Modded RPi4 Board for better SQ?

Jump to Latest
G

Ghoostknight

Member
Joined 2022
#1
Hello since i readed much about audio lately i wondered if it would make "sense" to mod a RPi4 board for "ultimate SQ" (kinda like the allo usb signature)

things i came across so far that do "in theory" make sense:

1. replace the DC/DC Converters, one step further would be to provide the different voltages directly but i think thats a step too far, and someone also posting here said good dc/dc converter do make nearly the same effect (maybe add filter capacitors too?)

2. filter capacitors for usb, maybe as a sort of decoupling for all usb devices drawing power?

3. remove the BT/WIFI modul or disable it in some way, since even if disabled in config it still actively transmits

4. replace the clock(s) for better ones

5. remove the headphone amp circuitry and maybe other unneeded stuff?

any other ideas/suggestions?

Tim Curtis pointed out in the moode forum (where i posted this thread too) that it may be a good idea to use the CM4 + I/O board, this would make it possible to buy the cm4 without wlan/bt to begin with and has emmc on board if wanted (and in the end it costs pretty much the same as a normal rpi4b+)

i wanna focus on usb playback and the i/o board seems to have a good layout, the pcie x1 could be used for a usb 3.0 board to drive a ssd and things like wlan/bt dongle (so the usb 2.0 can be used standalone for the dac, its also a good thing i think that the usb 2.0 chip is very close to the ports)
tho im not sure about the hat performace with this kind of board (official i/o board) since it has rather long lanes/traces

i also watched around for other boards (there are quite a few) but couldnt find any that looks better than the official one, most have just too much unnessacary stuff or too few usb ports... i specially checked those ones -> https://pipci.jeffgeerling.com/boards_cm (thanks for that jeff btw 🙂)

any ideas/tips are very welcome 🙂
 
#4
Ok, here’s another idea, leave it the way it is but how does the Topping D10 get its power to the usb chip, do you know?

With other usb DACs, it has shown to provide a decent improvement to have a separate, linear 5v supply for that so it isn’t relying on what the computer provides.

Am just trying to give you things to try so you don’t risk destroying your SBC with mods that likely won’t be noticed…
 
#5
If you plan to hook up an USB sound device, almost all SQ aspects on the analog side of that board will be governed by the board itself. But there are a few things you might want to consider:

  • Don't power the sound card from the Pi via USB - see to that it has a good low noise separate power (5v?) feed. This might entail USB "splitter" so that the 5V from the Pi is not in play.
  • Use an (optical) USB isolator in order to not inject any noise from the Pi via USB ground.

If you do the above you can use the Pi bog standard, your bullet list will be of no use - clocks and all (if you run i2s, clocks will come into play and so will all the other garbage from a computer).

Good luck!

//
 
#6
TNT said:
Don't power the sound card from the Pi via USB - see to that it has a good low noise separate power (5v?) feed. This might entail USB "splitter" so that the 5V from the Pi is not in play.
Click to expand...
i already do that 🙂
TNT said:
Use an (optical) USB isolator in order to not inject any noise from the Pi via USB ground.
Click to expand...
thats on my list 🙂 i just use the ifi isilencer+ so far, i will probably get this one https://de.shenzhenaudio.com/produc...d-low-latency-eliminate-the-ground-loop-noise

TNT said:
If you do the above you can use the Pi bog standard, your bullet list will be of no use - clocks and all (if you run i2s, clocks will come into play and so will all the other garbage from a computer).
Click to expand...
i mainly wanna do the changes above for a more "stable" Pi, so for example the cpu of the rpi runs with less jitter
so for example, filter capacitors on the usb not for the dac but to reject noise of other usb devices, better clocks for less jitter in the cpu (and the usb data for the dac) and so on
maybe this is really "snake oil" but i wanna try 🙂 someone already reported to get better SQ with better voltage regulators on the PI (and even better with seperate 5v and 3,3v (and 1,8v i think) power supplys, tho he reported not "much" difference to better regulators)
 
#7
CPU has noting to do with "jitter" (luckily) - especially not when using USB, as no clocks used for audio is transferred there. The worst thing is to use i2s from a Pi as it seem to deploy some sort of mixing of two frequencies to get one clock - talk about jitter :-D

Forget the Pi re: audio - use it as a processing platform but leave the audio to a separate board.

//
 
#8
TNT said:
CPU has noting to do with "jitter" (luckily) - especially not when using USB, as no clocks used for audio is transferred there. The worst thing is to use i2s from a Pi as it seem to deploy some sort of mixing of two frequencies to get one clock - talk about jitter :-D
Click to expand...
from my understanding jitter can happen everywhere where a clock is involved, so basicly every chip running at a certain frequency, cpu/ram/gpu (south/northbridge on normal mainboards) and all kind of bus`es and new clocks help reducing jitter, better/ more stable power can "support" clocks to give out better clocksignals beside all kind of interferences
tho it really is debatable if cpu clocks have influence since data gets reclocked in the usb chip in the end BUT i still wanna try and see what it does

tho people still claiming i2s to be superior to usb, even on a stock pi, i wonder why
atleast i noticed a difference with the ifi silencer+ and i dont use the power lines of the pi (also usb cables have a kinda great effect), so thats not it, but it tells me there is more going on then (most) people are claiming, for me, digital is NOT digital

i basicly just wanna try to get the maximum performance "in theory", if it will make an difference "in practice" i have to see 🙂 tho im not someone throwing hundred of euros on it, i just wanna get the maximum performance "on a budget" (and have a little fun tinkering around) 🙂
 
#10
TNT said:
Forget the Pi re: audio - use it as a processing platform but leave the audio to a separate board.
Click to expand...
Yes!

