So it wasn’t just me then…
Wow…this is an eye-opener for me
I suspect that linking this discussion to archimago will simply add fuel to those who want to fan flames (and that’s on both sides of the debate)! I remember contributing to this question very early on. I would never have guessed that it would have exploded the way it has.
Tim, I am full of questions??? Now I am diving in to this digital world. At heart I love math and my logic is simple but in the analog world I can hear the differences. An example would be changing a phono cartridge out. In some cases its huge. Graphs give me true answers too. I looked at THD, IM, and output using Ohms law for guidance too. But after a few months of reading about digital hardware I found myself buying products from so called experts without looking at math. I guess I needed Archimago to get me back on track. I guess I made a few mistakes and I want eveyone else to learn from mine! Too late for me but not for others.
But Bill, you would be hearing your own reproduction of the sound! If what you want worked, then you could save yourself a fortune by having someone with a $1,000,000 set up just email you a sound clip for you to listen to on your phone…
I’d like to go back to that post again and kindly ask you, if you’d differentiate between electronics and loudspeakers/rooms - you’re correcting your speakers and room acoustics, using measurements, after all.
You likely studied Floyd E. Toole’s book Sound Reproduction Loudspeakers and Rooms, where it’s the quintessence, that the limited set of possible measurements do very clearly show, how something will sound.
We’re looking at a lifetime of meticulous scientific work to prove that.
If you didn’t, I’d highly recommend it.
If you also had a chance to look into psychoacoustic research, you’d likely come to the conclusion, that a good portion of measurable electronic’s distortion characteristics are imperceptible to the auditory system, since they are several orders of magnitude smaller, than with electroacoustic devices.
Consequentially, there seems to be quite some hard evidence in favour of the “objectivist measurement camp”.
Most subjectivist audiophiles wouldn’t fall for equivalent marketing blah, when buying other technical appliances.
Why is that so?
Too many emotions, maybe, clouding the minds.
Peace to everybody,
To quote someone: “There are things you can measure that you can’t hear, and there are things you can hear that can’t be measured”.
There is no proof that everything we hear can be measured, which means all so called “proofs” about measurements with regards to what we hear is based on an assumption. Besides, if you have a reasonable transparent system, its not at all hard to hear difference between a Raspberry 3/4 as transport and a transport like opticalRendu or Sotm sms-200 ultra.
To sum it up: the only way to determine audible differences is to use your ear (yes, its that simple )
Here is a HiFi News lab report on a Project Stream Box S2 Ultra.
There was a measured 20db improvement in the jitter spectra of some DACs using it compared to the output of a pc. And virtually no improvement in the case of other DACs. So, I don’t think anyone, especially those who claim to have a technical approach to hifi, should generalize about these matters. And they certainly shouldn’t claim that all USB outputs are equal, because that is not true. Interestingly the S2 Ultra is based on a Pi, but the designer has gone to considerable lengths to clean up the USB output of the Pi, which you can read about here.
If someone were to say something just wrong like that, someone would be sure to remain anonymous!
I’m not saying this is wrong. Especially the first part is proven. The second part though… You cannot base such a conclusion on anecdotical evidence. If you want to use a human as measurement device, which in itself is perfectly fine, you need to conduct a double blind test. If you can then proof with statistical significance that the test group can hear things we cannot measure (yet), then I agree with this statement. Until then, I’m quite sceptical…
Yes, but only if I were to depend on fallible human ears! Being me, though, I would be much more likely to run them into a computer and compare analyses. That way we could see exactly what the distortions are that makes the poster think one sounds “better”.
When in doubt, measure with equipment.
As I have just pointed out in my previous post, it is perfectly possible to measure a 20dB difference in the jitter of some DACs depending on which USB source they are connected to. How can you doubt that this is possible?
I’m sorry, but I’ve lost the thread of your argument. We were talking about jitter from the USB port of a Raspberry Pi, to a modern DAC. I don’t see that the test you pointed to addresses that. Without knowing more about the unspecified “PC” used in that “lab report”, I don’t even see how to even treat that as evidence of anything in particular.
