[SOLVED] DAC Marketing and IMD (Also, Amps are too powerful rant)

[Vamp898]

I noticed that almost all DAC Makers advertise their Chips with THD+N and also the companies who make DAC and implement them advertise with THD+N.

For Example the ES9039PRO is advertised with -122db THD+N which is nice and cool but... it has -109db IMD. But ESS never tells that anywhere. As far as i could see also AKM and Cirrus don't give any IMD Values for their DAC Chip, only THD+N.

That is the first thing that annoys me, because the DAC is limited to -109, not -122db (unless you only output one single sine wave...)

Now i hear you say, who cares, both is inaudible right? Well yes, unless you use the DAC as Volume Control for sensitive IEM.

Because people love big numbers on paper, there is almost no Desktop Amp on the Market anymore where you can go beyond 5% Volume with most IEM without the risk of going deaf. Especially with Analog Volume Control, that is insanely annoying leading to big channel imbalance and sudden jumps in volume when trying to find the sweet spot in the 1~5% Volume area.

So i thought why not be clever. Set the Amp to Low Gain and Max Volume (still no audible Noise) and then just control the Volume on the DAC.

Well guess what, as soon i reach <=-80db, i hear distortion. Intermodular Distortion, i am not kidding. At first it is very very faint but as soon i reach around -90db things just start to sound very... harsh and hissy.

So no, i can not control the volume on the DAC, i have to find a mix of DAC Volume and Amp Volume which annoys. Its not a big issue, currently i have the Amp set to 30% (It then has an THD+N of around -105db) and then use the DAC in the range of -60db (-70db IMD) and -15db (<=-105db IMD). So i never hear either HD nor IMD

But my inner monk hates this Dual Volume Control and it annoys me that DAC Chip Makers and most DAC Makers seem to hide that value

 

[Whazzzup]

Set volume to max? What? Nooooooo.

Set correct impedance select correct gain, set volume to low. Adjust fixed line out volume from dac to correct exposure and depth. Not to loud and forward not to soft and distant. This is the preference. Micro adjustments to amp volume based on track exposure dynamics etc….

 

[Vamp898]

Set volume to max? What? Nooooooo.

On Low Gain. The Amp has the best performance at 100% Low Gain. That was the background for that idea.

Set correct impedance

I do not know any Amp who have an Impedance Selection. At least all the Amps i owned did that automatic, this Amp does it Automatic too.

select correct gain,

Low Gain^^ there is no lowest gain.

set volume to low.

It has highest Distortion at lowest Volume, are you sure? Shouldn't i operate the Amp where it has low distortion?

I doubt i will ever be able to hear it anyway, so this is speaking just theoretically, but using the Amp at an Volume where it performs good sounds more intuitive to me than use it at an volume where it performs bad.

Adjust fixed line out volume from dac to correct exposure and depth. Not to loud and forward not to soft and distant. This is the preference. Micro adjustments to amp volume based on track exposure dynamics etc….

If i do that, it is too loud, at least for BGM. Even at lowest possible Volume on the Amp, the volume is too high for BGM at work and stuff like that. So i do have to lower the volume on the DAC (or the input signal), there is no way around it.

If i do not lower the volume on the DAC, i can use like 1~7% on the Amp or something like that which is _very_ annoying with an analog volume knob. Also around 3~4% i have channel imbalance (Which is a common issue for analog volume control at very low volume).

At such an low volume, micro adjustments are simply not possible. The Volume Control on the Amp simply is not accurate enough for that.

 

[Whazzzup]

Ok imo and on my system. Amp has low or high impedance, hd800S is a high impedance hp. Low gain for me on hd800s is fine, while susvara og etc may want high gain. Set amp volume low for proper depth and exposure, for me and my amp, it’s 10 o’clock maybe 11. Set dac fixed line out to personal level. Then fire n forget. Maybe micro adjustments based on track exposure.

This for myself is non debatable, good luck. I would never set any parameter to max.

 

[Vamp898]

Ok imo and on my system. Amp has low or high impedance, hd800S is a high impedance hp. Low gain for me on hd800s is fine, while susvara og etc may want high gain. Set amp volume low for proper depth and exposure, for me and my amp, it’s 10 o’clock maybe 11. Set dac fixed line out to personal level. Then fire n forget. Maybe micro adjustments based on track exposure.

This for myself is non debatable, good luck. I would never set any parameter to max.

I usually try to operate my gear at the best possible sound quality. The DAC has the best sound quality at 0db (Max Volume) and the Amp has the best possible sound quality at Low Gain and Max Volume (And yes, i am well aware that i, as a human, are not able to hear the difference between -125db THD+N and -80db THD+N)

With the R70xa (470 Ω) i can go to 2 o'clock on Low Gain on the Amp with the DAC on line out. Not long, as this is pretty loud, but with the Headphone, i have zero issues and mostly i am around 10~12 o'clock. I only have issues with the IEM (9 Ω and 114db/mw)

I am currently listening to BGM while work. The DAC is set to -25db and the Amp is completely down, lowest possible volume on low gain and that volume combination is mostly fine for BGM.

