Project for standalone use of Keithley low current preamplifier.

[TiN]

Now with electrometer I have ability to measure tiny currents, but yet don't have versatile programmable source for such, other than K6221. Gotta fix this. 

To resolve this issue, I thought to reuse preamplifier from Keithley 4200/6430 system together with my existing SMUs like 2400.
So here it is, preamp in it's glory:



It takes power and input current from the SMU (100uA - 1uA ranges) and can amplify or attenuate the output current by factor of 10 to 1000000, depends on selected range.
Beauty of it is bidirectional operation possiblity, so it can be used either to amplify DUT current for measurement, or attenuate current from SMU into DUT for low-current source.

Per patent block diagrams, both modes of operation are shown as concept:

Measure:


Source:


DUT force and sense terminals are of course triaxial with guard and floating bootstrapped supply.

Problem is, I don't have 6430 to figure out interfacing protocol/connections between this preamp head and the SMU, so we need to figure out that part ourself. There is no service or detailed information about interconnect on Keithley documentation either, to my despair. 

After careful teardown, it become obvious why these little boxes are so expensive (ebay listing usually over 1K$ a piece  ).



Everything is packed very tight, with custom alumina substrate thin-film resistor networks on long legs.
These are used to set range of preamplifier. Little chip in PLCC socket is TMS370C702CWT microcontroller. There is Xicor 24164 I2C EEPROM on the back side.
Maxim chip is MAX509AC quad DAC in TSSOP.

Bottom side is busy as well:



I counted no less than ten TI LMC662 very low current bias amplifiers, bunch of NAIS V225N5's, three NAIS TQ2SS-L-5V latching relays.

Anybody have idea what are those SOT23-5's with marking? 

A00 A
AAAG
AAAH
D3E "05 on side"

And SOT23-3's? 


C9G
V2P6C
6X mirrored "L"
Top "CH on side"


Any help would be greatly appreciated.

I plan to restore full schematics of the thing, so we can adapt it for use with instruments other than rare Keithley 6430.
Board is 4-layer so it will be tricky. Either way I will have to desolder ceramic resistor networks to gain access on the front side. 

According to patent drawings, overall simplified schematics should resemble something like below:



So far I determined next pins on HD15 connector to SMU:

Pins 1,2,6,9,11 are not connected and unused.
Pin 10 is signal LO
Pin 4 is VCC power for I2C EEPROM/digital logic
Pin 12 is digital ground

Pins 3,5,7,8,13,14,15 are unknown.

[Andreas]

http://www.smdmark.com/en-US/search/code?id=aaah
says MAX8877 LDO regulator in 3.3 (G) or 3.6 V (H)
or FAN2508S adjustable for AAA

And RB411D for D3E (Schottky diode)

With best regards

Andreas

[Berni]

Who other than TiN to take one of these apart.

That PCB design tho . The thing is like swiss cheese, more isolation slots than actual PCB. I could imagine there board manufacturer and assembler not being too happy about such a delicate and floppy board.

What i am curious is how the actual range switching is done, the diagram doesn't seam too specific about it.

[VintageNut]

Hello TiN

I followed your link to the pictures of the preamps. What you have is the preamp for the 4200-SMU and 4210-SMU which are cards that go into a 4200-SCS and the now 4200A-SCS.

The preamp for the 6430 is different enough from the 4200-PA that they are not interchangeable. This is second hand info to me. I have not tried switching preamps between a 6430 and a 4200-SMU. I have used both systems and they both have  incredible low current performance for both measuring and sourcing.

[TiN]

Well, as I already found out there is little uC, which may tell host "i am 4200 preamp" or "i am 6430 preamp".
I doubt there is much hardware difference between either flavour, unless you have photos of guts of "true 6430" preamp?

Anyhow, I don't have 4200 or 6430, so it all irrelevant. Question is if we can make this useful with "normal" SMU like 2400 and some Raspberry Pi in between to talk with PA's MCU?

I removed magic resistors to gain access on front side. Resistor with 9 pins measured as "19.81K isolated element + 200K+2M+20M+200M+2G+20G network", which follows block-diagram nicely.
However second ceramic resistor does not give any reliable readings. Maybe it's dead?

