Quality audio opamp power supply

[sirhaggis]

Hi all,
I am working on a project that uses the TI PCM1795 32bit audio DAC, which has two differential current outputs.
The manual is here: http://www.ti.com/lit/ds/symlink/pcm1795.pdf
External opamp circuits on each output act as an I/V converter to generate differential voltage outputs (I am using the circuit on p37/39 of manual without the differential-to-single section). I intend to use these differential outputs to feed a differential line out (likely a balanced XLR line).  I'm quite comfortable with digital electronics so getting the PCM1795 to work has not been an issue, however everything after the analog outputs makes me somewhat uncomfortable. 
I have a couple of questions:
-From my understanding the outputs from the two I/V opamps (transimpedance amps) on (L-,L+) should be appropriate to directly output to an XLR line as long as I supply the appropriate line level by controlling VEE and VCC on the opamps? This circuit is the left hand section of p39. Do I need any additional support circuitry to output as a differential line output?
-That raises the question - I need to supply VEE/VCC with stable +V and -V voltage rails. I have clean 5V and AGND rails available, but how do I get a clean +ve, -ve and GND to supply the opamps? Is there a simple compact IC that can do this cleanly (I have very little room to work with)... and if so is there one that can also vary these voltages to provide adjustable line output levels?

Thanks,
Luke

[David Hess]

Hi all,
I am working on a project that uses the TI PCM1795 32bit audio DAC, which has two differential current outputs.
The manual is here: http://www.ti.com/lit/ds/symlink/pcm1795.pdf
External opamp circuits on each output act as an I/V converter to generate differential voltage outputs (I am using the circuit on p37/39 of manual without the differential-to-single section). I intend to use these differential outputs to feed a differential line out (likely a balanced XLR line).  I'm quite comfortable with digital electronics so getting the PCM1795 to work has not been an issue, however everything after the analog outputs makes me somewhat uncomfortable.

I am not sure how the PCM1795 is suppose to work.  Somehow it generates negative output currents to ground with a compliance of zero volts?  How is that even possible?  Does it have an internal charge pump?  Is there a negative supply connection that I missed?

The datasheet is incomplete and has obvious errors.  Table 2 and figure 38 make no sense unless all of the currents are actually positive.

Keep in mind that the differential output you propose will have a negative common mode output voltage of -2.87 volts.  Usually this could be adjusted by moving the operational amplifier ground point but the PCM1795 datasheets says nothing about compliance so I assume its outputs *must* operate into virtual ground.

I have a couple of questions:
-From my understanding the outputs from the two I/V opamps (transimpedance amps) on (L-,L+) should be appropriate to directly output to an XLR line as long as I supply the appropriate line level by controlling VEE and VCC on the opamps? This circuit is the left hand section of p39. Do I need any additional support circuitry to output as a differential line output?

The NE5534 shown in the application circuits is intended to be powered from +/-15 volts.  It will definitely not operate from 0 and 5 volts or even -5 and +5 volts in this application because of output voltage requirements unless the negative common mode output voltage is shifted closer to ground.

I would add output protection to protect the operational amplifiers from the wrong XLR cable being plugged in.

-That raises the question - I need to supply VEE/VCC with stable +V and -V voltage rails. I have clean 5V and AGND rails available, but how do I get a clean +ve, -ve and GND to supply the opamps? Is there a simple compact IC that can do this cleanly (I have very little room to work with)... and if so is there one that can also vary these voltages to provide adjustable line output levels?

Thanks,
Luke

Vcc and Vee do not have to be that stable or even clean because the PSRR (power supply rejection ratio) of the operational amplifiers will ignore variation and noise from its supply voltages.  These voltages have no effect on the line output levels.

If you do not have a suitable bipolar analog supply, then what you need is to convert +5 volts to between +/-5 and +/-15 volts.  Making some changes to the circuit would allow the operational amplifiers to run on just +5 and -5 volts in which case only -5 volts needs to be generated.

[2N3055]

David,

device sinks current on it's differential outputs. Center current ( 0 digital value) is to sink 3.5mA (hence -3.5mA) . Full scale negative is -1.5mA, full scale positive -5.5mA. Other diff out, of course will be reverse..
Datasheet is a bit, well, not user friendly...

