Why different roon bridge sounds different

May I ask a humble question here?

I have tried to stream through iMac and dietpi to my USB DAC. Here are the signal paths respectively

I found that dietpi sounds better (More powerful bass and 3D stage I would say). OK, sorry, let’s not saying “sound better”, but different “sound signature” to avoid another round of heat debate. In fact I always found the iMac to USB DAC sound signature is more bright. And, I also found the sound quality of iTunes to USB DAC was very bad, regardless of it should also go through the same coreaudio.


I am a computer studies graduate, and have studied some courses in data communication as well. Those courses didn’t cover the audio domain specifically, but I still understand some basic knowledge of digital transmission and how it’s design to tolerate noise or interferences. But in the digital audio domain my knowledge is very limited. Therefore, I tried to read through many posts in this forum and other audiophile websites. I also went through some articles by archimago (I like his scientific approach). There were many perspectives, terms and knowledges, but it didn’t help me to figure out why the differences I’ve encountered in this scenario.

There are many experts here. Anyone can help me out?

It’s a fascinating question Isiah. You have now experienced the gap between theory and reality. I’m afraid all I can do is make some comments, rather than tell you a precise answer.

For an older generation, this gap was often first experienced with electronics. The blackboard set out ideal circuits made from ideal components. But the real world contained only approximations of such devices within tolerances and subject to influences that were ignored for the sake of simplicity on the blackboard. The effect of vibrations from transformers on capacitors for example.

Digital circuits are described ideally. But they run on real world analog devices. Quite often those devices are doing things that are ignored for the sake of simplicity in the circuit descriptions. One of the amazing things about our hearing is that we can perceive such things.

So there are two issues when trying to differentiate real world performance from ideal specifications:

  • Firstly, not many reviewers actually run meaningful measurements. There is a lot of subjective opinion, which people find of varying degrees of utility. Archimago is one of the few reviewers to look carefully at signals;

  • Secondly, it’s not clear that we are measuring the right thing. Our hearing is extraordinarily sensitive to small disparities in timing that measurement systems may not capture properly. Sometimes Archimago’s measurements seem to show very little difference between different sounding device settings. See the recent HQ Player review for some examples;

So what are the possible differences that might account for a change in perceived SQ between a USB direct connection and an Ethernet to Dietpi RAAT network connection ? For starters, have a look at these KB pages:


The usual explanation is “noise” from processing or power supplies. The Ethernet leg provides some isolation from the noisy laptop. The Pi running Dietpi generates less noise into the USB connection.

In a worst case scenario such noise can be audible, but even when inaudible it appears to interact with real world devices in ways we may not be able to completely describe, but which have the effects you’ve noticed.


That Amanero is actually a converter module, right? Converts USB to I2S. How is it powered? If it is powered from the computer over the USB, I’m not surprised two different computers might sound differently. The other thing to note is that you are using two different sets of device drivers, Apple’s and Linux’s. Could be bugs in either which affect the actual bits delivered. Those would be my first two guesses. Of course, the thing to do would be to get an A/D converter, and measure the actual outputs from known test tones, see what the differences actually are.

Both Mac and Linux seem to use the standard drivers, no downloads.

1 Like

Quite a few DACs use the Amanero module as a USB receiver, as well as stand alone conversion devices.


I found the Hans Beekhuyzen chanel on YouTube very informative in the regards of your question(s).



Bottom line…everyone hears differently…so when it comes down to hearing differences from one setup to another decide which you prefer to listen to and stick with that until you have a desire/option to try something different and repeat the process.

Understanding the ins and outs of the technology can be fun, but sometime more distracting than finding what configuration you like and sticking with it. And as Hans always says, and I totally agree…enjoy the music.

@andybob @Bill_Janssen

Great sharing and thanks.
This is correct. Amanero is simply the USB module of my DAC.

My iMac is late 2015 model with 512G SSD.

The raspberry pi dietpi is powered by my iPad charger (5V 1A). I tried power bank, no sound difference (I read many people saying battery power provides better sound, but not to my case in this scenario)

iMac connects to ethernet. Dietpi connects by wifi.

In fact I am looking at the coreaudio. Maybe it does some sound processing before sending it to USB DAC (even in Exclusive mode). May dietpi also does it or in a different way?

Any knowledge about this coreaudio?