[VintageNut]

Well, as I already found out there is little uC, which may tell host "i am 4200 preamp" or "i am 6430 preamp".
I doubt there is much hardware difference between either flavour, unless you have photos of guts of "true 6430" preamp?

Anyhow, I don't have 4200 or 6430, so it all irrelevant. Question is if we can make this useful with "normal" SMU like 2400 and some Raspberry Pi in between to talk with PA's MCU?

I removed magic resistors to gain access on front side. Resistor with 9 pins measured as "19.81K isolated element + 200K+2M+20M+200M+2G+20G network", which follows block-diagram nicely.
However second ceramic resistor does not give any reliable readings. Maybe it's dead?

My guess is that the analog section of the 6430 preamp is identical to the analog section of the 4200-PA. The difference is probably MCU firmware.

I do not have a 6430 preamp. They are much more rare than a 4200-PA. I bought two of the 4200-PA from eBay as well as two of the SMU boards. Hoping for a working 4200A mainframe some day.

I agree that a Raspberry Pi talking with the 4200A preamp would be useful. I have two of these preamps and I would participate in a group funded project to make a working interface. Right now, they are expensive bricks.

[rsx90]

Hi TiN,

Thumbs up for the interesting project!

The markings could possibly correspond to the following parts:

A00 A -> LMC7101AIM5 opamp
AAAG -> MAX4516  NO analog switch
AAAH -> MAX4517 NC analog switch
D3E "05 on side" -> RB411D schottky diode as Andreas noted before

A03 A-> LM7121IM5 opamp

C9G -> CMPZ4684 3V3 zener
V2P6C -> Maybe V2p? If so, it's a BFQ67 NPN wideband transistor, not sure though. Another plausible option is V26C which stands for DZ23C30 dual 30V zener.
6X mirrored "L"-> MMBFJ176 P-ch JFET
Top "CH on side" -> could also be TOP,  T0P, or just OP/0P but could not find anything useful

Best regards,

rsx90

[TiN]

Great, thank you rsx90, this is helpful.

I managed to remove ceramic resistor networks from the module to gain access to the top side:



Time to start drawing some schematics... 

[AG7CK]

...
The preamp for the 6430 is different enough from the 4200-PA that they are not interchangeable. This is second hand info to me. I have not tried switching preamps between a 6430 and a 4200-SMU.
...

https://forum.tek.com/viewtopic.php?f=24&t=131561

[VintageNut]

...
The preamp for the 6430 is different enough from the 4200-PA that they are not interchangeable. This is second hand info to me. I have not tried switching preamps between a 6430 and a 4200-SMU.
...

https://forum.tek.com/viewtopic.php?f=24&t=131561

My opinion is that thread is not conclusive. ...and...since TiN does not have a 6430 nor a 4200, the point is moot. Call me the master of moot points. I deserve it.

[Brad O]

The 4200-PA is NOT the same as the 6430 preamp, though I understand the confusion because they seem quite similar from a user perspective.  Some parts are used between both of them, but they are not compatible or interchangeable.

Some of the main differences are that the 4200-PA has 100ohms between it's force and sense terminals, the 6430 preamp has something like 1Mohm between input/output high and sense.  The 4200-PA sense outer shield is also sense LO, while the 6430 preamp has input/output LO on the sense outer shield, so you need to use the back of the box for sense LO.  The 6430 preamp has higher output impedance and probably has lower current resolution, though I think they're spec'd about the same.  Even the communication is different between them, and they have different processors that control them internally.

Very interesting project TiN!  If anyone could figure it out, it would probably be you, please let me know if you get it working!  I think you have a long road ahead of you though, all the engineers here I talked to said it would be very difficult to adapt either preamp to work with another instrument.

[TiN]

Thank you, Brad.

I'd like to express my gratitude for your contributions here and number of other threads, that alone scored lot of points in my book for Keithley. 
Almost enough to make me want to buy one of those DMM6500 just in sake of completeness and comparisons purpose for our little Keithley Instruments collection.

