Digital Pot

[NeoY2k]

Hello everybody, my first post here! (I thought I already suscribed... but not.)

I'd like to ask you questions about digital pots.
I'm building (slowly!) a full digital system (it started on diyaudio, and I don't even have finished the DAC output so... there's time).
That means of course distant preamps. It also means you have to find a way to control gain...

Digital pots do exist but are very noisy. Is there any that could be used in a preamp gain control? (like that1512). If I'm right (i'm not sure, don't beat me!) "Cohen" topology preamps put the gain resistor in a current feedback loop. Digital pots have a high voltage noise, but lower current noise. So they *may* work not so bad in this place (depends on impedances etc).

Has anyone tried? Do you have some ideas about it?

What are the other ways to do it? I see I could use relays (too big and expensive), reed relays (less big, less expensive, but still big, expensive) or motorpots (you have to servo them and they may be a bit noisy too...).

Thank you,
Nicolas

 

[JohnRoberts]

I have looked at this problem for years and it isn't easy to switch gain at those impedances and levels.

Unless you are willing to undertake a major design project, there is a TI mic pre chip that accepts digital gain control that is respectable.

JR

 

[abbey road d enfer]

Hello everybody, my first post here! (I thought I already suscribed... but not.)

I'd like to ask you questions about digital pots.
I'm building (slowly!) a full digital system (it started on diyaudio, and I don't even have finished the DAC output so... there's time).
That means of course distant preamps. It also means you have to find a way to control gain...

Digital pots do exist but are very noisy. Is there any that could be used in a preamp gain control? (like that1512). If I'm right (i'm not sure, don't beat me!) "Cohen" topology preamps put the gain resistor in a current feedback loop. Digital pots have a high voltage noise, but lower current noise. So they *may* work not so bad in this place (depends on impedances etc).

Has anyone tried? Do you have some ideas about it?

What are the other ways to do it? I see I could use relays (too big and expensive), reed relays (less big, less expensive, but still big, expensive) or motorpots (you have to servo them and they may be a bit noisy too...).

Thank you,
Nicolas

The whole family of Maxim digital pots is essentially noiseless (at least not less than a conventional resistor of same value). Control port spillage is very low, and occurs only on gain changes. For a remote controlled mic pre, this shouldn't be a problem, because the gain is not supposed to vary once set. But they are operated from a single 5V supply, which makes them somewhat difficult to include in a mic pre that is subjected to higher voltages. Now, if you use a switchable pad that would ensure no signal higher than 5v p-p gets in, you could use them. Check the MAX 5484http://www.maxim-ic.com/quick_view2.cfm/qv_pk/4730. I've used the DS1801 some years back on a presentation mixer; performance was excellent, much better than actually needed. But I don't think it would end up being a top-class product. The PGA2500http://focus.ti.com/lit/ds/symlink/pga2500.pdf offers better measured performance, but it is not considered sounding good...
Motorpots is probably the solution that combines cost efficiency and signal integrity. You don't really need to servo them, you could use a +/- command and a return voltage from the second track (actually, it's your brains that servoes the pot!).
Assuming you use the LTP structure with gain resistor across the emitters, the pot would rotate in reverse.
I'm not a fan of relays, as you say, they are bulky, expensive and not terribly reliable in the long run.

 

[sahib]

This issue came up before and it will be a bit repeating myself I am afraid.

If you want to go down the route of motorized potentiometer use an R/C servo and couple it to a potentiometer. You do not have to bust your guts to try to find a reverse log pot. Reverse the control of the servo and off you go. Build a PWM with a 555 and use the digital pot for varying the pulse width. It would be nice with up-down buttons.

The R/C servos can move in very fine steps. I use them in 8 bits with 255 steps which is more than what is needed in audio. A standard ones cost only few bucks.

 

[NeoY2k]

Hi,
Thank you all. I see it's a common problem!

The TI part is PGA2500 and it's not that bad (-128dB EIN @ 150 ohms 30dB gain. Distortion isn't bad either).
However, +/- 5V supply doesn't give much headroom! If you want to run at the nominal +4 dBu, you only have... 4dB headrom (5V peak gives 8dbu peak).

