Tuesday, November 17, 2009

A Tale of Two Grounds

It was the best of rigs, it was the worst of rigs,
It was the age of 0.07 pA RMS, it was the age of 0.5 pA RMS, ±5 pA p-p at 60 Hz,
It was the epoch of separating signal ground from power ground, it was the epoch of connecting all power grounds to the signal ground,
It was one really annoying run on sentence written by some old white dude, it was a kick ass blog post written by a middle aged white d00d.
There's one thing I've learned here as postdoc that I've actually never seen discussed in other places. When I first heard it, I did a "Watch you talkin' about" head turn, thinking it was crazy.

That thing is, to get really good noise on your patch clamp rig (good meaning low here), you can make 2 grounds. One ground connects everything that is physically close to the headstage. This gets connected to the signal ground plug. The second ground connects everything else. That gets connected to power ground.

Figure 1: Nat's newly refurbished set up (which is tight if I do say so myself). Axopatch 200B and Sutter MP-285. Green As indicate things attached to signal ground. Magenta Bs are connected to power ground.

So what goes on the signal ground?
  • Microscope
  • chamber
  • condenser
  • bent piece of metal that can be put right in front of headstage/chamber to further shield (not shown)
That all gets connected to the gold pin at the back of the headstage (or equivalently at the signal ground input on the back of the amplifier). Note that you'll need to break the power ground connection (with a 3 to 2 adaptor). Doing this will get you most of the way there.

What goes on power ground?
  • Faraday cage
  • air table surface
  • manipulator
This is connected to power ground (either attach it to an exposed copper pipe or to power ground through the case of the manipulator). Doing this will get you to the super low RMS values specified in teh Axon manual.

Now I no longer fear denoiseing the setup.

The only tricky things I've run into are that most of the BNC inputs on the Digidata are connected to power ground. So usually I have to break the ground connection to connect the gain value output from the amplifier to the digitizer input (I've never had the amplifier signal output to digitizer analog in cause this).

I just turned on my amplifier, and with a pipette holder on, the PATCH mode RMS noise is 0.099, and WHOLE-CELL mode is 0.48. SHU-WEET!



14 comments:

Pinus said...

wow, very nice!

Are you available for contract denoising...I F*N hate doing it.

Dr. Jekyll and Mrs. Hyde said...

Are you available for contract denoising.

SERIOUSLY. You could make you some nice money.

When you say break the power ground, you really mean it? You plug the amplifier into a converter so it's 2-prong?? And it's only the amp and digitizer? This is crazy crazy stuff.

Dr. Jekyll and Mrs. Hyde said...

How is your noise with a real pipette in the bath?

Nat Blair said...

Thanks Pinus and DrJ. I'm definitely available for denoising services. Just bring your friends named Ben.

Yes, Dr. J, I mean plug the amplifier into a 2 prong adapter, make sure the converter's metal tab can't touch the screw on the outlet (which is usually connected to ground). Then everything else, including the digitizer, gets connected to a single long powerstrip, and plugged in normally with its 3 prong plug. Then everytime you connect anything to the amplifier, you'll have to check whether the signal ground and power ground are now connected. One by one addition of BNCs to the digitizer is good anyone, because some are old and crappy and will increase the noise themselves.

And yes, it's totally crazy. If my current boss weren't an electrical engineer, who did his post doc with one of the patch clamp developers, I likely wouldn't have even bothered trying it.

Nat Blair said...

Dr. J- I have the following noise with the current conditions (all patch mode):

Headstage only: 0.074 pA RMS
Headstage plus holder, no pipette: 0.088.
Headstage, holder, pipette (parafilmed near tip) with solution, right over the bath: 0.161.

In the bath the noise is high, since the headstage is passing so much current, and the headstage is resetting the capacitor very frequently. So to simulate a gigaseal, you can drive the pipette into a small coverslip that has a drop of sylgard dried on it (I always have loads of these sitting around, for just this purpose).

Doing this, I got a seal that was ~15 Gohm. That gave RMS noise of 0.498. I suspect this was high for a couple reasons:

the pipette was parafilmed at its tip, not sylgarded (or dental wax, or other coating technique).

The holding current was unstable in the sylgard, and that led to a lot of patch resets.

By eye though, I'd guess that I would be able to see a couple pA at 5 kHz BW. Not great, but not terrible (especially since this set up isn't specialized for single channel recording).

Dr. Jekyll and Mrs. Hyde said...

Wow, that's an interesting point about BNCs. I only think of them as "crappy" if they don't pass current, but I can see that higher standards exist. Next time I'm (re)building a rig, I'm trying this stuff out!

Dr. Jekyll and Mrs. Hyde said...

Wait, what about screwing the amp onto a rack? Do you do that? Doesn't grounding the carcase ground the amp?

Nat Blair said...

You'd think it would, but it doesn't. You can check this by taking all BNC connectors off your amp, and then check whether the outer ring on an amp BCN input (e.g. the scaled output) to the rack screw holding the amp. In all our amps, there's no connection (assuming the amp is plugged into the 3->2 adaptor). And yet, the case of the amp should still be connected to power ground, through the rack -> other components -> main power strip -> wall. That's important, so that if there were a voltage on the amp case, you would hopefully not get shocked when you touch it.

Nat Blair said...

since I can't edit my comment: check this by taking all BNC connectors off your amp, and then check whether the outer ring on an amp BCN input (e.g. the scaled output) to the rack screw holding the amp.

Here I meant to write, check if BNC out ring on amp scaled output is connected to amp rack using your multimeter. There shouldn't be a connection.

That's the problem with some of the outer ring BNCs on the Digidata 1320A I have. Some of the outer ring BNCs ARE connected to the digitizer power ground. That makes it the most likely place where signal ground can inadvertently connect to power ground.

Anonymous said...

Interesting post. Is there any danger of frying the amplifier by removing its ground connection?

Nat Blair said...

@ anony - Two things should make it ok: First, the outside of the amplifier case should be connected to the ground of the other things in the rack (check to make sure that is the case; if not, you can ground the rack to power ground).

That means that if there were a problem where a voltage were present on the external part of the amplifier, it'd be grounded, and shouldn't shock you if you touched it.

Also, since it ends up that really isolating signal and power grounds is hard with the current Digidatas, even the signal ground is connected to power ground.

Anonymous said...

I'm sorry
English is not a native language for me
plus, I'm totally new in setting this lab device.

now I'm facing a RMS 2.69 pA during patch state
,with a shape of sin wave.
I know it is probably caused by ground loop
but I just can't cancel this noise

can you give me some advices?

Your instruction is too hard for me
It would be great for you to use more photo next time



Anonymous said...

hey
I have posted a comment
but it appears to be nothing left

do you know what might happened?

Nat Blair said...

Sorry about that Anonymous, comments on old posts tend to get hung up in moderation. It is to prevent spam comments.

As for your question, I'd start by:

1) removing the pipette holder from the headstage.
2) shield the headstage with aluminum foil - connect the ground at the back of the headstage to the foil.
3) check the RMS noise at this point, and see if you still have a sine wave.

If you do, that means your ground loop is probably somewhere like the connections between the amplifier and the data acquisition board.

If the sine wave is gone after shielding with foil, that probably means there is a ground loop between the scope/bath/headstage, and the power ground.