Do you dispute the measurements? Do you think they are incorrect? In what way? Have you any of your own? What bit of science says that all DACs behave identically with all USB sources? The measurements are from an extremely well respected reviewer with decades of experience. You evidently didn’t read the post from John Westlake listing some of the weaknesses of the Pi architecture, so here it is in full.
“Basically its weakness are all based on Clock Phase Noise (Jitter) and local PSU noise (which directly impacts the Jitter performance) and also the resultant local RF emissions:-
The USB HUB / LAN IC is clocked by the CM3 SoC which has significantly higher jitter then an external Clock oscillator. More detrimental is that fact that clock phase noise is heavily correlated with the SoC software processing – resulting in Data correlated jitter.
The USB HUB / LAN IC PLL clock circuits are powered from the switching supply rail that also powers the SoC Memory, so as data is “processed” though the memory this causes modulation of the PSu Rail directly correlated with the Data patterns which then modulates the critical PLL in the USB HUB / LAN which is used as our USB “source” via this noisy PSU rail.
As one would expect little care has been taken with the quality of the 5V rail for USB DAC power – in fact its really pretty poor by “Audio” standards.
There are multiple “free running” switching regulators on the RPi PCB – whose clocks are not related to the USB clock Rate so there switching products are free to induced multiple / complex RF and Ground noise spurie in the system.
I could list more weakness, but you can start to appreciate the “Deficiencies” (not unexpected) of the simple Pi3”
I did indeed read the post, the entire thread, in fact.
First, about the “lab report”: I don’t “dispute the measurements”, but the documentation of the experimental design does indeed seem to be incomplete, so I’m not sure what he’s proving with them. In any case, this would seem to have nothing to do with measurements of a standard Raspberry Pi 3B+, which I thought this thread was about.
Then, about Westlake’s post: just parenthetically, I think this list of grievances by John Westlake, about the standard Pi, is more relevant to DACs powered by the master side of the USB connection. But even there, he talks about such concepts as “Audio” standards. Which are what? And how evidence-based are they? One might wonder if the proposed “fixes” are really more about propitiating the more clueless audio reviewers, than about actually making a useful improvement. More about bowing to the requirements of audio mythology than actually doing something important.
Again, Westlake offers no evidence, no measurements, showing that his improvements make a significant difference. It’s just argumentation. Hypotheses. And of course he has an ax to grind, a product to sell. I’m not saying he’s lying, just pointing out that it’s in his interest to argue that his product is somehow better for its purpose than the unimproved and much less expensive Pi (sales of which will not enrich him). Since we have no head-to-head measurements of the two, it’s an unproven hypothesis, I’d think.
If people wan’t to believe that measurements from Amir and Archimago tells everything about how it sounds, than they are free to do so. But notice the emphasis on “believe” because there is no proof involved. And thats one of the things I dislike about measurement discussions, some seem to think its all hard science vs bias-influenced listening, but as long as there is no proof that everything we hear can be measured its nothing but a theory at best.
Another thing: how many times have we read something like this from measurement descriptions: “there is a peak at 50Hz compared to XXX, but its not audible so nothing to worry about”. Not audible for whom and in what system? Because lets get one thing clear, what is audible or not depends a great deal on the HiFi system, room acoustics and even the listener.
I have the RME ADI-2 DAC (which does measure well), and I compared it to the Chord Qutest. I liked both, but they don’t sound the same. The treble is different, and Chord Qutest have a little wider (but less homogeneous) sound stage. Can someone point me to what difference in Amir’s measurements shows this? Nope, there is none, but they still do sound different
So much of what we hear is in our individual minds, it makes little sense to compare two pieces of equipment with any one person’s ears and brain. That’s why objective controlled measurement is so important.
In fairness, RME have been making studio grade audio interfaces that are designed to be connected to noisey powerful workstation class computers for a long time. I wouldnt expect anything less from them
OTOH, with a less expertly designed DAC, your mileage may vary - alot!
