OPA1632D heating up in a custom PCM1792A DAC

Hello,

As part of an open-source DJ controller/4ch audio interface I designed an I2S DAC using the PCM1792A DAC and OPA1632D op-amp. Schematics and board images are available here: https://imgur.com/a/m6g4z

It works perfectly (minus the heat issue) and sounds great (haven't measured yet but no noticeable difference from my MOTU UltraliteMk3). The OPA1632Ds get pretty hot - hot enough to be unpleasant to touch after a few seconds.
The PSU feeding the board measures 350mA at 12v so no huge current draw (there are 2 DACs and a 5 LEDs on this board).

A random guess would be some oscillation but wouldn't that result in some hearable noise? I also looked at the output of one of the channels with a 50MHz scope and didn't notice anything suspicious when outputting a sine wave using some tone generator.

Thanks!

Wouldnt coupling caps be a good idea both on the inputs and outputs of the balanced OPA1632  stage. There might be a dc offset at the outputs of the Dac ,this might be upsetting the bias on the opamps and as its all Dc coupled it might be effecting things down the line . Im no expert in dac design ,but after scanning the schematic that did jump out at me .Try a couple of electrolytics at the inputs and see if it reduces the heat dissapation.

keves said:

Hello,

As part of an open-source DJ controller/4ch audio interface I designed an I2S DAC using the PCM1792A DAC and OPA1632D op-amp. Schematics and board images are available here: https://imgur.com/a/m6g4z

It works perfectly (minus the heat issue) and sounds great (haven't measured yet but no noticeable difference from my MOTU UltraliteMk3). The OPA1632Ds get pretty hot - hot enough to be unpleasant to touch after a few seconds.
The PSU feeding the board measures 350mA at 12v so no huge current draw (there are 2 DACs and a 5 LEDs on this board).

A random guess would be some oscillation but wouldn't that result in some hearable noise? I also looked at the output of one of the channels with a 50MHz scope and didn't notice anything suspicious when outputting a sine wave using some tone generator.

Thanks!

Click to expand...

I would exhaust the oscillation angle first. It could be inaudible. Look at the output of the op amp directly. The oscillation might not make it downstream. And the extra capacitance of the probe may cause the oscillation to stop (if the scope doesn't show anything is the chip still hot w/ probe connected at power-on?).

As for fixing oscillation you might try stiffening up your power supply. In particular I would look at that switching IC for making +-15V from +12. It might not be up to the job.

Try adding 1n or so across the feedback resistors maybe. If that has any effect then it could very well be oscillation.

Otherwise, I suppose it is possible that a little heat is normal. At full tilt I think it will be doing +-50mA or so (2V / 356R). Does the heat persist without any input signal?

Thanks, these are all great suggestions. I'll try them next time I have access to the scope.
Do you have any recommendations for an alternative way of generating +/-15v from 9-12v DC?

keves said:

Thanks, these are all great suggestions. I'll try them next time I have access to the scope.
Do you have any recommendations for an alternative way of generating +/-15v from 9-12v DC?

Click to expand...

I did that with a switch-mode power supply based on the LT1371. That switcher is actually quiet, and if you follow it with LDO regulators, it's very quiet.

keves said:

The OPA1632Ds get pretty hot - hot enough to be unpleasant to touch after a few seconds.

Click to expand...

I assume that you're using the SOIC packaged 1632 and not the  MSOP-with-pad version.

OPA1632 is a very fast amplifier and needs proper decoupling, which you have not provided. It may indeed be oscillating.

To see high-frequency oscillations, you need to make sure your 'scope probe ground is good. That means don't use the 3" ground lead and clip it to any arbitrary point. Most probes come with a little kit which includes a metal spring curlicue. This guy has a point on it, and you remove the plastic cover of the probe tip and put the curlicue on it. Then when probing, put the curlicue point on the nearest ground location. And a 50 MHz 'scope might not be fast enough. (Remember, OPA1632 bandwidth is 180 MHz!)

