Topping DX5 II

[Dandelion]

It's not hard at all to measure with music too, I have, multiple times.
Instead of posting endless charts, have a look at the FSAF method.

It can be used to measure distortion with any signal, including music.
I'll let you do your homework and search here, we have posted lots of examples.

(hint: no difference than a nice multitone)

At the end of the day isn't this just a null with a given set of adaptive filters?

 

[Sokel]

At the end of the day isn't this just a null with a given set of adaptive filters?

It's exactly what you asked for, a measurements with music as the test signal.
Whatever music can do to any DAC is fully visible,.

Excuses like "yeah, but what about with music?" are long gone thanks to people who actually knew what to do to demonstrate it.

If you want to differentiate gear with that kind of performance search elsewhere, sound alone won't do it.

Here's one FSAF I have handy, that's a 20 yo interface:



Audibility of DACs are a long, LONG solved matter.
Wanna hear a difference? Focus on speakers, apply EQ, RC, etc.

 

[Dandelion]

It's exactly what you asked for, a measurements with music as the test signal.
Whatever music can do to any DAC is fully visible,.

Excuses like "yeah, but what about with music?" are long gone thanks to people who actually knew what to do to demonstrate it.

If you want to differentiate gear with that kind of performance search elsewhere, sound alone won't do it.

filters aren't invisible, that's my only issue with it. You still need to decompose the signal somehow to plot the different harmonics created.

Think about it, the single tone test showed odd harmonics created, if they were irrelevant and real world usage would confirm that they're both basically flat at audible range, why would the user for example perceive a midrange presence? Or the users of chord products (which have their set of harmonic creation) feel the difference in sound? Take for example the chord mojo, it has harmonics on multitone chart below -120db FS in the audible range, should be inaudible no? Yet people hear that color.

 

[Sokel]

filters aren't invisible, that's my only issue with it. You still need to decompose the signal somehow to plot the different harmonics created.

Think about it, the single tone test showed odd harmonics created, if they were irrelevant and real world usage would confirm that they're both basically flat at audible range, why would the user for example perceive a midrange presence? Or the users of chord products (which have their set of harmonic creation) feel the difference in sound? Take for example the chord mojo, it has harmonics on multitone chart below -120db FS in the audible range, should be inaudible no? Yet people hear that color.

We simply don't care about distortions at that level, any of it.
At the end of the day we measure it as Total distortion+Noise (see the chart, it's not only harmonic) or IMD, etc.

And no one has demonstrated a reliable difference at a DBT even with far inferior gear in terms of distortion.

 

[Dandelion]

We simply don't care about distortions at that level, any of it.
At the end of the day we measure it as Total distortion (see the chart, it's not only harmonic) or IMD, etc.

But if we don't care, why there's an audible difference then? And what is "That level"? I know you guys want to see that on a piece of paper and for that data or graph to be accurately representative of what would be heard in a real world scenario when throwing music at it. I don't think is possible.
Heresy you might scream, there's this test, that other test, etc. but pink/brown/white/whatever color noise is always that, same for a sine sweep. It's that thing, that sound. Songs aren't. They aren't static in amplitude or frequency response nor else, otherwise every song would sound the same. I think the test sounds we throw at machines are pretty representative and helpful, but they are a sample approximation of what happens when we listen to all sorts of material.

I give you another real world example of how saturation can be hard to detect: same stem, 2 different versions with 2 different levels of saturation, both show no difference in the frequency response graph and have same level across the spectrum and loudness matched, any eq plugin with matching eq function did exactly nothing, yet the more saturated one would sound denser and fuller. Now if I put the saturator plugin in something like plugin doctor (throws test tones and plots the differences in dynamics, transient, thd, imd, phase, etc.) I of course see the harmonic creation, and except 1 spike slightly above, the rest was below 100db fs on a graph. Now me nor my colleagues have bionic ears, in the real life scenario was audible to all of us. It compounds.

