EEVblog 1755 - Tolicore Femtoammeter TEARDOWN + REVERSE ENGINEER

[EEVblog]

Demo, teardown and reverse engineering of the Tolicore 3720 Femtoammeter and voltage source.
https://tolicore.com/products/3720-femtoammeter-voltage-source/
https://tolicore.com/product-documents/3720-Datasheet.pdf
Keithley low current measurement handbook: https://www.tek.com/en/documents/product-article/keithley-low-level-measurements-handbook---7th-edition
Counting LED photons video:

00:00 - Femtoammeter measurement
01:16 - Tolicore 3720 Femtoammeter unboxing
03:39 - Power on and UI
05:18 - 20pA full scale range
10:13 - Bluetooth and
11:35 - Humidity measurement is important
12:21 - Crazy sensitive!
13:45 - LED as a current source demo
15:28 - 1nA measurement test
16:43 - Burden voltage and electrometer input + self calibration
20:14 - Teardown
22:20 - Voiding the warranty
25:05 - Reverse engineering
27:32 - How a PCB guard ring works
33:52 - Humidity matters
34:33 - The voltage output
35:23 - Porcelain Multilayer Capacitors
36:22 - Voltmeter input
38:24 - Potentiostatic and Amperostatic Coulometry
40:15 - ADA4530-1 Internal schematic
42:25 - Contamination and PCB assembly cleaning matters
44:03 - Rogers PCB material

[Kleinstein]

It is a bit unusual that the autocal part actually adjusts the offset voltage at the current input, and is not just measuring gain parameters of some sort and do the math to correct the result.
So there must be some DAC to phsically set the offset voltage for the electrometer amplifier.

It looks like the circuit with the relays and resistors is actually a bit different than shown. It looks like the relay contacts are in series for the higher current ranges. This way, when in the lowest current range on the leakage from 1 relay is effective. More of the relay only get relevant when higher ranges are used and the higher current part can get away without the via fence. The highest current part can even get away with the solid state relay (can help with the protection).

In some acpects the old Keithley meter can still be superior: they allow for a higher voltage (AFAIK up to some 200 V for some meters)  at the feedback resistors and this way can use higher resistors and get lower noise in the medium to lower ranges. Only a 5 V OP-amp is somewhat limiting, but could be from power constraints.
The Keithley meters usually have the same amplifier input also for super high impedance voltage measurements. The separate voltage input can be nice, but is likely not all that high in impedance.

[David Hess]

The National LMC series of 15 volt CMOS operational amplifiers are competitive with the performance of the ada4530.  Eventually National provided the LMC6001 which is tested to 25 femtoamps; in my experience they are all below 2 femtoamps.  (1) The ada4530 is surely better, but at ten times the cost.  The ADA4530 package alone might be worth it because of the better pinout.

I am dubious of the ADA4530 input bias specifications because the simplified schematic shows the guard buffer driven by the non-inverting input, but the input bias current specifications show no difference between the inverting and non-inverting inputs.  Maybe their published simplified schematic is deceptive.  Didn't Burr-Brown have a similar part with the guard buffer driven after the input stage to avoid this?

(1) I built a quick and dirty electrometer using a Texas Instruments TLC2201 because that is what I had in reach, and even it was below 10 femtoamps despite its much worse datasheet specification.

It is a bit unusual that the autocal part actually adjusts the offset voltage at the current input, and is not just measuring gain parameters of some sort and do the math to correct the result.
So there must be some DAC to phsically set the offset voltage for the electrometer amplifier.

I was thinking of designing an extreme precision electrometer amplifier by using separate input offset and input bias current adjustments, with the input bias current influenced by adjusting the common mode input voltage of the bootstrapped follower.  Maybe they did something similar, or this would not reliably work.

In some aspects the old Keithley meter can still be superior: they allow for a higher voltage (AFAIK up to some 200 V for some meters)  at the feedback resistors and this way can use higher resistors and get lower noise in the medium to lower ranges. Only a 5 V OP-amp is somewhat limiting, but could be from power constraints.

The ADA4530 has specifications for 15 volt operation, like the old high voltage linear analog parts from National (LMC) and Texas Instruments (TLC).

[ballsystemlord]

@Dave , your video left me with more questions than answers.

Where is the 10Gohm resistor's capacitor?
Where are the ADCs for the Volt meter and current meter? What are they?
Is there anything special about the power supply (18650 battery) to avoid the switching noise coupling through?

Also, I didn't follow the Dave-CAD at all. I can see the electrometer in series with the AD8655 which feeds back into the electrometer's negative input. But I have no idea...

How does the circuit complete so that the DUT gets current into it if it just feeds into the elecrometer? Why does the amplifier (AD8655) feed into huge resistors and then back into the DUT's positive current input path? Like, you don't want to feed power backwards into the DUT! Maybe it's that I don't understand how electrometers work?

(See pictures.)
What are those cylindrical black things on the special board?
What is that one green cap-looking thing with a line down the middle? It looks like an SMD film cap, any reason for that?

Anyway, thanks for the video. It is an awesome tool!

[Kleinstein]

There is not much need for much capacitance in parallel to the 10 Gohms. This could be just parasitic capacitance. There may still be a capacitor on the backside (could be an air capacitor).

