Digital pots using relays

[Gregg]

I've always wondered about chip makers implementing an actual ohmic resistive element in combination with several back to back MOSFET's or any other semiconductor switching topologies (sort of like a linear rheostat... except with selectable resistance taps). I'd pay good money for one of those!

The last thing the manufacturers would think about when designing these things are large audio signals as found in guitar amps... On the other hand you have plenty of audio grade digital pots that can handle lower signals.

The other solution is to lower the signal after the last tube stage and use active TS with regular digital pots. What I did in my preamp is I loaded the CF with a fixed value TS which according to the TS Calculator produces flat response, attenuated the signal to -10dB and then used an active TS with regular digital pots. However this kind of solution is not suitable for the Gain pot.

One issue with relays is reliability. A circuit with 100 relays in it will give 20 times more relay trouble than a circuit with 5 relays. For that reason I'd not settle for anything less than hermetically sealed relays with gold contacts, which will bump the price up somewhat.

Well, you know what they say. You never know before you try.

[exclamationmark]

I guess I should have clarified.. I know digital pots exist, just that they, well... kind of suck. Mainly to do with the fact that they combine all the drawbacks of a solid state relay with a resistor that could barely be called ohmic (the I-V curves for most of these devices are horrible, though that might have more to do with the MOSFET switching or whatever they use). I've tried cramming a 3-bit device on a 40 pin DIP footprint using MOSFET switching, and it works somewhat successfully... except that it produces stupid amounts of crossover distortion. When I get enough time I might switch to SMT components and up the bit-count, provided I can solve some issues.

Ideally I'd like to end up with something I can just socket into my designs rather than allocating board space for it. I don't know why, but there's something so damn cool about having a hundred presets for a single tube amp, that you can just press a button, and hey presto!

[Steve Conner]

Exclamationmark, are you saying that you built your own digital pot out of discrete MOSFETs because the IC ones you can buy weren't good enough somehow?

I've mentioned this many times before, but MOSFET output solid state relays are getting smaller, cheaper and higher voltage rated all the time. I've seen forum members use them for channel switching in place of LDRs. Maybe one day they'll be small/cheap enough to make a 3- or 4-bit digital pot. Maybe they already are.

[exclamationmark]

Well, yes, I attempted to DIY digital pot, only that it didn't quite work haha... Not that conventional ones aren't good enough in terms of distortion, etc (as said before there are a few suitable for audio use), it's just that the voltages they tolerate are way too low.

I'd love to be able to use solid-state relays for a project I'm working on though! They wouldn't necessarily save me much money (they're around the same cost of mechanical relays), but the amount of board space and reliability would be worth it. I've tried a couple MOSFET based solid state relays but (I think?) the output resistance + capacitance was causing some strange things to happen. I hope to have better luck by making a SSR from scratch using MOSFETS so I can control the parameters a little better... Or I might just crack and shell out on a couple hundred mechanical relays.

[R.G.]

This is getting a very familiar ring to it. In the years-ago article in Audio Amateur, the author finally settled on JFETs as signal switches for their ability to switch high voltages and lack of artifacts, after considering other devices. It's possible to make a JFET switch that will switch tens of volts easily enough, and each switch is a TO-92 or similar and costs $0.20. I believe that easy, cheap JFETs will give you a switch that will do +/-30V peaks. Selecting other JFETs would probably better that.

[d95err]

I have two ideas for remote-controlled pots that should work with tube amp voltages. The first would be to use Alps motorized pots. They are not that expensive compared to things like transformers and tubes, so it should be OK for a DIY project. Using motorized stereo pots, one side would be in the tube circuit and the other used for a reference voltage. A microcontroller would read the reference voltage, store presets, and control the motor when doing preset recall. Two problems: they are slow, at least the ones I've seen on eBay. I think 10 seconds or so from min to max. Second, the maximum values available are 100k or even lower. I suppose you could get around that using cathode followers (or MOSFET source followers) in strategic places though.

