Product selection: digi pots

[obiwanjacobi]

Hi,

I'm looking for a digital potentiometer with a 128 (or 129) steps ranging from 10k to 250k.
Special consideration is that it (the resistor part at least) has to function on +/- 12V - it will be used in an audio project.

I have found the MAX DS3502U+ that seems pretty cool. But when I look for who is carying this chip I can only find mouser.

Question is, is it safe to go with this chip - when do you decide against it?
I expect to build only a few units of this project. Each project will need 20-30 of these chips (its a modulair design)...

Thanks.
Marc

[Bored@Work]

Question is, is it safe to go with this chip - when do you decide against it?

I always decide against any digital pot, unless one is absolutely necessary. Digital pots are the band aid of the inexperienced designer. What should almost always instead happen is that you want to drive a controlled current or a controlled voltage into the circuit. Which can be done with a DAC, and there is a much better choice of DACs than digital pots. Even if you need to control some amplification a multiplying DAC in the feedback loop of an opamp can do it.

[obiwanjacobi]

That is my feeling also - and I do not have very much experience in designing electronic circuits, but I love the creative trail of thought it can take you on 

I wish to replace normal (turning) pots with 'something' that is digitally controlled in EXISTING audio circuits (guitar effects). Here is an example.
A digital pot was the best thing I could come up with.

If some-one has other/better ideas, I would really like to hear from you.

Thanx,
Marc

[ejeffrey]

I always decide against any digital pot, unless one is absolutely necessary. Digital pots are the band aid of the inexperienced designer. What should almost always instead happen is that you want to drive a controlled current or a controlled voltage into the circuit.

 Which can be done with a DAC, and there is a much better choice of DACs than digital pots. Even if you need to control some amplification a multiplying DAC in the feedback loop of an opamp can do it.

I agree with you on the first part, but for an audio volume control I don't think there is any contest.  A digital pot is the way to go.  It is basically what they are designed for, and they do it well. A multiplying DAC will work, but there isn't exactly a great selection there either.

[SeanB]

What about motorised pots? They are audio rated devices after all, and have a audio law built in, along with good isolation between the digital and audio sides. Non volatile, and can handle very high voltage audio inputs.

[olsenn]

Be careful about how much power they are dissipating... you can get logic high by smoking pots

[obiwanjacobi]

What about motorised pots? They are audio rated devices after all, and have a audio law built in, along with good isolation between the digital and audio sides. Non volatile, and can handle very high voltage audio inputs.

The idea is to store the pot's values in memory (presets/patches) and recall them triggered by a Midi Program Change message.
If I can get rid of all pots I can mount it in a 1 unit high 19" case with just the LCD and some keys on the front...

EDIT: I just looked up motorized pots at mouser and the cheapest one was 5.68 euro. Way too expensive!

[typeglob]

The idea is to store the pot's values in memory (presets/patches) and recall them triggered by a Midi Program Change message.

I've been thinking about something similar and also found digi pots to (likely) be the most usable solution. In my case it concerns a small MCU controlled unit with 4 MIDI in, 4 MIDI out and 50k ohm 'expression' pedal in/out. If using a digi pot with enough precision, it would allow me to do things like convert a signal from a pedal with a linear pot (Fatar) to one with a tapered pot (Roland).

I wonder how well it would work if I would use a 100k ohm digi pot, and limit the range to 25k or 50k ohm. If the digi pot is linear, I'd still expect to have 25% (25k) and 50% (50k) resolution left so a 12-bit one should be plenty precise enough to yield a usable 7-bit MIDI value even at 25k.

[obiwanjacobi]

Resolution is not something I think will be a problem. Finding the correct value probably is.
The manual pots I replace can have any value (log or linear) but digi pots do not come in any value and are usually linear.

The conversion from linear to log can be done in software. But adapting the value is a bigger issue. My solution for now is to use normal resistors to make the digi pot operate in the 'sweet spot' of the range. I also thought of putting multiple digi pots in series but that is more expensive and requires software to make it work as one pot.

[ejeffrey]

Resolution is not something I think will be a problem. Finding the correct value probably is.

The resistance tolerance on pots, and especially digital pots is terrible.  You shouldn't be using them in circuits that are particularly sensitive to the end-to-end resistance in any case.  If used as a simple divider a 10k and 100k will perform similarly in the same circuit -- the value is of little consequence.  If you have an external resistor network connected you will need to rescale the values but that is easy.

The manual pots I replace can have any value (log or linear) but digi pots do not come in any value and are usually linear.

There are digital log pots if you really want them, but much less selection.  These days it is more common to use an external resistor network to give a psuedo-log law from a linear pot.

Each project will need 20-30 of these chips

What?

[obiwanjacobi]

What do you mean with:

If you have an external resistor network connected you will need to rescale the values but that is easy.

Yes I need a lot of them.

The idea is to rip out the PCB's of existing Guitar Effects (Stomp boxes) and control them digitally using an MCU, LCD and some push buttons.
Multiple effects will be housed in a single unit (max 16?) and each effect may have as many as 8 pots (equalizers) or as few as 1.
I expect that I will build a modulair system of banks of digi pots and assign the function of each (to what effect they belong) in the software.
I also want to be able to put them in any order (see also this thread).