Grumble grumble.
I don’t like the “everyone hears differently” talk.
This is supposed to be an engineering discipline.
I don’t believe in mysterious causes. Putting the DAC under a pyramid, or aligning it with Jupiter in the 12th house to gain a guiding divine force.
If it doesn’t sound right, something is broken.


I have many customers reporting a SQ improvement after migrating from a Mac as a source. I’m not surprised. Since you mentioned Archimago you’d see he has another article which measures USB ports - you see the graphs showing very much imperfections.

Even if the digital bits are perfect, noise and interference affects the analog part of your DAC and/or worsen the jitter, ultimately resulting in SQ degradation.

If you’re interested in such type of debate you can also check out the USB cleaning product threads in audiophilestyle.com forums.

Thanks for your video.

Some comments under the video caught my eyes immediately

There are few important points to clarify here

  1. Digital signals should be very invulnerable to ambient interferences. (I forgot all the formula and calculation I learned from data communication course, but I still remember the conclusion of the calculation very well.)

  2. Should the iMac and dietpi send the exact data stream to USB DAC? Or the audio driver, coreaudio etc would do any processing? Can the Roon engineer help me out on this question?

  3. Jittering should be no matter in this case too. I tried coaxial on my setup. “Sound signature” is different as its more sensitive to jittering.

Thanks wklie.

I did aware those USB cleaning products, but what those products claim are truly quite contradict with what I learned from school. Digital signal in this domain should not be so vulnerability to noise. And jittering should also not significant for async transmission. However, as mentioned by Andybob, it’s theory Vs reality. I think the most important point is I don’t have the money to buy all and try all, and then I don’t have the measurement equipment to buy all and compare all. That’s why I talk to people here and try to gain more insight.

Yes, they should send the same bits. But you are not taking into account the different electrical properties of the two connections. Those do not change the bits, but they can leak into the analog side of the DAC. The analog signal derived from the bit stream (typically a version of the bitstream upsampled by a digital filter) is nonlinearly affected by the electrical noise reaching the DAC through the USB cable that carries the electrically-encoded bits. That is, the same bitstream will in general convert to a different waveform thanks to the different electrical inference from different bitstream sources.


Thanks, the digital signal like this should be very immune from most kind of noise (theoretically). In fact there is formula to calculate how significant the SNR should be. I remember it’s very hard to occur. Sorry, I left college for a very long time and it was just one course. I know not much indeed so I think all your opinions may be right.

1 Like

The digital part is fine. But a physical DAC works in the physical, electrical world, not in the ideal Nyquist world. The electrical wiring that carries the electrical encoding of a digital signal also carries other electrical noise that does not affect the decoding of the digital part but interferes with the analog waveform created from the digital part. This is not an opinion, it is a statement of physical fact. Whether it makes an audible for a particular DAC depends on its electrical design and on the specific source.

BTW, electrical interference is responsible for at least one of the recently discovered security vulnerabilities of digital processors (row hammer). It is dangerous to take digital circuit idealizations as always valid.


This phrase is very important to my knowledge. Thanks a lot.

The digitally encoded bits are very robust, but they contain only half of the information (amplitude) that’s needed to reproduce the signal. The other half is the time domain which is provided externally from clock chips. They exist in different qualities (precision) and are sensitive to all kinds of influences (mechanical, thermal, electrical) that may add additional errors to the clock signal.

1 Like

This is not strictly true. The same bitstream will convert to an identical analogue waveform in both instances if the signal/noise ratio is high enough to be decoded properly by the DAC (which it will be over any half decent USB cable).

However… low level noise and RF interference may be passed down a USB cable and if the DAC doesn’t completely isolate this from it’s analogue circuitry, it may be added to the analogue waveform coming from the D/A.

Some D/As are reported to be more affected than others which is why they sound different when connected to a source which is doing less processing such as a rPi endpoint, rather than a full-blown computer such as an iMac.

Read again, I was saying exactly what you are saying. A “bitstream source” in that context is a chunk of hardware with electrical characteristics, not an idealized digital source subject only to Nyquist’s beutiful theorem.

Yep, we’re saying the same thing. You were writing your second response as I was writing mine. :slightly_smiling_face:

1 Like

I’ll take the staunch objectivist approach here: how many hours of psychology did you take ? :wink:

1 Like