I use a Raspberry Pi 4 + Intona High Speed USB-Isolator + Musical Fidelity M6srDAC with async USB mode. No noise, no clock issues, virtually no jitter. The sound is so much better and clean now than before w/o a bridge or w/o the isolator. I believe in that setup even the audiophile low noise power supply I use could be spared.
I’ve been lurking on these pages watching this whole thread with interest.
I felt compelled to write a pretty long missive - if only to stimulate others to go and study the subject matter they are commenting on - it’s aimed at the whole thread and no one individual.
I’m relatively new to Roon, only having started using it in January this year.
I am a convert, and have moved my various (3) systems to Roon from Volumio.
In what will turn out to be either a great move or folly - I have also paid for a lifetime subscription to Roon.
As what I like to think of as a “music first” audiophile - I’m more interested in what my own ears perceive and the music - than I am in what a marketing led journalist may tell me is better… but the important thing is the music - not just the kit that is used to get it.
I am an engineer (C.Eng and MIEEE) with over 130 patents granted during my career. In my early years worked in the recording industry repairing and updating mixing desks and all sorts of associated equipment.
I’ve made many of my own amplifiers and preamplifiers, and even built my own ladder DAC back in the early 90s.
I was around when the real change to mastering for CD rather than Vinyl started. That coincided with my engineering career being in digital signal processing (although my work is/was mostly in healthcare).
Audio-wise I am now very fortunate after many years of hard work to be able to afford a decent all digital setup - with a minimum box count, and the best bit being that others far brighter than me did all the hard work to make it so good.
I have a set of Kii Three speakers, Innuos Zenith SE and Phoenix with a Kii Controller as my primary system. All 3000+ of my CD’s are ripped (Red Book standard) to FLAC and stored both on a NAS and on the Zenith. The Zenith is also my Roon Core on a hardwired network with CAT6 cable. I also play some hi-res files that I picked up along the way, and stream both Tidal (including MQA) and Qobuz in their maximum resolution formats.
I also have a Raspberry Pi 3B+ with a HiFi Berry Digi +Pro running Ropieee which connects via SPDIF to a Graham Slee Majestic DAC and Solo Ultra Linear headphone amp that drives a pair of Beyerdynamic T1’s (2nd gen) and others including a pair of 1770’s and 770’s. I guess the habits of a studio life die hard, with Beyerdynamic headphones and AEG microphones being in the majority here.
I also have another more recent Raspberry Pi 4 running Ropieee which connects via USB to either an Audioquest Dragonfly Red or an Onkyo DAC-HA200 which can be used with a range of IEM’s or many of the other headphones from my relatively large collection (about 20 pairs).
I have not got the spectrum analyser out (yet) on any of this but these are my subjective conclusions from my ears (not golden by any means):
Using each of the sources direct to the Kii Three’s and with nothing else in the signal path there are (unsurprisingly) differences between a) The Zenith SE and Phoenix, b) the Zenith SE on its own, c) the Raspberry Pi 4 using USB and d) the Raspberry Pi 3B+ with the HiFi Berry Digi +Pro.
Set a) has perfectly “black” quiet points, a tall and very wide soundstage with excellent clarity, depth and what I would consider to be near perfect reproduction across the range of frequencies. Let’s face it - a system that retails for over £25k needs to be pretty special. It is.
The attack and decay particularly in complex orchestral or large band groups is excellent allowing you to pick out individual instruments in amongst what would be normally quite a confusing picture. It shows up poor recordings and bad mixes, being quite brutally honest in its reproduction. It’s occasionally like being there, but Although it’s incredibly engaging to listen to and very relaxing, it’s not what I would ever describe as a warm reproduction (often with harmonics) that some people seem to gravitate to. That’s not wrong - just different.
Set b) mirrors many of the points from set a) - but if there is a criticism the soundstage isn’t quite as tall or wide and the decay from very complex pieces may not be quite as fast. Otherwise, it’s still pretty special.
