The pcm1794a datasheet I/V converter and how to improve it.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Dirk, I must admit that I stopped experimenting with IC op-amp based I/V circuits quite some time ago. I achieved by far the best sound quality via a simple passive (resistor) I/V which is subsequently A.C. coupled directly (without any intervening active circuit) to my discrete FET based linestage for voltage amplification. Although, the linestage itself is functionally a discrete op-amp. The transparency of this configuration far surpassed the current-feedback IC op-amp based I/V converter I was using until then. My FET linestage has plenty of gain for this application (the fullscale signal from the DAC is about 0.4VRMS), so an active I/V stage located on board the DAC would be one active stage too many. So, I couldn't honestly offer you an opinion on the latest current-feedback IC op-amp based I/V implementations.

Hi Ken .
Can you give us details about the dac chip you use, and the value of the iv resistor?
I am now trying new possibilities in my discrete balanced iv. And I am thinking in using a 5ohms resistor iv. Do you think is to much ?
If I use the pcm1794 in mono I can use a 2,5ohm resistor.
 
I don't think the PCM1792/4s like driving a load impedance of anything other then something very low at all. I did experiment with using little resistors, placed just before the I/V opamps, on the output of the DAC and even with small values the performance was degraded.

I think there is probably a very good reason why most current output DACs either use a virtual ground, or tie the non inverting input to some common voltage, either supplies by the DAC or an external source.

Really though, there is very little wrong with the implementation in the TI datasheet. It might be boring, but if you choose opamps that are not fazed by a capacitive load and have enough output current to drive a 600 ohm load comfortably, then you're not going to go wrong. Due to the cap in the feedback path limiting the bandwidth that the opamp is asked to work with, you don't need one with a particularly high slew rate either.

Of course the gentle low pass filtering is necessary to remove the noise from the DAC too and should lower the overall EMI, so I don't see why people are interested in trying to get rid of it and then need fast opamps as a result to cope with the high frequencies that are then let through.

Now you can use opamps that aren't happy driving capacitive loads providing that you isolate the cap with a small series resistor (22-100R). In the given application the opamps are also working at unity gain, so I don't think you even need opamps that are that quiet either.

Having worked extensively with the PCM1792 I know that the PCB is extremely important in actually allowing it to provide the performance that it is capable of. Although looking extensively at the IV stage is a decent idea, I'd be interested at first, in just getting it to performance how it should given the TI application, that was, at the end of the day, used for the measurements given in the datasheet.
 
There is already a lot of designs that use passive iv resistors. And I want to try it too. WY do you think the noise will be higher, 5 ohms seems a very low value.

Actually, when I look at this more closely, I think I'm mistaken. If you adjust the IV resistor and the feedback resistor to obtain the same output of 1.41Vrms, the noise is much worse with a small size IV resistor and a large feedback resistor compared to a large IV resistor and a small feedback resistor. The one shown is pretty bad. If the IV resistor is changed to 1k and the feedback resistor is also changed to 1k, the output voltage is the same but the noise is far lower. Of course, we can't use a 1k iv resistor, so this is just an abstract test. The current generator was set to 1mA sine wave.
 

Attachments

  • passive IV.JPG
    passive IV.JPG
    25.4 KB · Views: 566
Last edited:
I don't think the PCM1792/4s like driving a load impedance of anything other then something very low at all. I did experiment with using little resistors, placed just before the I/V opamps, on the output of the DAC and even with small values the performance was degraded.

Yeah, I saw that in some of the application notes for other current output DACs. They had a resistor going from the - pin to either ground or a voltage reference. I tried that in Tina and the noise went sky high. Very bad idea.

I think you're right about the other points you make as well.
 
I don't think the PCM1792/4s like driving a load impedance of anything other then something very low at all. I did experiment with using little resistors, placed just before the I/V opamps, on the output of the DAC and even with small values the performance was degraded.

I think there is probably a very good reason why most current output DACs either use a virtual ground, or tie the non inverting input to some common voltage, either supplies by the DAC or an external source.

Really though, there is very little wrong with the implementation in the TI datasheet. It might be boring, but if you choose opamps that are not fazed by a capacitive load and have enough output current to drive a 600 ohm load comfortably, then you're not going to go wrong. Due to the cap in the feedback path limiting the bandwidth that the opamp is asked to work with, you don't need one with a particularly high slew rate either.

Of course the gentle low pass filtering is necessary to remove the noise from the DAC too and should lower the overall EMI, so I don't see why people are interested in trying to get rid of it and then need fast opamps as a result to cope with the high frequencies that are then let through.

Now you can use opamps that aren't happy driving capacitive loads providing that you isolate the cap with a small series resistor (22-100R). In the given application the opamps are also working at unity gain, so I don't think you even need opamps that are that quiet either.

Having worked extensively with the PCM1792 I know that the PCB is extremely important in actually allowing it to provide the performance that it is capable of. Although looking extensively at the IV stage is a decent idea, I'd be interested at first, in just getting it to performance how it should given the TI application, that was, at the end of the day, used for the measurements given in the datasheet.

Thanks very much for your answer. I really appreciate.
Can you please look at the circuit in post #3 , and tell me what you think, I really like to use buffers ween there is a lot of current swing.
they don't degrade the open loop phase margin. The OP amps can be different the ones shown.
 
Last edited:
Dirk is best to use buffers with low speed OP amp's. If you use buffers with high speed OP amps you can degrade the phase margin enough to make the circuit unstable. But as 5 element says is no need to use fast amplifiers in the iv section. The NE5534 is enough.
 
