PGA2311 high THD, why?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi all,

Would like some suggestions from you guys, if you have time!

I've made a small DAC PCB with an FPGA, PCM1754 DAC and PGA2311 for analogue volume control. Everything works well up to the PGA2311, at the input of the colume control (output of the DAC) I measure 0.005%THD (not bad for PC measurement). But after the PGA I measure 0.05%. The PGA2311U is rated to 0.0004%!

These are the conditions:
- 1.4Vrms Vin
- Gain is set to 0dB
- 1.4Vrms out
- 10k resistive load on each output.
- Supplied with +5VA, -5VA and 5VD from seperate regulators. All supplies work very very with noise below the scopes noise floor.
- AGND and DGND are seperated up until the filtering capacitors at the PSU (star GND) about 4cm away from the IC.
- Try tying AGND and DGND together underneath the IC but no difference.
- PCB is 4 layers with inner 1 and 2 being AGND layers, bottom and top being signal.
- SCLK and SDATA are only running when new data is sent to the PGA.

Here's the schematic.

Some notes/mods.

- All AC coupling capacitors are now 100nF film cap win parallel with 2.2uF WIMA film.
- R8, R11, R18 and R19 are 220k.
- All the caps in the low pass filter are now 470pF film and the resistors changed achieve the correct cutoff.
 

Attachments

  • 016-5_spdif_simple_DAC_reva_sch pcb.pdf
    21.9 KB · Views: 236
Good spot, checked the data sheet and the distortion gets worse with higher source impedance. I removed the 39R resistors and now I'm down to 0.02%.

I'm still thinking of other possible issues.

I don't buy it. The data sheet says below 600 ohms is recommended. Source impedances up to 2K ohms will cause minimal degradation. 39 ohms is not high impedance in anyone's book. There is something else going on. I suspect you have some oscillation going on that you can't hear but clouds your readings. This would be a good time for a scope. If that isn't a possibility, an attenuator back into the analog input of your PC and some spectrum analyze software might do. The attenuator is to ensure you don't damage or overload the PC input. You might also want to capture with the highest sample rate to get some of the out of band noise.

 
I don't buy it. The data sheet says below 600 ohms is recommended. Source impedances up to 2K ohms will cause minimal degradation. 39 ohms is not high impedance in anyone's book. There is something else going on. I suspect you have some oscillation going on that you can't hear but clouds your readings. This would be a good time for a scope. If that isn't a possibility, an attenuator back into the analog input of your PC and some spectrum analyze software might do. The attenuator is to ensure you don't damage or overload the PC input. You might also want to capture with the highest sample rate to get some of the out of band noise.


Hmmm, I know what you mean.

I've got a 100MHz scope, I went through the signal path before I did anything else - it's fine. I've been doing my THD readings at 192kHz but the PC only has a 20kHz bandwidth.
 
A 1.4V RMS output will have a peak voltage of 2V, and your power supplies are ±5V. That sounds like a wide margin, but the LME49720 is not a rail-to-rail device, so you're getting close to the output swing. In the datasheet, they say that with ±2.5V supplies, you get 1% THD+N at around 1V RMS output. By this logic, you should get 1% THD+N at 2V RMS with your ±5V supplies, but maybe 1.4V RMS is getting too close?

Try the measurements again with significantly lower signal levels (1V peak or so) and see if that changes things?

The LME49720 is a great amplifier, but for this circuit, it might make more sense to use a different amplifier that can swing larger output voltages. The OPA1612 has a rail-to-rail output, but not a rail to rail input - the common mode input range is still 2V smaller than the supply voltages, or ±3V for your circuit, which will be problematic for your noninverting stage. You could use two inverters, chained into each other instead, and then get the full voltage swing from the R-R output. since the opamp inputs will be near 0V

The other idea is to use something greater than ±5V for the output amplifier, and use extra regulators to make this into ±5V for the PGA2311

But, first, see if this matters at all. Other than that, I can't see why things aren't working...!
 
Or use the PGA2310(20) with its wider supply range and use the same supply for the opamps as well.

You show a balanced line driver, U7,8, I assume you measure your THD at the output of it, what test equipment are you using to do this balanced line measurement?
 
Last edited:
First off thank you for your replies. I've been busy lately so haven't been on the forum.

A 1.4V RMS output will have a peak voltage of 2V, and your power supplies are ±5V. That sounds like a wide margin, but the LME49720 is not a rail-to-rail device, so you're getting close to the output swing. In the datasheet, they say that with ±2.5V supplies, you get 1% THD+N at around 1V RMS output. By this logic, you should get 1% THD+N at 2V RMS with your ±5V supplies, but maybe 1.4V RMS is getting too close?