Usually you put in differential current to voltage converter, and that one, as you said, would have to have bipolar power supply...

Siniša

[dmills]

You really, really want the diff to single ended section, that is what cancels the common mode noise as well as providing the lions share of the reconstruction filter (There is a 2 pole MFB filer built around that opamp).

The usual approach would be to either place a single ended to differential driver after that filter, or to go with balanced (but not differential) output by connecting the output end of R9 to XLR pin 2, a 100 ohm resistor (closely matched to R9) between XLR pin 3 and circuit ground, and connecting XLR pin 1 directly to chassis.

There should be a connection between chassis and the circuit reference ground at ONE point and it should not share any common impedances with the XLR pin 1 connection to the chassis.

I am a little concerned by your mention of controlling something by means of VCC & VEE, those are just power supply pins (+15 and -15V with respect to ground is traditional with these parts, but it depends on what your designed full scale output is).

Regards, Dan.

[David Hess]

device sinks current on it's differential outputs. Center current ( 0 digital value) is to sink 3.5mA (hence -3.5mA) . Full scale negative is -1.5mA, full scale positive -5.5mA. Other diff out, of course will be reverse..
Datasheet is a bit, well, not user friendly...

In the past, TI had some of the worst datasheets (not to mention broken products) and I see they are still living up to their reputation.

Usually you put in differential current to voltage converter, and that one, as you said, would have to have bipolar power supply...

Siniša

That is pretty neat that it sinks current from a virtual ground with no negative supply.  I know at least two way they could have done that.  The odd part then is that they only use the NE5534 in their application example when a single supply amplifier could have been used like an OPA1688 simplifying the circuit.  It is even odder considering the applications the DAC is intended for and TI's proclivity to push newer (*) parts which often have a more limited supply voltage range.

(*) And more expensive parts - TI's online selection guide does not even bother listing a lot of older parts which are still in production and less expensive when they meet the required specifications.  I might think this is just an oversight but it has been this way for years now.  At Texas Instruments, if you are not marketing, then you are overhead.

[sirhaggis]

Thanks for your input guys... please bear with me as I am very much a noob when it comes to analog electronics. Building the digital circuit and writing firmware to drive this thing is easy for me, but looking at opamp and filter circuits makes me very uncomfortable. I have really only started learning about opamps very recently.

Usually you put in differential current to voltage converter, and that one, as you said, would have to have bipolar power supply...

Is it possible/recommended to offset the non inverting input on the opamps with a voltage divider and use a single ended power supply, say 12V? I could modify the design to have bipolar power inputs but it would be preferable to use existing power pins as inputs are getting a little crowded. I'm getting the impression I should just suck it up and design in a bipolar supply...

The design calls for balanced outputs with a variable line level - at least the two standard levels +4dBu and -10dBV (to be set by an MCU). Am I correct in my understanding that I should be able to control these line levels just by controlling the feedback resistor on the opamps?... so a design with variable feedback resistance (say a digital potentiometer) should allow me to control this? Am I missing something obvious?

You really, really want the diff to single ended section, that is what cancels the common mode noise as well as providing the lions share of the reconstruction filter (There is a 2 pole MFB filer built around that opamp).

Hmm I suspected as much but have not been able to identify/analyze the circuit until this comment. The issue here is I have very limited space available on the board, (enough for four NE5534's but no more than that). The NE5532 has a dual package but has an inferior slew rate... do you know of other quality opamps ICs with multiple per package? Audio quality is a major consideration here. The PCM1795 specifies NE5534 specifically to meet the specifications, but perhaps there are better opamps that I can substitute (I really don't know what specifications to look for and how they relate to audio quality on the output in this application).

The usual approach would be to either place a single ended to differential driver after that filter, or to go with balanced (but not differential) output by connecting the output end of R9 to XLR pin 2, a 100 ohm resistor (closely matched to R9) between XLR pin 3 and circuit ground, and connecting XLR pin 1 directly to chassis.

This sounds like a good option. Are there trade offs in terms of audio quality using a full differential circuit as opposed to this configuration? Is this a solution that would be applied in a commercial product?

I guess I really need to learn more about opamps and filters at this stage so I can better get a grasp of the required circuit design. I was hoping the application circuits in the manual would be enough for my purposes, but clearly I need to make some modifications. Any pointers to where I can learn more about opamp/filter circuits?

