DPOTs for audio control

[JR.]

In connection with a new commercial design I won't talk about, I have reviewed and dismissed DPOTs for audio gain control. Not for sonic quality at some fixed gain setting but for artifacts associated with switching them on the fly.

While I have not seen a complete characterization of the amplitude of this glitch in the data sheets, there appears to be a glitch mechanism associated with gain switching on the fly. The best I can guess it is associated with a similar mechanism to MSB transitions in DACs. The result in practice is that all internal switches do not connect and disconnect perfectly simultaneously, but connect and disconnect with some small time interval. This causes an output glitch that is proportional to the input voltage at that instant in time, but can be well larger than the target output signal as predicted by DPOT attenuation setting.

You will notice that better (sound) quality ICs using this internal switched resistance technology, like the gain control on the THAT mic preamp chip, use zero crossing detection to coordinate gain changes with near zero input voltage, to keep this glitch product small (this also mitigates the expected zipper noise too).

In my design I could have detected zero crossings and coordinated gain changes to them but this could cause another perceptual discontinuity as gain would zipper up and down around zero crossings. Perhaps fast enough with wide range program containing HF content and clearly less offensive than living with the glitches, but not very smooth either. I suspect the nature of any digitally controlled analog gain control will suffer some amount of sampling related quantization related to update rates.

I am not ruling out DPOTs for all audio control. In some ways they are superior to VCAs (potentially lower THD), but for continuous gain control adjustment, IMO they seem more work than they are worth. In addition to the premium charged for DPOTs with decent voltage swing capability, so back up on the shelf for now.

I have a lot more design stuff going on, but this is one thing I can share..

JR

[mediatechnology]

Thanks for posting this John. I know that in your application that the ADI LOGDACs might be too expensive. They also zip but IIRC are monotonic in their gain steps. Have you ever tried them and is ADI still in production of these? I remember them being used in the Studer 820 and MCI/Sony APR-24.

[JR.]

I never seriously investigated multiplying DACs, IIRC there were some positioned as digital faders, but my recollection is they were always challenged for dynamic range, power supply swing, etc...

Since it is the same technology, only older, I expect it to have the same problems or worse.

As far as I can tell the recent effort for higher audio performance has been put into some relatively high voltage, bipolar rail DPOTs from AD.

I notice that some modern DACs will accept a variable voltage on their VREF pin all the way down to ground, but this will still be challenged to make a very high performance audio path, and probably have the MSB, problem, and...... etc. being monotonic is mainly an issue when using one in a feedback path to make a DAC.

The MSB glitch I refer to, is not from the MSB having a scale error after settling, but switching out of sequence, so momentarily output is way high, before all lesser bits are set or cleared. Say you are switching between 0x7F and 0x80 (0111_111 and 1000_000), If bit 7 goes high, before the other 6 bits go low, momentarily the output will be 0xFF (6 dB louder) until the rest of LSBs clear. Kind of an internal timing thing.. It can glitch in either direction if the bits don't all switch at exact same instant.

JR

[mediatechnology]

VCAs. With a NVM DPOT on Ec.

Holy shit.

[JR.]

what's a NVM? Non volatile memory?

JR

[mediatechnology]

Yes. VCA, SPI interface for control, stored wiper (gain) setting, slewing between steps with an external C to eliminate zipping. Add an analog Ec for realtime control. The VCA always wakes up with the last gain setting and moves between steps with a smooth linear dB slope.

[JR.]

Yes, THAT has the technology to do that pulling bits from some existing parts. I'm not sure how large a market there are for those.

Cheap 16b codecs are like $1.50 so the VCA's days are numbered... IMO

I wouldn't invest serious money to tool up that part.

Their digital controlled mic pre should have legs, and analog VCAs as legacy parts will be around for decades. Hybrid digital controlled analog for middleware seems a kind of esoteric market, unless the VCA's get a bunch cheaper.


JR

[mediatechnology]

So what are you going to do JR?

[JR.]

I am not comfortable sharing details of what I am doing, since it is in a commercial design so proprietary. In broad strokes I am using a digital brain to crunch data,,and analog means to vary gain. The approach involves at least one new twist on using established technology to get improved performance. Maybe I can share this all later.

As I mentioned it would be cheaper to use the 16b codecs and do everything in the digital domain, while I expect to deliver better than 16b performance.


I posted this thread to save other the effort if they were considering DPOTs for audio control (as apparently some do). I have seen DPOTs used in at least one Dig controlled analog console (Gamble), but AFAIK he just lives with the glitches.

I should probably , post this into my old "shiny hammer" thread, but i am getting excited again about my uber-compressor where I proposed using a PIC to handle the side chain for a dynamics processor.. I am leaning toward a LCD display with a handful of rotary encoders to make EVERYTHING variable. A single channel dynamics processor could address separate compression, limiting, downward expansion, and de-essing, with ten or more adjustable variables in each sub function... all saved and recalled from memory.

But not here... and not this morning.

JR

[mediatechnology]

I am not comfortable sharing details of what I am doing, since it is in a commercial design so proprietary.

Kinda thought that's what you'd say.