I'm not sure that the OPA1632 is the best choice for a differential line driver. THAT1646 or DRV134 are better choices.

keves said:

It works perfectly (minus the heat issue) and sounds great (haven't measured yet but no noticeable difference from my MOTU UltraliteMk3). The OPA1632Ds get pretty hot - hot enough to be unpleasant to touch after a few seconds.
The PSU feeding the board measures 350mA at 12v so no huge current draw (there are 2 DACs and a 5 LEDs on this board).

Click to expand...

Your DC-DC converter is rated for only 67 mA per rail.  You might be overloading it.

I am indeed using the SOIC version. What would adequate decoupling consist of?
I will experiment with adding 1nF across the feedback resistors.
Do you have a picture of this probe device you're talking about - is it https://i.stack.imgur.com/PSo3N.jpg?

Re DC-DC - I'll try taking some measurements to see how much current I'm drawing from each rail.
Thanks!

keves said:

I am indeed using the SOIC version. What would adequate decoupling consist of?

Click to expand...

Generally 100 nF caps between each supply pin and ground.

I will experiment with adding 1nF across the feedback resistors.

Click to expand...

Read the data sheet for recommendations.

Do you have a picture of this probe device you're talking about - is it https://i.stack.imgur.com/PSo3N.jpg?

Click to expand...

That's exactly it. They usually come with the probes.

Re DC-DC - I'll try taking some measurements to see how much current I'm drawing from each rail.

Click to expand...

That's always worth doing.

Andy Peters said:

Generally 100 nF caps between each supply pin and ground.

Click to expand...

I have 0.1uF+2.2uF on each supply pin.

Andy Peters said:

Read the data sheet for recommendations.

Click to expand...

Will do

Andy Peters said:

That's exactly it. They usually come with the probes.

Click to expand...

I've never seen one of these in real life, but I just ordered a scope today so hopefully it's probes will come with that. I am not sure I have a way of actually holding a probe with this apparatus and have it touch both the pin and gnd, but I'll keep this in mind for my next design (or maybe leave room for a mini SMA connector...

Andy Peters said:

That's always worth doing.

Click to expand...

I agree. Will try doing it later tonight and report back. Thankfully I left jumpers so that I can easily do this.

Hm,
So I'm measuring a steady 60mA on the +15v rail, and between 50mA and 80mA on the -15v rail. This is after I removed the indicator LEDs which were consuming 10mA.

What could possibly be the cause for the difference? When measuring the voltages at the output of what is supposedly the 15v/-15v rails, I am seeing 10.8v/-10.8v. The output post LM317/337 measures at +9.3/-9.2v. I'm guessing the 10.8 could be explained by the regulator being used beyond it's 67mA limit? I'm surprised the rails are still pretty balanced given that.

Edit: I was accidentally powering the board with 9v instead of 12v. When powered with 12v the 15v regulator gives a nice +/-14.8v and the LM317/337 gives +12.8/-12.8v. Current measurements are still around +60ma / -50-80mA (sometimes even 90mA :/)

All measurements were made using a cheap multimeter so I would take their accuracy with a grain of salt.

I can't tell a difference between the opamps heating up more or less when no input is fed to the DAC.

Andy Peters said:

I did that with a switch-mode power supply based on the LT1371. That switcher is actually quiet, and if you follow it with LDO regulators, it's very quiet.

Click to expand...

Have you used it to generate both positive and negative rails? Did you follow the "Dual Output Flyback Converter with Overvoltage Protection" example from the datasheet? If yes, how hard is it to obtain an equivalent for that transformer? I have no experience with switching regulators, which is partly why I opted for the Murata module.

keves said:

Hm,
So I'm measuring a steady 60mA on the +15v rail, and between 50mA and 80mA on the -15v rail. This is after I removed the indicator LEDs which were consuming 10mA.

What could possibly be the cause for the difference? When measuring the voltages at the output of what is supposedly the 15v/-15v rails, I am seeing 10.8v/-10.8v. The output post LM317/337 measures at +9.3/-9.2v. I'm guessing the 10.8 could be explained by the regulator being used beyond it's 67mA limit? I'm surprised the rails are still pretty balanced given that.

Edit: I was accidentally powering the board with 9v instead of 12v. When powered with 12v the 15v regulator gives a nice +/-14.8v and the LM317/337 gives +12.8/-12.8v. Current measurements are still around +60ma / -50-80mA (sometimes even 90mA :/)

Click to expand...