Talking about which, I just got fresh delivery of 1801 head today:



It's easier, in relative terms (if you can say "easy" about 300 picovolt noise-level amplifier...  ). Another project to never-ending list...

[VintageNut]

Thank you, Brad.

I'd like to express my gratitude for your contributions here and number of other threads, that alone scored lot of points in my book for Keithley. 
Almost enough to make me want to buy one of those DMM6500 just in sake of completeness and comparisons purpose for our little Keithley Instruments collection.

Talking about which, I just got fresh delivery of 1801 head today:



It's easier, in relative terms (if you can say "easy" about 300 picovolt noise-level amplifier...  ). Another project to never-ending list...

Glad you scored that 1801. I heard that you were bidding on it so I stayed out...and... I do not own a 2001 nor a 2002.

[TiN]

Well, I couldn't resist and cheated on this project. Since I got bit stuck with 4200PA's onboard MSP microcontroller protocol, I've bought real deal Keithley 6430 preamplifier, and borrowed actual 6430 SMU unit from another fellow volt-nut for a test. Happy enough, this setup worked as expected, sourcing tiny currents with breeze.

Lowest range 1pA in action:



Pushing the limits, right picture shows +5 / -5 and 0 femtoampere current into Keysight B2987A electrometer.
That is 0.000000000000005 of an Ampere, so tiny! 



Actually B2987A is too noisy and have more noise than Keithley 6430 preamp itself.
I'll do some more tests with 6430PA and then try to carefully take it apart, so we all can finally see what is different.

This should hopefully enable us to modify 4200 PA's into 6430 version. 

[Alex Nikitin]

Just for fun of it, I've tried the same +/-5fA measurement on my Keithley 263 + 617 combination. Not quite as good but still a decent performance. The scale on the graph is 1fA/div vertical, 10s/div horizontal, 500ms sampling rate.

Cheers

Alex

[TiN]

My own Keithley 6430 fA-SMU repair update. Thanks to TheSteve, xDevs.com also acquired own broken 6430 unit (w/o PA), with diagnosis "no output". 
Received very well packed device, leaving no chance to easy damage of fragile instrument:



First smell check didn't reveal any issue, so turning on confirmed display operation, responsive buttons and only factory calibration, with count = 1.



Calibration dates as received and VFD segments brightness are good. Looks like box was never calibrated after factory. 

Check with DMM at the output indeed confirmed seller's diagnosis, no output for either current or voltage. Four screws and 5 minutes later:



Analog PCB is based from Keithley 2400 SMU, with exception of 1.05A range. Good old Keithley 2400 dual-slope integrating ADC, which is in fact Keithley 2000 ADC + additional digital stuff to control onboard I-DAC and V-DACs and handle source-meter functions like sweeps, guarding, etc. Sadly reference is not LM399, but a compensated zener diode.

Photo of K2400 analog PCB shown on next image to refresh memory.



There are additional modifications for external 6430 very low-current amplifier. These include isolated 1W DC-DC "Murata/Power Convertibles HL01R05D05Z":https://xdevs.com/doc/Keithley/6430/pdf/tdc_hl01rzc-55873.pdf and extra optocouplers for SPI interface to 6430 preamplifier. DC-DC specified for 100mV[~pk-pk~] max noise in DC to 10MHz bandwidth and isolation capacitance just 25 pF.



Closed look revealed the problem  :



Look closer in top right corner. See anything obvious? It can't be only that, right? It would be too easy. 



But yes, after restoring connection to analog board power supply input indeed made unit fully functional again.
So much for repair in The Signal Path style  . I almost feel violated, that we had no chance to dig into SMU analog board troubleshooting 



Since we in here anyway, let's remove analog board and see on the digital/power supply boards located under.



Keithley 6430 using special multicore planar transformer assembly on 16-layer PCB to provide low-noise high-isolation power to analog board. Anybody knows what is all this magic about?
I'd guess this transformer design used to provide low leakage in power supply DC/DC that feeds analog board output stage?



Standard Keithley 2400 does not have any of this fancy stuff, but overall design is very similar.

