Power supply, ground level and noise issues

[Christian Tatarau]

Hello and first of all Happy New Year!

We set up an acquisition box with one Intan headstage and we got a lot of electrical noise even after connecting everything to ground and after playing around with the ground cables. I was thinking about replacing the 5V power supply with a rechargeable battery and interrupting the power and ground lines of the USB cable. Like that I would completely disconnect acquisition box and Intan chip from the building and from the PCs power and ground. Is this flawed as a concept? Would that work and help with noise?

[Charles Latchoumane]

In our lab set up, even recording within a coulborn system (all metalic box that deliver a alternative electric shock), we were able to remove all source of noise through grounding. I assume the noise you are talking about is 60ish Hz right

We have a copper net that surrounds the recording area and that is grouded to the same point than the PC and the shock delivery machine, when all is to ground virtually 0 noise is visible except when electrical shock are delivered).

The method that describe should also work if you totally isolate your system, so it is worth trying, but a simple copper net might do it. Please let us know how it goes!

Charles

[Jon Newman]

Isolation might be a good idea, but the mains might not be the ultimate cause of your problem.

Often, in audio-frequency bandwidth systems like those used in ephys, low frequency noise (in the 10s to 1000's of Hz range) is actually high frequency noise that has parasitically exerted its effect on the sensitive analog portion by parasitically coupling from the power supply rails to the high-impedance analog inputs of the IC.  A good example is switching noise from a cheap switch mode power supply, like the wall warts that are often used to run the opal-kelly boards.  The linear regulators used to clean up the power supply rails on the OE board are essentially transparent to this noise. Switch mode supplies often have signals with rising edges that contain power out in the 10s-100's of MHz and linear regulators' PSRR drops off exponentially is this range. 

My suggestion is before going crazy with isolation is to try to use a  nice bench top power supply to power your opal kelly board. 

--
You received this message because you are subscribed to the Google Groups "Open Ephys" group.
To unsubscribe from this group and stop receiving emails from it, send an email to open-ephys+...@googlegroups.com.
To post to this group, send email to open-...@googlegroups.com.
Visit this group at https://groups.google.com/group/open-ephys.
To view this discussion on the web visit https://groups.google.com/d/msgid/open-ephys/d46f7da4-2843-4d5a-9633-8edbefb7c9ab%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

Jonathan Newman

Postdoctoral Fellow, MIT

http://www.mit.edu/~jpnewman/

[Aarón Cuevas]

Hi,

First of all do not, under any circumstace, cut the ground line on the USB while leaving the data lines connected, that would leave the data lines floating with no reference and could very easily damage your USB port or the FPGA USB interface. Should you really need to isolate the USB connection there are special devices available tha completely isolate all lines, removing this issue.

That said, have you connected the board itself to the ground plane your animal is in? Most of the grounding issues I've seen are solved by connecting the board itself to ground, either via the screw ports on its side (you have to open the board to acces those) or via the ground in the BNC connector.

Having a good power supply always helps, of course, but from what I've seen line noise can affect the boards' ADC but very rarely the headstages themselves (the long cables combined with the decoupling capacitor in the headstage create a rather efficient LC filter that removes high frequency components).

Best,

Aarón

On Friday, January 8, 2016 at 5:26:01 AM UTC+1, Jon Newman wrote:

Isolation might be a good idea, but the mains might not be the ultimate cause of your problem.

Often, in audio-frequency bandwidth systems like those used in ephys, low frequency noise (in the 10s to 1000's of Hz range) is actually high frequency noise that has parasitically exerted its effect on the sensitive analog portion by parasitically coupling from the power supply rails to the high-impedance analog inputs of the IC.  A good example is switching noise from a cheap switch mode power supply, like the wall warts that are often used to run the opal-kelly boards.  The linear regulators used to clean up the power supply rails on the OE board are essentially transparent to this noise. Switch mode supplies often have signals with rising edges that contain power out in the 10s-100's of MHz and linear regulators' PSRR drops off exponentially is this range. 

