SOLVED: Is USB audio quality resolved with the Raspberry Pi 4? i.e. No need for SPDIF Output HAT's

Audio Gear Talk Raspberry Pi
107

Hi Stephen,
Thanks for the clearer response and for the reference. I think it is important that we place this in context. I was referring to the iFi Power supply as a potential alternative for the £8 Raspberry Pi standard plugtop supply - and if swapping it out for something better (at under £50) was worth doing.
I was not looking to head into the territory of an audiophile quality linear PSU with shielded encapsulated or toroidal transformers, high quality voltage regulators and rectifiers with high grade capacitors. I have built many of those, and indeed have some circuit boards left from the last batch I made. The component cost alone is well over £180.
My thought was, could I determine an audible difference with the iFi or not. Over several weeks, I determined that yes - I believe I could. There is much more to making a good PSU than just keeping noise low: a reduced output impedance over a wide bandwidth, fast slew rates and rise-times and good hysteresis (the effective management of over/undershoot when the regulator output voltage is being loaded fast and hard with demand transients).
I would agree with you that there is a lot of “guff” out there in the so called Audiophile world - and that caters to many with more money than sense. The “Audiophools” as we know them in the professional audio world.
I reckon my comments were a simple review of the real world experience with my limited knowledge behind it. I had actually done some bench testing myself. How many here actually do that?
It’s interesting that my findings that the iFi was actually better than could be expected correlated with the findings by another amateur reviewer. I saw this in a different place - and had to translate it myself, but found another (better) translation of the original article for you - that’s here: Power Supply (8+) Group Test, LPS and SMPS - DAC - Digital to Analog Conversion - Audiophile Style
I’ve since used one of our HQ studio battery packs that works with our TASCAM digital field recorder (used on location) - a TASCAM BP-6AA as a USB supply - and that sounded a bit better than the iFi.
The only issue is I have to keep remembering to charge it. A cheap “powerbank” USB supply I have for recharging the phone when away from home really sounded awful - so am not sure what’s going on with that…
I think I’ll just stick to the iFi.
So to return to your references - it looks like someone slapped an iFi onto a spectrum analyser versus a Linear PSU and they got the following trace:

IMG_0950
IMG_09502181×1125 402 KB

One look at the left hand scale should tell you all you need to know, plus the “reference” PSU has an artifact at 12kHz.
Using a £10k spectrum analyser (I have access to similar - an APx515) doeesn’t necessarily mean that you will understand the results. I’d argue that even from this poorly referenced display screenshot - a noise floor that’s at -117.5dBV for a <£50 PSU isn’t too shabby. To the uninitiated this level calculates out at 0.000001334 V (RMS) or 1.334µV. and given the low noise level - it’s probably a long, long way below the point at which in a digital systems power rail that it could translate into an audible artifact.
The key message here from my point of view (and to return this thread more to topic) is:
Yes, I think USB audio quality IS resolved well with the Rasberry Pi 4, and however used - in combination with Riopeee they make great Roon endpoints. I reckon it would take a lot more cash to beat them.
That is just my opinion, and again I recommend everyone does their own research and testing.
My last comment is that we should all be careful of being drawn into pseudo-science. Trust your ears. The music is what’s important.

4 Likes
108

David, could you dumb this down for us? (Ok, maybe just me.)

Exactly what are we looking at? Is the iFi the “greener” graph, and the Linear PSU the “lighter” graph (the one below, with less volatility)?

Does the lower one then have less noise (lower noise floor)?

I see the big “spike” for both at 12kHz but struggling to see the “artifact” at 15kHz.

thanks!

109

Hi Chris,
I think my issue with commenting too much on this graph is that there is no real explanation of exactly what is being measured, how exactly, and under what conditions…
That’s the real problem.
This is all a pretty BIG topic, and I’m not sure I’m that good at explaining it…

That said, dBV as used here - usually refers to a signal level.
In this case I think the creator is trying to look at what we usually term “noise floor”
dBV is a voltage measurement relative to 1V – regardless of impedance.
Every step below that level in dBV gets correspondingly smaller and has a minus value.
Hugh Robjohns explains the dB differences between the various Pro standards here - I think that’s a good read.
https://www.soundonsound.com/sound-advice/q-are-all-decibels-equal
I have rarely used dBV - mostly the old dBm (for 600 ohm impedances) or the more common dBu (unterminated) - as this is what most pro audio folks refer to when just saying dB.
There are plenty of great tools online to convert between dB levels - this is one I use a lot: http://www.sengpielaudio.com/calculator-db-volt.htm

Then we need to define “noise floor…” again a big topic…
We used to measure the booth or the studio with all the gear in and switched on - but no people using a pro sound meter. Our aim was always to be at or below the 20dB (SpL) level on the meter. That background noise contributes to the inherent gear noise. (Condenser microphones actually create noise - see this as an example: https://service.shure.com/s/article/condenser-microphone-self-noise-or-noise-floor?language=en_US -but that’s for another day).

