Schiit Magnius Balanced Headphone Amp Review

[Doodski]

Can we estimate how much further distortion might rise with a particularly complex passage of music?

In some situations sine wave testing and multitone testing is more strenuous than a musical signal. A musical signal is ever changing and it goes from simple waveforms to more complex waveforms but at the end of the day a musical waveform can be represented by a instantaneous value plotted on a X-Y graph using the complex number system or as job operators just like a sine wave can be. The design of the circuitry and the parameters of that circuitry can reproduce musical waveforms or sine waves as we know. The difference is in what distortions and changes to the energy is induced onto that waveform and that's generally referred to as the linearity. The linearity being the change from the input to the output with the best scenario being no change at all to the electrical energy. The difficulty in measuring the musical waveform linearity is that there is no known original standardized value that can be used to reference to the output and therefore measure the linearity of the circuit. The metering of that process is at this time from what I know not practical and is not a process that is even available.

If we have DACs that can frequently clear 120dB in multitone testing today, would music maybe knock it down a few more dB? 10? 20?

Because music is comprised of ever changing sine waves that combine that means the distortion(s) can be more or less than that number you used. Meaning the linearity of the waveform can be better or worse than a multitone spec'd number or even a musical signal.

manufacturers optimizing gear for benchmarks, not for music.

When a circuit design is layed out and implemented the end result is a circuit that is designed for any instantaneous value of energy and in this case between the limits of your hearing range as a example. A designer can't design for a multitone test with the expectation that the musical linearity would be worse than the multitone linearity. It's just not possible because if a circuit can do multitone energy then it can do musical energy and vice versa. Because multitone is a steady combination of waveforms and because Fourier Transform Theory details the end signal is comprised of all the waveforms and odd and even order harmonics then in some ways a multitone test can be more difficult for a circuit to process just like a musical signal may be less difficult to process or more difficult depending on what energy it is comprised of. To think that the energy is different in a musical signal compared to a multitone test signal is to go against the Fourier Transform Theory and other theories that are proven. Energy is energy and in this case musical energy is not easier or more difficult energy to process it is just a ever changing energy and a multitone is a steady rate of energy that can be measured in a steady state that is between the limits of your hearing range. It's all energy and the circuitry just processes the energy and the designer just designs for the energy bandwidth that is being processed. To look for other non-linearities for the energy as if musical energy is different than multitone energy is looking for stuff that simply doesn't exist. This is pretty complex stuff and I've attempted to simplify and shorten as much as possible and I am not a master of the subject.

 

[TheTalbotHound]

Multitone testing is representative of a circuit handling multiple frequencies simultaneously.

If only amir actually showed the multitone test consistently for amps. Seems like a lot, if not most, of the amp reviews by him do not include it.

 

[Shazb0t]

I don't understand what the people who don't think the multitone test is representative of music think music actually is. Spoiler, It's sine waves. All signals can be represented by a sum of sine waves. This is basic electrical signal theory. Look up Fourier.

 

[KTN46]

If only amir actually showed the multitone test consistently for amps. Seems like a lot, if not most, of the amp reviews by him do not include it.

@amirm do you think it's necessary to do Multitone/IMD/Linearity testing for amps?

 

[GoldenOne]

The test involves recording the output of each dac playing a specific 30 second piece of music. Each one is recorded by the RME ADI-2 Pro FS R ADC. Once running at 44.1khz, once at 192khz, and once at full 24bit 768khz (The other two are just in case they are needed in future. The 768khz recordings are the ones that are used for the null-tests.)

Info on the ADI-2 Pro ADC performance can be seen here: http://archimago.blogspot.com/2020/09/measurements-rme-adi-2-pro-fs-r-black.html
(Also to note, when recording the ADI-2 pro itself, I did actually get a 2nd ADI-2, rather than using a cable loopback, as the ADC/DAC running from the same clock would be unfair.)

Null tests are then performed using the "DeltaWave" software developed by a member here:
https://deltaw.org/
https://www.audiosciencereview.com/...test-deltawave-null-comparison-software.6633/

The tests were run twice, once with clock drift correction enabled and once without. Though for the tests I did keep a 20khz filter on so that ultrasonic stuff did not affect/skew results. But of course the raw files will be provided so that people can run their own tests with their own configurations should they wish to do so.


The original recordings, full deltawave reports for each comparison, as well as difference .wavs will be provided.

All DACs are being fed from the same USB source (SMS200 Ultra) to ensure that there is no potential issue arising from feeding from a beefy gaming PC.
All products are using their stock power supplies, everything fed from roon, all cables identical, even using the same USB cable. I tried to make sure that absolutely everything that could be done consistently was done consistently even if I thought it'd make no difference.

And again I need to state, this test is NOT and CANNOT tell you what product is 'better'. That is not what i'm doing, that is not what I'm implying. I will NOT be making any sort of judgement about the products in the video. (apart from one particularly bad product that literally didn't even null properly with ITSELF).
This test is simply intended to demonstrate that two products that in terms of THD and most tests commonly done here measure excellently and to a degree that many would claim is inaudible, can still sound different, and behave differently when actually playing music/a non-steady-state signal.