Another idea is to make up the pot of two LED/LDRs (Vactrols) and have a microcontroller control the current through each LED. Because LDRs are have a non-linear response, one could use fixed steps, with each step trimmed individually to get the right resistance. It would probably be enough to specify 4-5 fixed steps and then interpolate between them, for a fairly continous sweep.

Just ideas though, I don't know if I'll ever get around to trying it.

[exclamationmark]

I've tried using vactrols driven by PWM to control the resistance before. A couple of things bothered me though... The vactrols varied immensely, in that I would literally have the change the duty cycle of the PWM for each vactrol I wanted to use, by using a multimeter to measure the resistance for every step I wanted... Needless to say this takes a lot of time. Also, their resistance never really reaches a steady sort of level (which is fine for volume control I guess).

[Steve Conner]

IIRC, the vactrol idea is patented by Mesa. It's what they use in the Triaxis preamp.

Again IIRC, they used dual vactrols, with one cell in the circuit and the other as a reference for a feedback loop, to get round the inconsistency issue. Like the motorized pots idea above. Matched (ish) dual vactrols are available, they are used for things like crossfaders in DJ mixers. I'm pretty sure a forum member has built this and got it working.

Fender used the motorized pots successfully in their Cyber Twin. It worked and people loved the gimmick. They had custom ones made, so maybe you can just buy them from the Fender spares department.

[Gregg]

After renewing my digging on the subject this old thread came to my attention. I found the article mentioned by loudthud but as you can see it lacks substantial details. I think 64 steps are to much for guitar amp pot, optical encoders do not come cheap, firmware is not available, relays switching order it's not clear etc.

Digital pot with relays.pdf

I did some additional digging to learn more about this and found this more detailed project:

Audio Volume Relay Attenuator with IR Control

It comes with a calculator giving you different log pot options:

Logarithmic Attenuator Calculator

You can design for a 4 bit pot (16 positions) and a 5 bit pot (32 positions) which sound like more suitable for guitar amp pots.
The main question for me still was about the relays switching sequence. After some boring calculations I made a table and found out that it corresponds to the truth table (the one in the middle) of this 4 bit encoder which is available from different sources:

http://www.grayhill.com/assets/1/7/Mech_Encoder_25L.pdf

Of course it's not suitable as direct control as it rotates continuously in both directions but since we already have the sequence it's not difficult to implement it in your own project/software.
Unfortunately I couldn't find a 5 bit truth table so I calculated one myself. Basically it looks like two 4 bit table on top of each other with one additional column where half of the column cells are "empty" the other half "full".
So what's your opinion on how many positions should such "digital pot" have? Would 16 be enough and are 32 too much?

[masco]

Is this same idea as a uPot? http://www.ti.com/lit/ds/symlink/lm1972.pdf

[Gregg]

No. LM1972 is a regular digital pot. The idea of the "analog relay digital pot" is to avoid SS switching and to allow for the high amplitude signals found in guitar amps.

[Steve Conner]

Neat stuff, Gregg! The stages are binary weighted, so the truth table is just a count from 0 to the number of steps minus one, in binary. You could use a BCD switch to control it directly for test purposes.

I think the minimum number of steps needed would depend on the application. A tone knob might be fine with 8 steps but a master volume might need 32. I once built a hi-fi amp with a 22 position switched attenuator in 3dB steps, and I found the steps too coarse at the high volume end of the range.

[Gregg]

From what I've seen some high end attenuators have 1dB steps in the beginning then 2, 4 etc. but with this design you can't get that. However this way the main problems concerning the signal amplitude and SS stuff are solved.

[Steve Conner]

The only drawback is, I think it's impossible to make a log taper variable resistor by this method. You can only have a log potential divider, the variable resistor has to be linear.

[R.G.]

At the risk of being repetitious, the only remote controlled pot I've ever really, really liked was the dual-shaft stepper motor with a knob on one end and a rubber-hose shaft coupler to an actual analog pot on the other end. It's smooth, quiet, has a nice feel on the knobs, and can be electrified into any position, then left either un-held or fixed in position by the motor.

Needs a uC to control the steppers, but that's getting downright cheap and easy these days. The real trick is getting a batch of small 200step/rev motors and doing the mechanical mounting.