I want the next stage (ADC buffer) to be able to switch between the preamp or direct +4dBu input with -18dBFS for 0dbVU (intended for live sound, so I want headroom! And I'll be using ultra low noise converters (in fact I'm still waiting for the chip that is announced for this month) so low noise matters too).

So FS is at 22dBu input. Vs 8dBu input for the TI chip. So to use it I would need a fixed 14 dB post stage: why not? But this chip has a gain start at 10dB, so my gain would start at 24dB and that is too high. So I'd need to make it switchable. I don't feel that solution to be ideal.

That will be the same problem with any 5V digital pot... But I might try this alternative. Or this PGA chip, or a That1200 with 5V digital pot.

Well I'm starting to think the most simple and cost effective way is to servo a 2-layers motorpot.

I still have to investigate/think more...

Thanks,
Nicolas

 

[jdbakker]

I want the next stage (ADC buffer) to be able to switch between the preamp or direct +4dBu input with -18dBFS for 0dbVU (intended for live sound, so I want headroom!

So have the PGA2500 drive your ADC directly (or through an OPA1632 level shifter if you feel so inclined) and put a switchable pad in front of the 2500 for line level signals. If you're going to be driving an ADC anyway there's little point in first amping your mic up to 4dBU + 18dB and then padding it back down, with all the noise and distortion that adds.

And I'll be using ultra low noise converters (in fact I'm still waiting for the chip that is announced for this month) so low noise matters too).

Which one's that, the new Sabre or somehing else?

JDB.

 

[NeoY2k]

Hi,
Yep I'm looking for the new Sabre. I already made boards for their DAC! (it has to be finished in an usable product, but it works).

In fact, my problem is that I work for theatre. That means that we mix "live" techniques with movie techniques - and these require very low noise and distortion, because you often record very low sounds with a lot of dynamics.

That's why I want to keep an ultra-clean switchable direct path to the ADC buffer, for use with really good preamps when possible/needed.

On the other hand, we can't make a living of only working for theatre, so we also do recordings, concerts... And then we need lot of preamps, they don't have to be the best, but the best they are for the money, the better.

Constraints on these are cost and space, but I want reasonable performance.

So from the cost and space point of view, the PGA2500 clearly is a winner. I would follow it with a "level shifter" to keep level compatibility with the direct input. But from the qualitative point of view, this is lame: to keep dynamics, I would need to run the PGA2500 at low gain, and low gain means more noise. So to escape the headroom problem I'm burrying the end into noise, and wether I'm further amplifying it or not doesn't change anything. I trade cliping for noise... But that may be the most cost effective way, so I can't dismiss it.

Another approach would be to run the That1200 with a high voltage digital pot. Thus we have headroom... But noise rises. How much?..

Another approach may be to use the That1200 with a motorpot, I think it is the solution that would allow the best quality. But I may be wrong, cermet noise can be high. It would cost the motorpot (alps) for each channel. After that, I would just switch with a digital switch which channel I'm talking to/sensing, use only one opamp to move the motor to follow the PWM filtered output of the µC, and we're done. But this approach coasts really more when you factor it for 8 channels... An R/C motor coupled to a pot is more expensive than an ALPS motorpot too (but may be more easily sourceable... but you have to mount it then).

Relays don't appear that expensive in the end, but their control for 8ch won't be easy (I'd need a latching shift reg and cheap transistor arrays to provide the power, that for each channel).

I'm still thinking about it and can't find an answer. I will probably lay a board for that1200 with plugs for a gain setting board: DS1801, AD's counterpart, motorpot, relays, and compare to the PGA2500 eval board. Well, at least I'll tell you the results lol. But it's not going to happen right now.

Why isn't there any motored rotary switch???

Thanks,
Nicolas

 

[jdbakker]

In fact, my problem is that I work for theatre. That means that we mix "live" techniques with movie techniques - and these require very low noise and distortion, because you often record very low sounds with a lot of dynamics.

