Objective Measured Audible Differences between two DAC/amp dongles

[GXAlan]

1) Well-designed DACs and AMPs when run linearly without clipping/distortion will sound the same. True.
2) Idiosyncracies representing unique circumstances can result in different sound. True.
3) Reproducibility can be hard. True.

Long post.

My Cambridge Audio DacMagic XS didn't sound as good as my Apple USB-C adapter. I've had it for weeks, listened to different music over different days and I found the DacMagic muddy compared to the Apple which leaner. When listening, I would just move my IEM from one connector to the another and correct volume using Foobar.

The Cambridge has more baseline hiss with my IEMs compared to the Apple DAC. 100% ABX'able. The AKG K3003 is 125 dB/V efficient. 8 ohm impedence (which is lower than the DacMagic's "minimum 12 ohm" recommendation). So this may be unfair to the DacMagic.

I just got around to measuring them this afternoon with a UMIK-1. I essentially aimed a UMIK-1 close to my earbuds (1-2 cm), using the foam packaging/box for the UMIK-1 as the test setup since the earbuds have a good friction fit with the foam.

I ran the first comparison sweep as is and got this:

wow! Right? You can see that when matching SPL at the lower frequencies, there is a mismatch at the higher frequencies. This supports the "lean" vs "bassy" feel.
So the measurements matched my subjective experience. From 150 Hz and below, you can see that the run-to-run variability is low so I don't think this is a technical error with the measurement, at least in the 60Hz and up range.

I put on my earbuds and tried to listen to various test tones to really determine if frequencies like 60Hz were audibile and they were. Remember that my UMIK will have lower bass than the actual experience since the mic does not seal the way the IEM does “in ear.”

I listened to music and the measurements matched the experience -- the Cambridge Dac FR tilt was recessed compared to the Apple dongle, esp. if the subjective volume matching was equal to the bass, or I just applied the 10db difference that I measured.

Now, I am intrigued. I wanted to see if the problem got worse at lower or higher volumes. I started with the Apple dongle at 40% and the DACMagic at 100% with the analog gain down.

So I tried Apple @ 90% and then matched the DACMagic. No difference— the DACs measured similarly. Then I tried 5%, no difference. So I increased the analog gain with the buttons on the DACMagic and also compared it to the USB-C single at high volume.

Now they measured identically. I then tried a whole series of various electronic attenuation and various analog amplifier settings of the DacMagic.

I was not able to get the two DACs to perform differently anymore.

And the crazy part, when I went to listen subjectively to see what happened, both DACs now sounded the same.

So my thought is that there must some combination of analog gain and digital gain that puts the DacMagic XS in its worst case scenario that I fortuitously stumbled upon early on. There is no easy way to match the analog volume as I suspect there are at least 128 steps for volume. It is possible that the ESS Sabre DAC implementation has an issue at whatever spec I am dealing with and going higher to lower digitally avoids that problem.

It is also possible that Windows was applying some sort of correction although it shouldn’t. The DACs were all configured to disable all enhancements. (But the built in Realtek sound card may apply changed).

In the first round, I manually selected the output device and in the second batch of tests, I used “Default Sound Output” since it was easier to electronically switch the output device in Windows. Perhaps this also reset some accidental setting in windows.

So in summary
1) I subjectively hear differences for weeks.
2) I can measure those differences.
3) I try to fiddle with the settings to maximize the differences.
4) It results in identical measurements.
5) And the subjective difference also disappears.

 

[restorer-john]

If you level match at 1KHz (pretty much a standard in audio) I think most of variations will go away.

 

[GXAlan]

If you level match at 1KHz (pretty much a standard in audio) I think most of variations will go away.

That’s the thing. In the first set of measurements, that would mean a 10dB boost which in turn would be boosting the bass that much.

After fiddling with the analog amp portion of the DacMagic, either by maxing the digital out and attenuating the volume or vice versa, I actually got the two devices to measure identically for the FR, broadly.

But a constellation of factors, can lead to some scenario where the DacMagic is dramatically altering the FR. I forgot to mention that this was a factory refurb from Cambridge Audio direct.

 

[Tim Link]

Over a year later I find this thread. Nice to see that you were able to correlate perception with measurement in a meaningful way. I suspect this sort of thing is going on a lot in the world of hi-end dac and other electronic device preferences. Impedance mismatches, overdriven or underdriven devices. That's why I think what you did is the right thing - measure at the speaker! That way we take into account all the possible interactions that might be happening between components and see the end result, which is what we are listening to. I'm having a similar situation pop up with my dacs in my Behringer UltraDrive digital crossover. I haven't measured anything yet but I stumbled upon an improved sense of dynamics if I digitally attenuate the input to the dacs by 6dB. I'm not sure what's going on there but I bypassed the op amps and run the dac chip outputs straight into the pre-amp. That ain't the right way to do it, but it sounded pretty good and solved the problem of the pro level output overdriving my consumer pre-amp. I'm thinking the dac chip might not be getting a good enough impedance load from the pre-amp, so reducing the output of the dac and turning up the pre-amp volume ends up sounding punchier, more dynamic. That's kind of funny because I could have just left the output stages alone and digitally attenuated, but I was so afraid of losing that precious bit depth! It's got me thinking I should get a MiniDsp 4x10 so at least I know the dac output stages are properly implemented.

 

[clearnfc]

Yes, i do agree that within hardware limitations it is possible to make adjustments to make different setups sound almost identical. Although, i have to say its very difficult (often impossible) to get different brands of speakers to sound the same.

However, i also want to point out that sometimes, its these differences that make pple prefer 1 thing over another.

You might have come across a thread in this forum talking about tone controls in amps. So, some pple have their own preferences. Its also true for recordings, some folks complained that some of the new recordings or streams killed the dynamic range.

 

[Tim Link]

Yes, i do agree that within hardware limitations it is possible to make adjustments to make different setups sound almost identical. Although, i have to say its very difficult (often impossible) to get different brands of speakers to sound the same.

However, i also want to point out that sometimes, its these differences that make pple prefer 1 thing over another.

You might have come across a thread in this forum talking about tone controls in amps. So, some pple have their own preferences. Its also true for recordings, some folks complained that some of the new recordings or streams killed the dynamic range.

Speakers have many parameters that are physical in nature so there's no way to electronically adjust the sound to be exactly the same when measuring the speaker from all directions. It may be possible at near field listening in a large room to make the on-axis sound very similar if neither speaker is pushed into distortion.

 

[clearnfc]

Speakers have many parameters that are physical in nature so there's no way to electronically adjust the sound to be exactly the same when measuring the speaker from all directions. It may be possible at near field listening in a large room to make the on-axis sound very similar if neither speaker is pushed into distortion.

Yes, i fully agree, esp. for the crossover!