Try the measurements again with significantly lower signal levels (1V peak or so) and see if that changes things?

The LME49720 is a great amplifier, but for this circuit, it might make more sense to use a different amplifier that can swing larger output voltages. The OPA1612 has a rail-to-rail output, but not a rail to rail input - the common mode input range is still 2V smaller than the supply voltages, or ±3V for your circuit, which will be problematic for your noninverting stage. You could use two inverters, chained into each other instead, and then get the full voltage swing from the R-R output. since the opamp inputs will be near 0V

The other idea is to use something greater than ±5V for the output amplifier, and use extra regulators to make this into ±5V for the PGA2311

But, first, see if this matters at all. Other than that, I can't see why things aren't working...!

I have tried having the input at lower levels all the way down to -60dB with no improvement.

I'm measuring 0.005% at the input pins of the PGA2311 and 0.05% directly on its output pins.

Or use the PGA2310(20) with its wider supply range and use the same supply for the opamps as well.

You show a balanced line driver, U7,8, I assume you measure your THD at the output of it, what test equipment are you using to do this balanced line measurement?

I'm using my PC as an analyser, I'm measuring the output differentially with YMEC RTA.
 
have you checked that there is not a silly error like the 10K resistors on the output op amps being far too low a value?

I would start by removing the output op amps and verifying the PGA by itself.

I have used these digital volume controls, they are pretty good and perform pretty much to spec. There is not enough around them other than what they are driving to cause what you are seeing...

As an aside, you might want to check the input impedance / capacitance of your test set. Try making a measurement with and without a 1K resistor in series with your distortion test set. If there is a significant change you have a clue to look into.

By the way, I assume you are leaving the ground (or negative input if your test set is balanced) at the same spot and only moving the measurement point from the input to output of the PGA...
 
First off thank you for your replies. I've been busy lately so haven't been on the forum.
I have tried having the input at lower levels all the way down to -60dB with no improvement.
I'm measuring 0.005% at the input pins of the PGA2311 and 0.05% directly on its output pins.
That's a bummer! Something else you might try is giving your PCB a good clean.
A few months ago I was putting together a class A headphone amp and making small tweaks to it while running THD measurements. After parallelling a resistor (and using quite a bit of solder in the process) suddenly the THD shot up from -115dB to something like -60dB. Some headscratching ensued, but then I remembered flux residue can be somewhat conductive. Turned out the flux remaining on the PCB was shorting something out. A good clean + scrub brought performance back to he original levels.
 
AX tech editor
Joined 2002
Paid Member
One or two possibilities: You are supposed to connect the relevant analog ground for each channel directly at each source ground. That also means that the separation of grounds needs to be respected at the source as well. PCB issue.

Other possible issue is the realtive large DC path to ground at the PGA input if I read the schematic correctly.

Jan
 
I think that you need to troubleshoot/check-out your test setup.
Test a known good low THD balanced source.
Why not measure the signal, single ended right at the PGA2311 o/p. b4 the balanced line driver.
I'm using my PC as an analyser, I'm measuring the output differentially with YMEC RTA.
That is software
What hardware is that? to take a balanced signal and do a A/D differentially?
 
I think that you need to troubleshoot/check-out your test setup.
Test a known good low THD balanced source.
Why not measure the signal, single ended right at the PGA2311 o/p. b4 the balanced line driver.

That is software
What hardware is that? to take a balanced signal and do a A/D differentially?


Originally Posted by Boscoe View Post
First off thank you for your replies. I've been busy lately so haven't been on the forum.
I have tried having the input at lower levels all the way down to -60dB with no improvement.
I'm measuring 0.005% at the input pins of the PGA2311 and 0.05% directly on its output pins.
 
Thank you everyone for your help - I have confirmed the issue.

First off I'll say that I'm using my PC as an analyser, I'm measuring the output differentially with YMEC RTA. This software is good and bad, I can get good measurements that make sense however a lot of the time it's hit and miss. I think this must have confused me somewhat.

Monte McGuire, you were on the right tracks however it was the op-amp in the filter that caused the issue, for some reason. I found the distortion was high at it's output when the source was driven above -18dBFS. I changed it to an LME49721 on +/-5V rails and now I get close to the best performance of the DAC of around 0.003% THD at the output of the analogue signal chain.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.