Keithley 263 Calbrator / Source

[Alex Nikitin]

It looks like my collection of Keithley gear is growing quickly. Here is the latest addition - 263 Calbrator / Source . A rather unusual instrument, combining some interesting features and capabilities. It can source voltage (from +/-5uV up to +/-20V on 20mV, 200mV, 2V and 20V ranges), current ( from 2pA full scale to 20mA full scale, the smallest step is 0.05fA !), charge and also can provide reference resistors in decades from 1K to 100G. There is also an internal electrometer amp that can accurately follow the voltage on the load in the current output mode, which means it is possible to measure resistances and voltage drop on diodes etc. The voltage source DAC is 16-bit (?) PWM based with a polarity switch, and the large reference resistors are provided with a temperature "compensated" reading (estimated from two measured during a calibration "cold" and "hot" values). The main reference is LM399.

I am waiting for Keithley 181 Nanovolmeter from the same range to arrive soon and also for some 2-lug triaxial connectors so I could operate the K263 and K617 together. I will post some measurements soon in this thread, but first impressions are very positive (measuring the reference resistors/voltages and currents sourced by K263 with K2015 and K617)

Cheers

Alex

[Alex Nikitin]

Here is the first measurement - 10V output from K263, measured by K2015, NPLC10 + 10averages, so ~ NPLC100. Vertical scale 0.5ppm/div.

Cheers

Alex

[Vgkid]

That seems like a rather versatile piece of equipment. It could almost double as an economy mfc.
Can't wait to see inside.

[VintageNut]

Congratulations! It is a very interesting instrument. I also recently acquired one of these.

On my instrument, the DC volts ranges are correct. The resistance ranges act like 100G is in parallel with every resistance. So, the 100G resistance range low by 2X. The 2pA current range is off by 2X.

One or more of the red Coto relays is the suspect. I am considering cutting out one or more of the relay wires since they are all wired above the analog board.

[guenthert]

Congratulations! It is a very interesting instrument. I also recently acquired one of these.

On my instrument, the DC volts ranges are correct. The resistance ranges act like 100G is in parallel with every resistance. So, the 100G resistance range low by 2X. The 2pA current range is off by 2X.

One or more of the red Coto relays is the suspect. I am considering cutting out one or more of the relay wires since they are all wired above the analog board.

100GOhm?  pA range?  I'd try cleaning first before cutting anythin' ...

[plesa]

Congratulations! It is a very interesting instrument. I also recently acquired one of these.

On my instrument, the DC volts ranges are correct. The resistance ranges act like 100G is in parallel with every resistance. So, the 100G resistance range low by 2X. The 2pA current range is off by 2X.

One or more of the red Coto relays is the suspect. I am considering cutting out one or more of the relay wires since they are all wired above the analog board.

Check if there is any input protection and if it is ok, this can explain high readout.

[VintageNut]

Congratulations! It is a very interesting instrument. I also recently acquired one of these.

On my instrument, the DC volts ranges are correct. The resistance ranges act like 100G is in parallel with every resistance. So, the 100G resistance range low by 2X. The 2pA current range is off by 2X.

One or more of the red Coto relays is the suspect. I am considering cutting out one or more of the relay wires since they are all wired above the analog board.

100GOhm?  pA range?  I'd try cleaning first before cutting anythin' ...

There is nothing to clean. All of the high resistance resistors are wired point-to-point to standoffs that are insulated with teflon spacers. The circuit board is just a convenient place to mount the standoffs. The only leakage path from the resistors to ground is through the relays. The relay control power is routed through the PCB but the relay SPST switch runs through the relay body above the PCB.

[Alex Nikitin]

Here is the overnight data - 16 hours in total. The drift looks like not temperature related, though the temperature did vary about 3-4 degrees over that time. The unit is quite old and was not powered for a considerable length of time, so hopefully this drift may settle after some days. In any case this -10ppm drift is very little considering the specification (0.0175% +500uV for 20V range).

Cheers

Alex

[plesa]

It seems to be more stable than recent 2600 series.

[VintageNut]

It seems to be more stable than recent 2600 series.

The voltage used for all of the voltage and current ranges of the 263 is generated by PWM. I am not sure how the 2600 series instruments generate voltage but I doubt that PWM is used. 

[Alex Nikitin]

Here is an interesting graph: 1V output generated by the K263 on 2V and 20V ranges. I was expecting 5% output on 20V range to be considerably noisier than 50% output on 2V range. But, apart from the (expected) lower accuracy on 20V range (right-hand scale), the (LF) noise (measured by K2015 on 1V range, NPLC10 +10av) is about the same. Vertical scale 1uV = 1ppm per division.

