Can RaspberryPi be truly high-end player in correct HWsetup?

Audio Gear DACs
21

[quote=“Jerry”]
This is subjective.

Of course- it is subjective. Everything with High Fidelity is subjective. I like, what pleases my ears and I take the way that leads me to listening pleasure.
It’s comparable with wines. A 100 times decorated bottle of wine will not automatically please my taste.

It’s something funny with this topic to discuss- I had it a thousand times. Sometimes it seems to have the potential to start a war :mrgreen:

So long as it feeds fun what we are trying out to reach listening pleasure, it is ok- even if the efforts carry some nonsense- isn’t it? :smiley:

Greetingz

22

[/quote]
I too like neutral sound the most. Then every record won’t sound the same.

What means ‘neutral sound’? I can’t imagine that anyone of us has the chance to be present at the recording session of an album. So no one knows about the acoustical circumstances during the recordings.

If I’m on a live concert it may occur that the sound I hear doesn’t fit me. The guy beside of me tells me, that he enjoys the sound.

So again- this is all subjective. I think the ability to hear is something individually different.

23

Neutrally colored sound. Or sound that isn’t “warm” or “cold” or “heavy bass” or “shimmering treble” or whatever. I like to get as close to the original recording as possible. Absolute correctness is impossible to achieve, but i like to strive towards it, with simple means.

You are correct that it’s highly subjective. That is why i said that “I like” and not “the objective truth is”. If the original recording sounds like crap then i want it to sound crap on my speakers (“Metallica - Death Magnetic” for instance).

24

Ok, I know what you mean- and I know ‘Death Magnetic’ - would call it ‘Digital Distortion’ :mrgreen:
But someone needs courage to produce it like this :mrgreen:

25

It’s funny that audiophiles are into mystery and magic to get better sound when it doesn’t matter as long as producers are releasing crap sounding records. See http://en.wikipedia.org/wiki/Loudness_war
However it seems that the trend is reversing. Perhaps the future is bright :slight_smile:

26

We hope so! Not every mastering engineer got to the limit.
The Golden Age of tape machine dynamic limiting is gone :frowning:
In a digital recording situation, it is well done with levels from -18 to to -12dBFS.
The last 3 bits not to use, or better left it for best headroom, needs particularly good converter with high dynamic range and also ultra low noise floor. That’s also done with good power supply.
In the mastering or producer chain there is some outboard equipment, which will always get the signal from DACs and back to ADCs. So the Signal is converted several times. When a converter fits those requirement, then it is neutral, or high end for me.
In the Sabres 9018, the conversion is the best I ever heard. And so we designed a very neutral DAC with this chip. But on the other hand, we can get real high levesl with -0.1dBFs without distortion. So it will play also the loudness optimized stuff.
When you hear classical music, or vocals and acoustic instruments live, then you can get a picture from instruments, room reverberation and timing.
The ears or better our brain has a good memory for those effects, and you can get this picture back home to evaluate what your audio equipment can reproduce.
And you can trust your ear.
Let it be subjective, but not wrong.

27

Well, yes you are, but this is not the point.
A non-perfect mic will not record the input perfectly, but the “changes” it introduces are (unless you have a 5 €/$ mic) constant. It’s a matter of transfer function.
If you record two sounds with the same mic, they may not be perfect, but if the recordings are the same, then the sound that produced that recording is also the same.
And this is the goal of the measurement I proposed: I don’t judge based on one recorded sound alone, I judge the difference between recordings. If they are the same, then the two sounds that were recorded are also the same. And if they are, the power supplies are equivalent.

And no, manufacturers were not wrong, but they were not dealing with DACs until recently. Old audio amplifiers benefit from a better power supply, newer DACs will benefit much less.
Have you opened the links I gave you?
Read this and the following answer:
crazy-audio.com/2013/09/rasp … mment-1473

I help you by pasting here:

They used the worst power supply and the sound was still fine. Surprised? I am not.
Bring evidence in favor of your statements now, I did it.