My suggestion is before going crazy with isolation is to try to use a  nice bench top power supply to power your opal kelly board. 

On Thu, Jan 7, 2016 at 9:31 PM, Charles Latchoumane <kskus...@gmail.com> wrote:

In our lab set up, even recording within a coulborn system (all metalic box that deliver a alternative electric shock), we were able to remove all source of noise through grounding. I assume the noise you are talking about is 60ish Hz right

We have a copper net that surrounds the recording area and that is grouded to the same point than the PC and the shock delivery machine, when all is to ground virtually 0 noise is visible except when electrical shock are delivered).

The method that describe should also work if you totally isolate your system, so it is worth trying, but a simple copper net might do it. Please let us know how it goes!

Charles

On Friday, January 8, 2016 at 5:21:18 AM UTC+9, Christian Tatarau wrote:

Hello and first of all Happy New Year!

We set up an acquisition box with one Intan headstage and we got a lot of electrical noise even after connecting everything to ground and after playing around with the ground cables. I was thinking about replacing the 5V power supply with a rechargeable battery and interrupting the power and ground lines of the USB cable. Like that I would completely disconnect acquisition box and Intan chip from the building and from the PCs power and ground. Is this flawed as a concept? Would that work and help with noise?

--
You received this message because you are subscribed to the Google Groups "Open Ephys" group.

To unsubscribe from this group and stop receiving emails from it, send an email to open-ephys+unsubscribe@googlegroups.com.

To post to this group, send email to open-...@googlegroups.com.
Visit this group at https://groups.google.com/group/open-ephys.

To view this discussion on the web visit https://groups.google.com/d/msgid/open-ephys/d46f7da4-2843-4d5a-9633-8edbefb7c9ab%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Christian Tatarau]

Thanks for your suggestions! The noise is not line noise, it is rather 100 -1000 Hz. We will try now to identify the device producing it and check again all ground connections. But I've actually connected to ground the faraday cage (copper net), the metal table on which we measure, the headstage via GND and REF pin (which are both connected on the Intan headstage) and the acquisition box via outer BNC connector. I will then try a lab current source instead the standard 5V converter.

[Christian Tatarau]

We got now a lab power supply and during a 2h recording we caught several artefacts which you can see in the image. Measurements were made in saline solution in a grounded copper Faraday cage. The artefacts appear only in the channel connected to an electrode, not in the other channels. The noise I was talking about in my previous post is not a big problem any more and is probably due to devices in our building. Did anyone encounter such shapes in recordings? What can that be?

[Jon Newman]

What is the sample rate? Are these going through a digital filter? 

On Thu, Feb 25, 2016 at 6:27 AM, Christian Tatarau <ctat...@gmail.com> wrote:

We got now a lab power supply and during a 2h recording we caught several artefacts which you can see in the image. Measurements were made in saline solution in a grounded copper Faraday cage. The artefacts appear only in the channel connected to an electrode, not in the other channels. The noise I was talking about in my previous post is not a big problem any more and is probably due to devices in our building. Did anyone encounter such shapes in recordings? What can that be?

--

You received this message because you are subscribed to the Google Groups "Open Ephys" group.

To unsubscribe from this group and stop receiving emails from it, send an email to open-ephys+...@googlegroups.com.

To post to this group, send email to open-...@googlegroups.com.
Visit this group at https://groups.google.com/group/open-ephys.

To view this discussion on the web visit https://groups.google.com/d/msgid/open-ephys/7e9d1095-963d-4f81-bd53-57e892d6d50b%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Christian Tatarau]

The sample rate is 1 kHz and the signal is going through the acquisition box DSP where I set a bandpass filter from 0.1 to 350 Hz. The Intan chip has a highpass filter itself, I set this one also to 0.1 Hz. So we might be seeing the impulse response of the filters. The question is where that might come from.