[T4P]

What do you mean with:

If you have an external resistor network connected you will need to rescale the values but that is easy.

Yes I need a lot of them.

The idea is to rip out the PCB's of existing Guitar Effects (Stomp boxes) and control them digitally using an MCU, LCD and some push buttons.
Multiple effects will be housed in a single unit (max 16?) and each effect may have as many as 8 pots (equalizers) or as little as 1.
I expect that I will build a modulair system of banks of digi pots and assign the function of each (to what effect they belong) in the software.
I also want to be able to put them in any order (see also this thread).

I do not know if The Tone God has shared what he uses but you can get an idea from his website's technical articles
http://www.thetonegod.com/

[obiwanjacobi]

Thanks for the link. I see he's using the led/ldr combo.

[TerminalJack505]

That is my feeling also - and I do not have very much experience in designing electronic circuits, but I love the creative trail of thought it can take you on 

I wish to replace normal (turning) pots with 'something' that is digitally controlled in EXISTING audio circuits (guitar effects). Here is an example.
A digital pot was the best thing I could come up with.

If some-one has other/better ideas, I would really like to hear from you.

Thanx,
Marc

Have you looked into using optocoupled photo FETs (for example, $1.80 in quantities of 25 at Digikey)?  Or maybe good ole JFETs?  (FETs As Voltage-Controlled Resistors.)

The photo FET is basically the same concept as the LED/LDR combo.  It will require more power than a digital pot.  It doesn't suffer from the common mode voltage issue that the digital pot will have, however.  That's something that could come in handy since you are replacing pots in existing systems.  (One less thing to worry about, anyway.)  They also have a wider voltage range and wider resistance range than digital pots.

You'll need some way to drive them, obviously, so they're not going to be as simple as a digital pot.  They are unlikely to be as precise as a digital pot either but that will largely depend on how you drive them.

Just one more idea.

[obiwanjacobi]

Thank you for the suggestion. I stumbled across the "FET as variable resistor" while searching the net. I just wasn't sure if/how that would work.

Bottom line will probably be that I need to do some testing and experimenting. Most guitar effects run on a single 9V battery so I think the MAXIM ic I posted at the start will work. It might even work for circuits that run on a symmetric supply (max +/-7.5V).

Yesterday I order 2 hi powered leds and 2 'normal' LDR's and some heat-shrink tubing (I hope will fit). I hope to do some experimenting with them. Because, although this solution will require a bigger power supply and drivers to drive the LEDs, It will also probably be the most successful for the circuit in question (I hope). I will do some measurements (how hard to drive the leds) and try to simulate one pot in a real circuit (like the example I gave earlier - that's a unit I own). I also find the isolation this solution provides very appealing.

I've also requested samples with MAXIM  Hee, you can try  Otherwise I will buy a couple of those digi pots and experiment with them as well.

After that, if I still am not satisfied I can try the FET option. Would useful for others to do a comparison (say on YouTube). I am not sure if I have the knowledge to do it right. We'll see.

Meantime I am also trying to finish two other projects  This is just research for the next project. Although I think that if I can pull it off, a lot of guitar players are going to be very happy. 

[TerminalJack505]

Yes, the isolation is a nice feature of the LED/LDR and photo FET solutions.

Regarding photo FETs, I found this article on EDN:

Use a photoelectric-FET optocoupler as a linear voltage-controlled potentiometer

Figure 1 shows the resistance vs. IRED current.  It looks like you would need a fairly precise current source to get the desired resistance.  On the plus side, it doesn't look like they will need a lot of current to achieve useful resistance values.

The article shows a circuit using two of these optocouplers to get an excellent linear response.  Of course your circuit wouldn't need a second optocoupler since you'll be using a MCU.  You could just characterize each optocoupler in your system and write these characteristics to the MCU's EEPROM or flash and use these when driving them.  You might have to factor-in the temperature too.

I've never used these things.  I might have to get some just to play around with them.

Have fun with your project!

[obiwanjacobi]

Hee that's great info there. Thanks very much.

Yeah, the range is about 0.75 mA to about 4.5 mA (Figure 1). I have no knowledge of how to make a simple current source and if the whole thing will be cheap /simple enough - remember I need a lot of them.

Yeah, I read people trying to match up LDR's and stuff, but I think I will just design a test-bed that measures the range and linearity and output the numbers I need for the software to correct it. Takes a little bit more runtime processing, but saves manual labor. Additionally I might also put in some space on the board for fixed resistors (or wire jumpers if not used) to offset the range to the 'sweet spot' of the pot. The 'sweat spot' would be the range of values for that pot that would be most usable in the effect.