Otherwise this is just usual Keithley SMU and it works exactly same way (minus the 1.05A current range) as 2400 when Keithley 6430 amplifier is not connected.
There are no front terminals available with K6430, so output/input provided at the rear ports only.

Quick calibration and checks for 10V and 1V source:





Noise is not great, so perhaps we should pimp this unit with LTZ1000A reference, aye? (Not that K6430 is designed to be low-noise voltage source, it's performance marvels lie in picoampere current domain with preamplifier head).

Calibraiton of the box is "simple", with help of 3458A and set of 100Meg, 1Gohm, 10Gohm and 100Gohm resistors (usually part of Keithley 5156 kit, which I do not have (yet?)).
I didn't tried it with PA yet.

[antintedo]

Keithley 6430 using special multicore planar transformer assembly on 16-layer PCB to provide low-noise high-isolation power to analog board. Anybody knows what is all this magic about?
I'd guess this transformer design used to provide low leakage in power supply DC/DC that feeds analog board output stage?

I can't tell from the photos what the layout is, but if one core is used for primary and others for secondaries, the resulting coupling capacitance can be much lower than with traditional design. "Design and Evaluation of a 10-mA DC Reference Standard" by G. Fernqvist mentions this architecture briefly.

[Kleinstein]

Keithley 6430 using special multicore planar transformer assembly on 16-layer PCB to provide low-noise high-isolation power to analog board. Anybody knows what is all this magic about?
I'd guess this transformer design used to provide low leakage in power supply DC/DC that feeds analog board output stage?

I can't tell from the photos what the layout is, but if one core is used for primary and others for secondaries, the resulting coupling capacitance can be much lower than with traditional design. "Design and Evaluation of a 10-mA DC Reference Standard" by G. Fernqvist mentions this architecture briefly.

With separate cores for primary and secondary it's not a transformer anymore. There is also too little magnetic coupling.

One could sometimes use several cores instead of one larger one to get higher power and still keep the low form factor.  It is possible, but would be a little odd.  For very low coupling one might use several transformers in series.  It is easier to get low coupling with low voltage and thus fewer, though thicker windings. So a possible choice would be to first go to a low secondary (e.g. 1 V range) than a low voltage to low voltage  (e.g. only something like 2 turns - 2 turns) and than in a 3rd transformer to the final voltages.

One advantage of  the planar transformer is that the coupling capacitance is predictable and reproducible.  So one can find a good balance to get coupling from the right, neutral spots, effectively like having extra shields. Capacitve coupling can be low, but no real extra advanatage there over a more traditional transformer.

[TiN]

I think output of this transformer should be in ballpark of 210-250V bipolar, because SMU have highest range 200V.

[antintedo]

With separate cores for primary and secondary it's not a transformer anymore. There is also too little magnetic coupling.

Quote from the paper in case my explanation was confusing:

The double-screened transformer uses two toroidal cores, independently wound, one each for primary and for secondary. Each is (...) coupled an external “shorted turn” winding around both cores. The resulting coupling capacitance from primary winding to screen is less than 0.1 pF and in use 10 nA of -pk AC coupling current is achieved with at least times more in most commercial units.

So yes, effectively what you described as two transformers in series. They mention a patent but it's pretty hard to find its number. There is probably something more to the physical construction than stated in the publication.

[arlo_g]

Hello TiN,

I realize that this is an old thread, but the fantastic photos that you posted of the Keithley preamps here and on xdevs.com have me wondering how they implemented the shunt swithching.

There seems to be an LMC7101 with its inverting input connected to the switched end of each shunt, as well as an LMC662 with one of its non-inverting inputs connected to each shunt(on opposite faces of the PCB. I attached a copy of one of your photos with opamps on two shunt pins sketched in).  This pattern is also seen around the low current shunts in 2612 and 2636 SMU's.  It seems very likely that the LMC662's are used as a voltage follower to monitor the voltage on the swithched end of each shunt, and they are visibly driving guard rings.  The LMC7101's might be there as a voltage follower to drive the output current into the shunt in use.

