Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Avoiding Ground Loop With Multiple Inputs to DAQ Board

Status
Not open for further replies.

Hi-Q

Junior Member level 2
Joined
Oct 2, 2011
Messages
24
Helped
0
Reputation
0
Reaction score
1
Trophy points
1,283
Activity points
1,653
Hey,

I have a DAQ board with several analog inputs. The inputs are differential and are connected to the DAQ board by coaxial connectors (in+ shielded with GND and in- shielded with GND).

I have several input amplifier/detector/whatever modules which feed those DAQ inputs, all of them in separate metal cases but put together into a rack. The inputs of the amplifier modules are unbalanced (SMA or 3.5mm rf connectors). The outputs are quasi-differential, meaning that the output signal is actually unbalanced (coaxial connector), but there is a ground reference (sense) signal provided (another coaxial connector), so it seems reasonable to connect these two lines (signal and sense) to the differential input of the DAQ board.

The power supplies of the individual amplifier modules are isolated from each other (at least I plan to do so) in order to avoid ground loops if the inputs of several amplifier modules are connected to a single external module where they then share the same ground (via the external module's case as the rf connectors are obviously all grounded to the module case).

At that point I have a star ground scheme, at least as long as the external module's case is small enough to be seen as the center of a star connection (nice).

Now I have to connect the outputs of the amplifier modules to the inputs of the DAQ board/module.
If I just use two coaxial cables per amplifier/DAQ pair, I will create another star ground point at the DAQ board where the outer conductors of the coaxial connectors are connected.
Unfortunately, something having two star center points is no longer a true star ground scheme. I assume that now I have quite some ground loops, which I wanted to avoid in the first place.

What would be the best way to deal with such a situation?
Should I just connect all the shields (use standard coaxial cables)?
Should I only connect the shields to the connector on one side, and if so, which one (amplifier output or DAQ input) ?
Should I have the shields completely flowing on the other side or should I at least connect the shields of each differential pair on the other side, but not connect it to the outer conductor of the coaxial connectors on that side?

If I would not just connect all shields to ground on both sides, how would I make the one necessary ground connection to the DAQ board?
Would I just connect one of the shields, providing a single ground connection without generating a loop?
Or would it be best to somehow connect the DAQ ground directly to the case of the external module (if this was possible at all) in order to have a true star ground instead of a star with one elongated beam (chain connection of ground from the external module to the DAQ board via one of the aplifier modules)?

Regarding the connections from the amplifier modules to the DAQ board, we are talking about 3 to 18 inches (the DAQ board would be in the same 19" rack as the amplifier modules).

Thanks in advance for all comments.
Regards,
Hi-Q
 

Hi,

could you upload a sketch of your setup, including the DAQ board at least two amplifier boards as well as the power supply of the amplifier boards and those of the DAQ board. Please also mark each cable with its lines/wires and potential/signal.
What kind of DAQ board are you using? Maybe the board has a dedicated input and output ground plane. How is the DAQ board connected to the PC (USB, Ethernet, Wireless, RS232)? How long are your coaxial cables, and what is your maximum signal frequency?

greets
 

Hi,

I also vote for a sketch of your setup.

Why two coax?
Why not one shielded pair? (this avoids ground currents between the shields of the two COAX.)

***
Since you already use differential signalling you don´t need GND as clean signal reference.
Thus you may use a solid GND wire along with the two COAX ... and connect the shiedls of the COAX only at one side (DAQ side?). All GND current should flow through the solid GND wire.

Klaus
 

Ok, I am very bad at sketching.
And I don't exactly know how I would make the power supply thingy. Maybe imagine just a few separate wall plug adapters, each one followed by some serious low pass filtering and a low noise linear regulator (I think the DAQ board is supplied that way)

The DAQ Board (don't know the exact make from my mind) already has the differential inputs with the two coax connectors.
All the outer contacts of the coaxial connectors are connected to ground planes on several layers, which are common for all of the connectors (and all of the complete DAQ board).

The cables between the amplifier boards and the DAQ board are between 3 and 18 inches, depending on where in the rack the modules are placed.
Frequency range is DC to a few MHz.

The DAQ board is connected to computer via an USB chip and between the USB chip and the DAQ board there are isolators.