There is very little between c) and d) with the latter seeming to just get ahead on clarity and handling of complex “big” soundscapes as generated in some electronic music and movie themes.
Both Raspberry Pi iterations perform way, way better than the price point might lead you to expect, and as in most things where we seek perfection, there is a law of diminishing returns as you get further and further towards Nirvana (the place not the band). So the Pi’s are really good, but there is a very clear difference between a)/b) and c)/d) that you’d probably expect.
The Kii threes do intrinsically tidy up their input signal and reduce any jitter, so I can’t say that I’ve heard anything untoward from the Raspberry Pi’s as they do their stuff, no matter whether they are using SPDIF, optical or USB.
The sound stage from both c) and d) isn’t anywhere near as expansive, and nor does either resolve the picture to place musicians within it accurately - in comparison to system a). However, make no mistake, both RPi setups are far better at doing all of that than either a relatively new Lenovo Laptop or a Mac Mini - both of which I have also tried on the same system.
The Pi’s are in my opinion also almost an order of magnitude better than the excellent Chromecast Audio - a fantastically convenient device for some applications.
I think it’s fair to say that the system sounds better on USB using the Raspberry Pi 4 rather than the Raspberry Pi 3 on USB (no HAT) but it is hard to pinpoint exactly why. To my ears it just felt that the Raspberry Pi 3 was battling through some “mush” in comparison to the ease at which the Raspberry Pi 4 was operating.
My conclusion is that it’s hard to go wrong at almost any system level by using a Raspberry Pi as a streamer.
They’re really very good, and probably far better than they ought to be considering the low price of putting together a reasonable one. In my view it has made other streamer and DAC manufacturers step up their game to try and differentiate from this $130 upstart. That’s a good thing.
Using something better than the standard wall wart power supply as used with most Raspberry Pi implementations will almost certainly bring you additional sonic benefits, and I recommend the 5V 3A iFi power supply, because yes, I could hear an improvement - notably a clearer background.
So a penultimate comment from me is on the “bits are bits” statements we have seen here and elsewhere.
So aren’t bits just bits?
Well, yes they are - technically but in this engineer’s view (and it’s only my view) the issue is not only the bits, but the TIMING of when exactly those bits are supposed to be there. That’s a measure both of how clean the waveform is (nice square edges, limited hysteresis and a fast rise and fall) and just how much noise is interfering with the “threshold” (the point at which a voltage rise or fall is recognised as a 1 or a 0).
The point at which that threshold is reached is when the digital signal processor actually recognises it and it is this timing that in my view is CRITICAL.
More importantly than my opinion, psycho acoustic studies tell us that the human ear/brain is INCREDIBLY sensitive to these timing differences - particularly when both ears are in use.
It’s what helped us survive being eaten at some point in the dim past.
For those who want to know more, a quick Google of “Intraaural Time Differences” will open a whole new world. A long time ago, I worked on some hearing test equipment for those with cochlear implants and I did a fair bit of reading around this topic. Some people have spent a lifetime studying just one small aspect of psychoacoustics - so a few comments in a forum barely scratches the surface…
Should small factors like jitter, electrical noise, processing/clock/timing variables combine to create or exaggerate even a small delay effect - which causes the timing to be off - we will probably notice it.
Some studies have demonstrated that at certain frequencies the human ear/brain can detect frequencies with a difference of just 1Hz, and timing differences of under 50 micro seconds or 50 x 10 to the -6. It may be smaller, but physiological measurements at that speed get more difficult.
So yes, bits are bits, but the accurate delivery and processing of these bits and their relationship to the clock signal and how they are perceived by the human ear is why digital audio engineering is just so fascinating and why such debates and myths carry on on forums like this.
My advice - do your own research and make your own mind up.
A last word here for Ropieee, a superb piece of free software that makes creating a Roon endpoint and maintaining it an absolute breeze.
Thanks to the diligence of its creator, it just keeps getting better with added features.
If you use it, I urge you to donate something to the creator so that it can just keep getting better, and so that support can continue.
Whatever happens - just enjoy listening to music.