[Jon Newman]

Right, and that would be a lot easier to answer with the full bandwidth signal. Why not sample at 30 kHz and get rid of the filtering as much as possible. If this is due to a static discharge in the room or something like that then that still won't be enough, but it will give us some idea as to which part of the spectrum the transient occupies and that will give us some hints about where its coming from. 

These artifacts are not periodic right? 

Finally,  can you put an oscope probe on the 5V lines going to the board, AC couple to the scope, and set your trigger pretty close to the signal, and if possible, persist the scope trace indefinitely. That will give you an idea if there are transients on your power lines.

--
You received this message because you are subscribed to the Google Groups "Open Ephys" group.
To unsubscribe from this group and stop receiving emails from it, send an email to open-ephys+...@googlegroups.com.
To post to this group, send email to open-...@googlegroups.com.
Visit this group at https://groups.google.com/group/open-ephys.

To view this discussion on the web visit https://groups.google.com/d/msgid/open-ephys/4a0d65c9-534d-4edd-992f-4e61c7f39089%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Christian Tatarau]

I can do a longer measurement on Monday with the full bandwidth and without acquisition box DSP. The intan chip filter cannot be disabled, right?

Yes, intervals between artefacts are irregular. I had 7 during a 2h measurement session. What is striking is that the events have changing polarities.

We also got an USB isolator and a rechargeable battery and we repeated the recording with a galvanically isolated system. The artefacts are still there but they have a much reduced amplitude as you can see in the picture. Same settings as the last picture.

[Christian Tatarau]

Thanks for the idea with the oscilloscope, I set the trigger and compared the triggered events on the 5V supply to the artefacts. They do not occur at the same time.

Also, as you suggested, I measured with disabled filters and at 30 kHz, you can see the results in the images. I got 2 large artifacts. Both plots show the same 2 events, one shows a 12s interval, the other 1s.

Contrary to what I wrote above, this time the artifacts occured in all recorded channels, not only on the one connected to the electrode. One idea is that it might be an electrostatic discharge or a damaged capacitor somwhere...

[Jon Newman]

Is there a device that randomly draws a lot of current all of a sudden in your room somewhere -- e.g. the a freezer or fridge (more specifically, its compressor motor)? This looks like it might be something like that. Can you zoom in on the falling edge so we can see how fast it is? 

How much cabling is there between your specimen and the Intan chip?

What does grounding configuration look like?

If its static discharge, then we still need more temporal resolution to tell whats going on. Static discharge will manifest itself as wideband (like out into the hundreds of MHz) noise for 10's of microseconds. This would result in large deflections of random polarity, so I'm not sure thats what is going on. Although, i'm sure there is a way  it might be made to look like this given the details of the A/D circuitry on the chips.

--

You received this message because you are subscribed to the Google Groups "Open Ephys" group.
To unsubscribe from this group and stop receiving emails from it, send an email to open-ephys+...@googlegroups.com.
To post to this group, send email to open-...@googlegroups.com.
Visit this group at https://groups.google.com/group/open-ephys.

To view this discussion on the web visit https://groups.google.com/d/msgid/open-ephys/f00ab4eb-3f5b-4b3e-8eef-c48f8e95119f%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Christian Tatarau]

We don't have any fridge in our rooms nor in the neighbouring rooms. I asked the Engineer of the hospital to have a look and maybe he knows which devices might produce such an effect.

Between saline solution and Intan chip there is a concentric bipolar electrode, the connector soldered on the electrode interface board from Intan, that is all.

Here is a zommed image. I basically jumps from one sampling point to the next. That is 1/30 ms.

[Jon Newman]

1. Yeah, so it would much better to have more temporal resolution on the trace. All manner of things could be happening in that signal sample period that would be very useful to know when debugging this issue, but are not visible at this time scale. I now recall that it is very hard to get this information because the amplified analog signal is not exposed on any pin of the intan chip so you can't just probe it with a scope. hmm.

Do you have an "old school" differential amplifier with wider bandwidth that you could use to sample the signal on a scope to see if this artifact persists and to see what its specific frequency characteristics are?