You may want to put a number to that first before you start optimizing. How low is the ambient noise floor in your environment? As the stage manager of one recent venue said to me recently: "How many dB does your mike's noise floor really need to be below the sound from our HVAC systems ?"

So from the cost and space point of view, the PGA2500 clearly is a winner. I would follow it with a "level shifter" to keep level compatibility with the direct input. But from the qualitative point of view, this is lame: to keep dynamics, I would need to run the PGA2500 at low gain, and low gain means more noise. So to escape the headroom problem I'm burrying the end into noise, and wether I'm further amplifying it or not doesn't change anything. I trade cliping for noise... But that may be the most cost effective way, so I can't dismiss it.

I suspect that in live/theater work the 2500 will not be the limiting factor.

Relays don't appear that expensive in the end, but their control for 8ch won't be easy (I'd need a latching shift reg and cheap transistor arrays to provide the power, that for each channel).

http://focus.ti.com/docs/prod/folders/print/tpic6595.html

I'm still thinking about it and can't find an answer. I will probably lay a board for that1200 with plugs for a gain setting board: DS1801, AD's counterpart, motorpot, relays, and compare to the PGA2500 eval board. Well, at least I'll tell you the results lol. But it's not going to happen right now.

You mean the THAT1500, right? Remember that the gain set connections are inputs too; all RF junk injected there ends up in your output, and any parasitic capacitance will lead to undesired results.

Why isn't there any motored rotary switch???

They used to be very common in telephone exchanges, driven by the dialing pulses generated by the telephones of the day. They were LOUD and not entirely reliable, and were eventually replaced by (reed) relays and later by digital routing.

JDB.

 

[NeoY2k]

In fact, my problem is that I work for theatre. That means that we mix "live" techniques with movie techniques - and these require very low noise and distortion, because you often record very low sounds with a lot of dynamics.

You may want to put a number to that first before you start optimizing. How low is the ambient noise floor in your environment? As the stage manager of one recent venue said to me recently: "How many dB does your mike's noise floor really need to be below the sound from our HVAC systems ?"

So from the cost and space point of view, the PGA2500 clearly is a winner. I would follow it with a "level shifter" to keep level compatibility with the direct input. But from the qualitative point of view, this is lame: to keep dynamics, I would need to run the PGA2500 at low gain, and low gain means more noise. So to escape the headroom problem I'm burrying the end into noise, and wether I'm further amplifying it or not doesn't change anything. I trade cliping for noise... But that may be the most cost effective way, so I can't dismiss it.

I suspect that in live/theater work the 2500 will not be the limiting factor.

Relays don't appear that expensive in the end, but their control for 8ch won't be easy (I'd need a latching shift reg and cheap transistor arrays to provide the power, that for each channel).

http://focus.ti.com/docs/prod/folders/print/tpic6595.html

I'm still thinking about it and can't find an answer. I will probably lay a board for that1200 with plugs for a gain setting board: DS1801, AD's counterpart, motorpot, relays, and compare to the PGA2500 eval board. Well, at least I'll tell you the results lol. But it's not going to happen right now.

You mean the THAT1500, right? Remember that the gain set connections are inputs too; all RF junk injected there ends up in your output, and any parasitic capacitance will lead to undesired results.

Why isn't there any motored rotary switch???

They used to be very common in telephone exchanges, driven by the dialing pulses generated by the telephones of the day. They were LOUD and not entirely reliable, and were eventually replaced by (reed) relays and later by digital routing.

JDB.

Yep, you're right. In fact, I'm looking at the PGA2500 thread and am starting to think that it must be quite good, indeed.
For the THAT1512 test board (yep the 1200 is the line receiver... need coffee) I'll put that gain circuitry on a "shield" that would plug directly in. But you're right, the cleaner the better.
Thank you very much for the ref.
And for the rotary switches history! I didn't knew that. I found some on internet... Well. Too big. Funny though.
Thank you
Nicolas

P.S: I won't start prototyping these very soon, I want to finish my DAC box first. But I was thinking about the next step... Well let's finish this fast. I want a ready-to-go complete unit in less than 2 month. I can do it. xD.

 

[PRR]

> low gain means more noise.

That's over-simplified?