Sourcing nanoamps and below like a Samurai, or KEI 6221 DC/AC SMU teardown

[TiN]

Well, yes, I've got yet another Keithley piece. Need a cure already!

This time it's 6221, DC/AC precision low-noise current source. Essentially it's an 11W 4-quadrant SMU, but without measurement functionality and much improved current source performance.

To my memory, I never saw anyone posted guts of 6220/6221 anywhere online before, so enjoy!

Exterior, controls and initial condition

   
Keithley 6221 front panel

   
Keithley 6221 rear connectors

Internal construction, design features

   
Cover removed, analog board overview

   
Analog board without shields

   
Analog domain FPGA, rear section and interlock circuitry


ADC/DAC and main reference section

        
Analog board power supply

        
Common mode inductor, perhaps? I wonder what this thing does exactly.

   
PCB cutouts to avoid leakage

   
High-current range relays and resistors

   
Medium current range relays and resistors

   
Low current range high value resistors

   
Output power stage

        
High impedance, low leakage amplifier front-end

        
High impedance low current circut design

   
Auxilary bias?

        
Fully shielded low-noise power supplies

        
Mu-metal shielding and board label

        
Triaxial 3-lug connector and input circutry

        
Voltage reference section and auxilary circuts

   
Relay drivers

Diagnostics and secret menu

By holding PULSE+RECALL during power on we can enter secret menu section.
This also unlocks self-diagnostics in TEST menu. I think they hide self-diagnostics for safety reasons, as 6221 can generate potentially lethal voltage levels on it's output.

   
Vacuum display burn-in mode for segment brightness correction

   
Test with 10pA current with Keithley 6485

Repair

No repair was required.     

Calibration

Calibration procedure in manual calls for HP 3458A 8.5-digit multimeter and Keithley 6487 picoammeter. Unlucky me, these ones I have, with little deviation by using Keithley 6485 instead of 6487. DMM is used to verify current ranges from 100mA to 100nA and picoammeter take care of lowest 10 nA and 1 nA ranges.

Performance test

TBD

Application example, measuring resistance

TBD. I will use this source for one of it's typical purposes, source current thru resistors to measure tempco. Before I was using Time 9823 calibrator or HP 3245A in current source mode, but these guys can't provide compliance higher than 10-something volts, which limits resistance measurements to about 50-80 KOhm. With 6221 I can extend voltages up to 100V, and that should let me measure few Meg resistances with better uncertainty. I don't have 2182A for delta measurements, but I have Keithley 2002 + EM A10 preamp which is better if I go milliohm-microohm regions

[e61_phil]

I used this unit a lot at the university to characterize transimpedance amplifiers (1G and above with several kHz of bandwidth). The 6221 is capable of a few kHz but the output signal isn't very pure. Therefore, I used it only for DC in the end and a photodiode for the AC measurements.
Nice unit

[ap]

Tin, what is your plan in measuring high-ohms resistors with it? The accuracy of these (as well as their picoamp-meters) is relatively low. Depending on the range, as bad as 0.3 to 0.4% (nanoampere ranges). The calibration with your 6485 would lead to a simliar error that adds. So you will be close to about 1% in current accuracy for the current. So will you be measuring current at a higher accuracy by using a lower value shunt and use the 6221 as a source only?
Also, not sure what you paid, but these tend to not be really cheap. If you wanted to measure say 100Meg, and you had access to e.g. a Fluke 5440 (or bought it at a probably much lower price than the 6221), you would be able to use your 3458A in the 10/1uA range (depending on voltage applied), where it has a 1-year accuracy of about 30/60ppm. Combined with the 5440 uncertainty, that would still be pretty precise.

[TiN]

I've got this box for around 1.3k, shipping included. I think it's decent price for box made in 2008, also given no repairs required.

Likely I'll use just as a source of stable current, as for TCR measurements absolute current does not matter, as far it stays stable and low noise. Will be interesting to see how current output compare to other sources, such as Time 9823, HP 3245A, Keithley 2400 SMU, Agilent 4142B(41424B) SMU, HP 3458A ohm source and also DIY current source using opamp, foil resistor and LTZ1000-based ref.

F5440 is out of reach for me, due to it's shipping weight, even if I get one for great price, shipping cost and difficulty would kill the deal. Otherwise I wouldn't bother with Time electronics 9823 MFC. Low current ranges on 3458A are pretty noisy. Also I'll be recalibrating both 6485 and 6221 anyway.

[ap]

Sounds like a pretty good price, and if you are only measuring trends, then the fact that its not very precise or even not calibrated will indeed not hurt you.

