First time poster (long time lurker 😁)
I’m trying to drive a 3e tpa3255 module with a differential signal from an adau1701 DSP via I2S by feeding a pcm5102 module (the common purple one) with a mono signal on the left channel and the same mono signal (but inverted in the DSP) on the right channel. It seems to work but sometimes when I remove the inversion I can still hear some signal, when I’d expect none (except some noise from slight variations in the output stages). Maybe interference on the I2S?
Anyone tried this? Should it work? Any downsides? Any tips?
Keen to use this approach to keep weight and noise down and quality up in a portable PA party speaker.
Thanks for any help!
I’m trying to drive a 3e tpa3255 module with a differential signal from an adau1701 DSP via I2S by feeding a pcm5102 module (the common purple one) with a mono signal on the left channel and the same mono signal (but inverted in the DSP) on the right channel. It seems to work but sometimes when I remove the inversion I can still hear some signal, when I’d expect none (except some noise from slight variations in the output stages). Maybe interference on the I2S?
Anyone tried this? Should it work? Any downsides? Any tips?
Keen to use this approach to keep weight and noise down and quality up in a portable PA party speaker.
Thanks for any help!
So you're creating a differential signal through re-purposing a stereo DAC to handle a mono signal in both positive and negative phases. I do that a lot, seems to sound better at least with the DACs I use. It improves the SNR too, by about 3dB.
When you feed the same signal to both phases you'll be unlikely to get cancellation so precise that you'll hear nothing at all. The relevant parameter is 'CMRR' for common-mode rejection ratio and 60dB is considered a reasonably good number for this at the receiver. But you would also need the gain matching on the L and R channels of your DAC to be a lot better than it is (2% from the DS, typically) to hear nothing at all. If you want to get closer to nothing you'll need to add a fine gain control to one of the channels and trim that by ear.
When you feed the same signal to both phases you'll be unlikely to get cancellation so precise that you'll hear nothing at all. The relevant parameter is 'CMRR' for common-mode rejection ratio and 60dB is considered a reasonably good number for this at the receiver. But you would also need the gain matching on the L and R channels of your DAC to be a lot better than it is (2% from the DS, typically) to hear nothing at all. If you want to get closer to nothing you'll need to add a fine gain control to one of the channels and trim that by ear.
Member
Joined 2018
Additional information, you can also apply this technic in the ADC stage. It reduces ADC large signal distortion. It was drastically reduced In my design FreeDSP Catamaran. Also, the SNR is as well.
ADAU1701 embedded Codec has excellent potential for Audio performance.
Ordinary ADAU1701 Single-End ADC Input through to the embedded DAC(Single Ended)
Differential ADAU1701 ADC Input through to the embedded DAC(Single Ended)
CyberPit
ADAU1701 embedded Codec has excellent potential for Audio performance.
Ordinary ADAU1701 Single-End ADC Input through to the embedded DAC(Single Ended)
Differential ADAU1701 ADC Input through to the embedded DAC(Single Ended)
CyberPit
Thanks for the quick and reassuring replies abraxalito and cyberpit (wondering now if I should have been more creative with my username!)
Not sure if I’m thinking about it right, but it seems that I shouldn’t be too worried about a positive and negative phase gain mismatch… i could trim the gain as you suggest, but I can’t see how that would improve output quality and it might reduce output gain. Is that how you’d see it?
One of the reasons I went this route is because the output stage of adau1701 has low signal output, so trimming input gain might be counterproductive.
Interesting what you say about good input stage… and not what I’d heard (nor reported in the spec sheet for my wondom adau1701 module)
Looks like you’re seeing a 5db improvement… and sorry if I’m missing something obvious but what exactly are you graphing above? If the input and output is so good on the inbuilt adc/Dac I’m wondering if I’m adding unnecessary complexity with 3x external DACs… but output level maybe still a problem. What do you think?
Not sure if I’m thinking about it right, but it seems that I shouldn’t be too worried about a positive and negative phase gain mismatch… i could trim the gain as you suggest, but I can’t see how that would improve output quality and it might reduce output gain. Is that how you’d see it?
One of the reasons I went this route is because the output stage of adau1701 has low signal output, so trimming input gain might be counterproductive.
Interesting what you say about good input stage… and not what I’d heard (nor reported in the spec sheet for my wondom adau1701 module)
Looks like you’re seeing a 5db improvement… and sorry if I’m missing something obvious but what exactly are you graphing above? If the input and output is so good on the inbuilt adc/Dac I’m wondering if I’m adding unnecessary complexity with 3x external DACs… but output level maybe still a problem. What do you think?
Miniscule as in inaudible changes in level of one channel can make a significant difference to CMRR. Going from the single ended output of the 1701 to a differential output in the 5102 you will have added approx 12dB of gain, so a fair amount to play with and I'm guessing you have way more than enough to clip the amps output.
Also try adding a 'Delay' block to one of the outputs and try adding one sample of delay.
Member
Joined 2018
Well, nobody did like this way except me 😎
The above two output spectrum was actually measured on the same board. (I needed to replace measurement ADC's op-amp to LME4999 for this trial. Because the ADAU1701's canceled distortion was below the measurement ADC's preamp itself...)
The second harmonics was reduced to over 20dB! This means the non-linear factor was canceled by this method.
If your source sound equipment has an XLR-balanced output, it is easy to try this technic.
CyberPit