The module is specified for the 12 V input, so with the reduced input, and the load you applied, it can't get its output up to what is expected. Raising the input voltage means that it can work properly. (A fun test that you can do with non-canned DC-DC converters is to monitor the switch node on a 'scope as you raise and lower the input voltage. At some point when the input is too low, it'll stop working as the duty cycle approaches 100%.

Anyway, so what happens is that the load will always draw the current it requires, unless there is a current limiter preventing it from doing so.  As you try to draw more current than spec, the output rail starts to collapse as the switcher tries to keep up.

I can't tell a difference between the opamps heating up more or less when no input is fed to the DAC.

Click to expand...

My guess is that power consumption is dominated by the op-amp's quiescent draw, spec'ed at 14 mA. But you have to add active draw, which is basically output voltage divided by the load impedance, and if it's oscillating, then that can be significant (even if it's so high frequency you can't see it on your 'scope).

keves said:

Have you used it to generate both positive and negative rails? Did you follow the "Dual Output Flyback Converter with Overvoltage Protection" example from the datasheet? If yes, how hard is it to obtain an equivalent for that transformer? I have no experience with switching regulators, which is partly why I opted for the Murata module.

Click to expand...

Yep, basically that's it. For the transformer I used one of the Coiltronics (now part of Eaton) Versa-Pac[/quote] jobs, and I simulated the design using LTSpice with inductor values chosen from available transformer configurations.

I followed the switcher with LM317/LM337 regulators, which I'm not sure I need, but if I respin it I will replace those regulators with LDOs.

Thanks for the explanations. Stuff regarding the switching regulator makes perfect sense. I ran it with 9v by mistake, but back to 12v now

Tomorrow I am receiving a (supposedly) 100MHz scope, and next week I'll receive the 1nf capacitors so I'll dig in further. Hoping that I can improve things. Measuring properly with the scope will likely be hard due to the tight layout and lack of a good way to attach the probe/gnd thingie

keves said:

Have you used it to generate both positive and negative rails? Did you follow the "Dual Output Flyback Converter with Overvoltage Protection" example from the datasheet? If yes, how hard is it to obtain an equivalent for that transformer? I have no experience with switching regulators, which is partly why I opted for the Murata module.

Click to expand...

keves said:

Have you used it to generate both positive and negative rails? Did you follow the "Dual Output Flyback Converter with Overvoltage Protection" example from the datasheet? If yes, how hard is it to obtain an equivalent for that transformer? I have no experience with switching regulators, which is partly why I opted for the Murata module.

Click to expand...

Yes, that app-note design was the basis for mine. I added LM317/LM337 regulators to the switcher outputs; I am not entirely convinced they are necessary. That said, if I respin or reuse the design elsewhere, I'll replace those regulators with LDOs.

The transformer is an Eaton (Coiltronics) Versa-Pac part. The supply will do half an amp on each rail. I use a 4 A line lump for input.

I guess I'll need to overcome my fear of both switching regulators and LTSpice and start playing with them for a bit.
I am however surprised at how expensive this transformer and switcher are. Seems like this will end up being >$20 (~$7.5 for the transformer, $10 for the switcher)

Hmm, so I got my scope today but couldn't find anything obvious. What kind of a voltage swing should I expect for these potential oscillations?

I added some cheap 1nF capacitors I have across the feedback resistors of the first opamp in the signal change (IC4, added over R17/R18) as well as 2.2nF across the inputs and outputs (between Vin+ and Vin-, and Out- and Out+). The thing still gets super hot.
Any ideas on how to proceed from here?

Is it possible this has anything to do with having Vocm connected to ground?

Thanks!

Hello,

i think you have a design flaw in the outputfilter section, you forgot the  capacitors. please reed the pdf of the PCM with the recommendations.

The OPA1632 has a tendency for parasitic oscillation which does not show up on the scope. In my experience, the only way to detect it is increased THD, particularly at high amplitudes.

Of course this amplifier has relatively high quiescent current, so will become quite hot anyway.

Samuel