Unsure about analog ground / ground loop (GND and AGND)

[Markk]

Hi

thank you all for the great work!

I'm trying to figure out the right way to use GND and AGND in a project. There is some info in another post, but I'm still unsure.

I'm not formally schooled and new in electronics, so I read a book about PCB design. I learned about ground loops, when and why they happen and what you can (and cannot) do about them.

My project involves a ground loop. It's the classic example of the supply going out to a sensor (possibly a bit away) and the signal coming back. In my case it's a high voltage supply (400V to 1500V photo multiplier tube). The voltage across the tube must be quite stable, as the tube's gain varies with the voltage to the power of 6 .. 9. So it's not so easy to shift things around there. The signal is a current pulse from (or rather into) the anode. It is decoupled through a cap to be conditioned and then digitized by the teensy. Obviously any sudden ground shifting adds error currents.

First I thought, brilliant, the Teensy has a separate AGND that I can connect the returning ground to and everything is nicely referenced to anythink picked up along the way. I wondered if I needed some dampening to keep the two (GND and AGND) reasonably close but I thought this was the way to go. Especially since the K20 datasheet (p. 11) suggests you can safely have +/-0.1V difference on VSS-VSSA.

But then I saw the ferrite in the Teensy 3.x schematic, tying AGND to GND. Ferrites are exotic beasts to me

A: Should I still go ahead and create the GND to AGND ground loop?
(And is the ferrite actually the kind of dampening one needs?)

B: Or should I remove the ferrite?
(And where is it? Any suggestions if I need to and how to damp instead?)

C: Or should I extend both GND and AGND pincers-like to the sensor, decoupling the high voltage to AGND out there (if this is the right terminology) and run all the signal conditioning curcuitry against AGND?
(And can AGND safely sink the max ~60 milliamps needed on startup?)

D: Or do I have to go the hard way, creating and working with a differential signal all the way back?
(the tricky part of the whole project is actually the signal conditioning, so I fear this would complicate things massively!)

E: Other options anyone?


This is still on the drawing board. I habe no practical experience. So I can't actually be sure if it's a real problem or if I'm overly cautious. I would just like to understand a bit more before I order PCBs.

Thanks for all your help.
-Markk

 

[Markk]

Anyone? Please?

 

[whollender]

Hi Markk,

In the most ideal situation, you'd like your input signal referenced to AGND, which would mean connecting your high voltage supply ground to it. Unfortunately, you can only safely connect the two grounds if either one or the other supply (high voltage supply or the teensy supply) is isolated from mains ground. If they're both connected to mains ground (ie the high voltage supply through its power cord, and the teensy through a computer's usb port), then you could have very high current flow through the ground connection.

It would be easier to say whether or not you'll have a problem if you can provide more details regarding your power supplies.

William

 

[Markk]

Hello William

thanks!

It would be easier to say whether or not you'll have a problem if you can provide more details regarding your power supplies.

I guess it's difficult to communicate. This is the "guidance" forum, so things are not yet all worked out and I can't show you finished schematics etc. Hen and egg.

This is where my concerns come from:

http://books.google.ch/books?id=0ZW...page&q=figure 1.9 input signal wiring&f=false
(hope this link will work permanently)


Source: Peter Wilson, The Curcuit Designer's Companion

Some background:

The Teensy and the HV generator will have the same supply and are very close together, probably on the same PCB (ground plane). They are normally battery powered. The battery may be charged via USB (which is cut from VIN) or some other source, but this is "behind" the battery.

The loop only really starts afterwards, through the HV generator out to the sensor. I would like to get some flexibility there: this could be only a short connection to the PMT socket right next to the PCB (some centimeters of i.e. twisted wires, or even a 90° daughter PCB with the PMT socket). Or it could go through a coax cable up to a few meters.

The sensor has its own casing which is connected to the cathode (ground) of the PMT for shielding and HV safety. The casing need not necessarily be in direct electrical contact with any other casing. In fact it will most likely be insulated in a plastic enclosure (wheather proof), sometimes together with the main PCB's case, all things floating when on battery.

