tho i should also say that i tinkered a bit software wise with moode (to reduce jitter, and overlcocking the cpu is a easy way to reduce cpu jitter (tested with jitterdebugger) but that was clearly not the way togo sound-wise, underclocking did way more) and i actually believe you guys that cpu/os jitter isnt the main concern but i actually think noise/power/shielding is a main concern here (well relatively speaking) i just wanna optimize it as far i can and have a little fun while still being on the cheap side (compared to the allo usb bridge, which has the big disadvantage of still using the cm3 imo tho)
tho maybe i change my mind after using a usb isolator, i will test this first
tho maybe i change my mind after using a usb isolator, i will test this first
If you fund that noise seem to be the main contributor and that reducing it seem to improve SQ, then your whole topology is the wrong one.
The lowest possible noise is to move any switching unit as far away as possible from the DAC and use a signal conveyor that dont pick up noise and is of low bandwidth. Toslink is such an interface but it must be used in conjunction with a DAC that recreates a clock locally (requires a PLL) for low jitter.
Some report the sound when using an opto link to be "boring / dead"... If one have tuned a system to sound "good" with a noise infected DAC, once this is removed by e.g. using Toslink (opto), the experienced "deadness" stem from the previous need of tuning (amps, speaker ...) to tolerate the sound.
Its all a system game...
//
The lowest possible noise is to move any switching unit as far away as possible from the DAC and use a signal conveyor that dont pick up noise and is of low bandwidth. Toslink is such an interface but it must be used in conjunction with a DAC that recreates a clock locally (requires a PLL) for low jitter.
Some report the sound when using an opto link to be "boring / dead"... If one have tuned a system to sound "good" with a noise infected DAC, once this is removed by e.g. using Toslink (opto), the experienced "deadness" stem from the previous need of tuning (amps, speaker ...) to tolerate the sound.
Its all a system game...
//
actually i still wonder about this, since many are afraid of jitter from toslink (tho i couldnt test it so far with my topping d10) but i think toslink would be "far" superior to usb on my windows machine, maybe a usb isolator will help here too i cant "really" confirm yet but i just cant bear the windows sound compared to moode with raspberry pi, i always contributed that to a too poluted pc, tho on the same pc windows also sounds worse compared to a tweaked linux system (tho pulseaudio sucks too... >.< for convenient its great, for sq not so much) but with my soundblaster Z and "cheap" headphones i also dont hear that stuff, just with my main setup (which still is fairly cheap... topping d10+presonus eris e8)
the eris e8 also show polluted systems very easy, with a nice audible hiss, a DI-box helps (with ground-lift) tho with my rpi setup i dont need it anymore and i found the DI-box actually changing the sound somewhat to the worse
the first thing i get is the topping hs01 usb isolator, i will report how it changes sound before i start this project here (and will test it windows vs moode)
i think a good test would be also a underclocked vs overclocked rpi, since i can kinda clearly hear changes here, would be great if the sound stays the same with the isolator, tho i kinda dont hope so or i wasted my time so far x)
i also wonder about people who claim a "too analytic"/non musical sound and so on with isolators, emi filters etc personally with my studio monitors as main setup i always strive for a "neutral" sound (with slight (like 1,5-2,5db) attentuated highs for less fatique/sibiliance)
but i kinda had a similar expierence with too many ferrite beads on power lines of my active speakers, some of these kind of changes "do something" but you (or atleast me) cant put your finger on what is actually the "true sound", its really more of a preference kind of thing
atleast what i noticed, kinda specially with usb filters/usb cables (and side-injected power instead of usb power) is that overall the sound gets more pleasent to listen to specially in long sessions (atleast i think thats why people also say it sounds more analogue compared to digital, me included), you dont get that "ah i had enough for today" feeling (specially with "higher" volumes like 75-80db) or even ear/brain fatique (beside the usual better soundstage, clearer highs/bass/voices etc)
to be honest i was just to cheap so far to buy one of those 300€ usb isolators for my 600€ setup x) but topping finally released a affordable galvanic isolator, which supports high speed usb 2.0, actually now saying it, i did test one of the 1.1 isolators, tho windows still sounded inferiour (i did test it with wasapi in the qobuz client) and on linux the 1.1 isolator doesnt work with the topping d10, since it uses 32bit by default on linux and i couldnt find a way to set it to 24bit, well you can set pulseaudio to 24bit for example but alsa still communicates over 32bit with the dac
this would indeed explain some of the "unlogical" (subjective) expierences
tl:dr i will see what the usb isolator does on my rpi setup first
Last edited:
Ghoostnight reading through this it sounds like you have a bunch of experiments lined up that will be fun. I don't want to rain on that parade.
