PSU considerations for a design around CS4272 audio codec

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


I'm making my first steps in this domain and I thought this chip https://statics.cirrus.com/pubs/proDatasheet/CS4272_F1.pdf is an easy way to start. Still, I could use some advise regarding the psu and power on sequence.



There will be some kind of usb to I2S converter which most likely won't be an isolated type meaning the PC ground will connect to the audio codec's digital ground at least. This chip has separate pins for digital and analog ground which do not present electrical continuity between them. So, there is an option to keep digital and analog ground completely separated if that would be meaningful to do. The plan is to use usb Vbus for the microcontroller and the audio codec's digital supply -via a 3,3V LDO- and an isolated DC booster for the analog supply of the chip and the input/output analog buffers. The question is if this is the right thing to do or should I just tie all grounds on one plane and be done with that?


Another problem that arises is power consumption. I think that the whole device would be too much for one usb 2.0 port. I'm planning to use one usb for the digital section and feed the DC booster from another usb or external psu. But simply doing that means two usb cables could be connected in any sequence. Generally, I know that powering on these chips before the clocks are engaged is not a good thing. But is the digital or the analog supply that needs to wait for the clock? In the above mentioned scheme, digital supply and signal go always together. According to CS4272 datasheet: "When using the CS4272 with an external MCLK, hold RST low until the power supply, MCLK and LRCK are stable." In my case, RST comes from the usb microcontroller.



I hope it's understood that my knowledge is very basic. I would appreciate any advise. Thanks for taking the time to read this.
 
Even on a chip where the analog and digital GNDs aren't showing a direct short between them it would be strongly advisable to keep them at the same potential as the datasheet almost certainly has a spec for the maximum voltage between the two (typically less than a diode drop, protective diodes might be fitted internally).
 
So, I moved all grounds on the same plane and I abandoned the DC converter idea as well because I realized that it would need something like 10W input, not compatible with usb. It will be an external psu for everything but the usb to I2S module, followed by linear regulators. And here another question came up. What regulators to use? CS4272 supposed to do -114dB DNR and -100dB THD+N. I've seen a noise floor at -140dB just feeding it with usb Vbus and some inductors in series. And what about its buffers psu? Do I really need LT3042 or could I get away with something cheaper and easier to implement? Any opinions would be helpful.


PS. About the power on sequence, RST control wouldn't deal with all bad scenarios after all. Instead, I'm going for a relays chain that makes it completely fool proof.
 
Your -140dB figure is likely read from an FFT plot where they don't usually tell you the bandwidth, making the figure rather meaningless. I couldn't see any spec in the codec DS for PSRR on the rails so perhaps if there's an eval board from the manufacturer just copy what they do? I have no experience with the CS4272 so don't have practical suggestions other than don't just rely on inductors in filtering Vbus unless what's being powered is purely digital. On my own PC there's no shortage of LF noise out of USB which goes straight through a typical LC filter.
 
Yes, I understand. My thought was to stay within the limitations of this chip as stated in the datasheet and achieve a decent noise floor across audio spectrum only. Just trying to make my life easier and not having to chase ghosts. Thus the question about the regulators' output noise. What would be the threshold of diminishing returns in this case?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.