Noise floors are therefore the minimum level of a system. Any acoustic or audio system has a noise floor, right down to and including the thermal noise inherent in the wires (interesting that’s given in my audio reference book as approximately -125 dBm). The noise floor of any recording is the sum of all the noise floors encountered during the recording process. If you want to maximise headroom for transients, be able to adequately compress audio or do other interesting stuff when mastering to red-book or greater - you need the scope to be able to do so. A high noise floor limits that - and it’s why room treatments and soundproof booths to put the animal (sorry - drummer) into are so important.

Acceptable noise floor for any studio is -65dB or lower. Above -65db and the noise really gets in the way. We always used to aim for -65dB to be safe with all the kit hooked up - and if we could get to it -70dB would be great.

Back to the graph.
The noise floor of any measurement system is also limited by the ADC systems ability to resolve. I calculate that the noise floor of a 16bit ADC system can’t ever resolve below -96dB and even a 24bit ADC can only go down to -144 dB. We don’t know exactly what system or settings were used in measurement here. Real world experience tells me that the noise floor will always be higher than the theroetical maximum because of electronic noise within the measurement system.

So what exactly is this graph trying to show?
with my engineering hat on, I only see several low noise devices all well below the noise floor of a studio… and the point is?
If this was in my studio, I’d probably ignore it - except maybe for the 12kHz spike…

The way this was presented is that the iFi is the green trace and is therefore noisier than the (two?) references - again not clear. It is true - the iFi is noisier, but is that comparing a switch mode with a linear PSU again not 100% clear, but we’re now down into the uV range and this is PSU noise that’s way, way down and I’d argue very close to the real noise floor (which isn’t even referenced).
I get what you are saying - but perhaps the word “volatility” here isn’t right - because actually if you look at the noise band - then it’s pretty consistent and very common for switched mode supplies to look like this.
Again, I’d argue that this trace is so low - I wouldn’t worry about it.
As for the artifact - yes it was indeed the big spike at 12kHz - apologies - that was my typo (now corrected).

That’s all a bit of a ramble (it is the weekend) but hopefully explains a bit and gives you places to go and research/learn stuff too… I’ve never stopped learning. Roon and Ropieee are two recent examples… and I’m a newbie here.

110

So true. Many years ago I bought an USB cable with silver plated conductors. Reviewers had reported hearing better “air”, which made me hear it too. One day I decided to do a blind test with help from my daughter. Guess what, after several tries, I could not pick out the “better” cable more than approximately 50% of the time, so basically I wasn’t doing any better than guessing.

2 Likes
111

I agree with your comments. Maybe a case of paralysis by analysis.
The measurements are so low in the spectrum, maybe the noise anomalies are simply not important.
Not sure…

Full disclosure, I disagree with many comments and observations made by the AS site.
Measurements are great, but they only go so far. I think measurements help provide an overview of performance, and then the ear must judge from there.

For instance a frequency response sweep of a speaker simply cannot tell you how that speaker sounds. You must listen for yourself.

At the end of the day, we are all on the quest for ‘our’ preferred sound, and this process/outcome is 100% subjective.

I also agree with your comments regarding RPI4’s with Ropieee. Its a great endpoint for the money. Running into my RME ADI 2 dac, music sounds excellent. I feel little motivation to add i2s to AES/SPDIF hats etc.

One day I might add one of these super capacitor filters, from Ian Canada.

I hear good things.

My understanding is the super capacitors act like a battery, both filtering and providing consistent voltage to the PI. This methodology saves on expensive LPS’s or battery solutions.

Thanks for your comments.

1 Like
112

… comes back to the old bits are bits argument.
I think in some instances bits are indeed bits, but IME implementation really matters.

I recently completed an A/B comparison against my RPI4 and my Laptop running Roon bridge.
Not a blind test. Every time I thought I heard a difference I retested repeatedly.
I found the differences I thought were there, weren’t.

I still feel the laptop sounds slightly better (more detail, front vs second row soundstage - maybe) but this could just be my bias. If indeed there are any differences between my Laptop and RPI4 they are so small, I’m not sure it actually matters. As the laptop was running on battery, I suspect a better power supply for the RPI4 would go a long way.

System is: Source > RME ADI 2 (generic USB) > Belden interconnects > Firstwatt F7 > Speakers.