D1 and D2 are differential input amplifiers on the DAQ board.
The amplifier in the amplifier module is just an example. The real circuit is much more complex, but the sketch shows how the ground of the output signal is referenced to where the shield of the input coax connection is entering the amplifier module.
Please excuse the quality of the two coaxial cables. They seem to have been bent around a lot.
 

Attachments

  • Rack.png
    Rack.png
    25.2 KB · Views: 98

Hi,

Ok, I am very bad at sketching.
And I don't exactly know how I would make the power supply thingy. Maybe imagine just a few separate wall plug adapters, each one followed by some serious low pass filtering and a low noise linear regulator (I think the DAQ board is supplied that way)

It seems sketching isn't really your strength, at least by doing it with a pc based painting program :wink:. A hand drawn sketch would be certainly better (and easier/faster to draw). The regulation is not of concern for your question, the sketch should serve as illustration for everyone to explain your application. The grounding/isolation of your power supply may be of interest.

The DAQ Board (don't know the exact make from my mind) already has the differential inputs with the two coax connectors.

The specific model number of the DAQ board would be interesting, so we could have a look at the user/reference manual, and how the inputs are defined/labeled and their intended function.

All the outer contacts of the coaxial connectors are connected to ground planes on several layers, which are common for all of the connectors (and all of the complete DAQ board).

That's how coaxial cables are usually used.

The cables between the amplifier boards and the DAQ board are between 3 and 18 inches, depending on where in the rack the modules are placed.
Frequency range is DC to a few MHz.

Assuming by a "few" MHz a frequency of about 10 MHz is meant, transmission line effects should not bothering you (18 inch << λ/10 • VF = (c/f)/10 • VF).

The DAQ board is connected to computer via an USB chip and between the USB chip and the DAQ board there are isolators.

Do you mean your USB connection is galvanic isolated e.g. by an IC like [1]?

D1 and D2 are differential input amplifiers on the DAQ board.

Your sketch shows two inverting (transimpedance) amplifiers, which are ground referenced, so the input is single ended. And also the output of your shown opamps is referenced to ground, thus also single ended.

The amplifier in the amplifier module is just an example. The real circuit is much more complex, but the sketch shows how the ground of the output signal is referenced to where the shield of the input coax connection is entering the amplifier module.

To answer your question we need at least a partially accurate representation of your opamp circuitry. The shown circuitry does not match with your demands at all.

Please excuse the quality of the two coaxial cables. They seem to have been bent around a lot.

That does not matter, but what I'm missing is the mentioned SENSE line.

FURTHER:

What is your extenal module?
What is its purpose?
How is it supplied?

There are some aspects left in your sketch, especially the supply of the external module.

According to your sketch each opamp module has two star points, but they are connected to different boards. This would not be a problem in my opinion so far BUT I think the sketch is not complete at all (USB isolation? External module supply?).

greets

[1] https://www.analog.com/en/products/adum4160.html
 

The DAQ board is based on this evaluation board design from AD:

https://wiki.analog.com/resources/eval/ad9684-500ebz

There will be more than one AD9684 on the board, but I think four or six channels and amplifier modules would be dealt with the same way as two of them, because the only trivial case would be a single channel.

The galvanic isolation uses technology comparable to the one you linked to, but it is not done at the USB level because there simply is no isolator chip fast enough for the serial transfer of the required amount of data per time.
Therefore the isolation is done at the parallel signal level. If you look at the evaluation kit, the isolators would be located in between the two boards.

As mentioned before, I plan on using separate isolated supplies for all the modules. Maybe it would be wise to connect the grounds of all the modules by some 10k resistors so they do not float away from each other when not connected to the external module.

The external module is something that may change, just like a DUT within a test environment. It's power supply would also be independent from the other modules.

Yes, a few means from one to up to or under ten.
I wonder if it would be good to connect the shields to ground with small capacitors on the otherwise unconnected side (if I would not connect them directly on that one side) in order to reduce reflections if signal parts would be present that are high enough to make transmission line effects work on them.


Yes, the signals are single ended ground referenced.
The sense line is the one on the lower left of the amplifier modules that shows the center conductor of the coaxial connector going to the right to the ground. So it effectively senses the ground where the amplifier also is referenced to.
This would then be connected to one of the VIN- pins of the AD9684.

I don't see two star points for each opamp. each one has a single star point, just where the positive input joins the ground of the amplifier module's input connector and the sense line (center contact of the lower left coaxial connector).
 