FWIW, I'll share my conclusion. USB is evil. What? As some have suggested, skipping USB and getting the sound out of a SBC like the Pi using I2S gets you where you need to be. The data is fine. It must be isolated and clocked. Often that's done in the DAC. There are also discrete options. My point to this comment is if you find yourself spending a lot to clean up USB, it might be time to rethink and eliminate it.
Here is a cheap experiment. No doubt you've found USB cables make a difference. Take a reasonable example and cut it down to 6". Just chop it and resolder the wires. Sound will be different and maybe for the better. I ultimately built my own 1" for best sound. Took lot's of flak from the engineers, but just ignore. Nothing about this should work. But it does. It is supposed to be a lossless data stream. The cable should be 75ohms. I came to question these assumptions when employed to connect a computer to a DAC. Maybe it's just evil.
When you take the leap to I2S your Pi will be largely out of the critical path. Clocking, choice of software and power supplies on the pi will still make a small difference, but tiny compared to eliminating USB. Ultimately if the isolator was perfect, nothing on the Pi would matter. Just the data.
FWIW, I'll share my conclusion. USB is evil.
What? As some have suggested, skipping USB and getting the sound out of a SBC like the Pi using I2S gets you where you need to be. The data is fine. It must be isolated and clocked. Often that's done in the DAC. There are also discrete options. My point to this comment is if you find yourself spending a lot to clean up USB, it might be time to rethink and eliminate it.Here is a cheap experiment. No doubt you've found USB cables make a difference. Take a reasonable example and cut it down to 6". Just chop it and resolder the wires. Sound will be different and maybe for the better. I ultimately built my own 1" for best sound. Took lot's of flak from the engineers, but just ignore. Nothing about this should work. But it does. It is supposed to be a lossless data stream. The cable should be 75ohms. I came to question these assumptions when employed to connect a computer to a DAC. Maybe it's just evil.
When you take the leap to I2S your Pi will be largely out of the critical path. Clocking, choice of software and power supplies on the pi will still make a small difference, but tiny compared to eliminating USB. Ultimately if the isolator was perfect, nothing on the Pi would matter. Just the data.
I am interested in improving PSU but Pi3 and Pi4 looks not same.
http://www.yokohama-alice-factory.jp/kiji/RasPi/RasPi log2.htm
http://www.yokohama-alice-factory.jp/kiji/RasPi/RasPi log2.htm
Here is a little update
i didnt much progress so far but im reading alot again lately and wanna definitly still do it! im just still in search of the most (cost)effective way todo things i actually didnt report here what the usb isolator (Topping HS01) did, it sounded clearer, overall a little improvement, i liked it but the ifi isilencer placed behind it still changed the sound nearly the same as without isolator, so its a little strange imo
the isolator was defective after 3-4 months but i got recently the replacement, i hope i was just unlucky but the casing also gets quite warm (like 35-40°C), i will see how long it lasts this time i guess
testing many little things is actually quite tiring and errors cant be excluded (objectivits have a point with that the brain can also fool you), i had the best results with testing things like the ifi isilencer if i know the current system very well, if you listened to the same setup for a long time changes are just more apparent instead of switching many things around one after another, so i guess i get the best results with doing it kinda slowly
yes slowly i get also to the same conclusion, i will definetly try one of ian canada`s digi boards to compare it to USBI'll share my conclusion. USB is evil.