 

[RHO]

But if we don't care, why there's an audible difference then? And what is "That level"? I know you guys want to see that on a piece of paper and for that data or graph to be accurately representative of what would be heard in a real world scenario when throwing music at it. I don't think is possible.
Heresy you might scream, there's this test, that other test, etc. but pink/brown/white/whatever color noise is always that, same for a sine sweep. It's that thing, that sound. Songs aren't. They aren't static in amplitude or frequency response nor else, otherwise every song would sound the same. I think the test sounds we throw at machines are pretty representative and helpful, but they are a sample approximation of what happens when we listen to all sorts of material.

I give you another real world example of how saturation can be hard to detect: same stem, 2 different versions with 2 different levels of saturation, both show no difference in the frequency response graph and have same level across the spectrum and loudness matched, any eq plugin with matching eq function did exactly nothing, yet the more saturated one would sound denser and fuller. Now if I put the saturator plugin in something like plugin doctor I of course see the harmonic creation, and except 1 spike slightly above, the rest was below 100db fs on a graph. Now me nor my colleagues have bionic ears, in the real life scenario was audible to all of us. It compounds.

I think you forgot (ignored) Sokels last sentence.
Answers all your questions.

And no one has demonstrated a reliable difference at a DBT even with far inferior gear in terms of distortion.

 

[malaire]

Changelog:
V1.49 (July 17, 2025)

*Fixed: After detecting abnormal headphone amp DC, the main interface no longer displays "HP DC abnormal".

That doesn't make any sense to me - why would they remove that text display? And user experience shows that that was added in 1.49 not removed.

If you check Chinese changelog, it says "修复：耳放直流量检测到异常后，主界面不显示"HP DC abnormal"" which when translated to English with Google (as I don't understand Chinese) is "Fixed: After the headphone amplifier DC current is detected to be abnormal, the main interface does not display "HP DC abnormal"".

That makes more sense: i.e. there was a problem that the text wasn't being displayed and that problem has now been fixed. Seems to me that whoever wrote English changelog doesn't know English that well and made an error there, i.e. "no longer displays" instead of "does not display".

 

[Dandelion]

I think you forgot (ignored) Sokels last sentence.
Answers all your questions.

I guess we can kindly ask the user with both the topping and the rme to do a blind test and see if it spots the difference more than 60% of the times, what else can I tell you? Or even somebody with a chord and a topping, or whatever colored amp.

 

[Dandelion]

Btw for context I opened the project and saved the preset to throw it under plugin doctor, this is the saturation level I was talking about

 

[RHO]

I guess we can kindly ask the user with both the topping and the rme to do a blind test and see if it spots the difference more than 60% of the times, what else can I tell you? Or even somebody with a chord and a topping, or whatever colored amp.

60% doesn't cut it.
It will have to be over 90% to be statistically sound.
60% is way too close to be an absolute guess.

 

[Dandelion]

60% doesn't cut it.
It will have to be over 90% to be statistically sound.
60% is way too close to be an absolute guess.

That's what sample size is for, if you guess 2 out of 2 correctly but it was just luck that would be 100% but still luck.

 

[olieb]

this is the saturation level I was talking about

You were talking about -100dB, but this is a THD of -68dB! That is way, way more than any dac we are talking about here.
And this with a signal at ≈-27dBFS. How does it look with levels that are actually interesting for saturation, like -10/-6/-1dBFS?
And why 27Hz?? How does it look like with 200Hz or anything midrange?

 

[Sokel]

Btw for context I opened the project and saved the preset to throw it under plugin doctor, this is the saturation level I was talking aboutView attachment 465543

Don't know how you captured this but its clearly clipped.

And as we talk about music, here's Diary of a Madman as a tribute to Ozzy through the same 20 yo interface.

Here's the settings:



..and here's the result:



The subject has been beaten to death.
Take my advise, search elsewhere.

 

[RHO]

That's what sample size is for, if you guess 2 out of 2 correctly but it was just luck that would be 100% but still luck.

Yes, of course. No disagreement there.
I would at least go for 20 samples to be reliable. Get 18 (or 2) out of 20 correct and we're talking.

 

[yorkieaudio]

Just had dispatch notice from Shenzenaudio for my White DX5 II, so the new stock has finally turned up of at least the white models.

 

[Dandelion]

You were talking about -100dB, but this is a THD of -68dB! That is way, way more than any dac we are talking about here.
And this with a signal at ≈-27dBFS. How does it look with levels that are actually interesting for saturation, like -10/-6/-1dBFS?
And why 27Hz?? How does it look like with 200Hz or anything midrange?

