This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

PCM1792A: Resistor values for OPA134 family for minimal distortion.

Part Number: PCM1792A
Other Parts Discussed in Thread: PCM1792, OPA134, , OPA1612

Figure 37 in the datasheet for the PCM1792 shows a typical OP-AMP setup to first convert the current output of the DAC to a voltage (U1 and U2) and then to add the two differential paths together (U3).

 

Based on the datasheet for the OPA134 family, it would appear higher resistor values are clearly better in terms of distortion. Low pass filtering can be set at the right 3dB point by suitable selection of the capacitors. Of course, the resistor values must not be too excessive, otherwise offset issues and possibly also noise will start to become an issue.

 

The choice of resistor 820R for R1 and R2 is necessary to reach the output voltage swing (4,5V). There is little room for further increase here – otherwise, the large swing will itself result in increased distortion.

 

However, when it comes to the summing amplifier U2, why use such low values for R3 to R6? Surely these could be set to 2k2 for example, with C3 and C4 adjusted to 330pF to keep the 3dB at 220kHz.

 

This leads to the following questions:

1. In general: What resistor values are optimal in an ultra low distortion / ultra low noise audio application using the OPA134 family?

2. Would the 2k2 / 330pF combination around U2 not result in a lower overall distortion  – at least on paper?

  • Hi Mark,

    In general, we favor the low values to minimize noise, especially on the PCM1792A so we can achieve the very low noise needed for the SNR specification.  Another thing to consider is the balance of the gain of the I/V stage and the differential stage.  We can achieve better noise by maximizing the gain on the first stage and minimizing the gain of the second stage.  This is a little hard to explain, but if you would like to take a look at the attached TINA simulation you can see for yourself.  In the simulation I even take the supply rails out a bit to prevent over-driving the output of the IV stage.

    I suppose like many engineering challenges, we are forced to compromise! By increasing the value of your resistors you will improve distortion but increase noise! My advice would be to target a minimum SNR value and maximum THD+N value, then try to find a value that satisfies both (if possible).

    My colleague Wayne Liu wrote an application note on this subject if you would like to take a look at it.

    Thanks!

    Paul

    I-V Stage Gain Experiment.TSC

  • U3 in figure 37 presents a varying load to DAC I to V converters U2 and U1. The nominal impedance seen by the U1/U2 opamps is about -0.7K (that is, current flowing out of second stage and being sinked by first stage). But the impedance will grow to many times that during the cycle of a sine wave--roughly 4:1 in spice. Round numbers: If the output impedance of the driving opamp is 1 ohm, then the voltage drop from the changing input impedance sets up a voltage-dependent divider with 1 ohm on top and 1K to 4K on the bottom.

    If the bottom R is 1K, then the divider output is 0.9990 Vin. If the bottom R is 4K, then the divider output is 0.9998 Vin. So, you have your sine wave undergoing amplitude distortion--roughly 6.4mV over a 8Vpp cycle. This is roughly 13,000 lsb of amplitude error on a 24-bit converter. This is enormous.

    If great THD and THD+N is your goal, then you should avoid opamp topologies that present relatively small changing loads to driving opamps. A second order sallen-key active butterworth will provide very high input impedance (hundreds of Mohms to Gohms) lessening the distortion isuse considerably. However, that topology has limited gain options. You can't get much more than a gain of 1 from that topology before damping gets wonky. If you look at AKM DAC reference designs, they always use the active sallen-key butterworth (albeit without explaination).

    I'm not submitting the above as fact, but something I've thought a lot about and spent a lot of $ on prototyping. I'd love to get corrected.
  • PS. I should add that the driving op-amp is closed loop, and will compensate automatically for the changing loads, lessening the error considerably....
  • Hi Paul,

    Thanks for the response. Actually, I suspected that the reasoning is just as you say (i.e. lower resistor values means lower noise). I discarded the presence of noise due to the presence of the filter capacitors. Your response prompted me to revisit this point. A quick look at the noise measurements in the datasheet of the OPA134, it seems that there is little difference if the resistance remains below 2kOhms.

    Thanks a lot for the reply - great help!!

    PS: I noted the response from SeattleEE (see below). I have added further comments there.
  • Hi

    Thanks for the details. At first, it was a bit hard (for me) to follow as I was expecting the majority of such errors to get ironed out by the negative feedback (closed loop) which seems to be what you say in the added comment. In any case, I did some investigations using PSpice and the results were interesting:

    1. If the summing amplifier uses resistors around 2k, the added distortion from this stage is virtually zero.

    2. However, reducing the values to 3-400 Ohms caused a dramatic increase in distortion. I will look a bit further to find out why but I suspect that it is the effect that you originally described. Again, negative feedback helps but it can only do so much.

    Thanks for the help!
  • Sorry, after I posted I realized I forgot to preface the point about about the opamp being closed loop. The aim of the post was to help conceptualize it with big numbers.

    A killer audio opamp like the OPA1612 has an open loop gain of 3M. When the loop is closed, the output resistance is reduced by the OL versus CL gain ratio. So, if you have a gain of 3, then the output resistance is reduced by 1M.

    But, the opamp folks never really talk about open loop output impedance in specs. it's a very complex topic if you look at Zo curves. I suspect the opamp folks understand it intimately, because they must stabilize the opamp to make it sell-able. But they don't talk much about it ;)

    Anyway, it'd be helpful if someone from TI could comment here. As DAC harmonics fall to -125 dbc on new DAC parts, the impact from output circuits begins to matter a lot. In fact, DACs seem to be limiting overall loopback performance today (eg if generating a sine and capturing it with an ADC, the DAC is the weak link).

    PS. On the front page of the OPA1612 spec you can see a plot of the THD+N versus load. Yes, 600 ohms is much worse than 2K at 5-10Vrms. Those are current peaks below 20mA, well below the output capability of the opamp. And probably due to closed loop output impedance. TI? Is that correct?
  • Quick comment - there are some folks over at the precision opamp forum that we consider our stability experts! You can check out some of their content here: e2e.ti.com/.../375041
  • Hello again (Paul and SeattleEE),

    I had a look at another 'application report' to brush up on some fundamentals. Good paper! As you (Paul) seem to be a TI employee, I wanted to mention that there seems to be a typo in fig.1 where I think 4kT should be 16E-21 and not 16E-20. It is correct later in the paper. Anyway, using the OPA134 family, it really seems that any increase in noise is insignificant below around 2k Ohms. Above 2k, the noise from Rs starts to take over. In conclusion, I will go for 2k2's.

    Again, thanks to both for the feedback (no pun intended :-) )
  • Cool catch, I'll let Wayne know.

    Thanks!