[sdiy] Audio taper pots - shunt resistors and contact resistance

[Neil Johnson]

Hi,

(Sorry for the lateness of my reply... busy with family and work)

If I may offer some thoughts on Robin's points....

Robin Whittle wrote:

[skipping summary]

> Using a shunt resistor to ground will create a reasonably good
> "logarithmic" response from a linear pot.  I think this response is
> much more suitable to audio level control than a linear response.

It is not bad over a certain range, say 30-40 or so dB, but is far from
ideal.  But for certain applications it can actually be better than a
typical commercial-grade "logarithmic" pot (unless you buy really
expensive high-conformance log pots) since "log" pots typically use a
two-piece linear approximation, and there is a definite (and some say
audible) join between the two regions.

High-quality audio faders (Penny & Giles) have special resistance pads 
to construct a smooth audio log track.  But that's expensive.

One advantage of bending a linear pot is that you can define the region
of interest - for example many mixing desks use this technique on the
pan pots to achieve the desired mid-pan attenuation (-3dB, -4.5dB, -6dB,
etc).

> However the resulting shunt resistor circuit will only work well as
> long as the wiper's contact resistance with the resistive track (plus
> the wiper's contact with its metal track) does not vary
> significantly, since this "contact resistance" is in series with the
> shunt resistor, or whatever load is placed on the wiper of the pot.

I guess it depends on your definition of "well".  But considering the
shunt resistor method is a crude approximation in the first place the
static effect of the contact resistance is minimal.

All good manufacturers specify the characteristics of the contact 
resistance, both in absolute terms and also how it varies as the wiper 
is moved. The latter is often quoted as a "noise voltage".  For example, 
both the Panasonic and Alps pots mentioned below are
specified to noise voltage ("sliding noise" as Alps say) of no more than
100mV.

> I believe the best way to achieve an audio attenuator circuit it is
> to have a pot with a suitable "audio taper" or "log" response, and
> then to drive the wiper voltage into a high impedance amplifier. The
> output of the amplifier then would drive a mixing state or whatever
> else follows. The idea is to eliminate any significant current
> flowing in or out of the wiper.

Well, in many opinions the ideal audio attenuator is a switched
attenuator, as found on high-end pre-amps and other audio equipment
(e.g., Neve pre-amps), with many advantages.

And actually, you *want* some wiper current.  If you haven't read it I
strongly suggest the following:

http://www.bourns.com/data/global/pdfs/OnlinePotentiometerHandbook.pdf

In the section on contact resistance you'll note that contact resistance
appears to be inversely proportional to contact current.

> A second best approach in a mixer is to have he wipers of multiple
> audio taper pots for the different input channels each feeding a
> relatively high value resistor, with the other end of all the
> resistors going to a summing point.  This could be a high impedance
> input of an amplifier which follows the changing voltage at that
> point.

I thought voltage mixing went out in the 70s?  All those noisy high 
resistances, and too much influence between channels as you change 
levels.  And having physically-long high-impedance circuits is not
really conducive to low-noise design.

> More likely it will be the inverting input of an op-amp which has its
> non-inverting input grounded, with a resistor from the output to the
> inverting input to turn the summed input current into a suitably high
> voltage.

Yep, the typical virtual earth summer.  Although for really
low noise you'd use a balanced bus with a low-noise transistor front-end 
summing-amp.

> If the second approach (in voltage or current mode) uses relatively
> low value resistors, then we are back to the initial problem of a
> shunt resistors: the circuit only works if the contact resistance is
> very low compared to the resistance of the pot and the shunt
> resistor.  How low is "low"?  It depends upon how fussy you are about
> noise when turning the pot.

That's why we keep DC away from the pot!

> The contact resistance of interest is the variation in contact
> resistance when turning the pot.

Yes, the "Contact Resistance Variation" (CRV).

> If you don't intend turning the pot when listening to the audio, then
> most or all of these concerns do not apply.

Indeed.

> Pots are imperfect devices.

Is there any device which is not imperfect?

> One imperfection is the variation in contact resistance as the pot
> is turned.  This is the big problem with the shunt resistor approach
> (or the second approach above, with relatively low resistances).  If
> you have any DC component in the drive to the pot, contact
> resistance variations when turning will generate noise on the output.
> A good circuit will eliminate all such DC components.

Well, audio is AC, so DC has no business being on a pot, so that
dramatically reduces any CRV noise.  A pot in a DC circuit --  that's a
different application, in which we're interested in the DC value, so we
should filter as required to remove any CRV noise.

> Still, the problem remains - variation in contact resistance imposes
> itself as a gain change in the output signal, but one which varies at
> audio rates when the pot is turned.

I think overall that as long as you keep DC away from pots in the audio
path then CRV noise is the least of your worries.

> This means that when any audio signal is present, there will be
> noise imposed when the pot is turned.

No doubt, but is it audible?

> If the signal is all mid-range, you may not hear the noise because
> the signal masks the noise.  If the signal is purely bass, you will
> hear contact resistance variation noise clearly as low to mid-range
> noise, because the low frequency bass waveforms are no longer smooth
> curves on the output - the more the voltage deviates from zero at
> that point in time, the greater the noise due to the the varying
> contact resistance.

That depends what you mean by "low frequency".  Audio typically implies
20Hz to 20kHz or thereabouts.

> I think it is probably not good enough to test the pot and decide
> its contact resistance variation is acceptably low for a shunt
> resistor circuit and whatever fussiness you might have about audio
> signals, depending on what signal you are testing it with.  Someone
> else, or perhaps yourself at a later time with different signals, may
> be fussier.