2. With regard to your comment: Between saline solution and Intan chip there is a concentric bipolar electrode, the connector soldered on the electrode interface board from Intan, that is all:

The connector has a ground/reference, right? What is the specific configuration for how ground a signal are attached to to the headstage: length of wire etc. this can matter quite a lot because it can increase the ability of the passive portion of the circuit (your electrodes and the saline thing) to sense stray electric and magnetic fields.

3. Did you probe your power lines yet to see if the there is crap on them that is locked to these artifacts?

--
You received this message because you are subscribed to the Google Groups "Open Ephys" group.
To unsubscribe from this group and stop receiving emails from it, send an email to open-ephys+...@googlegroups.com.
To post to this group, send email to open-...@googlegroups.com.
Visit this group at https://groups.google.com/group/open-ephys.

To view this discussion on the web visit https://groups.google.com/d/msgid/open-ephys/293eeb07-60f1-44b1-ba30-e526fcbfbd7c%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Jakob Voigts]

Hi guys,

this is a tricky one - intermittent noise is the worst..

 

2. With regard to your comment: Between saline solution and Intan chip there is a concentric bipolar electrode, the connector soldered on the electrode interface board from Intan, that is all:

The connector has a ground/reference, right? What is the specific configuration for how ground a signal are attached to to the headstage: length of wire etc. this can matter quite a lot because it can increase the ability of the passive portion of the circuit (your electrodes and the saline thing) to sense stray electric and magnetic fields.

you might have tried this already, but just in case, it might be helpful to try just keeping the simplest configuration that picks up the noise (beaker with saline and electode, or even bare headstage) in a secondary faraday cage, like a plastic box wrapped in tin foil or something that is grounded to the headstage gnd&ref but nowhere else. I've seen plenty of cases where weird HF noise made it trough otherwise working cages, if this helps then it might be worthwhile checking the shielding again. 

3. Did you probe your power lines yet to see if the there is crap on them that is locked to these artifacts?

 This seems like the best guess to me - when i had a similar issue a while ago it ended up being specific to one outlet that had something causing spikes on the same power circuit. You could also try using one of these batteries to be independent form the building power, or at least to debug this: http://www.amazon.com/CyberPower-CP1500AVRLCD-Intelligent-1500VA-Mini-Tower/dp/B000FBK3QK/ref=sr_1_2?s=electronics&ie=UTF8&qid=1457674540&sr=1-2&keywords=ups+battery

 

[Christian Tatarau]

Before I go on: what kind of connector do you have on the electrode interface board? I guess our might be problematic. We got this electrode and socket board. I did not use the standard 16 pin connector which comes with the EIB from Intan because we will need all 32 channels in the future.

[Jakob Voigts]

I dont think the connector should make too much of a difference. Of course the longer the electrode / connection to the intan chip is the more of an antenna you get, but you can always remove that issue from the equation by locally shielding the connection.

In my experience in mouse cortex recordings I always had the worst noise issues when doing acute recordings where electrodes end up somewhat exposed, but its almost always enough to just wrap as much of the electrode as possible, and the entire EIB and intan headstage in a bunch of copper foil / aluminum foil. 

In drive implants, the whole drive body typically acts as a shield making the length of exposed electrodes/connections somewhat less relevant.

[Christian Tatarau]

Thanks for the many ideas. I was able to find & eliminate one nasty source of noise which was rather surprising: electrostatic discharge between human body and the metal door frame through door knob and latch. It was hard to find because it happens only when several conditions are met and often even when you've already taken the hand away from the metallic parts of the door.

[Jakob Voigts]

awesome,

in case it hasnt come up before in this thread, ill utilize this google-able spot to plug staticide - i find that wiping floors and/or lightly spraying this:

https://www.amazon.com/ACL-Staticide-AntiStatic-Liquid-32/dp/B002WWNLVO/ref=sr_1_7?ie=UTF8&qid=1476706010&sr=8-7&keywords=staticide

eliminates these issues, the spray is also very useful for reducing motion artifacts in acrylic behavior chambers.