The electrometer type currentmesurement may actually drive some voltage to the DUT. The would be the offset of the amplifier(s). So to get zero current one should not shorten the input, have an open input.
Why they use the 2 amplifier version here is not so clear to me for this case. It would make sense if one targets a large output voltage (like +-10 V or +-100 V), as this way the critical input amplifier can run on low voltage (like +-3 V) and the driver could use higher voltage. This also makes sense when there are other functions, like read voltage from the same input.

The small cylindircal black parts are diodes. They are marked as D... and likely there for the protection of the higher current ranges ( 20 mA, 2 mA, maybe 200 µA).

[gamalot]

What are those cylindrical black things on the special board?
What is that one green cap-looking thing with a line down the middle? It looks like an SMD film cap, any reason for that?

The black ones are mini‑MELF diodes, and the green ones are likely acrylic resin film capacitors.

https://www.onsemi.com/download/data-sheet/pdf/fdh300-d.pdf

[Hydron]

Keithley 617 runs the TIA resistors up to +/- 20V, but bootstraps the input/gain stage so it sees less voltage than this. Each resistor also covers 3 ranges, IIRC with x10, x1 and x0.1 at the ADC input to handle the scaling. It does have a higher voltage amplifier section (200V) but that's for Volts/Ohms.

People have dabbled with using the ADA4530-1 as a replacement/improvement for the unobtanium dual JFETs on the input of the K617. TI has a similar part (OPA928) that I'm going to have a go with myself.

As for capacitor across the 10Gresistor, there's hints that this might be part of the PCB layout itself - there's a couple of flood fills that seem to be designed to add some controlled capacitance to this node, one linked via a resistor, the other unused (maybe to allow for further tweaking if needed?). I wish we could see the bottom of the board.

[ballsystemlord]

What are those cylindrical black things on the special board?
What is that one green cap-looking thing with a line down the middle? It looks like an SMD film cap, any reason for that?

The black ones are mini‑MELF diodes, and the green ones are likely acrylic resin film capacitors.

https://www.onsemi.com/download/data-sheet/pdf/fdh300-d.pdf

Just to be clear, I wasn't asking if the green one was a film cap so much as I was asking what is it doing? Normally, AFAIK, you wouldn't need film caps on non-power supply stuff.

[gamalot]

What are those cylindrical black things on the special board?
What is that one green cap-looking thing with a line down the middle? It looks like an SMD film cap, any reason for that?

The black ones are mini‑MELF diodes, and the green ones are likely acrylic resin film capacitors.

https://www.onsemi.com/download/data-sheet/pdf/fdh300-d.pdf

Just to be clear, I wasn't asking if the green one was a film cap so much as I was asking what is it doing? Normally, AFAIK, you wouldn't need film caps on non-power supply stuff.

That chip by the arrow in your pic (U20) is the well known low noise LDO LT3094.

[EEVblog]

As for capacitor across the 10Gresistor, there's hints that this might be part of the PCB layout itself - there's a couple of flood fills that seem to be designed to add some controlled capacitance to this node, one linked via a resistor, the other unused (maybe to allow for further tweaking if needed?).

Yes, I noticed that too after the edit.

[Kleinstein]

Using the PCB matrial for the capacitor makes some sense, if the material is really good. It would explain why they have a relatively large area for the critical input node. Normally one would want to keep the area small to reduce leakage.
The not connected copper island is however normally a no go - it kind of adds a slow settling part with the capacitance and leakage.

[David Hess]

Using the PCB matrial for the capacitor makes some sense, if the material is really good. It would explain why they have a relatively large area for the critical input node. Normally one would want to keep the area small to reduce leakage.
The not connected copper island is however normally a no go - it kind of adds a slow settling part with the capacitance and leakage.

That can work, however most printed circuit board material is not qualified for that application and can suffer from hook and crippling amounts of dielectric absorption.  Robert Pease discussed high value carbon composition resistors which had inherent dielectric absorption.

I have occasionally seen schematics which show a dotted line on one or both sides of a high value resistor to indicate electrostatic shielding, which might be just a thick copper trace under the resistor connected to the low impedance end.

This gets into the realm where I have occasionally used a gimmick capacitor, which is a short length of twisted wires, or where shielding was important, a short length of Teflon coaxial cable.

[Hydron]

I think the rogers material used is PTFE based, so maybe this is an instance where it's good enough. Without seeing the back of the board it's hard to see exactly what has been done, there are a couple of other things I'm curious about too (like where the vias go from the second biggest range resistor, and what is going on with the higher current ranges after the 1M feedback resistor).

[BD1QMP]

What role do the reverse-parallel diodes play when placed in series with the feedback resistor?

[David Hess]

What role do the reverse-parallel diodes play when placed in series with the feedback resistor?

So they are between the output and feedback resistor, with the true output taken between them?

[thm_w]

I assume the diodes on the left side are first, then ones on the right are connected when the SSR is on.
The ones on the left are a soft clip, and the right side a harder clip?

https://electronics.stackexchange.com/questions/655570/how-does-anti-parallel-diode-for-limit-current-work