Thanks again,
Luke

[sirhaggis]

I'm actually rethinking my use of the PCM1795 as it seems the support circuitry required is a bit excessive for my tight space constraints. I'm thinking I might go for the PCM5142 instead which has inferior SNR but supports higher bitrates, has built in programmable DSP and has single ended line level outputs.
http://www.ti.com/lit/ds/symlink/pcm5142.pdf

I guess I then just need external circuitry to allow for an adjustable line level, and differential signal output for XLR. How would one go about designing circuitry to adjust the line level and enable it to be changed by an MCU accordingly? Negative feedback opamp with digital potentiometer on the feedback resister?

Luke

[dmills]

I commend a copy of Doug Selfs "Small signal audio design" to your attention.

And no, digipots in the feedback loop are almost always a horrible idea, a jfet switching in an extra resistor is usually going to be better, but you must ensure you have enough bias to hold the thing off.

TI have some interesting fully differential opamps that would let you do the filter and line driver in one chip, not cheap, but good analogue seldom is.

There  are a large number of more modern opamps then the 553x, just nothing that combines the cheapness with the goodness in quite the same way.
While I would avoid the 5532 for the I/V stage just because of the very fast edges due to the lack of filtering at that point, but they do just fine once you have reconstructed audio.

Alternative parts:
LM4562, OPA2132, plenty of others.

Regards, Dan.

[2N3055]

I  agree with Dan, buy Douglas Self books.. Realy good source of info for audio..

And 553x is probably most underappreciated opamp ever... There are not many that are better, and an none with that price performance..

You could have single supply circuit, but hard to do right and with components costing much more...

I would suggest a dual supply, and 5534. like in datasheet and a prototype.. And then optimize if and what is needed..

[dmills]

Yep, the only reason the audiophools are down on the 553x for audio applications is that is is cheap and has been around a while (Strange since pretty much every recording has been thru dozens of the things, think mixing desks and studio gear, the designers of which are all in love with that part, quite rightly).

The I/V stage right on the DAC output is however NOT an audio application but an RF one, and in that location slew rate does matter somewhat more then in normal audio, personally I have never noticed a problem with a 5532 here, but I could maybe see the potential for one to exist. 

You can do marginally better on any given parameter, but that particular mix is very hard to beat for low impedance line level audio.

73 Dan.

[2N3055]

Dan,

You are, of course, right, but there is an interesting side effect.. That stage practically serves as integrating current to voltage converter, and acts as a filter stage itself...
So it doesn't convert I/U at full digital out stream speed, contributing to filtering effort...
And it works just fine, avoiding multistage filtering with pretty much same effect..

And I agree it can be marginally improved, at 100x the money.. And no perceptible audio difference...

As you said, you can listen through the golden speakers, all you are listening passed through hundreds of 553x in recording studios..

Cheers.

[dmills]

Yep, all agreed, but that market often requires you to use something (almost anything) else.
We need to convince them that the 5532 is 'Vintage' then it will suddenly become desirable in audio kit...

To the OP, the resistor to match the drive impedance trick costs you 6dB of output level compared to a proper line driver, and hence 6dB more interference susceptibility, but the balanced output is so much better then any unbalanced one that that is not really a major consideration most of the time.

Read also AES48 for details on the right thing to do with the screen connection.

Regards, Dan.

[David Hess]

The design calls for balanced outputs with a variable line level - at least the two standard levels +4dBu and -10dBV (to be set by an MCU). Am I correct in my understanding that I should be able to control these line levels just by controlling the feedback resistor on the opamps?... so a design with variable feedback resistance (say a digital potentiometer) should allow me to control this? Am I missing something obvious?

There may be a simpler way to control the output level.  It is not clear from the PCM1795 datasheet but it may be that the output level can be controlled via the Iref current.  Other DACs would certainly allow this.  The compromise is that distortion and noise will likely degrade somewhat at lower output levels but the difference may be insignificant in your application over a 14dbV range.

[sirhaggis]

All great ideas guys, thanks for the input. I'm going to take some time looking at Doug Selfs "Small signal audio design" as recommended and think about how to proceed.

Much appreciated
Luke