Hi,

The DAQ board is based on this evaluation board design from AD:

https://wiki.analog.com/resources/eval/ad9684-500ebz

There will be more than one AD9684 on the board, but I think four or six channels and amplifier modules would be dealt with the same way as two of them, because the only trivial case would be a single channel.

The galvanic isolation uses technology comparable to the one you linked to, but it is not done at the USB level because there simply is no isolator chip fast enough for the serial transfer of the required amount of data per time.
Therefore the isolation is done at the parallel signal level. If you look at the evaluation kit, the isolators would be located in between the two boards.

Ok, 500 MSPs are quite fast, which cannot be achieved by a capacitive isolation IC (as far as I know).

As mentioned before, I plan on using separate isolated supplies for all the modules. Maybe it would be wise to connect the grounds of all the modules by some 10k resistors so they do not float away from each other when not connected to the external module.

I agree with that, I think I would also omit the 10 k resistors and connect them directly as close as possible at the opamp board side (not only connection the power supplies ground banana connectors). This avaoids a potential different between the individual opamp boards, which would result in an undesired current flow.

The external module is something that may change, just like a DUT within a test environment. It's power supply would also be independent from the other modules.

I assume it shares a common ground, where all connector-grounds are connected!?

Yes, a few means from one to up to or under ten.
I wonder if it would be good to connect the shields to ground with small capacitors on the otherwise unconnected side (if I would not connect them directly on that one side) in order to reduce reflections if signal parts would be present that are high enough to make transmission line effects work on them.

The shields of which cable? Those attached to the external module has to be connected at both sides, as you need a return path of the current traveling towards the inner conductor. The connection by a capacitor is common for the shield of an USB connector, but there is a dedicated GND line too. The connection of the USB shield is pretty controversial and if you are performing a web search on this topic you will find different solutions/suggestions also from connector/semiconductor manufacturers (direct attached to GND; connected by a resistor, capacitor or ferrite bead).

Yes, the signals are single ended ground referenced.
The sense line is the one on the lower left of the amplifier modules that shows the center conductor of the coaxial connector going to the right to the ground. So it effectively senses the ground where the amplifier also is referenced to.
This would then be connected to one of the VIN- pins of the AD9684.

Maybe you could update your sketch, this information is not obvious in your sketch.

I don't see two star points for each opamp. each one has a single star point, just where the positive input joins the ground of the amplifier module's input connector and the sense line (center contact of the lower left coaxial connector).

I assume there will be a ground plane on your opamp board, which will also be connected to your SMA connectors. One connector is on the right side, and two on the left side of your opamp board. As each of this SMA connector will be attached to the same ground plane you will have two "star" points (by assuming the two left SMA connectors are close to each other).

Please update your sketch, so we can see how the ground of your opamp boards is actually used/wired especially how/which ground is used by your SMA connectors.

greets
 

I see inconsistencies in the specification that make it hard to answer your questions. You are initially talking about "DC to a few MHz" (post #4) and are finally switching to a 500 MSPS ADC. The ADC has 2x50 ohm differential input equipped with SMA jacks. A real design would use shielded 100 ohm differential pairs and a differential driver amplifier to achieve wideband common mode rejection.

"Quasi-differential" or pseudo-differential signaling makes no sense for high MHz signals, because it can reject common mode signals only in the low frequency range. For the MHz range, promising methods of common mode suppression are strictly symmetrical differential circuits and common mode chokes. Or do you expect common mode noise to be restricted to the low frequency range?

The isolation point is yet rather vague. There are methods to isolate high MSPS and GSPS data streams, really demanding.
See e.g. the information about Cleverscope CS448 design lined in this thread https://www.edaboard.com/showthread.php?388481-Best-isolated-or-differential-scopes-probes
 

I agree with that, I think I would also omit the 10 k resistors and connect them directly as close as possible at the opamp board side (not only connection the power supplies ground banana connectors). This avaoids a potential different between the individual opamp boards, which would result in an undesired current flow.

I would like to avoid connecting them directly at the opamp boards in order to avoid multiple sites where the different grounds are connected. The 10k resistors would serve as to not completely let the grounds flow away from each others, like maybe hundreds of volts, which might be dangerous for the components when the inputs are eventually connected to the external modules.


I assume it shares a common ground, where all connector-grounds are connected!?

Yes, all the coaxial connectors at the external module are connected to the module's metal case, which would be the common ground for all amplifier modules.


