setup.

[MadTux]

Nice instrument, 20V/10A SMU is a nice thing to have.
Inside has quite a few similarities with Keithley 236/237, staged Mosfet output stage, even similar connectors on it.

[dietert1]

If response handshaking does not work properly, you could try to insert "Sleep(100)" between the command asking for data and reading the response. In case it works, you can try and reduce the delay. Anyway it's not about the response time of the meter but about threads to run.

Regards, Dieter

[42Khz]

If response handshaking does not work properly, you could try to insert "Sleep(100)" between the command asking for data and reading the response. In case it works, you can try and reduce the delay. Anyway it's not about the response time of the meter but about threads to run.

Regards, Dieter

Hi Dieter,

Thanks for your suggestion, I gave it a try. Unfortunately the instrument is still very eager to tell me its name and EEPROM versions, but not much more. I changed the delay up a bit but still no dice.

To confirm that I can both write to the R6246, and read data from other instruments (in this case the HP3456A), I wrote a little test script. It sweeps the SMU output voltage from -62V to 62V in 1V increments, making the HP3456A measure the output after each step. The attached plot shows the difference between the value reported by the HP3456A and the value sent to the R6246 in mV.
For a more serious measurement I would of course use a second DMM, this is just a demonstration.

[dietert1]

The sequence in the GPIB example of the calibration manual translated by leighcorrigall looks like this

...
1360 PRINT @R6245A;"XE 1"
1370 INPUT WAIT 1, “”, A$
1380 INPUT @R6245A; B$
...

Don't know why there are two input statements. In that example the DV command is also followed by a

INPUT WAIT 10, “”, A$

There is a table with explanations, e.g. line 200 "FMT..." to set the format on the 6245. And "input wait.." appears to be a delay.

Regards, Dieter

[42Khz]

The sequence in the GPIB example of the calibration manual translated by leighcorrigall looks like this

...
1360 PRINT @R6245A;"XE 1"
1370 INPUT WAIT 1, “”, A$
1380 INPUT @R6245A; B$
...

Don't know why there are two input statements. In that example the DV command is also followed by a

INPUT WAIT 10, “”, A$

There is a table with explanations, e.g. line 200 "FMT..." to set the format on the 6245. And "input wait.." appears to be a delay.

Regards, Dieter

Thank you very much for pointing me in the direction of that manual: adding a delimiter format command solved the issue!

Attached is a quickly produced IV curve of a yellow LED measured using channel A. The current limit is set to 10mA, as can be seen.

[wichers]

Bought a late ADCMT R6246 as well, it was rough on the outside but the inside is in pristine condition. All it needed was a new power button.

For those interested, I dumped the motherboard firmware.

[wichers]

Hi,

May I ask what your GPIB read speed is? I'm using an AR488 (prologix compatible) adapter. As mine is very slow (reading at 1 line per second), I'm trying to find out what the problem is.. I'm currently double sweeping in advance and I'm picking up the samples later, the sweep itself is very fast, fetching the measurements isn't.

This is my python R624X instrument wrapper:
https://github.com/wichers/pymeasure/blob/master/pymeasure/instruments/advantest/advantestR624X.py

And here I'm using the wrapper in a small example:
https://github.com/wichers/pymeasure/blob/master/examples/Current-Voltage%20Measurements/iv_advantest.py

Fetching the measurements (this seems fastest) is basically done using:

Code: [Select]

instrument.write('nub_01?')
count = instrument.read()
instrument.write(rmm_01?')
for i in range(0, count):
    measurement = instrument.read()

Kind regards,
Alexander

[42Khz]

Hi Alexander,

Congratz on your score and thanks for dumping the firmware! I'll do the same for mine when I get around to it.

Unfortunately I can't really help you: I ran my sweep in MATLAB. The SMU was reprogrammed and data was collected one point at a time and quite slowly at that. Maybe half a second per point?
I'll probably pick this back up when I have some free time again in a month or so to look into running sweeps on the SMU itself.

Regards, Jeroen

[aronake]

I happened to buy a R6246 very cheap. Fascinating piece of equipment and Japanese 1990ies electronics design!

[serg-el]

https://cloud.mail.ru/public/wugJ/ognyErGRf

Full information

[aronake]

https://cloud.mail.ru/public/wugJ/ognyErGRf

Full information

Great collection of documents! Thanks!