Guard ring problem very low current bias INA

[Morgoroth]

Hi all, well, hope you can help me with this problem.

First all I'm electronic enginener with little experience on professional layout, except for very simple circuits as hobbist and few real job (simple designs), now  I'm on an startup that require to design a pH meter with precision of 0.0X, as far I know pH probes have very high impedance (I've readed all documentation I found) so a very low current bias INA is requires to mesure the signal properly. Searching on google I founded this design.

I decided to reply the reference circuit in Altium by few reasons:

1) I love doing this.
2) I'm not sure to understand all the complexities related with INA and I think buying the development tools and comparing my own layout and software I could learn that I need to feel that my design is well done.
3) This design will be the base of other designs that will come later and I need to be very sure of what I'm doing.
4) Practice my Altium skills.

Now the doubs:

1) Recomendations related to guard rings could be sometimes confusing and I'm not sure how implement it with AD0683.

The ring guard should keep traped all the current that could probably been liked sorrounding the input signal with the lower potencial pin to trap thous signals.

Next image show how I have my design in this moment, and the next one show how I think it should be implemented (red line in paint-lab). But the line is so tiny, that I think I could have problems in production and makes me think that the chip is not designed getting that in mind, as for example INA116 that have dedicated pins for that porpouse.





Some recomandations advice to use this layout for unity gain configuration, don't know why and I want to know.

2) In few technical notes I readed that power planes (4 layer PCB) are required for precise measurement, however, the high impedance of the probe makes me think that this don't apply to this design given the FR4 impedance could leak current on signal path, and maybe a good idea is to put a layer under the path of the signal at same potential as guard ring to recover all leaking currents, but with risk to get noise (I think), I don't wkow what to do about it.



I need some advise and if someone recomed a good book talking about this will be great.

Thanks in advance.

 JP

[wraper]

Some recomandations advice to use this layout for unity gain configuration, don't know why and I want to know.

Because if the guard ring is not following potential of input signal, it will actually cause a leak by itself = worse than nothing at all.

[tree]

Make sure you put a solder mask "trace" on top of the copper trace. This will prevent the solder mask from covering up the copper during manufacturing. You want to do this so that stray leakage currents can leak to this bare copper, where it wouldn't have been able to if you had solder mask over it.

[moffy]

I have had some experience with building a PH meter and the problem I ran into with using just DC was offset voltages(battery type) at the electrodes. The electrodes and the sample can produce mV level offsets which are quite unstable. The better solution is to use an AC signal and demodulate it, look at chopper demodulators, which will ignore the offsets.

[Morgoroth]

Make sure you put a solder mask "trace" on top of the copper trace. This will prevent the solder mask from covering up the copper during manufacturing. You want to do this so that stray leakage currents can leak to this bare copper, where it wouldn't have been able to if you had solder mask over it.

Ty tree, there is any way in Altium to void solder mask in some copper lines ?

JP

[tree]

Make sure you put a solder mask "trace" on top of the copper trace. This will prevent the solder mask from covering up the copper during manufacturing. You want to do this so that stray leakage currents can leak to this bare copper, where it wouldn't have been able to if you had solder mask over it.

Ty tree, there is any way in Altium to void solder mask in some copper lines ?

JP

My quick solution is to select the guard trace copy it and place the copy on the "Top Solder" or "Bottom Solder" layer depending on which layer you have your copper.

[Morgoroth]

Nice and easy, I'll do it, thanks for the tip.

 JP

[tree]

When you view the board in 3D, after you have placed your solder mask, you should see the bare copper and that's how you know it'll come out right during manufacturing.

[Morgoroth]

When you view the board in 3D, after you have placed your solder mask, you should see the bare copper and that's how you know it'll come out right during manufacturing.

I use 3D all time, so will be easy to check as you mention this.

Anyway, the mayor problem I have is that the guard ring is not closed in my PCB, given the shape of the chip I can't find a way to make it as I understand that must work (totally closed over the input signal), so if I keep it open, as in the first picture, that should be able to recover almost all the leaked current,'but' there is a GND pin between the input and the output of the INA and maybe that could be a problem, I'm not sure, that's my main problem. Seems like I'll goin got make several PCBs and test each one of them :/, will be expensive.

JP

[Morgoroth]

I have had some experience with building a PH meter and the problem I ran into with using just DC was offset voltages(battery type) at the electrodes. The electrodes and the sample can produce mV level offsets which are quite unstable. The better solution is to use an AC signal and demodulate it, look at chopper demodulators, which will ignore the offsets.

Thanks Moffy, I have no time or resouces to develop a new strategy for this, given your experience, that noise you mention could be corrected by a kalman filter and calibration ?, because if that noise is white, the kalman filter must work well if I tune it properly, and, if that noise have a DC component, the calibration should be able to fix it until I make a new design with AC and shopper demodulator.

JP

[T3sl4co1l]

Soldermask is an attribute for copper primitives, it can be tagged on pretty much anything selectable.  This way you don't need to keep multiple copies of geometry (and forget to move them all if you change the layout), and the expansion follows the design rule.

Don't know if this is in the free tool.

Tim

[Morgoroth]

Works nicely in free tool



TY

  JP

[tree]

Soldermask is an attribute for copper primitives, it can be tagged on pretty much anything selectable.  This way you don't need to keep multiple copies of geometry (and forget to move them all if you change the layout), and the expansion follows the design rule.

Don't know if this is in the free tool.

Tim

I've been doing it inefficiently all this time.

[Moriniman]

I did a design for custom made Ion selective electrodes (we were vulcanising our own porous membranes).

No solder pad for the input pin; lift the pin and make a direct connection to the ultra low loss co-ax.

[Morgoroth]

I did a design for custom made Ion selective electrodes (we were vulcanising our own porous membranes).

No solder pad for the input pin; lift the pin and make a direct connection to the ultra low loss co-ax.

so,  no solder mask and direct connection for the low loss coaxial.

but I need to replace probe time to time,  maybe a terminal block be a good choice?

btw,  then thats why hanna probes come without connectors....

JP