Cheers

Alex

P.S. Obviously, the measurements were consecutive, I've just used the time scale from one run to combine the data on a single graph.

[Kleinstein]

The two rather similar curves suggest that the main noise source could be the internal voltage reference and not the DAC circuit or output amplifier / divider.
The noise level about what you expect from an LM399 or similar reference. So it is not that bad. I would not expect better from such an instrument.

Also the meter to measure that curve could be an issue. The reference in the K2015 might not be that much better.

[Alex Nikitin]

The two rather similar curves suggest that the main noise source could be the internal voltage reference and not the DAC circuit or output amplifier / divider.
The noise level about what you expect from an LM399 or similar reference. So it is not that bad. I would not expect better from such an instrument.

Also the meter to measure that curve could be an issue. The reference in the K2015 might not be that much better.

No, this graph rather suggest it is not the reference noise, as the outputs are at 5% and 50% of FS and the reference noise should influence the output proportionally. I will check the noise level of 2015 used later today, but my suspicion is that noise might be from the filter and/or output amplifier. In any case, a very good performance from this unit.

Cheers

Alex

[Kleinstein]

The source is using a setup of  reference - PWM stage - filter - variable output amplifier.

Any noise from the filter section would be scaled with the amplification. So the 20 V range would show more of the filter noise. Most of the noise of the amplifier is also scaling with the amplification, as there is one DC loop from the output and not separate stages. Also noise from the PWM stage (e.g. clock jitter) would be about the same at the PWM stage output and thus scale with amplification. There might be a little PWM residual coming through. Here the 1 V from the 2 V range has 50% PWM and thus highest AC amplitude, but low amplification after than. The 1 V from 20 V range has less AC amplitude, but more than 1/10  from the 50% PWM. So residual AC should be slightly higher for the 20 V range.

The main noise source that does not change with amplification is the voltage reference. Output voltage is proportional to that reference, and the PWM stage acts like an multiplying DAC. So it make no difference getting a factor of 1 from PWM and 0.1 from the amplifier or the other way round.

As the 2015 is using the same LM399 type reference as the 263, I would expect about half the noise from each.

[Alex Nikitin]

The source is using a setup of  reference - PWM stage - filter - variable output amplifier.

Any noise from the filter section would be scaled with the amplification. So the 20 V range would show more of the filter noise. Most of the noise of the amplifier is also scaling with the amplification, as there is one DC loop from the output and not separate stages. Also noise from the PWM stage (e.g. clock jitter) would be about the same at the PWM stage output and thus scale with amplification. There might be a little PWM residual coming through. Here the 1 V from the 2 V range has 50% PWM and thus highest AC amplitude, but low amplification after than. The 1 V from 20 V range has less AC amplitude, but more than 1/10  from the 50% PWM. So residual AC should be slightly higher for the 20 V range.

The main noise source that does not change with amplification is the voltage reference. Output voltage is proportional to that reference, and the PWM stage acts like an multiplying DAC. So it make no difference getting a factor of 1 from PWM and 0.1 from the amplifier or the other way round.

As the 2015 is using the same LM399 type reference as the 263, I would expect about half the noise from each.

Yes, you are quite right about the reference noise, and my only excuse is that I have a rather bad cold at the moment    . On the amplifier noise side it is not as clear as the scaling amp in the 263 works with a low inverting gain (x2.5 , x0.25 x0.025) and as such the amplifier noise would not change proportionally to that gain. The output amplifier is a composite built from ICL7650 chopper amp and NE5534 as an output amp. The ICL7650 has 2uV p-p (DC-10Hz) noise and that is what we probably see on the output.

Cheers

Alex

P.S. - I've measured 100mV output on 200mV range and got only about 0.7-0.8uV p-p maximum deviation and mostly about only 0.4uV  . Much better than I've expected.

The 2015 I am using measures about 1uV p-p on 1V output from the Fluke 731B (with the same measurement parameters, NPLC10 + 10av).

[Alex Nikitin]

Some more Keithley fun  .

The first graph is 100mV output on 200mV (left scale) and 20V ranges (right scale), 1uV/div. Very good performance from the voltage source.

The second graph is 10pA current supplied into 100G 1% resistor (covered by a bit of foil and suspended in alligator clips over the bench). The 2015 measures the output of the internal electrometer, so for 10pA and 100G it should measure 1V. It took about 20min to settle, but a 10min run on the graph is quite impressive. Essentially a stable reading of 99.4G .