Sample rate is 44.1 kHz. Test tone that inputs the plugin is at -4.5db and as far as frequency it was just setup like this from the start and kept it there, the lowest the more you see how such saturation develops across the spectrum, if I sweep it up everything just scales up, it's the same blend of even/odd harmonics and nothing happens below.

You're not seeing a graph you'd usually see around here for machines, rather what the saturation circuit of the plugin and what it does in this production setting, so that's why it might look confusing to you guys and one person sees clipping and the other interprets it at -27db FS, what you see is simply the output of the plugin with the same settings it had in the position of the chain it was.

About the cumulative THD print produced by this tool that you read, don't compare it to the data format you see around here, just look at the single spikes of the harmonics created (and btw if I sweep the test signal to 1 kHz the cumulative print for THD obviously lowers since nothing below happens).

 

[Sokel]

Sample rate is 44.1 kHz. Test tone that inputs the plugin is at -4.5db and as far as frequency it was just setup like this from the start and kept it there, the lowest the more you see how such saturation develops across the spectrum, if I sweep it up everything just scales up, it's the same blend of even/odd harmonics and nothing happens below.

You're not seeing a graph you'd usually see around here for machines, rather what the saturation circuit of the plugin and what it does in this production setting, so that's why it might look confusing to you guys and one person sees clipping and the other interprets it at -27db FS, what you see is simply the output of the plugin with the same settings it had in the position of the chain it was.

About the cumulative THD print produced by this tool that you read, don't compare it to the data format you see around here, just look at the single spikes of the harmonics created (and btw if I sweep the test signal to 1 kHz the cumulative print for THD obviously lowers since nothing below happens).

Let me understand here: What's the original test signal of this chart and at what level?
I might be able to replicate it.

Remember, the level of TD+N or any kind of noise and distortion is better be expressed at dBr (relative to the fundamental's level) or else can be deceiving.

 

[Dandelion]

Let me understand here: What's the original test signal of this chart and at what level?
I might be able to replicate it.

Remember, the level of TD+N or any kind of noise and distortion is better be expressed at dBr (relative to the fundamental's level) or else can be deceiving.

Sine wave, -4.5 dbFS.
What you have to replicate? We're kind of going OT, my point was that harmonic distortion and its psychoacoustic effect on sound can be more pronounced than what standardized tests and graphs can make it appear, hence why people hear coloration in some preamps.
This practical example was just to show a real life scenario of how a blend of harmonics added to a signal can make it sound denser and thicker even when overall level and frequency response show no difference. Needless to say *of course* a headphones preamp is not as colored as a distortion box and we're talking about subtle coloration. There's this kind of rme cult where that machine can't be touched somehow, I'm not saying is bad, I was just saying it can make sense if the user reports hearing more midrange on it, or that chord users hear a color there as well in other parts of the spectrum, and the fact that graphs show such color below -100dbFS doesn't necessarily mean is inaudible.

 

[Sokel]

sine wave, -4.5db FS

Can't see a sine on your chart at -4.5dBFS.
Can't be the 20Hz one as no H2 and H3 are at 40Hz and 60Hz.

Also there's no normal visible noise floor.
Is it some kind of cross-correlation method used?

The whole pattern looks weird, are you sure is not some kind of triangle or square wave here?

 

[olieb]

Test tone that inputs the plugin is at -4.5db and as far as frequency
... and the other interprets it at -27db FS,

Well, if you show a plot where the fundamental is at -27dB (assuming the 23Hz peak is actually the fundamental and not a distortion peak) and give no further information whatsoever, what were you thinking?
How is anybody supposed to know what your saturation plugin is doing? And how you reference your axis? And you do not show what happens with 0dBFS or -10dBFS. Saturation is supposed to be non linear, isn't it?
Still the 3rd harmonic is at -70dB relative to the fundamental (not -100dB).
This is more or less 50-60dB(!) higher than the ADI that you claimed had "fuller midrange" because of distortion.
That is like hearing hand clapping from 100m away and then concluding hand clapping from 20km away will have a similar effect.