The analog output on the RPi crap, no matter what you do. The I2S output has a poor clock, but the data is fine. Use a reclocker board if you use the I2S. USB just transmits data, and the audio clock is done by the USB device (let's not consider crappy old USB audio devices that use a clock over the USB here). In other words: the data coming out of the RPi is good, so there is no need to mod the board.
 
#12
Ghoostknight said:
from my understanding jitter can happen everywhere where a clock is involved, .... 🙂
Click to expand...
It happens for every clock but the only location it matters is at the point of D/A conversion. This is why a Fifo cleans out all previous clock performance and why upgrading a clock in a router is just waste of money and time. This requires a bit of communication technology knowledge to realise - guessing and "feel facts" dont help.

//
 
#13
Well, you two sound confident that a asynchronus usb dac leaves no option to improve the sound, but still i clearly hear a difference with the ifi isilencer+ and different usb cables (and i also heared a difference going from rpi3+ to rpi4 and specially with my crap main pc usb output (i kinda abounded my main pc for audio at this point, crap headphones and a soundblaster z does the job but my main setup reveals how crap it sounds compared to a rpi)),

i guess the next logical thing to tell me is i hear placebo... and i kinda dont wanna go into this discussion here, i just wanna discuss how to improve the board, if it matters or not "in theory"
 
#15
mbrennwa said:
Yes!

The analog output on the RPi crap, no matter what you do. The I2S output has a poor clock, but the data is fine. Use a reclocker board if you use the I2S. USB just transmits data, and the audio clock is done by the USB device (let's not consider crappy old USB audio devices that use a clock over the USB here). In other words: the data coming out of the RPi is good, so there is no need to mod the board.
Click to expand...
I'm using a RPi3 with a Hifiberry Digi clone that uses a Wolfson chip, coupled to a modified Cambridge Dacmagic 100 that also uses a Wolfson chip. If you look at the spec sheets, both Wolfson chips have a built-in jitter-correction, so the signal is being corrected for jitter on both sides. It sounds pretty good to me.
 
#16
mbrennwa said:
If the ifi thingy makes a difference, then there is a problem with your USB setup. Does your DAC have a proper galvanic isolation of the USB input? Does your DAC use power from the USB system?
Click to expand...
no galvanic isolation (i use a topping d10) but i use a Y-usb cable which injects power from a independent source, so atleast i can 100% say its not a power issue

tho i used one of those cheap isolators before but my main windows system i was still using back then still sounded inferiour with it, unfortunaly i couldnt use it with linux as the isolator just supported usb 1.1 and linux drivers just send 32bit audio, with windows i was able to set it to 24bit/96khz
 
#20
Well because the description suggests custom frequencys i just asked the supplier if it would be possible to exchange one 25mhz "rail" with a 24mhz "rail", since the usb 2.0 hub of the cm4 io board has a 24mhz clock
tho im unsure, is it even possible to exchange crystal clocks with those ocxo clock signals? and would it be possible to feed one mhz clock of the 3 to 2 devices? (the usb 3.0 hub of the pcie card i plan to buy and ethernet use 25 mhz)

i dont think anyone here has tried this clock card but i would appreciate any ideas/thoughts 🙂 (just from a how good is the clock standpoint, not a "that stuff doesnt matter" standpoint plz)

Also the CM4 io board needs to be powered by 12v+, so the pcie slot is compliant (also the provided 3,3v rail seem to be just for pcie express card), unfortunaly the cm4 modul dont has schematics but i think this will generate its own 3,3v and 1,8v purely from the 5v rail, so my goal is to just power the cm4 io board (and the pcie 3.0 hub) purely from a 5v rail (and the cm4 from 1,8v and 3,3v rail and i think the pcie 3.0 card needs 3,3v too ( 5v which gets normally downconverted from the 12v and 3,3v )

so i probably can skip the whole cm4 io board power section and just inject 5v for the remaining board (and the pcie 3.0 usb card), idk yet what the POE voltage is and how i can still support it with the above plan but i dont really plan to use it anyway

any help/ideas are greatly appreciated 🙂

I also plan to tinker alot with filter capacitors and shielding, for example shielding the usb hub/cm4 module, maybe even a whole ground layer on top/bottom (or maybe just grounded heatsinks, alot to try i think 🙂 )

but overall i think cm4 io board + cm4 module is the way to go here, it has alot of benefits imo (not cramped like the 4b+, standard sized hdmi, emmc, no need to remove bt/wlan module, adding shielding like the stationPi, no unnessarcy headphone circuit, all i/o on one side)

i probably also will look for a air soldering iron (and get the pinecil before i start, great cheap soldering iron 🙂 my old one needs a replacement anyway), i struggled to often with removing smd parts or just didnt even try
 
You must log in or register to reply here.
Top Bottom