I think this demonstrates how sensitive this IEM is and that i have no choice but lowering the volume on the DAC. So i try to find the best possible combination of DAC/Amp Volume that results in the highest possible quality.

 

[danadam]

Because people love big numbers on paper, there is almost no

... IEMs with sensible sensitivity

Well, honestly I don't know if that's true, I only remember I had some IEMs in the past with some insane (IMO) values.

Well guess what, as soon i reach <=-80db, i hear distortion. Intermodular Distortion, i am not kidding. At first it is very very faint but as soon i reach around -90db things just start to sound very... harsh and hissy.

<= -80 dB? Won't that leave with like less than 40 dB SNR (with very good DACs). That will sound kind of hissy.

 

[Vamp898]

... IEMs with sensible sensitivity

Well, honestly I don't know if that's true, I only remember I had some IEMs in the past with some insane (IMO) values.


<= -80 dB? Won't that leave with like less than 40 dB SNR (with very good DACs). That will sound kind of hissy.

First of all, i can no longer hear the distortion at -80db. Surprise surprise. When i tested it earlier, i forgot to increase the PC volume back to 100%.

Now the PC Volume is back to 100% and i hear no distortion at -80db on the DAC, it sounds perfectly fine.

But the SNR seems to be much better. This here is THD+N at -80db on the DAC

This here is IMD

I assume both measurements are limited by my DI

For Comparison, this is how it looks at 0db

And for the Ultimate, definitive fun, here is -95db (Again, pretty sure SNR is limited by my DI)

So i take everything back, there is no distortion issue, even at -95db...

 

[2leftears]

I know the problem has been solved, but FWIW:

Channel imbalance is indeed an issue with traditional analog volume pot controls at very low volume settings. Some premium pots are a bit better in that respect, like the ALPS blue pots, but you won't find those in all amps, especially small desktop or portable amps. One solution is to make your own cable with built-in attenuation, which I have done to attach a modern 2V line level CD player to a vintage 150mV line level amp.

Another problem in the past was that the simple analog volume controls on some (older) amps generate the highest noise floor somewhere in the third quadrant of the pot control. This is (I think) because at minimum volume the base of the differential input pair transistor sees zero impedance to the system ground, at maximum volume it sees the low(ish) output impedance of the source, but somewhere in the middle setting of the pot (resistance-wise) it sees a high impedance to both the system ground and source which means the base/emitter junction noise of the input transistor is less attenuated.

Modern transistors and amps have such low noise levels that this isn't really much of a problem anymore (in my experience).

 

[danadam]

But the SNR seems to be much better. This here is THD+N at -80db on the DAC

Hard to tell from FFT but if it is 32K point FFT then it seems to be about 8 or 9 bits of dynamic range/SNR. Here's 32K FFT of full scale signal at 8-bit / 44.1k sampling rate, for comparison:

I have here Tanchjim Space (ASR review). With Cosmos ADCiso it shows:

I reduced the volume by about 80 dB using the hardware buttons:

and used Cosmos Scaler with full amplification (+26.7 dB), which improved SNR by 9 dB:

Then I captured some things. The first 2 seconds contains shorted inputs, to get the self noise. Then it is the output of the DAC playing:

• 2 seconds of silence (16-bit dither actually), so you can compare ADC+Scaler self-noise with DAC noise
• 1 second of full scale 1 kHz
• first 5 seconds of three tracks from "Daft Punk / Random Access Memory"

I digitally amplified the captured signal by 54.5 dB to normalize it to about -1 dBFS. The result is in the attachment.

To me that would be too much noise (but maybe I don't understand what you are doing exactly).

 

[Vamp898]

Hard to tell from FFT but if it is 32K point FFT then it seems to be about 8 or 9 bits of dynamic range/SNR. Here's 32K FFT of full scale signal at 8-bit / 44.1k sampling rate, for comparison:


I have here Tanchjim Space (ASR review). With Cosmos ADCiso it shows:


I reduced the volume by about 80 dB using the hardware buttons:


and used Cosmos Scaler with full amplification (+26.7 dB), which improved SNR by 9 dB:


Then I captured some things. The first 2 seconds contains shorted inputs, to get the self noise. Then it is the output of the DAC playing:

• 2 seconds of silence (16-bit dither actually), so you can compare ADC+Scaler self-noise with DAC noise
• 1 second of full scale 1 kHz
• first 5 seconds of three tracks from "Daft Punk / Random Access Memory"

I digitally amplified the captured signal by 54.5 dB to normalize it to about -1 dBFS. The result is in the attachment.

To me that would be too much noise (but maybe I don't understand what you are doing exactly).