The signal returns from the PMT back on the original PCB to be conditioned and digitized by the Teensy. The problem is whatever the lines back and forth pick up. I have no idea how severe this can get in the field, i.e. if it will be a real problem at all.

The signal (as currently observed on the breadboard) is a current signal pulse of roughly 1.5 us with a 20 uA to 2'000 uA peak. I use a transimpedance (current to voltage) amplifier in an integrator setup to convert the fast pulse into a comfy voltage step. The output will be digitized by the Teensy, then the integrator will be reset. A second channel through a comparator triggers an interrupt. I plan to program low power modes later, with the ADC DMA-ringbuffering so when I wake up from the interrupt I can look back into the ringbuffer and calculate the voltage step (and the underlying base line slope).

From what I understand, it is best to run the signal back differentially, this means looping back not only the signal, but ground (or some level coupled to it) too. I kind of hoped this would be easy to reference by connecting to AGND. Given the +/-100mV allowable swing of the K20 I thought this should be possible.

In this case, it would actually correspond to (a) in the posted figure above (i.e. the best solution). This would also involve grounding the signal opamps und the comparator to AGND. I will use low power components so this should all only amount to one or two milliamps. However on startup the HV cascade will draw/sink up to 60 mA. Some of that might flow back through AGND so it should at least stand that.

If connecting AGND is not suitable then I would have to go for (e) in the above figure which is obviously much more complicated.

Perhaps I should break my question down into simpler subquestions and go from there:

* What are the properties of the ferrite?
* How much current can AGND sink?
* can the ferrite be removed?

Thank you very much.
-Markk

 

[PaulStoffregen]

You could try removing the ferrite. As long as AGND and GND remain within about 0.3V of each other, things should be fine.

 

[Jp3141]

The reason to bing AGND back separately would be because the current consumed by the remote sensor causes too much I*R drop in the long GND wire. However, if it consumes a (steady ?) few mA, this is unlikely to be an issue -- say the wire resistance is 0.1 Ω/m and you have 3 mA -- that gives a ground shift of 0.3 mV/m length. Would that upset your accuracy ?

For long wires, I'd be more concerned about RF and AC power line noise pickup. Be sure to use twisted pair, or a coax line. You might put a ferrite around both wires (like you might see on a computer cable) to keep the RF out.

 

[whollender]

It sounds like you're on the right track. Based on the low max current of the HV supply, I don't think you'll have any issues running the HV supply off of AGND instead of your nominal ground. If you do have issues with the ferrite on board the teensy, you could easily remove it, and replace it with another of your choosing, or just short it all together. Paul may be able to comment on whether the AGND ferrite can handle those sorts of currents.

Your signal pulse isn't super fast, so if you have noise problems with longer cables, you could add a ferrite around both wires like Jp3141 suggests to reject common mode noise, and even a series ferrite on the signal line. A ferrite around both wires (or a common mode choke) will reject high and low frequency common mode noise, and a series ferrite will reject high frequency differential noise (you shouldn't see any LF differential noise as long as you use twisted pair or coax for signal/ground).

 

[Markk]

You could try removing the ferrite. As long as AGND and GND remain within about 0.3V of each other, things should be fine.

Before I throw it over board I definitely need to learn more about it. Is it possible to get the part number or spec?

Thank you very much .. for everything!
Markk

 

[Markk]

For long wires, I'd be more concerned about RF and AC power line noise pickup. Be sure to use twisted pair, or a coax line. You might put a ferrite around both wires (like you might see on a computer cable) to keep the RF out.

I think I misused the term "ground loop". It was always RF and AC noise pickup that I was concerned about. Will have to learn much more about those terms, about ferrites, "chokes" etc. Thank you Jp3141 for the pointers!

 

[Markk]

It sounds like you're on the right track. ... Your signal pulse isn't super fast, so ... A ferrite around both wires (or a common mode choke) will reject high and low frequency common mode noise, and a series ferrite will reject high frequency differential noise (you shouldn't see any LF differential noise as long as you use twisted pair or coax for signal/ground).

Thank you, I will try to follow up on this suggestions. The "common mode choke" seems to be exactly the creature that I need. Nice tip.

-Markk