or maybe another digi board but ian canada seems to be one of us "subjectivists" and i kinda trust him in that regard, a little recent story to that:
i read through the UcConditioner and Ians Power supply stuff and found a cheap china "supercapacitor filter" to kinda "test" his findings before spending more money (https://de.aliexpress.com/item/1005004825653850.html?) and this china thing (which is very much probably inferior to ians products) to power the raspberry pi in combination with a Ifi iPower X 5v already made quite the change, i would say it did overall more good than the topping Isolator and the Ifi Isilencer - combined-, its definetly easy audible to hear the change, so i think he is onto something with his boards
specially for us/me "cheapasses" which dont wanna spend 1-2k on a linear power supply alone Yup i definitly agree, i did this test already a while ago and shorter cables ARE better! tho for convienent i just keep it under 1 meter so far, maybe i will shorten this if i have setup my 19" rack which is also planned
i also recently bought this cable: https://www.audiophonics.fr/en/usb-...opper-silver-quad-shielding-075m-p-16825.html because for 20€, why not? it seemed well build and i also heared a little change to the better (well, compared to some generic 5€ cable) BUT i recently found out that it seems to be quite out of spec, if i add 1 meter "generic" cable to it the usb connection straight up doesnt work anymore, i will maybe go for Supra cables if i stay in usb territory, supra seems to specially address signal integrity (like you say 75 ohm spec) which is a good thing i think
i heared quite the difference with the usb isolator in place and adding the supercapacitor filter to the raspberry pi, so im not that sure about this, but i also didnt power the Topping HS01 from a external source, i probably should try this, coming to my mind now as i write this
i dont understand chinese but from the pictures it actually looks like the Pi2, with the Pi2 it was easy to replace the voltage regulators but the universal voltage regulator chip of the Rpi3 and 4 cant be easly replaced, see my posts here: https://www.diyaudio.com/community/...er-board-for-raspberry-pi.316150/post-7187898 , the best thing that can be done is to add filter capacitors to the existing voltage rails i think
One thing i bought around last week was this board: https://www.waveshare.com/cm4-to-pi4-adapter.htm to get the project going, it was somewhat recently released and seems to be the perfect minimalistic board for the CM4 module, if i would have gone for the CM4 IO Board i would have added the same usb hub to the pcie express board and its already included on the waveshare board in an overall minimalistic approach (and im kinda a fan of minimalistic ) while still offering the advantages of the cm4 module
also one advantage of this board is this: the CM4 IO board has a whole power section of different voltage rails for the pcie express slot while the waveshare board runs natively from 5v, the only regulator that is on the waveshare board as far as i saw is a 1,2v voltage rail for the usb 3.0 hub
i think the waveshare board is overall just better for my planned setup, shorter traces, no whole power section i have to modify, usb 3.0 hub build in (which will be mainly just used for peripheral (ssd / bt&wlan usb stick etc) but not the dac, so quality shouldnt matter that much here)
my plan to go further with usb right now looks like this: testing the usb c otg port (which is directly connected to the SOC without any hub in between) compared to the usb 3.0 hub, but the raspberry pie needs to be powered over the GPIO pins but ultimatily i will probably settle with spdif over i2s since many report that usb is just inferior over good i2s/spdif implementation and they probably have a point there
One thing that also bugs me for some time now is not being able to connect my pi to the pc directly and im still in search for the optimal way to connect the two, so i can send audio from pc to raspberry pi where camillaDSP handles correction and outputs it to my dac to my speaker/headphones
But back to the CM4 and adapter board
I will probably test first how it compares to the stock raspberry Pi i still use right now specially the Cm4 without wlan/bt module and with emmc
beside this:
1. to add filter capacitors to the cm4 voltage rails since there is no easy way to replace the regulators
2. exchanging the clocks of the cm4
3. maybe replacing the 1,2v rail for the usb hub (just to clean up things, since i believe if its a crappy voltage regulator it could also effect other stuff)
4. add shielding (maybe just grounding the cm4 heatsink)
5. testing the usb otg port compared to the usb 3.0 hub for the dac
6. figuring out which ian canada modules could fit this project, since now for example the StationPi could become a benefit again with the little waveshare board, im thinking also the battery power supply could become handy since it means running off the grid, tho overall i try to keep it simple
) while still offering the advantages of the cm4 modulealso one advantage of this board is this: the CM4 IO board has a whole power section of different voltage rails for the pcie express slot while the waveshare board runs natively from 5v, the only regulator that is on the waveshare board as far as i saw is a 1,2v voltage rail for the usb 3.0 hub
i think the waveshare board is overall just better for my planned setup, shorter traces, no whole power section i have to modify, usb 3.0 hub build in (which will be mainly just used for peripheral (ssd / bt&wlan usb stick etc) but not the dac, so quality shouldnt matter that much here)
my plan to go further with usb right now looks like this: testing the usb c otg port (which is directly connected to the SOC without any hub in between) compared to the usb 3.0 hub, but the raspberry pie needs to be powered over the GPIO pins but ultimatily i will probably settle with spdif over i2s since many report that usb is just inferior over good i2s/spdif implementation and they probably have a point there
One thing that also bugs me for some time now is not being able to connect my pi to the pc directly and im still in search for the optimal way to connect the two, so i can send audio from pc to raspberry pi where camillaDSP handles correction and outputs it to my dac to my speaker/headphones
But back to the CM4 and adapter board
I will probably test first how it compares to the stock raspberry Pi i still use right now specially the Cm4 without wlan/bt module and with emmc
beside this:
1. to add filter capacitors to the cm4 voltage rails since there is no easy way to replace the regulators
2. exchanging the clocks of the cm4
3. maybe replacing the 1,2v rail for the usb hub (just to clean up things, since i believe if its a crappy voltage regulator it could also effect other stuff)
4. add shielding (maybe just grounding the cm4 heatsink)
5. testing the usb otg port compared to the usb 3.0 hub for the dac
6. figuring out which ian canada modules could fit this project, since now for example the StationPi could become a benefit again with the little waveshare board, im thinking also the battery power supply could become handy since it means running off the grid, tho overall i try to keep it simple
"One thing that also bugs me for some time now is not being able to connect my pi to the pc directly and im still in search for the optimal way to connect the two, so i can send audio from pc to raspberry pi where camillaDSP handles correction and outputs it to my dac to my speaker/headphones"
Why not do this over a network? Directly connecting to the PC seems like a step backwards.