Thanks,
Marc

[T4P]

Yup photoelectric-fet couplers are cheaper and work better and you don't need to go manual and do the led's and ldr's together
Less man hours, lower cost, lesser effort

BUT there's a catch ... i can't find any part except for 1 photo fet optocoupler that is obsoletium on E14 but i can find quite aplenty on RS

Best one to use is the H11F1SM

[obiwanjacobi]

BUT there's a catch ... i can't find any part except for 1 photo fet optocoupler that is obsoletium on E14 but i can find quite aplenty on RS

I have no idea what you just said... 

[T4P]

BUT there's a catch ... i can't find any part except for 1 photo fet optocoupler that is obsoletium on E14 but i can find quite aplenty on RS

I have no idea what you just said... 

Basically  i can only find 1 photo fet optocoupler on my favourite site and that is obsolete ...
I tend to speak in riddles sometimes and that's what makes BravoV hate me so much

[BravoV]

BUT there's a catch ... i can't find any part except for 1 photo fet optocoupler that is obsoletium on E14 but i can find quite aplenty on RS

I have no idea what you just said... 

LOL .. you're not supposed to, cause that garbage post is intended or supposed to increase his "reputation", trust me, once you've spent enough time around, eventually you will bump into his junk posts everywhere at almost every thread, and seems like he is hallucinating that he is sort jack of all trades in electronic. 

[T4P]

BUT there's a catch ... i can't find any part except for 1 photo fet optocoupler that is obsoletium on E14 but i can find quite aplenty on RS

I have no idea what you just said... 

LOL .. you're not supposed to, cause that garbage post is intended or supposed to increase his "reputation", trust me, once you've spent enough time around, eventually you will bump into his junk posts everywhere at almost every thread, and seems like he is hallucinating that he is sort jack of all trades in electronic. 

HAH. Bullshit, reputation? REPUTATION? You've gotta be kidding me ... I saying that the optocoupler is obsoletium on element14 now is a garbage post?
It was related to being able to get photoelectric fets easily or not ...

[TerminalJack505]

I'm guessing you already knew this but it was news to me...

It looks like there is a name for what you are making with the LED/LCR pair.  They're called resistive optocouplers/optoisolators or vactrols.  I found a couple of companies that make them--Silonex, Inc and Excelitas Tech.

This webpage was very informative.  It shows some audio applications using vactrols.

Also, it looks like the JFET/photo FET approach will only work for signals that are just a couple of volts, peak-to-peak.  Apparently, the linearity suffers beyond that.  So it sounds like you're on the right track.

[TerminalJack505]

I found another good resource on vactrols:

PerkinElmer Optoelectronics Literature

Start on page 27.

Even if you roll your own there should be some good information there.  Of interest is the information on distortion.

[obiwanjacobi]

@TerminalJack505: wow, that is terrific! Thank you so much. And I didin't know they were called Vactrols. This will take me some time to read and absorb.

I want to thank everybody that contributed. You've dug up resources I would have never found myself. Thank you all.

[lpc32]

So what did you end up doing?

[linux-works]

I've spent a few yrs developing and testing analog audio attenuators.

first, let me point you to my (partner's) website that has kits and schematics for controller based vol controls: www.amb.org (look for 'lcduino' and 'delta1' as products).  the code is open-source and we support users on the amb.org forum.  what we have now is a relay atten and I'm happy to answer any questions about that.

the code also supports 2 other attens; one is a TI pga23xx chip and the other is a cirrus cs3318 chip.

the pga is a pain to deal with and its fussy about layout and power.  but its cheap and it can be an ok vol control for stereo.

my current fave is the more expensive (and harder to solder chip) cs3318.  its 8ch of analog vol control and its pretty much noise-free and distortion-free.  I have one in my current DIY home stereo and I plan to do a board for it, someday (right now its perf-board since I have not designed a pcb for it).

I would avoid digital pots.  they are not meant for audio.  I tried a few simple ones but you really want a buffer and that buffer to be gain controlled.  the pga/cs chips have both input atten stepped R's and also variable gain (feedback R) based on whether you are below 0db or in the gain region.

I also have done my share of motor pots and my code will support that, too (as will the lcduino) but running audio thru a mech pot is not so great.  l/r balance always sucks toward the low end of the pot and there's crosstalk as you go down in levels, too.  I now use digital control of analog-domain vol controls and swear by them.  or relays, if you can deal with the relay clicking


HTH

[linux-works]

oh, and as for LDR's, you don't want them.  they have distortion, they are a HUGE pita to match (and keep matched) and they are more of a 'fad' than true high end atten.

what they excel at is muting!  if you needed a soft mute (hard off/on, no middle states) they are great for that.

but to run LDR's for vol control is a mess.  no one seriously uses them (well, one company that I know of does; and they charge upwards of $20k for their preamp due to the HUGE complexity of making it all work well).  its the old audio alchemy guys doing that and I have no idea how big their sales are, but I've never known anyone to own a $20k preamp before

[magetoo]

I've been thinking along the same lines as the OP did (two years ago..) so this was an interesting read.

l-w, it seems like you are doing something pretty different (volume control/fading) from what he wanted to do.  I'd be interested to hear what you would use for a generic pot replacement that isn't a volume control.

I guess the correct answer for this sort of thing is "redesign the thing from scratch" and use DACs for control, but...