My plot was an 131k Point FFT. I use the Gauss Windows with a=4.5

What i did in my last test (where i no longer heard any noise nor distortion) was

PC: 0db
DAC: -80db
Amp: High Gain and roughly 80% Volume

But this was just for the sake of testing, that is not real world use. Real World use is

PC: 0db
DAC: -80db
Amp: Low Gain and roughly 33% Volume

At this Volume, the music is barely loud enough, that i can hear it, but it is _very_ low. I can only give a feeling as i have nothing to measure it accurately enough, but i would say it feels like 5~10db Volume or so. I have to work my way down. When i start listening music, i would hear nothing at -80db. I start at -50db and slowly/steadily decrease the volume over time and i reach my hearing threshold at about -85db where its almost gone (maybe a single peak here and there, at most). So it would be impossible to hear any noise ever, at all, at that Volume.

When i listen to a Live BluRay for example and i go to -15db on the DAC and 33% Low Gain on the Amp, it is pretty loud. I can not watch a full concert on that volume.

So i am moving in the range of -80db (Almost Hearing Threshold) and -15db (Pretty loud).

As this is only a difference of 65db... oh man i am way too sensitive for noise so maybe don't take me as an example

Especially with the SE846 i have a lot of issues with Volume. For example on my DAWN PRO when using it for BGM, i am mostly at 001/100~002/100 on Low Gain. And the issue is that at these volumes, i can hear Noise coming out of the device.

It is not my USB Port, i have an DSD Isolator which, after measuring, is excellent and performs pretty much as good as it gets (even though only costing 15$, cheap and simple solution) and the same issue happens on my Smartphone. EMD maybe? Not sure, but there is (sometimes) an annoying, erratic, "electric" sounding noise which can be heard when listening with such an sensitive IEM at such low volumes.

One of the reasons i got the Amp now. <0.1Ω Output Resistance, <0.3µVrms Noise

So to sum up a bit (sorry for the long posts^^): When i am at -80db, i will be at _very_ low volume levels where noise does no longer matter, no matter how big it is^^

 

[Davesrose]

First of all, i can no longer hear the distortion at -80db. Surprise surprise. When i tested it earlier, i forgot to increase the PC volume back to 100%.

Now the PC Volume is back to 100% and i hear no distortion at -80db on the DAC, it sounds perfectly fine.

But the SNR seems to be much better. This here is THD+N at -80db on the DAC

Wait, so your OP is complaining about specs of the DAC chip used on a DAC: and your issues with hearing hiss is the volume of your source (your PC)? This doesn't seem to be an issue with pots when it comes to different analog stages, or chip specs with an external DAC. It's very possible on the PC side that there was some added clipping/noise with whatever processing that could have occurred. PCs are notorious for not having apps that are "exclusive" and would be influenced by whatever audio settings you set for Windows (or their own EQ stages).

What's "Live BluRay"? Do you mean watching blu-ray concerts on PC? That also depends on your app, its settings, and if it has an exclusive mode for Windows API.

 

[gregorio]

The Amp has the best performance at 100% Low Gain. That was the background for that idea.

Where did you get that background? It’s a particularly bad idea as amps will typically have best performance at around 60%, a good amp will have a wide range of optimum performance, say 20% - 80%, few will even be good at 100%, let alone best. At a low gain setting though it might be dependant on the topology, so again, how did you determine best performance at 100%?

If i do not lower the volume on the DAC, i can use like 1~7% on the Amp or something like that which is _very_ annoying with an analog volume knob.

Then it’s the wrong amp for the task.

Real World use is
PC: 0db
DAC: -80db
Amp: Low Gain and roughly 33% Volume

That’s mad, why on earth would you deliberately reduce the resolution/dynamic range of your DAC by a factor of 10,000? That’s got to be close to a record for the worst gain-staging in history, even for a consumer, but I thought you said you were an engineer or did some audio engineering?

I really hope I’ve misunderstood something and this isn’t actually your “real world use”?

G

 

[2leftears]

That’s mad, why on earth would you deliberately reduce the resolution/dynamic range of your DAC by a factor of 10,000? That’s got to be close to a record for the worst gain-staging in history, even for a consumer, but I thought you said you were an engineer or did some audio engineering?

Would that not depend on the way the digital volume control is implemented in the DAC? A separate digital volume control chip after the DAC chip, employing a switched resistor ladder network wouldn't reduce resolution surely, albeit that you still lose a little bit S/N ratio due to one or two extra semiconductor junctions?

(-80dB on the DAC might still be a bad idea from the perspective of picking up more amped audible noise in the DAC/amp interconnect though).

 

[Davesrose]

Would that not depend on the way the digital volume control is implemented in the DAC? A separate digital volume control chip after the DAC chip, employing a switched resistor ladder network wouldn't reduce resolution surely, albeit that you still lose a little bit S/N ratio due to one or two extra semiconductor junctions?

And the OP said it was solved before the DAC stage: setting the source PC to 100% volume. Why convolute matters, when it's very easy that a source PC will have its own source digital processing before it's output to a USB DAC?

 

[danadam]

So i am moving in the range of -80db (Almost Hearing Threshold) and -15db (Pretty loud).

Ah, ok. From the way you wrote it initially, I somehow understood that -80 dB is your nominal level, a level you usually listen at.