Why not do this over a network? Directly connecting to the PC seems like a step backwards.
yup there are a few options, i kinda have to try them to be sure which is the "best"
which holds me back on the network is that i use wlan at the moment, i dont think it will work that well but i like the idea of either connecting the raspberry pi directly via usb or network without conversion steps, i will probably test things in the coming months
Is the WLAN to avoid noise on the network lines or is it a placement issue that you're trying to solve?
I think it's awesome that you're testing so many variables. I'm excited to get to that point, but I have to make some hardware decisions first.
I'm still trying to decide if the pi4 is the best tool for the job. Each day I'm deep diving on a different SBC researching the various pros and cons. Today I've been reading a lot about Zimaboard which seems like a great platform for this. Personally I like the idea of having a dedicated PCIE USB lane that can be given it's own external power supply.
I think it's awesome that you're testing so many variables. I'm excited to get to that point, but I have to make some hardware decisions first.
I'm still trying to decide if the pi4 is the best tool for the job. Each day I'm deep diving on a different SBC researching the various pros and cons. Today I've been reading a lot about Zimaboard which seems like a great platform for this. Personally I like the idea of having a dedicated PCIE USB lane that can be given it's own external power supply.
its more of a placement issue with the router right now, but i definitly heared that the rpi onboard wlan degrades sq over a usb wlan stick
(you can disable the wlan/bt modul in pretty much all rpi distros)
but i also like the idea of being not "connected/wired" at all even if theoretically ethernet should be galvanic isolated anyway but i need to compare the two again to give a good statement here
hardware decisions is kinda the hardest part since you keep tapping in the dark unless you try many different things or "count on" other opinions
yes i also had that idea, for example volumio supports the asus tinkerboard too but i went for the cm4 since there are many distros already made for it (specially my goto moode-audio) with other sbc`s you need to setup your own os and i could imagine that driver issues is also a thing with i2s hats (if they are even pin combatible to the sbc) and you dont know how long the "niche" sbc`s are supported and so on, the RPI is really a great economy imo
so i thought its best to stay with it and just try to squeeze out as much as i can and a stock rpi is already far from worse imo, it already sounds way better then my main pc 
Usually to determine something like that, you have to work "in the industry" as most hobbiests arent going to afford equipment that can measure how good a clock is. Like, if you worked at some hardware validation group at a place like Intel or Dell; they'd have the quality of oscilloscope that would show how consistent those clock edges are rising and falling.
@Ghoostknight do you have to get a new PSU for the CM4? I think I read the 5V 5A isn't going to cut it right? (12V?)
yees, i agree, while being a subjectivists i still believe measurements are important to get a good "foundation", even if things that change sq cant be all directly measured
i probably just have to try my "luck" or dont try it at all in terms of clocks
speaking of clocks:
i also found another rpi clock board (which showed up the last few months i guess) -> http://home.teradak.com/products/118.html
which is "probably" better then the ocxo clock which derives the 2 clocks (25mhz and 54mhz) from the 10mhz clock or?
im also kinda wondering if i could just hook up two txco clocks directly without much circuitry but my understanding of electronics is kinda limited, i either have to ask here for that or hire someone of fiverr
the cm4 itself needs 5V/2A (if you underclock probably just 1A but specs say 2A), the waveshare carrier board i ordered needs 5V/3A overall, just like the normal rpi 4