Removing C50 and C52 should indeed stop the oscillation. But why did the Behringer engineer(s) decide to add them? Noise suppression on Vcom? But with the noise being a common mode signal, how much effect would those capacitors have anyway? I don't know and that's why I wanted to keep them in and wanted to kill the oscillation by means of the resistor. But you can always give it a try. It's at least a far more easy to do mod than I did.
Jan
Jan
In my diy soundcard first pcb, I had provided an option to feed the DC buffer either from CS4272 pin#15 -like Behringer does- or from a voltage divider, selected with a jumper. All other things equal, the CS4272 Vcom was noisy. So, it might be a combined problem partially solved with my early mod in this thread -post#18- which relaxed the input of IC16 and now with this mod that works at the output. Perhaps both are needed.
Hi all.
I have done many experiments:
Boosting the LDO voltage up to 5V with battery power, replacing the 4066 to HCT - all this did not produce any significant improvements.
The best solution for measurements was separation of input beffer VCOM line from rest of the consumers. I spied the schematic in the evaluation board for CS4272
An even greater increase in productivity was given by cleaning out the supply lines. Especially the VA pin of chip - it was separated from the common 5V.
I have done many experiments:
Boosting the LDO voltage up to 5V with battery power, replacing the 4066 to HCT - all this did not produce any significant improvements.
The best solution for measurements was separation of input beffer VCOM line from rest of the consumers. I spied the schematic in the evaluation board for CS4272
An even greater increase in productivity was given by cleaning out the supply lines. Especially the VA pin of chip - it was separated from the common 5V.
The clip indicator is controlled by XMOS and has a clear digital threshold of exactly -1dB
@bobun , can you elaborate a bit more on what exactly you have changed? You said you disconnected Vcom from other circuits, but where exactly did you make the cut? And those disconnected circuits are now left floating? And how exactly did you supply the VA pin? With a separate power supply? A circuit diagram would be helpful, just as the pictures MagiBus already asked for. This would greatly help to reproduce what you've done.
About those two FFT plots: the lower one with highest THD and noise floor is made with the original circuit? For a good before/after comparison of THD + N, you should take measurements at exactly the same level, but the levels are too different (0 dBFs vs -5.27 dBFs). Knowing the THD of the CS4272 starts to rise sharply from about -4dBFs and up, you should measure at the same level. You can move the level where the THD rise starts up by several dBFS by increasing the supply voltage of the circuit to 4.9V. See one of my earlier posts. At 4.5V, it is operating below the minimum specified voltage of 4.75V, which is not a good idea and which I consider a flawed design choice from the Behringer engineers. Noise is not so much affected by this, so when you said that raising the supply voltage to 5V did not help, did you look at the noise floor only? And signals below -4 dBFs?
Cheers, Jan
About those two FFT plots: the lower one with highest THD and noise floor is made with the original circuit? For a good before/after comparison of THD + N, you should take measurements at exactly the same level, but the levels are too different (0 dBFs vs -5.27 dBFs). Knowing the THD of the CS4272 starts to rise sharply from about -4dBFs and up, you should measure at the same level. You can move the level where the THD rise starts up by several dBFS by increasing the supply voltage of the circuit to 4.9V. See one of my earlier posts. At 4.5V, it is operating below the minimum specified voltage of 4.75V, which is not a good idea and which I consider a flawed design choice from the Behringer engineers. Noise is not so much affected by this, so when you said that raising the supply voltage to 5V did not help, did you look at the noise floor only? And signals below -4 dBFs?
Cheers, Jan
Thanks! I have a few comments. First, this 10μF electrolytic at CS4272 pin #15 is way too big. Manufacturer suggests 1μF max. I don't know if this could affect reliability, just keep it in mind. Next, it remains to find out the origin of this Vcom instability. I understand you use CS4272 Vcom for the buffer but with this 10μF cap. What if you remove it? What if you use the output of IC16B with its modified input? C50/52 are only 10nF in your picture, most likely some high frequency oscillation fix. Do you think they could be removed now it became more stable?
I cannot be responsible for the manufacturer. This same manufacturer uses a 10uF cap in their own evaluation board
I tried use 1, 10, 47 and 470 uF elec caps - there was no effect with more than 10 uF. Perhaps not the best, but more right choice would be to use a 1uF PPS film instead of a 10uF electrolytic. Or both together (like as I did with the VA pin). I can check it later..I didn't quite understand your question. The MC4580 doesn't handle capacitive load anyway, therefore I use pin#15 of CS directly for ADC buffer.
As I wrote in my previous post, without the C50/52 noise and distortion increase slightly. This circuit is completely copied from the evaluation board.
A reasonable conclusion should be that CS4272 pin #15 is a high impedance output for reference DC thus the necessity for a DC buffer. It might be a compromise to use pin #15 Vcom without IC16 or other buffer. It might also explain why the manufacturer suggests 1μF. A bigger cap might introduce low frequency oscillation. But then it's theoretical. If it works keep it this way! C50/52 are a band aid for a problem earlier at the circuit, so my question was if they could be removed since the Vcom problem is solved.
At 0dBFS hard digital clipping occurs regardless psu rails. Try to stay at least at 0,1dB.
At 0dBFS hard digital clipping occurs regardless psu rails. Try to stay at least at 0,1dB.
I can't agree with you about the band aid. This is a digital HF device and any DC line must be decoupled at the end.
If the Behringer engineers had placed the IC16 between the CS and the IC8 without those capacitors, there would be no problem. Very long traces do their dirty work. As I wrote, the absence of C50/52 removes oscillations but adds distortion.
The output impedance of VCOM is 25 kΩ. Of course, it would be wise to force this output signal with an opamp buffer or with a large capacitor (as in my case)
The output impedance of VCOM is 25 kΩ. Of course, it would be wise to force this output signal with an opamp buffer or with a large capacitor (as in my case)