1 Like
28

Thanks, I understand now what your meaning.
I haven’t see the measurement of the hifi berry DAC.
Maybe that on this performance level , moreover on the way trough your recommended measurement set-up, DAC - AMP-Speaker- room - and back with 5€ microphone- microamp-ADC , you can’t see only a little impact, but do you really don’t hear any difference.
I think Hifiberry did a god job with this low cost DAC and hide nothing on the published measurement. (measured open with 96k Samplingrate)

I’m dealing with THD+N down to 0,0002% (all spectral lines under -128dB) and I had to develop my one 32 Bit ADC to get these measurements.
I think, the DAC measurement are limited by my ADC. But I could do it not even better with a 10000€ Stanfort Anlysers (Sr785).
You can see some graphs on akouo.at/dac.html.

The Chip we use costs more than some ready to play DACS for the Raspbi. Why should we not spend money in analogue stages and power supply, to get out the theoretical limits. And most design decision are not made only based on measurements, but also on listening tests.
When we take measurements, we want to see values for a single device. Sometimes it is also good to compare, but why should we take the longest signal way for measurement.

When you want to hear difference, then it could be a way. Compare two recorded files with for example the AB, - XY compare with foobar. I like that, it is funny and you can hear differences in sound also with a LOW-FI computer-audio set-up ( z.B my desk radio - 48kHz USB Soundcard and Yamaha NS5 Monitor in the Nearfield).
And also here is the rule ,the better sound reproduction brings up more differences.

I can’t not bring evidence for that all, its my opinion and my design rule. My satisfied customers are the proof for it.

Let’s make a little calculation, we said in German ‘Milchmädchenrechnung’ :slight_smile: . (not too technical and a little bit naive)
When you get PSSR of 80db, assumed high enough for few $ DAC and low cost OPV
We have a Noise level of a bad power supply with 60mV ,
The effective noise level is then 6uV (80db below)
A 16 bit audio signal (96db theoretical dynamic) with 1V reproduce the lowest signal of 16uV.
You see, the noise is near to the lowest signal. ( use a digital volume control in the software, it will be much worse)
And we want to hear high Res audio with 24bit. (144dB dynamic) Where are all that bits gone. Behind the noise floor. :open_mouth:
So lower the power supply noise and you get more details in the music.

We want to talk about a truly high end player.

29

… in order to come back to topic :wink:

30

The Question being Can a Raspberry Pi be Truly a High end player is subjective to what you see ah High end player being. I have seen many subjective comments and opinions on what hight end players are and what they can do. Even the debate on Power Supplies. The notion that a quality power supply is not necessary is far from any real world experience I have. With reasons that the op amps rejections within the DAC is why. There is no DAC in the RASPI, and since I am a computer engineer with a major computer firm, I can tell you, don’t use a cheap power supply on a computer because you can’t hear the difference, use one of quality to protect the computer from damage. It is one thing to have a disposable computer for education, but another if you plan on relying on this computer for sometime.
If you download the schematics of the Raspi you will notice that the 5VDC power supply feeds your USB almost directly with the exception of a 47uF electrolytic cap with a 100n bypass cap. If the power supply is poor you will distribute that to what ever actual USB device you connect to it. Wall warts are subject to all sorts of noise and ripple. Your dish washer or washing machine might even send some nice noise down that path. This is not good for the electronics, worse for audio. In the competitive world class enterprise servers, were cost is everything, we don’t skimp on a good power supply.