Getting the leakage current from LMC7101's on unused shunts down into the fA range seems like a challenge.  I think that driving the SMU output (guard) potential into the switched end of unused shunts is a non-starter because the 6 decade range of shunts would lead to unacceptable leakage current in the lowest current range from even uV offset errors across the lowest value shunts. Analog switches between opamp outputs and shunts might need guarding (bootstrapping to 0V across switch channels) to achieve fA leakage.

Switching both supply pins (and likely the non-inverting input) of the LMC7101 on each unused shunt to (SMU output) guard potential would set their output transistors' channels and gates as well as ESD diodes all with 1mV or less potentials across them: this could easily drop the 7101 output currents to pA range, and maybe it's enough to get down into fA's.

Is there any chance that you traced the supply connections of any of the LMC7101's connected to shunts?

[pattant]

hello,

there is lots of interesting information in this thread! I am in a different position from most it seems: I have a couple 6430 source meters, but no remote pre-amps to go with them. Recently some noise issues made me think of the pre-amps again.  It seems like the 6430 remote pre-amp is hard to obtain, while the 4200 pre-amp is available...

so my big question is: has anybody tried the 4200 pre-amp with the 6430 source meter?

[42Khz]

Hello TiN,

I realize that this is an old thread, but the fantastic photos that you posted of the Keithley preamps here and on xdevs.com have me wondering how they implemented the shunt swithching.

There seems to be an LMC7101 with its inverting input connected to the switched end of each shunt, as well as an LMC662 with one of its non-inverting inputs connected to each shunt(on opposite faces of the PCB. I attached a copy of one of your photos with opamps on two shunt pins sketched in).  This pattern is also seen around the low current shunts in 2612 and 2636 SMU's.  It seems very likely that the LMC662's are used as a voltage follower to monitor the voltage on the swithched end of each shunt, and they are visibly driving guard rings.  The LMC7101's might be there as a voltage follower to drive the output current into the shunt in use.

Getting the leakage current from LMC7101's on unused shunts down into the fA range seems like a challenge.  I think that driving the SMU output (guard) potential into the switched end of unused shunts is a non-starter because the 6 decade range of shunts would lead to unacceptable leakage current in the lowest current range from even uV offset errors across the lowest value shunts. Analog switches between opamp outputs and shunts might need guarding (bootstrapping to 0V across switch channels) to achieve fA leakage.

Switching both supply pins (and likely the non-inverting input) of the LMC7101 on each unused shunt to (SMU output) guard potential would set their output transistors' channels and gates as well as ESD diodes all with 1mV or less potentials across them: this could easily drop the 7101 output currents to pA range, and maybe it's enough to get down into fA's.

Is there any chance that you traced the supply connections of any of the LMC7101's connected to shunts?

Hi arlo_g,

I just spent some time staring at TiN's excellent PCB photos trying to figure out how Keithley did the current ranging in at least the 6430 preamp (primarly this one). The 1k and 10M resistors caught my attention and suddenly it hit me, they're using (a modified version of) the arrangement in their US patent 5,994,947!

The gist of the patent is that they use a nice non-linear impedance (diodes in the patent) which they can either guard to achieve extremely low leakage when the switch is supposed to be off, or include in a feedback loop to compensate for their forward voltage when the switch is on.

The photograph clearly shows that pins 1,2 and 3 of the LMC7101 are shorted together, so it cannot possibly be configured in a traditional opamp configuration (2 supply contacts and then either an input or an output?). That fact coupled with the patent description must mean Keithley are using the input protection diodes on the LMC7101 as diodes 38, 40 in the patent. I would guess that they're fairly tiny and reliably low leakage.
From that point it just takes matching the patent to the PCB, although I believe the node shared between the 176 P-JFET and Maxim switch gives away that Keithley deviated slightly from their patent circuit. I've indicated the difference in my schematic diagram below.

I hope that helps!


PCB with some scribbles:



Schematic. I didn't check whether the circuit is really the same for the 4200 preamp but I would be surprised if the current ranging was different.

[Noopy]

If you are interested in the internals of the LMC662, I have posted them here:

https://www.eevblog.com/forum/projects/opamps-die-pictures/msg4651233/#msg4651233

https://www.richis-lab.de/Opamp62.htm