Soon I should receive triaxial bits and cables, and then will build a proper enclosure for this kind of measurements.

Cheers

Alex

[Alex Nikitin]

More I work with the Keithley 263 and learn about it,  more I like it. It can use a self-calibration ("Ladder") technique for calibrating high-Ohm ranges (1M to 100G) internally, though a direct external calibration/verification is also possible. It is also using a temperature compensation for high-Ohm and low current ranges, where the calibration is done at two temperatures ("cold" and "hot" values) and the actual value used is calculated by the microprocessor from internal temperature readings continuously monitored  by a special sensor. In the resistance mode the microprocessor displays the calculated value and in the current mode it adjusts the voltage source appropriately to correct for the resistance change.

Cheers

Alex

[Alex Nikitin]

It looks that the K263 is stabilizing now as it is continuously switched on. Here is a 3hours run today for 10V output. The temperature in my home lab was quite constant during this time (less than 2 degrees variations). Measured by the Keithley 2015, NPLC10 + 10 averages. After the New Year I plan to take the K263 to my work lab and do some better measurements with the HP3458A, including some linearity tests.

Cheers

Alex

[Alex Nikitin]

Here is 10V output of the Keithley 263, measured over 96 hours. The temperature variations in the lab here were about 5 degrees C, from 19C to 24C, and it looks like the voltage variations are due to the temperature changes only, with about 1ppm/C tempco.

Cheers

Alex

[Alex Nikitin]

Here is a nice Christmas present for me (and for my Keithley 617 + 263). I've bought this lot a couple of weeks ago from the US and it was delivered this morning  . I have few meters of Belden 9222 waiting and will have some proper triaxial cables soon!

Cheers

Alex

[VintageNut]

Here is a nice Christmas present for me (and for my Keithley 617 + 263). I've bought this lot a couple of weeks ago from the US and it was delivered this morning  . I have few meters of Belden 9222 waiting and will have some proper triaxial cables soon!

Cheers

Alex

Please take and post pictures of the build process.  Merry Christmas!

[Alex Nikitin]

Please take and post pictures of the build process.  Merry Christmas!

Merry Christmas to you too!

Next time I will try to take some pictures. Unfortunately, this lot is quite old, and the assembly diagram for the current Pomona 5056 has nothing to do with it! So it was not easy, but I've built one triax to triax cable and connected the 617 with 263. Works quite well, actually (with the cable connected and the end screened, the 617 shows no increase in the leakage current <1fA) . Below are some photos of the results (apologies for the quality, it was fairly dark in the lab).

Cheers

Alex

Here 10fA is sourced and measured (the residual "zero" current from the 263 was suppressed on the 617)



The same for 100fA



For 19pA



100G resistance sourced and measured.

[CalMachine]

The 263 seems like a pretty nice unit with decent stability.  Here at the shop we are in need of something that can source a charge and this Keithley 263 is one of the candidates we've been talking about getting to fulfill that hole in our services.

[Alex Nikitin]

As the K263 performs rather better than expected, my plan is to improve the weakest point in this unit - the electrometer. The chip used - AD549L - was probably the best available at the time as a single chip solution, however the input current is in the area of 30-50fA and this creates a noticeable offset (and noise) in the active current sourcing mode. Also the AD549 does drift with temperature much more than the latest electrometer chips. From the look of it, the circuit allows for a simple upgrade with the ADA4530-1 . The error current can be reduced to less than 1fA and the temperature drift by over ten times. It also should improve the accuracy of the self-calibration routine and remove the null adjustment requirement (as the initial offset of the ADA chip is less than the temperature drift of the AD549 for 1C).

Cheers

Alex

[VintageNut]

As the K263 performs rather better than expected, my plan is to improve the weakest point in this unit - the electrometer. The chip used - AD549L - was probably the best available at the time as a single chip solution, however the input current is in the area of 30-50fA and this creates a noticeable offset (and noise) in the active current sourcing mode. Also the AD549 does drift with temperature much more than the latest electrometer chips. From the look of it, the circuit allows for a simple upgrade with the ADA4530-1 . The error current can be reduced to less than 1fA and the temperature drift by over ten times. It also should improve the accuracy of the self-calibration routine and remove the null adjustment requirement (as the initial offset of the ADA chip is less than the temperature drift of the AD549 for 1C).

Cheers

Alex

The 263 is a very good and useful instrument for low current ranges and high resistances. You appear to have an excellent specimen. Looking forward to your results from the modification.