All professional audio designers will tell you, “Quality of any high end sound system starts with the power supply!” From the choice of capacitors and the size to the bypass capacitors they use. I have seen so many poor amps that you can literally hear the hiss of the power supply at relatively low volumes. I have also restored many audio systems in my life and never seen a high end sound system, digital or analog with the components in a wall wart, inside their chassis. They are usually of much higher quality. Yes, you might have a DAC that can reject some noise, but how about that IV amp or the transport and clocks that are sensitive to noise.
On the digital side, the noise has major issues with system clocks, and increase Jitter and other nasty things. I personally build power supplies that have lower noise, and I CAN hear the difference. Not with some 1Khz signal, but with the detail of the sound as stated by others.
On the Raspi 5vDC, I have used that supply to power up a CM6631A (via a USB interface) with its own 3.3 voltage regulator. I took great lengths to use a very love noise regulator here, as to improve the 12 Mhz transport clock performance on the SPDIF interface. I would not feed that effort some cheap power supply, for remember that garbage in always means garbage out.
In my opinion, I do not think the Raspi has what it takes to deliver, what I think is, the best high-fi, I personally want something that will play anything I put in it, without converting it to something else. I want bit perfect, and the Raspi is a cool device, but don’t think it can handle the higher definition, uncompressed studio music, in my collection. That is why I upgraded to the cuboxI-4pro, yes it cost more, but provided more power to provide for higher res music. No, I can’t tell the difference between a saw tooth wave and sine (difference in sampling at high frequencies of 44.1-192khz). With a good two speaker system 16 bit and 24 bit changes the dynamic range enough to improve the staging and makes the music dance around the room. Of cause those who test this on a surround sound system, really don’t understand two speaker systems, or a good mono recording.

Today, I am in a video recording studio, as the SME in a video. This room has nice sound dampeners only one the wall, ironically not places for sonic characteristics, but in a nice pattern, from someone who hasn’t got a clue, but liked the way they looked. No wall warts here either, at least they have that part right.
If it sounds good to you, then it must be right for you, but if it sounds bad, you have to make it right, not for anyone else, but yourself, so keep on listening and enjoy.

31

After your posts I tried to look in Internet to check for studies investigating the actual minimum THD the ear can discern and I found that people in controlled environments can barely discern music with 1% THD difference. Let’s suppose those people were terribly bad at that, let’s consider than 0.01% THD the minimum level the ear can recognize.
No need to go down to 0.0002% THD, it’s a waste of money and time that could be spent better in other parts of the system.
Before I read the usual answer (“I can hear the difference”): those were in controlled environments, an apartment is not.

Anyway, I admit I never thought about the Milchmädchenrechnung about noise rejection and noise left, thanks. Of course, noise is not THD, I consider that calculation as an additional info you wanted to provide.

Back to PSU: I actually gave a look at mine. I found this (the probe was set to 10X, readings should be multiplied by 10):
dl.dropboxusercontent.com/u/544 … G_2377.JPG
and the PSU is:
dl.dropboxusercontent.com/u/544 … G_2379.JPG
The line on the oscilloscope is measured on the USB contacts, it’s what the USB audio card actually gets.
You can see small bumps in the otherwise basically perfectly line: they are caused by the LNK ethernet activity LED. Now I disabled it, I found a small utility for the R-Pi, and they are gone.

Summary: the oscilloscope clearly shows that a normal USB power supply (the one sold together with the R-Pi for 5-10 dollars), after the R-Pi capacitors, has only 10 mV p-p noise left (or less, it’s difficult to judge from this photo).
And the USB audio card I use, as basically every USB audio card, doesn’t work at 5V inside but at 3.3V (I also checked mine) and the lower voltage is obtained though a linear regulator and capacitor. The noise at the pin of the audio DAC is therefore even lower than the input noise on the USB power/data connector.

As I said, before worrying about the PSU, I would worry about SO MANY other more important things.

32

My first question is why are you all measuring THD for a PSU to a RASPI? This isn’t the issue, neither is the voltages described, it is more about the effect on the internal clocks and functions of USB and I2S busses. The pops and clicks heard by people in the older release and the need to over clock the board to get around it, as well as the need to performance tune the kernel to make this stop is indicative of the lack of computer power the Raspi has.
Based on Universal Serial Bus Specification Revision 2.0 section 5 on USB transport and the USB Device Class Definition for Audio Devices 2.0 specs I see this issue a bit different.