There are people far fussier than you can imagine who design audio
mixers and pre-amps who have debated this topic and have working
solutions.  I suggest grabbing the schematics of a high-end mixing desk
and trying to piece together the various design decisions and 
trade-offs.  I am merely an interested observer.

> Pot contact resistance varies with the wear on the pot, dust, the
> creep of grease which was intended to lubricate the shaft but which
> over years or decades creeps across to the resistance track and/or
> the metal ring which the other part of the wiper goes onto.

Don't forget bits of metal from the wiper that embeds itself in the
track as it wears off the wiper.

> While it would be possible to rig a pot up to a reciprocating system
> to give it a hammering and try to wear it out, there is no way you
> can reliably predict its behaviour years from now, with the effects
> of moisture, dust, usage with forces on the shaft you don't
> anticipate, the drying of oil into something thicker and more likely
> to lift the wiper from the surfaces etc.

Manufacturers already do this.

> Assuming you care about audio quality and want your machine to sound
> good in years to come, I suggest you avoid all shunt resistor
> approaches and use very light loading indeed on the wipers of all
> pots which handle audio signals.

I suggest the opposite: low-value sealed pots.  Or for very high end 
applications use stepped attenuators, metal foil resistors, and 
high-quality switches.

But wait ... oh my ... they cost money!!!

> Another problem with pots is that the resistance element has a
> non-zero width and the wiper fingers are not all touching a point on
> the resistive track with exactly the same voltage divider factor.
> Therefore, as you turn the pot, the various wiper fingers are
> traversing their own little tracks, each of which has variations in
> voltage divider factor with respect to the other.

Yep.  But then the wiper assembly normalises all those little points
into one electrical potential.  Unless you are pumping 100s of volts 
across a pot the current flow through the track is going to be minimal.

> This is particularly the case where the "log" or "audio" taper is
> achieved by printing a thicker layer of carbon on top of a thinner
> layer, with a diagonal border to smooth out the transition.  Then,
> an outer wiper finger can be in the thick area and an inner wiper
> finger on the thin area, with these points of contact having
> radically different voltage divider factors, despite the fingers
> being correctly on-axis with the rotational centre of the shaft.
> Then, for any input voltage to the CW terminal, significant voltage
> differences would occur between these contact points, so current
> would flow from one wiper finger to another, with the final wiper
> voltage being some mix of the four contact voltages, depending on the
> impedance of those contact points and the contact resistance of each
> of the wiper fingers.

The transition point is roughly midway on a log pot and by the time the 
wipers are in the transition region most of the track resistance is 
already between the input (CW) and the wiper.  Again, looking at Alps 
pots about 30% of the track resistance remains at the transition region.

> The output voltage of the pot, even with a infinite impedance
> voltage following amplifier, represents the sum of, for instance,
> four contact wipers, each on their own imperfectly linear (or
> logarithmic or whatever) voltage divider curve, each with their own
> varying contact resistance (which varies at a potentially audio
> frequency rate especially when turning), in series with the impedance
> of that part of the track itself, which would be difficult to
> calculate, due to resistance along the track and across it to the
> four points of contact by the wiper.

I think you're imaging effects that will be inaudible.  Or do you have 
actual evidence that this is the case?  Cited papers perhaps?

> Furthermore, the contact wiper has a non-zero length - it shorts out
> some length of the resistive track, which varies with the topology of
> the surface, the pressure on the wiper and how worn the wiper is (the
> wiper will get flatter over time, with less contact pressure at any
> one point, and so will be more subject to being lifted from the
> resistive surface by grease, dust or whatever).  There's not much
> which can be done about this.  Fortunately, as far as I know, this is
> unlikely to produce any audible noise unless you go looking for it,
> with a DC voltage across the pot.

All of this is covered by the CRV and noise voltage specs.  Of course, 
over time as the track gets dirty, any audible noise produced will 
increase, hence the recommendation to use sealed pots.  Or regular 
maintenance to keep your pots clean.
And, again, in audio circuits keep DC away from the pot.

> I have a Behringer Eurorack MX802A mixer.  It is compact and very
> neat looking.  The faders are rotary pots - Panasonic 9mm pots EVUE
> or EVUF if I remember correctly, similar to these:

Ummm... right.... Behringer you say....

> ...This use of a 100k pot with a
> 3.3k resistor is an extreme violation of the guidance I have given
> above.

If it works then why not go with it?

 > ALPS suggests:
 > "[Impedance on the Output Side] As when the impedance on the output
> side for the voltage adjusting circuit is low, it may receive the
> affect of contact resistance between the resistor and the movable
> slice, please set the impedance to the value as much as 100 times of
> the total resistance value or more."

They are referring to the industry-standard method for measuring a pot's 
performance.  And by quoting it in their datasheet they're trying to 
protect themselves against idiots who connect the wiper to something 
significantly less and then complain when they don't get the advertised 
behaviour.

> So for a 100k pot, according to ALPS I should have a load impedance
> of 10 megohm.

In a precision DC ratiometric circuit you may well do that, for example 
the input of a FET op-amp.

 > Instead, my load resistance is ~3.3k ohm.  This is a
> 3000:1 violation of the above guidance!

Well, 3k3 on a 100k pot does on the face of it seem a little extreme. 
But if it gives you the control response that you need then there is no 
issue at all.
Just don't go crying to Alps that the in-circuit behaviour it is not 
conformant to the pot's advertised response curve.

[lots of discussion about a specific product]

Cheers,
Neil
-- 
Modules and more :: www.cesyg.com
Homepage :: www.njohnson.co.uk ::