Regardless of if this test setup is 'perfect' or not, the fact that results are repeatable, consistent, and show differences at surprisingly high levels, does show that there is something more to the story. Exactly what/why that is would require deeper investigation from someone with more knowledge and access to better equipment than myself.

I'm not doing this to go "see! You're wrong!", this is nothing to do with being right or wrong. If I wanted to be 'right' i'd go argue on reddit.

I'm just interested about what is actually causing these differences, and i'm frustrated that the subjectivist side of the market is saying "just use your ears", when I'd much rather get a proper, scientific answer as to what design/engineering choices affect what.
And the objectivist side of the market is claiming these differences don't exist at all and seems completely loathe to even test or investigate them at any level because THD is low and that's that!

I'm doing this to hopefully start a reasonable discussion that can then be built upon. That's all. I'm interested in this, I'm curious about it, and am simply frustrated that no one else is willing to try it.
And as previously mentioned, I am MORE than open to critique and suggestions for how to do this in the best way. Cause that's exactly what I want.

So if there are genuine suggestions then I would love to hear them.
But "your review is rubbish", "sinad is high shut up", "you don't even have an APx555 why bother" etc are not particularly useful

I've just received my ADI-2 pro back and am unfortunately now running into a couple other issues.
The first being intermittent USB drop outs which is making recording uninterrupted basically impossible. And the second being that my ADC isn't meeting the spec for the product, or performance that I was previously getting (and others like archimago are getting).

for example archimago is getting about -115dB THD+N on loopback, and with my setup configured identically i'm only getting -100dB now. (-107dB if I lock freq to FFT and use rectangular windowing). Spec of summed performance for the DAC+ADC should be -111dB.
This figure does not change regardless of what source device (PC, pi etc) I am using, PSU, or anything else i've tried.

Given as i'd been waiting TWELVE WEEKS for this to get back from RME i'm less than happy about this.
I've emailed my dealer and RME to see what can be done, but given the AKM situation and the fact that there are no ADI-2's in stock anywhere, I'm not sure what will be able to be done. (I'm also not exactly keen to wait another several months to MAYBE have the issues sorted).

I'll run some tests to see how high I can get nulls of the same dac as a baseline, but I'm a bit concerned I may need to start this project from scratch again (or continue as is but have results separated by 'new' and 'old' ADC)

 

[Alec Kinnear]

Not a THX 789 killer alone but together with Modius and balanced usage this schiit is killing it. But man shipping cost is ridiculous with these. These are almost USA only products.

US is a huge market. Operating at capacity, living the dream, why would Schiit start in with all the hassle of EU paperwork and filing? I live in Europe and hence Schiit is not an option for me either. There's a few clowns on eBay who premium import Schiit for fetishists and collectors. Without some kind of EU infrastructure, there's no realistic warranty service. Have fun trying to export and import a broken amp for service.

With the anti-China drum beating going on in the US, Europeans have better options to buy from Topping or Aune with minimum headache than Yanks. We also have our own good manufacturers like Pro-Ject and Musical Fidelity (one company now), as well as RME.

No Schiit for Europe for now.

 

[QuantumNet]

I don't understand what the people who don't think the multitone test is representative of music think music actually is. Spoiler, It's sine waves. All signals can be represented by a sum of sine waves. This is basic electrical signal theory. Look up Fourier.

To be pedantic, anything that has discontinuities (square waves, etc.) in the signal cannot be perfectly represented by sine waves, however, you can get relatively close. With that being said, I imagine the vast majority of music does not contain these discontinuities.

 

[charleski]

To be pedantic, anything that has discontinuities (square waves, etc.) in the signal cannot be perfectly represented by sine waves, however, you can get relatively close. With that being said, I imagine the vast majority of music does not contain these discontinuities.

A square wave is a sum of an infinite number of sine waves. But our hearing is band-limited, so all those high-frequency components don’t matter since they can’t be heard.

 

[Shazb0t]

To be pedantic, anything that has discontinuities (square waves, etc.) in the signal cannot be perfectly represented by sine waves, however, you can get relatively close. With that being said, I imagine the vast majority of music does not contain these discontinuities.

To be even more pedantic, yes that's exactly what Fourier proved. You can create an equivalent square wave with a summation of sine waves. That's the entire point.

 

[AdamG]

To be pedantic, anything that has discontinuities (square waves, etc.) in the signal cannot be perfectly represented by sine waves, however, you can get relatively close. With that being said, I imagine the vast majority of music does not contain these discontinuities.

Welcome Aboard @QuantumNet.

 

[QuantumNet]

To be even more pedantic, yes that's exactly what Fourier proved. You can create an equivalent square wave with a summation of sine waves. That's the entire point.

Yes and no, depending on how strict your definition is.

In reality, your square wave made from a finite-limited fourier series will always be plagued by the gibbs phenomenon, causing an overshoot and undershoot near your discontinuity.