In short that cheap wall wart power supply is probably loaded with substandard Capacitors that over a short time and heat, will change its function. Yes it is only a $50 computer, so it might not make much sense to spend a bunch on a quality Linear power supply. Ripple and all sorts of other things, make clocks nasty, and do bad things. The TCXO on the USB controller is not the best, but most computers are limited too. The biggest disappointment for me in computer based digital sound, it just that, poor clocks. That is why a high quality CD player always sound better. Don’t take my word on that, read reviews. If I have a low very Jitter 25ppm (red book says 100ppm is all that is needed) 25Mhz TCXO, I will install it, and use a battery just to what happens, and compare it to another Raspi via ABX.
The CPU power of my Raspi was very limited and doesn’t do well with DSD, or HD audio unless over clocked yet, even with over clocking it to remove the pops and clicks do to Async issue. I hate OC computers, they get hot, heat is a computer killer. Even with the tweaks has anybody compare the original to the output based on what audio information is being delivered and skipped. High end audio for me it about the detail and getting all the music from the source. If I am going to make a computer an audio appliance for high end music, I am going to start with one that has some CPU power.
I don’t use the I2S and have not listened to it on the Raspi so am not making any issue on this, but the USB port on the Raspi is shared and for good streaming should not be, HOG mode is what many software programs call it. In USB terms, the USB wire in interleaved between endpoints and not dedicated. This isn’t good, because Asynchronous isochronous audio endpoints produce or consume data at a rate that is locked either to a clock external to the USB or to a free-running internal clock and function, it functions best when the entire Host is dedicated to streaming. Poor CPU power and poor power supplies, bad cables low quality clocks (common on computer IO and sound cards) can cause a whole lot of damage to the streaming audio. This is NOT a HARD DISK, for they use more sophisticated protocols (SCSI based) . Any changes in clock sync isn’t handled by USB hardware, but host software. Current standards for host software for audio is explicitly simple, as to be universal and easy to manage, without rewriting the stack or requiring special drivers. USB does have Error handling/fault recovery mechanism built in protocol, and the protocol includes separate CRCs for control and data fields of each packet. The issue is that stream pipes deliver data in the data packet portion of bus transactions with no USB-required structure on the data content. All corrections are then done by Interpolations, via the host software. Interpolations correction is not bit perfect, and might sound good, but definitely looses detailed output and isn’t the same as the original recording. The worse the defects that the streaming pipe sees the more detail will be lost.
You can’t measure this with THD. When using high end DACs, should have a high end source. Can a Raspi do it, all depends on what is expectable to you.
I use a CuboxI-4pro currently, my goals and objectives are very different then many here, but I really do not see the Raspi, via USB being my device of choice.

33

If the Raspi does not allow good transfer over the USB port , which is a system or better hardware weakness and not directly comparable to the poor computing power, then you better don’t use it for audio streaming. I expect that the data stream is perfect error free, without plops or recalculated or interpolated corrections.
I use USB only with the Laptop, not with the Raspi
After the USB transfer, the audio data stream must be asynchronously regenerated with a good stable clock. For this use a good DAC with a fine low jitter PLL and for that also a good power supply .
If it is not possible to get high end sound out of the Raspbi USB, then use the I2S port.

THD and THD+N graphs are only to show, how good it is possible , especially with good power supply. If increased jitter is present, if the data transferred defect or the power supply is not good, then you can see it in the spectrum.
In this graph ( akouo.at/raspberry-pi.html ) you can clearly see the 50, 150,200 Hz from my power supply. I used just a poor filtered and very simple regulated power supply with LM317 on the breadboard. But all spectral lines are lower then the usual noise level from other DACs.
The Rasbpi was powered with a normal mobile phone supply and I was really surprised of that little noise.
Compared with a perfect power supply on the other measurements of my DAC (akouo.at/akouo/images/ADDA-L … ectrum.jpg) we can see, that there is no part of the line frequency.
But here you can see other effects, caused by jitter on the USB port, but also low enough.
These measurements were made with an XMOS USB to AES3 adapter.

When the noise level is generally higher, created by high jitter and poor power supply, then you do not see these parts of the disorders, this will be a part of the noise.
When the power supply is modulated by a blinking low current LED, then I think this also disturb the music reproduction, not the digital data processing.
Maybe we don’t hear this low THD, or someone likes higher ones from tubes etc., but I know, that the conversion from digital to analog works nearly perfect. That is what I want from a high end source and that’s why I use my High End DACs with good power supply.

