Should I return a Keithley 238?

[leighcorrigall]

Hi Friends,

I want to discuss how I can diagnose a suspected physical defect of the 100 V measurement range of the Keithley 238 I (A10) model I have.

A calibration procedure was conducted with a Keithley 6500 DMM with 6.5 digits of precision that was calibrated in August 2020. Let's assume that the accuracy is better than 0.035 % for this instrument, which would be sufficient to calibrate that Keithley 238. I have already calibrated another Keithley 238 this way, so I think it is reasonable to assume that this method works.

After calibration, the instrument was able to source voltages between 15 V to 110 V with high accuracy, according to my Keithley 6500, another Keithley 238, and a Keithley 2200-32-3. However, when measuring voltages at this range, the instrument is well outside of specifications. The table below represents these voltage measurements:



I performed another calibration procedure afterwards in an attempt to correct the shortcomings, but without success. It does not appear that results are a consequence of human error.

What I noticed while verifying the measurement ranges is that if I set the voltage compliance to 15 V, the Keithley 238 can measure 10 V precisely. However, setting the compliance limit to anything greater than 15 V will cause the instrument to use the 100 V range. The same 10 V measurement will read out of tolerance. I believe that the 100 V measurement range of this device is faulty.

On inspection of the service manual, it appears that the voltage sourcing and measuring are separated between two boards. The following simplified schematic appears to be what defines the voltage measurements. Can anyone confirm this? I am a humble materials scientist and not electrically trained but I am trying to learn.



I also notice that there are strong resemblances between the Keithley 238 and Keithley 236/7 design in this context, so I think that the following electrical schematic applies. The major differences being the R18, R16, R 17 assembly for the 1000 V range, and the K7 relay.




If I had to guess the cause of the 100 V range measurement problem, I would say that it has to do with either R4 or R15. I know this to be true because the 1.5 V measurement range (K6) seems to be in working order. Before I start tearing the board apart, does anyone have advice for me on where I could start? I believe I have most equipment necessary to diagnose and repair the board, I just need help on where to begin. 

Thanks for your thoughts and input.

Regards.

[garrettm]

I'm a bit confused by how you are testing the SMU. But it’s probably not your fault, as I can be dense sometimes.

So if only for my own clarification, are you in Vsource/Imeasure mode or Isource/Vmeasure mode?

Voltage can only be a compliance parameter if you are in Isource mode.

If you want to measure an external voltage, you would need to be in Isource mode with the source current set to 0.0000 on the 1nA range.

To test the readback of the SMU's Vmeasure you would then connect an external voltage source to the SMU and compare the programmed voltage at the external voltage source (likely measured by your DMM) with the measured voltage at the SMU.

If you want to compare the programmed compliance limit vs. the SMU's readback with your DMM directly, then you would need to set the source current to something large enough that you can reliably reach and stay at the compliance limit, say 100uA to be safe (110V/10Mohm = 11uA). You may observe small differences when setting the source current to another range, say 1mA to 1A.

It's important to distinguish between error in the readback of Vmeasure (as seen on the display) and error in the programed compliance limit. They are not the same. The 238's datasheet states that the compliance programming error is the same as the voltage source. So the worst case error in readback of the compliance limit is the sum of the source error and measurement error: for the 110V range, that would be (0.033% +24mV) + (0.025% +10mV) or 0.058% +34mV. It's likely much better than this in practice.

[leighcorrigall]

Hi Garretm,

So if only for my own clarification, are you in Vsource/Imeasure mode or Isouce/Vmeasure mode?

The Keithley 238 that I am verifying was set to measure voltage: I_source = 0 nA and V_measure.

To verify the 100 V measurement range, I measured a voltage source with a Keithley 6500 before measuring it with the Keithley 238. The difference between these numbers is well outside this tolerance as indicated by the spreadsheet screenshot attached to my last post. By comparison, this voltage measurement range has a significantly greater % error than the 1.5 V range, which should be more difficult to calibrate. However, voltage sourcing with the Keithley 238 at the 100 V range is spot on when measured by the Keithley 6500.

Clearly, there is something problematic with the board because the other Keithley 238 was calibrated exactly the same and does not act in this way. The other Keithley 238 can both measure and source voltage with a high degree of accuracy -- much less than the tolerance.

---

Also, my comment about compliance had nothing to do with measuring it. I was explaining how I was diagnosing the problem. If the Keithley 238 is set to measure voltage, it works great when the compliance voltage is set to 15 V. Anything greater allows the instrument to switch to the 100 V range, where it has no ability to stay within measurement tolerance. No matter what I do, the 100 V range is terrible. I suspect that it might have been caused by the transformer failure that occurred before I purchased it, which began this post.


Regards.

[leighcorrigall]

False alarm.

The problem can be identified by inspecting the difference in measurements between the positive and the negative voltages of the same magnitude from +/- 1 to +/- 100 V. The offset voltage at 000.00 was the same as this difference at +/- 30, +/- 70, and +/- 100. I probably added this error by using triax adaptors, while converting between the Keithley 6500 and the Keithley 238 while calibrating. Even more so, the short procedure should be performed as close as possible to the instrument being calibrated.

Either I did not properly configure the offset voltages (C18X and C22X) and/or perform common mode calibration (C1X). All readings are well within specifications. I used a brand new Keithley 2200-32-3, that was not part of the calibration procedure, to verify the measuring capabilities of the Keithley 238 I (A10) that I had troubles with.

I suppose the only problem is that this foolish mistake is now immortalized on the internet. Thanks for the sanity check, Garrettm. I would have not bothered with calibrating it again had you not responded to my plea for help.

Oh well, at least now I know all my Keithley 238's are in good operational condition. 

[garrettm]

At least it wasn't a hardware fault! The main analog board is a pain to get to.

Are you manually entering the commands and data, or did you write a partially automated script? I could see it being pretty easy to accidentally send the wrong value if using manual entry. Even when partially or fully automated, calibration routines can be hard to get right, so don't feel bad.

It's possible that the connectors and adapters are adding some thermal emfs to the measurement, though this shouldn't be much more than a millivolt. When shorting, you can use your female BNC to male banana adapter in tandem with the female 3-lug to male BNC adapter to plug directly into the input LO banana jack. That's about as direct as you can get without modifying a standard triax cable. Ironically, Keithley has a document that shows how to make a custom 3-lug triax to 4mm banana cable for the input HI, which wouldn't be a bad idea to have for calibration and general purpose lab work. They sell the 237-BAN-3A premade, but it's absurdly priced at 380 USD.

I think the suppress function should have nulled out that offset during testing. A stable offset can be easily fixed with calibration (assuming it isn't absurdly large). It's when the offset is noisy or drifts erratically that a repair is in order. My 237's 1.1kV Vsource range may need repair as it's zero offset jumps around erratically from 2 to 7 mV. The 110V source range has 6 mV of zero offset, but it's nice and stable. So calibration should buff that out nicely. The unit in question had a shorted Q19 on the output board and overheated 2.2Mohm resistors on the resonant converter. But it looks like I might need to dig a little deeper to find the remaining gremlins.

[jogri]

Ironically, Keithley has a document that shows how to make a custom 3-lug triax to 4mm banana cable for the input HI, which wouldn't be a bad idea to have for calibration and general purpose lab work. They sell the 237-BAN-3A premade, but it's absurdly priced at 380 USD.

Do you happen to have a link to that document? I'm currently in the process of making such cables for my 237, would be nice to know the proper way of doing that.

[Qw3rtzuiop]

Here it is