The shields of which cable?

Of the cables from the amplifier modules to the DAQ board.

Those attached to the external module has to be connected at both sides, as you need a return path of the current traveling towards the inner conductor.

Yes, of course. That's why I sketched the outer conductors of those cables being connected to the outer conductor of the coaxial connectors of the external module and the amplifier modules.

The connection by a capacitor is common for the shield of an USB connector, but there is a dedicated GND line too. The connection of the USB shield is pretty controversial and if you are performing a web search on this topic you will find different solutions/suggestions also from connector/semiconductor manufacturers (direct attached to GND; connected by a resistor, capacitor or ferrite bead).



Maybe you could update your sketch, this information is not obvious in your sketch.



I assume there will be a ground plane on your opamp board, which will also be connected to your SMA connectors.

Yes, you assume correctly.

One connector is on the right side, and two on the left side of your opamp board. As each of this SMA connector will be attached to the same ground plane you will have two "star" points (by assuming the two left SMA connectors are close to each other).

That's why i don't want the body of those SMA connectors to be the ground reference for the DAQ input. The shields of the coaxial cables are just meant to shield them from noise and external interference signals.

That's also why the center conductor of the lower left SMA connector is meant to carry the ground reference to the IN- of the DAQ board.

Just imagine that D1 and D2 have their positive input drawn towards the top and the negative towards the bottom.
And the center conductors are connected horizontally.

Ok, I updated the sketch and inserted the cables from the amplifier modules to the DAQ board.
I left the shield connections open, because that's what I don't know: Where and how to connect them.

If I connect more than one I have a ground loop like, for instance, from the ground plane of the DAQ board via the shield of the cable to the amplifier module one, via the shield of the cable to the external module's ground, via the shield of the cable to the amplifier module two via the shield of the cable to the DAQ board back to that ones ground plane again.
 

Attachments

  • Rack2.png
    Rack2.png
    14.4 KB · Views: 65

Sorry, I just saw that post.

I see inconsistencies in the specification that make it hard to answer your questions. You are initially talking about "DC to a few MHz" (post #4) and are finally switching to a 500 MSPS ADC.

The signals are in the single digit MHz range, indeed.
I see no failure in heavy oversampling, though. Even if the ADC can do 500 MSPS it can also do less. Sampling a 5 to 8 MHz signal with 80 or 100 MSPS avoids aliasing quite good. And it is always good to have some headroom. Too often I have designed something just to specification only to find out later that a bit more effort would have saved me from doing an upgraded design almost immediately after the first one was up and running.

The ADC has 2x50 ohm differential input equipped with SMA jacks. A real design would use shielded 100 ohm differential pairs and a differential driver amplifier to achieve wideband common mode rejection.

Well, sometimes you take what you are given...

"Quasi-differential" or pseudo-differential signaling makes no sense for high MHz signals, because it can reject common mode signals only in the low frequency range. For the MHz range, promising methods of common mode suppression are strictly symmetrical differential circuits and common mode chokes. Or do you expect common mode noise to be restricted to the low frequency range?

Yes, I expect noise and disturbing signals mostly in the sub-MHz range. Mains hum is also a huge issue in the environment the rack is going to be working in. And switch mode power supply interferences.
The isolation point is yet rather vague. There are methods to isolate high MSPS and GSPS data streams, really demanding.
See e.g. the information about Cleverscope CS448 design lined in this thread https://www.edaboard.com/showthread.php?388481-Best-isolated-or-differential-scopes-probes

At the currently planned data rate the isolators in the parallel data path can bear the data stream. It's just too much at the serial layer.
 

Hi,

something that would concern me would be the return path of your sampled signal (by the AD9684). If you are not connecting the shield of the coaxial cable attached to the + input of your ADC, the return path will be the inner conductor of the coaxial cable attached to the - input of the ADC. This leads to the creation of a frame antenna, thus your measurement system will be more susceptible against EMI and will also radiate more.

greets
 

You are right. With the 50 ohms terminations in place this would be the case. I thought about either removing the terminations (we don't intend to use frequencies over a few MHz) if I would not connect the shields where the cables meet the DAQ board.

I thought about connecting the shields at least via capacitors, so I might also keep the 50 ohms terminations but AC couple them too. Then I would have termination for the higher frequency parts while not generating a ground loop for low frequency signals.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top