34

If you can measure Jitter in a THD check, then you have major clock issues, even if it is dropped below the audio spectrum. High end electronics might have some form of rejection circuit, or cleaning process, but we never want to rely on them, we want good source to start with, clean clocks and low noise. There is no ECC to fix damaged signals and the things implemented to get around faults, are not high end quality sound. Multi-rate digital signal processing is more complex then what can be heard. More over, if it is heard, you are starting with a very a very poor device, for it is what you can’t hear or measure in the analog world that trashes the sound quality. ABx a Walmart special DVD with a CD, differentiate the sound from it to the same CD playing in a high quality CD player, the difference isn’t what you can here, but what you can’t hear or what is missing. Interpolation, and aliasing corrections, create sound or coverup what is not making it through transport. It is funny but when you re-clock a poor DAC or CD player, the change in sound is dramatic. Redbook standard only spec out a 100ppm clock, that leaves a lot of options, and even the best clock is only as good as the power supply being delivered to it. Anyway, a few really good books for those who want to understand more about this is “Multi-rate Digital Signal Processing” and “Discrete-Time Signal Processing.” Audio streaming protocols aren’t really very good, they are implemented because they are cheap. Even I2S have its own issues, but I see the point you make about the Raspi USB. I concur 100% with you saying it is poor. I was looking at the clock on the board, and even though is a TCXO, it isn’t really a very good one. Most computers don’t use low jitter devices on USB ports either. If I was to change this clock, I would then be committed to a quality liner power source or battery to get the best use out of it, then again, why, if the CPU can’t keep up. I2S then becomes the only option. I think I will re-provison my Raspi as something else, and use the cuboxI for music. Then again this challenge has been presented and that is alway fun. Thank you guys for an interesting subject.

35

Just for info, we are talking about extremely small values like THD measurable in scientific notation, S/N over 120 dB, few ppm jitter.
For my personal curiosity, could someone try this test and post or send the results?
tonometric.com/adaptivepitch/
Just to get an idea… many people cannot distinguish sounds 3 dB apart…

36

Many people also don’t see the deeper sense of ‘high end’ audio. There must be a relation between the inability for sensitive listening and the desinterest for high end audio :wink:
Some years ago the inventors of the cd and adequate devices presented their products as perfect. They have made ‘serious’ measurements for an evidence. Today every serious listener knows that digital audio was far away of being perfect in these times. No one talked about jitter, noise and the need of clear and precise digital signal for proper da- conversion- because this facts were definitely UNKNOWN. My explanation for everything that is noticeable by listening but not measureable is the lack of knowledge. I’m sure, that the time will come when we will know about more facts that take influence on our listening experience. Hearing is believing. No one is listening through his oscilloscope but with his ears- probably the most sensitive instruments we have.

Greetingz, Robert

37

Lag-na, I tend to agree.

A brilliant power supply will be masked by so many other “system” variables, difference in L / R audio circuit track lengths, component choice and placement on the DAC etc. etc. etc.

G

38

Ok but I was just curious: you have a better musical ear than me (probably), and surely better equipment. Independently from the R-Pi as high-end, I would be interested in knowing how people score in that test. I also asked a musician but I am waiting for an answer.

39

I assumed that all things which mask a good power supply are already done, otherwise we need not speak of the high end.
All that brings us closer to the true music reproduction is welcome.
And if you can distinguish pitch differences or can recognize a rhythm or a music sequence again has absolute nothing to do with listening to music.
This can be trained to perfection and tested even with very bad music systems.
I tried it, it is stressful and it is keeping me away from listening to music.
Adaptive pitch: 2Hz, rhythm:80%, tonedeaf 77,5%, musicl visual 77,8%

40

Why not in the Thd + N plot ? Where else ? Jitter shows always a higher noise level or some spectral lines. If the jitter artifacts in the measurement is in the limit of the 24bit resolution, it’s hard to hear it. The reed book is 40 years old and none cares about the Jitter in first hours of the CD.