[plesa]

Thanks for teardown. It is pitty it is so expensive. I suppose that 26xx series killed marked for this instrument.
Strange is how yellow is the wheel compare to rest of front panel. The most recent gear do not have print on wheel, they needs to save few cents

[SeanB]

Different plastics, panel probably is polycarbonate and wheel is a PVC formulation, so they age at different rates, and have different failure modes.

[Kleinstein]

They have a quite impressive DAC (16 Bits 50 MSPS - likely used for waveform generation). So chances are it could produce a really pretty, low distortion sine.

What is that large toroid core ?  It looks a little like a transformer, but very few turns.

[e61_phil]

My experience was a quite bad AC behaviour

[julian1]

Great pics. I saw that one on Ebay!

That inductor/choke looks like a piece of minimalist modern at.

Googling the components and there are quite a few high-speed/video bandwidth devices. Several buf634 buffers with 180MHz bandwidth, and the maxim4545 which are RF switches from DC to 300MHz with controlled impedance. The ltc1668 is a DAC with 50MSPS. The current ranging looks to be all switched with relays - perhaps to better handle the AC?

[e61_phil]

I found a picture which I measured five years ago. I'm not sure, but I think it was around 1nA at 1kHz. Measured by a Femto DLPCA-200. I tried another 6221 from another department. That unit showed the same results. Therefore, I went with photodiodes to measure low AC currents.

Is it possible to measure something like that for comparison?

[Kleinstein]

That low current curve looks like it has quite some switching noise. I would expect the source more at one of the power supplies, not the signal generation itself.

[e61_phil]

That low current curve looks like it has quite some switching noise. I would expect the source more at one of the power supplies, not the signal generation itself.

Wherever it may come from, it is quite ugly for TIA characterization.

[TiN]

I'd expect 1nA @ 1kHz level measurement is itself already challenging, very outside of usual approach "hook wires together, plonk buttons, enable output and bob's ur uncle".
Based on previous posts and tests with Fluke 5700/Datron 4808 even microamp currents at AC are already difficult and depend a lot on cabling, shielding, isolation, etc.

Anyhow, once I get to performance tests, I plan also to test AC performance.

I've also bought same DAC devboard while ago for calculable AC reference project idea.



So it was fun to discover same chip used in 6221.

[VintageNut]

The output impedance of a current source is very high. The HI conductor is the high impedance of the two HI and LO. If this HI conductor is not shielded, electrical noise WILL be impinged on this conductor. If there is not a proper guard, leakage from HI to LO will happen.

This is why electrometers measure current on the LO side and use a low impedance voltage source. Much less prone to noise problems.

The Achilles heel of a current source is that you have to guard and shield the output otherwise you will have leakage and/or noise problems.

Sourcing mA or uA is easy-breezy. Sourcing nA with low noise and low leakage requires some know-how and very good cables as well as proper guarding and shielding of the device being powered. 

[CalMachine]

I would like to join the nanoamp sourcing club.     My new beauty I got earlier this week is going to be a great addition to my home lab.  I, first, had to take her in to get a fresh adjustment    Though, low current is one of our weak spots where I work...  I know DC is not nearly as difficult to achieve the low levels you have with AC.  What kind of guarding was implemented?  Can you take some pictures of the cabling and test setup?

[TiN]

Whatever you do, keep those nice pics coming

[VintageNut]

Figure 2-2 is an artist's rendering of guarding a low current force/measure from the 6430 instruction manual.

Next is an implementation. The three alligators at the end of the cable are HI, LO and GUARD. HI and LO connect to the resistor. GUARD is connected to enclosure.

Finally, the enclosure is sealed and the measurement is taken. Settling time is hours to charge all of the air in the enclosure to GUARD potential. A stable reading is achieved but after a long wait.

[CalMachine]

I will do some more rigorous testing on the 237 here in the next few days, with lots of pics! 


Here we go!  Some more hurdles for the 237.

[CalMachine]

Let's check out the pA performance.

Both the 3458A and 237 were performing surprisingly well 

[CalMachine]

Now for the nA hurdles! 

[Vtile]

This is severe. TiN you need to get a metrology grade thermometer to measure the illness.

[plesa]

For one project I needs to compare the K6485 and 6517B @ 100pA. As a source was used K2602B.
As you can see the K6485 is little bit noisier ( second graph with NPLC25) compare to K6517B ( first graph, blue line NPLC 10). Green and yellow line is burden voltage, where K6485 has 2 times higher burden voltage compare to 6517B.

So the name Picoammeter for K6485 is little bit misleading
Any ideas what to compare with this setup?

[voltsandjolts]

Sorry to wake this sleeping beauty...
Just adding some info.

Keithley 6221 Interlock Connector

The interlock connector within the unit is Phoenix Contact 1825129
The mating part is Phoenix Contact 1825310

[bsw_m]

10aA source&measurement