This seems to work good. I'm seeing something like 1.5uVrms at the - input pins for 7.8mA p-p current output from the generator. So, they are really low impedance. Distortion and noise are both very good. Obviously, the lower we can make the feedback resistor in the IV stage, the lower the input impedance and noise will be in the circuit. Buffers!!! we need MORE buffers!!!! hahahaha.
 

Attachments

  • LME49990 DAC.JPG
    LME49990 DAC.JPG
    70.9 KB · Views: 416
Last edited:
Dirk is best to use buffers with low speed OP amp's. If you use buffers with high speed OP amps you can degrade the phase margin enough to make the circuit unstable. But as 5 element says is no need to use fast amplifiers in the iv section. The NE5534 is enough.

I think the LME49990 is a better choice because of lower noise and distortion, at least I think that's right.
 
The NE5534 has a surprising good transfer linearity in shunt mode so distortion may not be lower when you use LME49990. It has other advantages though like very good PSRR, lower voltage noise, higher speed and a better output circuit.

OK, maybe in real life it's better, but not in my simulator. The NE5534 is slightly worse, though very close, to the performance of the LME49990.

I like the idea of using buffers because we can then lower the feedback resistor even more. One thing I just realized is that the FB resistor is already so low that we are basically driving a headphone. So, we need a buffer that can drive a headphone like the LME49600 in order to have a FB resistor of 150 ohms or so. Unfortunately there are no spice models for that particular chip.
 
Last edited:
Dirk don't forget to use a buffer in the output opamp , its important.
Try to simulate with the discrete buffers I use in post #3.
But use the php ones at input and npn ones at the output opamp like the circuit.

Buffers also decrease output impedance and distortion of the opamp as they increase open loop gain. So you get a better opamp for cents.
 
Last edited:
I think the schematic in post three is essentially what Dirk has done in post 92 - follow the opamp with a buffer stage of some sort that will isolate the opamps output from the feedback resistor and the capacitive load of the cap in parallel. This allows you to lower the value of the feedback resistor and keep the performance high and perhaps lower the overall system noise.

I do not see why this wouldn't work unless the very fast current feedback opamp has a problem with the added complexity around it and it might oscillate. That's never stopped anyone in the past mind you and building it might be the only way of testing it properly.

As I mentioned before though, looking at the distortion numbers of a simulator is all well and good, but in this actual situation I'd be more interested in seeing people building and measuring. This is mostly because, as I said before, you need to get things just right, otherwise the performance of the DAC itself will be compromised and it's that that could very well be the bottleneck and not the surrounding circuitry.

You may very well find that the added complexity yields nothing in terms of the actual measurements. Kind of like swapping in an ultra low noise opamp only for the Johnson noise to keep dominating for no real net improvement. I do like the idea of providing a buffer that can allow the opamp to function purely in class A, of course the buffer needs to be biased into class A too otherwise what's the point :D
 
Dirk don't forget to use a buffer in the output opamp , its important.
Try to simulate with the discrete buffers I use in post #3.
But use the php ones at input and npn ones at the output opamp like the circuit.

Buffers also decrease output impedance and distortion of the opamp as they increase open loop gain. So you get a better opamp for cents.

Can your buffer drive 150 ohms cleanly? That's what I need to know.

Thanks for the reminder about the output opamp. Will do chief!
 
I think the schematic in post three is essentially what Dirk has done in post 92 - follow the opamp with a buffer stage of some sort that will isolate the opamps output from the feedback resistor and the capacitive load of the cap in parallel. This allows you to lower the value of the feedback resistor and keep the performance high and perhaps lower the overall system noise.

I do not see why this wouldn't work unless the very fast current feedback opamp has a problem with the added complexity around it and it might oscillate. That's never stopped anyone in the past mind you and building it might be the only way of testing it properly.

As I mentioned before though, looking at the distortion numbers of a simulator is all well and good, but in this actual situation I'd be more interested in seeing people building and measuring. This is mostly because, as I said before, you need to get things just right, otherwise the performance of the DAC itself will be compromised and it's that that could very well be the bottleneck and not the surrounding circuitry.

You may very well find that the added complexity yields nothing in terms of the actual measurements. Kind of like swapping in an ultra low noise opamp only for the Johnson noise to keep dominating for no real net improvement. I do like the idea of providing a buffer that can allow the opamp to function purely in class A, of course the buffer needs to be biased into class A too otherwise what's the point :D

I've switched to VFB opamps on the advice of Joachim.

I'm slowly gathering the gear I need to do really good measurements. It takes time and it's expensive.

Thanks for your input!
 
I think the schematic in post three is essentially what Dirk has done in post 92 - follow the opamp with a buffer stage of some sort that will isolate the opamps output from the feedback resistor and the capacitive load of the cap in parallel. This allows you to lower the value of the feedback resistor and keep the performance high and perhaps lower the overall system noise.

I do not see why this wouldn't work unless the very fast current feedback opamp has a problem with the added complexity around it and it might oscillate. That's never stopped anyone in the past mind you and building it might be the only way of testing it properly.

As I mentioned before though, looking at the distortion numbers of a simulator is all well and good, but in this actual situation I'd be more interested in seeing people building and measuring. This is mostly because, as I said before, you need to get things just right, otherwise the performance of the DAC itself will be compromised and it's that that could very well be the bottleneck and not the surrounding circuitry.

You may very well find that the added complexity yields nothing in terms of the actual measurements. Kind of like swapping in an ultra low noise opamp only for the Johnson noise to keep dominating for no real net improvement. I do like the idea of providing a buffer that can allow the opamp to function purely in class A, of course the buffer needs to be biased into class A too otherwise what's the point :D

Thanks very much for your time.
Believe me the buffers are working in class A .:D

As I say before is not very wise to mess the close loop of a fast amplifier.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.