Even theoretically though, and more formally, the limit of the partial sums for the fourier series does converge which is what I'm assuming you're referencing. However, this convergence is considered a pointwise convergence and does not meet stricter definition of uniform convergence, essentially meaning that the fourier series converges at every point _except_ for the point at which the discontinuity exists. If you look at the point of the discontinuity, the function converges to half the height of the square wave. More over, you can also prove that the gibbs overshoot/undershoot converges to ~9% of the square waves height as number of terms tends to infinity. In my view, this shows that you cannot perfectly replicate a square wave, even if the partial sums of the fourier series converge at infinity. And in reality, it's even more damning since you can't ever get rid of the gibbs phenomenon since you'll always be limited in the number of terms you have due to technical constraints.

I would like to be crystal clear though and say this is a completely theoretical discussion and has no bearing on my views on the multitone test and its relation to music, which is what I originally replied to.

Welcome Aboard @QuantumNet.

Thanks for the warm welcome!

 

[gino1961]

Strangely enough since digital sources appeared people have started complaining with sound quality. With analog sources all equipment sound at least decent. More like music i mean.

 

[Veri]

Strangely enough since digital sources appeared people have started complaining with sound quality. With analog sources all equipment sound at least decent. More like music i mean.

That's a reaaaally broad statement to make, though.

 

[RHO]

Strangely enough since digital sources appeared people have started complaining with sound quality.

Because studio technicians treated it like analog while they shouldn't. Not because there is anything wrong or inferior about digital sound reproduction. Certainly not since the advancements made in the technology in the last decades.
So, yeah, maybe digital did not sound as good as analog in the early years of the technology. But I think we are waaaaaaaaaaaaayyyyyyyyyy past that already.

 

[gino1961]

That's a reaaaally broad statement to make, though.

absolutely.

 

[gino1961]

Because studio technicians treated it like analog while they shouldn't. Not because there is anything wrong or inferior about digital sound reproduction. Certainly not since the advancements made in the technology in the last decades.
So, yeah, maybe digital did not sound as good as analog in the early years of the technology. But I think we are waaaaaaaaaaaaayyyyyyyyyy past that already.

Hi i think you are very right. It could be a result of overprocessing used for producing digital masters. There is a very simple way to check that ... to make a decent digital copy of a high quality analog tape and compare the original to the copy on a high revealing system. If the copy cannot be distinguished from the original is a good point for digital.

 

[RHO]

Hi i think you are very right. It could be a result of overprocessing used for producing digital masters. There is a very simple way to check that ... to make a decent digital copy of a high quality analog tape and compare the original to the copy on a high revealing system. If the copy cannot be distinguished from the original is a good point for digital.

You can just measure it. No need to listen actually.

As far as I know, most of the processing was related to compensations needed for vinyl playback that could be omitted with digital playback. But studio technicians still did them (boosting treble, attenuating bass) while producing for CD.

 

[gino1961]

You can just measure it. No need to listen actually.

Thanks. But measure what ? i said listening because it is in the end what we do. Issues in bad digital recordings are very evident. I wonder if they listen carefully to what they produce. They have multimillion equipment ... what are they doing ?

As far as I know, most of the processing was related to compensations needed for vinyl playback that could be omitted with digital playback. But studio technicians still did them (boosting treble, attenuating bass) while producing for CD

why ??? they are not sane ... imho of course
thanks again.

 

[RHO]

Thanks. But measure what ? i said listening because it is in the end what we do. Issues in bad digital recordings are very evident. I wonder if they listen carefully to what they produce. They have multimillion equipment ... what are they doing ?


why ??? they are not sane ... imho of course
thanks again.

You can simply measure the difference between the analog signal of the original and the analog signal of the digitized audio after digital to analog conversion.
It depends on how the audio is digitized. But it could well be that the digital version is cleaner. Just running a ADC followed by DAC on the analog original will not yield any audible differences when done right.
You could blind test it.

Why they did those manipulations to the audio? They didn't know better.

 

[MRC01]

To be pedantic, anything that has discontinuities (square waves, etc.) in the signal cannot be perfectly represented by sine waves, however, you can get relatively close. With that being said, I imagine the vast majority of music does not contain these discontinuities.

To be even more pedantic, yes that's exactly what Fourier proved. You can create an equivalent square wave with a summation of sine waves. That's the entire point.

Yes and no, depending on how strict your definition is.
In reality, your square wave made from a finite-limited fourier series will always be plagued by the gibbs phenomenon, causing an overshoot and undershoot near your discontinuity. ...

At the risk of being master of the obvious, the emphasis I added above is true, yet understates the facts. Signals having discontinuities (like square wave, discontinuous 1st derivative) cannot exist in nature. A perfect square wave has an instantaneous rate of change, which as a sound, means an instantaneous rate of change of air pressure, which is not physically possible. Sound can only travel through a medium, and that medium (in our case, air) is inherently bandwidth limited due to its physical properties. The notion of a perfect square wave exists only in our minds, not in nature.

Since every sound that actually exists in the real world is bandwidth limited, it can be perfectly represented by a finite Fourier series.