I suspect in a blind test, I would have a 0% chance picking a difference between PC and RPI4.
Maybe with a different DAC, my findings/opinions would be different.

In saying that, my RPI4 with Ropieee sounds quite different to my Pioneer N70a streamer running via Spdif without Roon.

113

Audio is sent digitally over USB isn’t it? As in zeros and ones… is electrical noise something that can affect zeros and ones?

114

If the error correction is working, No. The ones and zeros either get there or they don’t. There is no such thing as a “degraded one or zero” :smiley:

I think the issue is that electric noise might affect the analog part of the device that is receiving the signal if it is not well designed with adequate noise filters.

1 Like
115

Things aren’t always as simple, as they might appear at first glance, since there are several switch mode voltage regulators inside your laptop, generating all the different voltages required.
So in the real world, running a laptop on battery doesn’t clean up it’s noise spectrum a lot.
Just saying and ducking away…
:kissing_smiling_eyes:

1 Like
116

I think this is the REALLY interesting stuff. The science of psychoacoustics has demonstrated that we all perceive sound slightly differently - especially at the extremes of the spectrum (but not limited to those extremes). Timing differences are acutely heard by the human ear.
Knowing too that confirmation bias and expectation bias are also real factors, it’s actually a wonder there is any agreement at all.
I and others here can only give an opinion with mine based on 30+ years experience and some formal engineering training - but it’s still only an opinion.
I have heard degradation of quality and measured low level baseline noise (particularly from 8kHz up) from a laptop USB source. Was it my DAC that was sensitive/being affected?
A change of DAC suggests it wasn’t.
It is actually that experience which led me to look at a second hand Innuos Zenith after I decided to look for something else as a better source. The change (including blind testing) was very noticeable. Later, testing another laptop the difference wasn’t quite as marked - but was still very clearly there. Conclusion: That differences occurred between different sources - that shouldn’t be there…
Why? Could be USB implementation or poor wiring, electrical noise - I really don’t know, but electrical noise was at least measurable - but often so low as to be of questionable effect.
Then I needed to listen in different places, and I discovered Roon. That then led to the need for more Roon endpoints - and I tried the RPi3 with a HiFi Berry Digi+ Pro HAT. It appears (and measured) to have lower noise than the laptops with both DAC’s and the difference between it and Zenith were (again in a blind test) noticeable - because I could pick out which I was listening to 5/5. It was similar for the RPi4 alone and USB which is what got me here to this thread.
So am I any further ahead?
Well, I hear a difference and measure a negligibly low noise floor with the Zenith. I can blind test define if it’s that which is the source whichever DAC I use.
An exceptional source for the money is an RPi3 with a HiFi Berry Digi+ Pro HAT or an RPi4 for USB, noting that I cannot reliably pick between the two in a blind test. I get an impression that the RPi3 with a HiFi Berry Digi+ Pro HAT is better - but again that has absolutely no scientific evidence.
I can still reliably distinguish either RPi in blind tests from the Zenith. It’s not huge, but the clarity and impression of the recording environment and “the room” that just seems to be there with the Zenith - that’s my giveaway.
So what is the difference (and why) between your experience and mine?
I don’t doubt you for a minute. But what is the difference here?
I think it is the resolution of the rest of the system. My system and setup is my expensive indulgence at the culmination of a good career…
I have a separate free standing single storey brick barn/building that was converted specifically as only my studio and listening room.
It was converted to studio construction standards and uses the same materials and techniques.
It is a room constructed on a steel framework on isolators from the inner skin of the building (mainly for damp - but acoustically it’s great too). The dimensions are deliberately out of square to defeat standing waves, and there is a vaulted ceiling with slightly different pitches. All surfaces are constructed using studio quality insulated sound deadening acoustic board.
The floor is tiled concrete (with underfloor heating!) but has rugs on it, and the windows/door are triple glazed. There is a heavy curtain in front of the door. There are soft furnishings and book cases that help make the room pretty dead acoustically and the background SPL is around 8dB. Very quiet.
The speakers are Kii Three active speakers driven directly using the Kii Control interface using USB or SPDIF over coax.
The Kii’s are renowned for being unforgiving and brutally honest in their reproduction. They’re not everyone’s taste, but I really like the clarity they bring and I never tire of hearing them.
I’d suggest if anyone were to test equipment here in this environment - if there are differences then they might have more chance of hearing them or measuring them.
In any case, the important thing is that RPi 4 and USB works well, and I write this sat in the very studio listening to Beth Hart on a new RPi4 endpoint destined to be used in my kitchen…
Enjoy the music.

1 Like
117

Any recommendations for a good price/performance linear power supply fir the Rpi4 running Ropieee?

