Resistance synthesizer

[Tacotatop]

Hi humans of EEVBlog,

I have designed a resistance synthesizing test cell PCB for a project at work that test our chronoamperometry system, where a fixed voltage (0.3V) is held and actively driven to maintain the voltage and the current across the system is measured. The current that we are measuring is in the 0.1nA to 100nA range. The PCB is based on the ADG714 octal CMOS switch, and a bunch of large resistors. The PCB is a hat for a nRF52832 Dev Kit.

Currently I am facing an issue are
- the ADG714 is not switching properly
- Excessive leakage current (1uA)

What I have tried so far:
- Reflow everything in a reflow oven, and cleaned with IPA & ultrasonic cleaner
- Checked every solder joint using microscope + multimeter probing
- Check resistance using multimeter
- Testing using Palmsens4 and running the firmware/python code for the nRF52832DK (encountering same issue as above)
- Removing the ADG714 from the board, and manually shunting each pin in resistance value (everything checks out)

I am at my wits end here. Please help

Attached is the altium project documents for it

edit: added PDFs of schematic and PCB file after feedback

[ArdWar]

Are you really asking someone to download a random ZIP file on the internet? From a new user with zero post history?
Come on, post a snapshot image or at least PDF export of your design. Better chance at getting someone to look at your problem.

[jonpaul]

post jpg of schema

Suspect poor design

No time to risk your ZIP   3M

j

[moffy]

You are trying to maintain better than 0.1nA leakage current total across 3 ICs with 8 inputs and outputs each, with a ground plane and no special leakage guarding techniques on a standard PCB?! Low leakage requires appropriate techniques, design and sometimes materials such as teflon.

[shakalnokturn]

To add to the above: A good start would be to read the data sheets before designing, the ADG714 is specified for 0.1nA typical leakage per transmission gate.
I wouldn't be surprised if the "'not switching properly" was a side effect of the leakage current.

If I understand correctly your test setup's purpose is to simulate the presence of the electrodes and sample to analyze for sanity / calibration check of chronoamperometry equipment?

Wouldn't a current sense resistor and single controlled  FET to simulate the resistance be simpler? Add in digital control as required.
If it's a single prototype or small series you could stick to your original idea but use relays in place of the ADG714.
How many resistance values need to be simulated and how specific in value must they be? Isn't there a simpler setup (R-2R) that would get you the number of measurement points needed using less switches?

[Postal2]

I have designed a resistance synthesizing test cell PCB .......... The PCB is based on the ADG714 octal CMOS switch, and a bunch of large resistors. ......

It must be developed with photoresistor.

Wouldn't a current sense resistor and single controlled  FET to simulate the resistance be simpler?

Nonlinear.

[shakalnokturn]

Wouldn't a current sense resistor and single controlled  FET to simulate the resistance be simpler?

Nonlinear.

True, but then if you're sensing current and have a fast enough feedback loop (I'm sure this application isn't that fast...) it can pretend to be linear for the purpose.

[Postal2]

... if you're sensing current ....

See this one:

The current that we are measuring is in the 0.1nA to 100nA range. ...

pH-meter as example:

[Tacotatop]

The purpose of the test setup was to check linearity of our prototype system and a sanity check, so I was asked to design this test cell to automate the testing.

We had a previous design similar to this using DIP switches instead of the ADG714, and testing required a lot of manual work.

The chronoamperometry chip is the EmStat Pico Core based on the ADCuM355, which applies the voltage and measures the current on itself.

I am not that familiar with low leakage current designs so I will research design considerations for low leakage designs and ask my supervisors for help.

[Postal2]

I am not that familiar with low leakage current designs.....

Since the design is already ready, the easiest way is to remake it using golden or specialized relays.
Semiconductor switches cannot be used.

added PDFs of schematic and PCB file after feedback

The PCB is really crazy to achieve 0.1nA leakage! You can throw it away, nothing can be done with it.

[ArdWar]

With nanoampere signals you really need to take more care with component selection and board layout. Pick physically larger component for increased clearance (E.g. SOIC or DIP instead of TSSOP, 1206 resistors etc). Pull back ground plane further away or maybe don't use ground plane at all around the signals, you don't need it here. Keep voltage gradient between signals uniform and low, don't interleave them. The advice on using [reed] relays are indeed valid here, it's proven good tech when you don't need speed, and the physically larger size helps with signal isolation too. On more extreme case you may need to consider guards and cutouts, but I don't think it's necessary here.

CMOS switch are not generally recommended, usually due to charge injection in addition of potential leakage. In any case it should not be that (1uA) bad. Are you sure they're good parts? Static mishandling for example are good way to blow leakage spec away.

[jwet]

There are better CMOS switches for an application like this too.  You don't need a 2.5 ohm switch to switch 3G loads.  You pay a big penalty in leakage with these "large" switches.  The lowest leakage switch that I recall is the MAX328 which is  a pA with 3500 ohm Ron.  Keithley used to have a great app note on designing low current measurement devices, clean layout, air gaps, guards are all required at these levels.  You're almost to the air wiring on teflon standoffs with reed relay levels.  Good luck.

[Postal2]

.... the air wiring on teflon standoffs with reed relay levels. ....

Calibrate and use:

[Positron Enthusiast]

Second on the reed switches for the first batch, they are excellent. I recommend standex-meder or an RF-rated reed as they will be more fully characterized than the cheaper brands like littelfuse etc.. You didn't really specify, but is the measurement entirely DC? If so, capacitance of an alternate analog switch will not matter. 

if you stick with solid state / semiconductor switches they will all have much worse figure of merit (R_on * C_s). Typically leakage is inversely proportional to R_on and proportional to the parasitic capacitance. If you need low charge injection that should be doable. Analog devices has a good parametric search tool for their analog switches. maybe their ADG5xxx series would be a good place to start for low leakage, most of those are latchup immune to boot. If you put the switch in the correct place in your amplifier you can remove it's gain error for gain switching applications. Analog devices has some good example circuits for that.

I think postal was recommending some sort of electro-optic solution, not entirely sure what they had in mind. I guess maybe omron has some good opto-fets. Omron G3VM Series comes to mind but I think those are all digital/binary mode? I tried looking for alternate manufacturers with good figure of merit but some of their catalogs were just way too much work to navigate.

Still, I would stick with reed relays/ reed switches. Not a low-leakage expert but the layout/ board material recommendations are also good places to start.