Capacitor size after zero-feedback PS

[rayma]

Rayma, thanks for your response. You are right, the external DC enters the DAC at 2 x 470uF electrolytics (obviously SMT because the entire PCB is in that technology) and an internal regulator after. I still have to figure out why are the dynamics gone after removing the 6800uF output capacitor. Teddy Pardo says that no smoothing capacitors are required after the reg, however, I see people putting additional caps anyway. The only explanation I see is the already mentioned high output impedance of the circuit. But as Jan's calculation shows, an output cap would reduce it just slightly, which means that the output cap may compensate for cable length in transient response. But then again, the 1000uF at DAC input would be just enough for the same purpose in my experience.

This is very interesting. Are the extra capacitors at the external regulator end of the cable, or at the DAC end?
Theoretically, the regulator should perform better with just the 10uF, but the long cable complicates matters.
Could you try a very short cable, just to see what happens?

 

[Willi Studer]

- 10uF+0,1uF(C5+C6) are part of regulator circuit and are in place
- 2 x 470uF are at DAC end (original design) and are also in place
- 6800uF at PS end, after the regulator, was removed and that flattened the dynamics
I think I will experiment with shorter cable. Thanks for your efforts.

 

[rayma]

- 10uF+0,1uF(C5+C6) are part of regulator circuit and are in place
- 2 x 470uF are at DAC end (original design) and are also in place
- 6800uF at PS end, after the regulator, was removed and that flattened the dynamics
I think I will experiment with shorter cable. Thanks for your efforts.

The short cable should improve matters. I'd also try the long cable with the 6800uF at the DAC end, if you can rig that up somehow.

 

[AndrewT]

An analogy I saw some other Member use is to imagine the on board decoupling as a battery.

This works for me.
I can see the battery (HF decoupling cap) supplying the fast changes in current demanded by the very fast changes in state of the digital switches.
The MF decoupling (typically 4u7F to 22uF) recharges the batteries during the short period of "no change in state" ready for the next fast pulse.
Each digital chip needs it's own battery (typically 47nF to 220nF but we are seeing more now that MLCC is making bigger capacitances in 805 packages).
Each battery needs a nearby source to recharge itself. So we see lots of small electros spread around the PCB.

Now go to the edge of the PCB.
All the MF decoupling needs to be recharged. That is done with bigger decoupling that due to their size can only recharge at a slow rate of change. These could be 220uF or 1mF.

A bigger on board electro may have a much bigger influence in recharging all those tiny rechargable batteries than a remote giant electro that is separated from the PCB by many thousands of nH of cable inductance, can ever manage.

It comes down to "how quickly" the recharging can be done. The "how big" is probably not important at the frequencies seen inside fast digital electronics.