118

@David_Ford I myself have provided my experiences of building stuff (to this forum) where I spend £$ in the hope less $£=More or equal on the principles of ‘measure & listen’ .

e.g. I have experience of throwing together a botch of a streamer as a temp measure that surpassed in SQ a $3,000+ streamer in repair (then sold the repaired streamer). This spurred me to build various full on streamer that stopped me upgrading my hi-end DAC as SQ saw an obvious increase = I did not mind offering my ‘research, measure, listen, tinker’ output experience on this matter and yes was always a little surprised by some of the responses mostly negative and some out of context … As you I value the experiences of what people do, have tried, have learnt.

Yet, it is forums like this you can find some gems, some insights, some passed-on knowledge. So, forum keep these coming form the doers, knowers, what-if-ers …

2 Likes
119

That’s jealousy inducing. I’ve measured my room at 2 AM to avoid background noise when setting up speaker DSP. In a standard living room I obviously can’t get anywhere near this figure. Genuinely interesting stuff.
Enjoy your oasis of calm, well until you turn the Kiis on…

2 Likes
120

Maybe:

Seems to get good reviews.

1 Like
121

I hear from others that for €90 - this isn’t a bad low cost option.
https://www.audiophonics.fr/en/hifi-power-supply/audiophonics-lpsu25-linear-regulated-low-noise-power-supply-usb-220v-to-5v-2a-25va-p-11364.html
I have not tried it myself though. I believe Hans Beekhuyzen did a video review a while back and it came out OK.

I’d also concur with Stephen’s earlier post in which he suggested:
“Check out SBooster PS’s. They are clean and provide excellent results. Not the same money obviously. https://www.sbooster.com/“
I have only had one of these on the bench briefly - it was a friends so didn’t want to hack it about - but it seemed to give a very clean 5v.

My second edit - in that if you’re prepared to do some wiring yourself (and you’re OK dealing with properly insulating mains connections/grounding etc…) then there are many sources of nice linear PSU’s - the advantages being you can create one to your own spec… Audiophonics have some good ones - but there are other industrial supplies (look for ones designed for powering sensors) that are cheaper and probably do just as well.
Find a friend with a decent oscilloscope & frequency analyser.

122

There is a reason I had to build it in a separate building… strangely the significant other isn’t into Gary Numan - Savage played at the “proper” levels… :rofl:
That tight, fast reflex and directional bass delivered by the Kii’s is what won me over.
So close to my studio experiences with some very good kit, including PMC monitors that cost more than my entire building and contents!!!
This is the building:

image
image2016×1512 1.61 MB

3 Likes
123

:joy: :+1:

I was told off yesterday when my own partner came down to complain they could feel the Apollo 11 launch 2 floors up…

The building looks just right, and way beyond my pay scale. Very nice.

1 Like
124

Thanks David.

I appreciate the considered response. Sounds like you have a very nice system and a well treated, quiet room. :wink: 8db back ground noise is incredible!

I agree with your comments.

The engineer in you might find this link interesting. Archimago's Musings: MEASUREMENTS: Intel i7 PC and Raspberry Pi 3 B+ Audio Streamer - XLR / RCA, Noise and Jitter. Do digital transports / streamers really make a difference? Do USB cables?
It covers this exact topic. Archimago compares an i7 laptop and RPI.
The site has a number of really interesting posts on MQA, test measurements of product, etc.

But fair warning, he comes from the bits are bits camp. Nevertheless its interesting data, take from it what you will.

My experience with the RME ADI 2 Dac is that USB PC based sources sound more similar than not.
However when comparing sources via SPDIF, I have found the sources I have tested to have larger and more quantifiable differences.

Obviously there are many factors are at play (As you well described); including my hearing, system hardware (I run vintage speakers/higher distortion), amplification (Firstwatt F7/more distortion), the room (I have some treatment), etc but the USB jitter rejection and ultra low noise floor of the DAC do provide some scientific basis for this observation.

Measurements of RME ADI-2 DAC and Headphone Amp | Audio Science Review (ASR) Forum .

Extracted from the Archimago link:
Newer DACs like the Oppo UDP-205 and RME ADI-2 Pro FS have excellent S/PDIF jitter rejection so even this difference is minimal. The better your DAC, the more “bits are bits”.

At the end of the day, you just need to be happy with what you hear. I’m happy with the RPI4, for now.
I just ordered the Ian Canada Super Cap Conditioner to see what it does.

125

I have used Teradak PS’s. They are decent IMO:
https://www.ebay.com.au/itm/TeraDak-SB3-DC-hifi-linear-power-supply-DC-5V-3A-/252934019341

1 Like
126

Nice listening space…

1 Like