Increasing noise performance by replacing mechanical pots

"Adrian Nievergelt"

That is more like voodoo science not engineering.

You MUST explain what the app is and not just assume that others accept your un-stated and probably false assumptions.

Kindly get real.

.... Phil

Sorry, didn't want to confuse people with my brabbling. So the application is essentially a q-control circuit for dynamic atomic force microscopy. It takes an oscillation signal and feeds back a delayed version into the oscillator, thus adjusting the damping by actively driving or actively damping. In electronics that means taking the input signal, using a manually adjustable phase shifter (first pot) to shift the phase, then running it through a variable gain stage (second pot), and finally adding the result to the original input and outputting that. The typical signals are in the ~100mV range at frequencies of 100kHz to

Sincerely Adrian

Any analogue alternative will have serious temperature drift problems and /or non linearity.

I suspect that fine adjustment, resettability and good temp stability are important in your device - so consider the use of multi-turn wire wound types.

.... Phil

Does this need to be infinite resolution? If not, perhaps you can use a MDAC. Sadly, the MDAC is probably cheaper than a good 10 turn pot. [I've got a nice selection of 10 turn pots with dials, pulled from old test instruments and scrapped biomedical gear. I started that treasure hunt when it occurred to me that multiturn pots were on the endangered species list, at least if you want them on the cheap.]

I don't have IEEE online access, but jfets as resistors are kind of temperature sensitive. I don't think I've seen them used in circuits without feedback control.

Noise how? Are you worried about picking up EMI? Are you worried that the pots themselves will generate thermal noise? What?

If it's thermal or excess noise in the pots, choose a pot material that's not subject to that problem. If it's EMI, consider shielding the leads going to and from the pot -- if you're really concerned, build a little housing for the back of each pot, that surrounds it and makes a good solid connection to the shielding on the wire. Better yet, just use a good enclosure.

Or, try reposting without any restrictions against digital electronics, and see what answers you get.

......or mount the pots on the board and mount the board to the rear of the user panel, eliminating wiring to/from the pots.

I don't think the pots will generate much thermal noise, but they might pick up EMI. At the back of my head is the design paradigma that you should make signal paths between op-amps as short as possible. cabling out to a pot seems somehow wrong. I'm not sure if I don't worry about something that is not a problem. Apart from that the wiring and the larger pot are quite a parasitic inductance/capacitance and might cause problems and oscillations with a 1MHz signal. again I might completely misjudge the problem.

With the answers I have been getting it will probably turn out like this yes, shielded cables and housing.

I might use a combination of a digital pot driven by an analog signal. I was trying to avoid this route as I'm a bit limited on board space and this adds quite a bit of complexity.

I was mostly curious if that other thing might work, and as I said if some people have maybe tried it. I saw floating resistors with OTA's, but from what i could judge those were also massively thermally not stable.

Thanks Adrian

Or use extension rods like the Tek scopes?

Pots go bad at one point. Back when I designed audio and telecom gear I always used a VGA (voltage controlled gain amplifier) to control signal levels. The pot supplied the controlling voltage (with some filtering). There are many VGA chips available with analog and digital control interfaces.

Haha - I nearly added that! My first job was reparing TEQ for a telecoms co.

"Adrian Nievergelt"

Like to play your cards real close to the chest - right ?

Lawyers will be having orgasisms ...........

Particularly with multi-turn WW pots.

- you always gotta watch out for the Plimsoll line ...

.... Phil

ame

f my

ept

Crank up the (fixed) gain as much as possible in first few gain stages. Then send this signal off for signal processing. (digital or analog) Then if there is, for some reason, too much gain in the front end you can attenuate it to the desired level. But do all the signal processing at a high level. Then you don't care (too much) about noise and interference.

George H.

ace

y a

that

t's

ads

the

The one problem I've had with switches to low level signals is that the metal shaft can* act like a little antenna and couple crud in the room down into the signal.

so non-conductive extension rods.

George H.

So you're mostly concerned about EMI and cable capacitance running a cable out to a panel pot, is that right?

The usual way to fix that in your frequency range is an optical encoder, a small microcontroller, and a low-value digital pot. High value dpots are slow on account of the switch capacitance--check out the AD8403, which gives you 256 taps down to 2k resistance, and works out to a few megahertz. You can combine three dpots in a pseudo-Kelvin-Varley divider to get smaller steps if you need to--connect two pots in parallel, and run the third off the two wipers. Run the first stage pots a couple of codes apart to reduce the effect of differential nonlinearity, and you can get ~15 bit resolution. You don't get anything like 15-bit accuracy, but at least it'll be monotonic, and for a manual adjustment that's often all you need.

The JFET trick is also okay but becomes nonlinear fairly rapidly at higher amplitude, and has a not-too-nice temperature coefficient. (The trick of feeding back 50% of V_DS to the gate helps but is far from perfect.)

Another approach would be to use a pair of varactors as a variable voltage divider. If you use a pair of duals, you can get rid of the nonlinearity to first order. For a purely analogue method, that would probably get my vote.

The old-school approach would be to mount the pot somewhere in the middle of the board, wherever it needs to be, and run a shaft extension to a panel bearing. ;)

Cheers

Phil Hobbs

My little JVC FS-6000 "hi-fi" in my office has a motor-driven volume control pot (if operated via the remote). ...Jim Thompson

People really should understand the "whys" -- then the "wherefores" just make sense, instead of being dimly remembered rote learning.

As long as your op-amp can drive the cable, and as long as you're not going to excessively attenuate or phase shift the signal over your frequency range of interest, putting long lengths of cable with pots at the ends _between_ op amps isn't going to cause problems.

Putting long cables to pots in the _feedback_ path of an op-amp _could_ cause problems, because of cable capacitance and (if you really screw up) coupling to other circuits. So your challenge is to not do that -- make sure that your circuits will work well with low-impedance outputs launching the signal onto the cable, a pot at the end to attenuate it, and a return. Then you'll be fine.

On a sunny day (Thu, 05 Apr 2012 08:45:18 -0700) it happened Jim Thompson wrote in :

That must be very very old.

Philips used to make a whole range of analog audio ICs where they had a DC control voltage for gain, bass and treble too IIRC. So if your 'signal' is in the audio range.... Do not remember any part numbers though. But I likely still have some in the attic.

It could cause problems if he's talking about putting pots at the end of a long cable on the other side of the room vs. inside the Faraday shield of a chassis. This probably isn't the case, but it's not entirely certain from the information provided.

I'd just add that "it depends" if the cabling to the pots is going to a high EMI area, or remaining within an EMI controlled area such as inside the chassis. Even then is it going past some SMPS, deflection coils, or other noise radiating parts, etc.?

That is a very useful addition to the discussion also. For someone learning the ropes, they should gain understanding of why a feedback path is a different animal than the path between independent stages.

Also consider that the scope makers frequently used a mechanical extension from the chassis exterior to a pot mounted on a board inside. This solves various sorts of problems such as a pot at a high potential offset from ground, keeping signal paths short for a low level signal within/between stages, etc.

I don't worry too much about adding noise with a pot from the wiper itself unless there is DC current superimposed on the low level signal. If that is the case, then I figure out a different approach.

Bipolar VCA off the shelf. You can also roll your own with some TI chips and a bit of glue.