Metrum Octave Dac - What are the Chips used

[ThorstenL]

Hi,

Originally Posted by SoNic_real_one

Do you listen the headphones at 102dB? Because I don't.

I don't listen to headphones.

The 102dB come from the peak demands of music signals and would correspond to around 85dB average SPL and equally to "0dBFS" peaks in the signal. The crest factor in this case is 17dB, which is often exceeded by truly un-manipulated recordings.

Of course much modern "popular" music is compressed to a point where the dynamic range is barely 6dB, but I personally rarely listen to that sort of stuff seriously.

At any extent, if you worry about "x" distortion at digital full scale, look at the distortion of your transducers at the corresponding SPL...

And if they do not offer less than "x" distortion then any thesis that "x amount of distortion makes high quality impossible"is instantly falsified.

Ciao T

[ThorstenL]

Hi,

Originally Posted by SoNic_real_one

Exactly... nobody listen for long term at the nominal levels.

Nor does music contain sinewaves with 0dBFS. So your point is?

Originally Posted by SoNic_real_one

At medium sound levels the distortions of good, open-back, headphones are smaller than the usual graphs for 90-100dB.

We do not know how the Metrum DAC performs at -12dBFS (equal to 90dB SPL if 0dBFS peaks are at 102dB), unless someone carries out the measurements.

Originally Posted by SoNic_real_one

And all those are harmonic distortions, not random aliases.

Do not try to change the topic, we where discussing the Metrum's measured HD, nothing else. If you want to make a separate case that the ultrasonic images constitute distortion (they don't, in the traditional sense) or that they are random (which they are not of course - that being a crucial point BTW) you are welcome to do so separatly.

Originally Posted by SoNic_real_one

A few comparative measurements:

These measurements correspond to a single frequency and AFAICT to 90dB SPL. In other words, these are at around -12dBFS if we want normalise them.

Here alternative data on headphone distortion BTW...

Headphone non-linear distortion measurements

What I can see is between -75dB/0.02% H3 (the headroom measurement at I believe 500Hz) and -40dB/1% H3 (the Headfi thread) at 90dB. No measurements for higher SPL's are readily available, but we can predict H3 levels will be much higher and easily exceed your stated -60dB/0.1% that you claim make high sound quality impossible.

Ciao T

[qusp]

I cant find the measurements of the jh13 right now, the FR plot would no doubt sets some jaws wagging as they are heavily Eq'd with the XO to produce something that is flat at the eardrum. there are some somewhere though, perhaps at Tyll's forum.
how about feeding that dac a sinewave or music with -60db distortion then..... what do the measurements look like now

[qusp]

and before anyone says anything, yes these particular headphones (I have several and speakers) roll off at 20k, you'de have a hard time finding a balanced armature monitor that goes much higher

[SoNic_real_one]

Originally Posted by qusp

how about feeding that dac a sinewave or music with -60db distortion then..... what do the measurements look like now

Exactly. If the headphones add distortion (mostly 2nd and 3rd), why don't feed them as clean as possible? Input all that garbage and sure they will make even more distortion and IM from that and on a wider range. We will have IM products made in headphones that where never there in a filtered situation.
And I don't know about the rest of you, but I have a volume pot between DAC and headphones. So while the DAC ouptut is always at odB FS, my headphones are probably at -20dB FS (can't measure now). So all those graphs of headphones distortion should be taken at that level too - I bet will be lower than at the full scale.

[qusp]

agreed, I actually meant to say feed those headphones with -60db distortion, but same difference... it compounds. not all the distortion products will be correlated, but i'm sure it will be an additive process that will sporn new spikes.

my customs would blow my eardrums at that level; I know many argue it has to be sized for the dynamic peaks at a live event, but who listens to music at that level? if you sat in the front row at events all the time youde get hearing damage pretty quick too

anyway I think we better leave these guys to it Sonic, i've made my point and as long as we're talking NOS rather than this dac, others not really interested in the dac will come in to argue as well and i'm not interested in further conflict, i'm still interested in hearing the dac, just to see what the fuss is about, but if I owned one i'd be upsampling externally

[jitter]

Originally Posted by qusp

{zip} i'm still interested in hearing the dac, just to see what the fuss is about, but if I owned one i'd be upsampling externally

IMHO, hearing it for yourself and making up your own mind is the only way to go. Chances are you'll love it, but I'm also willing to accept that maybe this DAC isn't for everyone. Hope you can audition one, around where you live...

[qusp]

Originally Posted by jitter

IMHO, hearing it for yourself and making up your own mind is the only way to go. Chances are you'll love it, but I'm also willing to accept that maybe this DAC isn't for everyone. Hope you can audition one, around where you live...

yeah it'll never work with my system (I need i2s input and my system is multichannel for digital XO duties) but i'm still interested to hear it. i'll send you a PM tomorrow as our conversation was a bit unfinished, i'm part way through a reply, but i'd rather not post in the thread and start things off all over again

[jitter]

I've been looking at the shape of the 5 kHz sinewave at 44.1 kHz fs of the Octave again, and I realized I'm not looking at IM here. These are the steps from DA-conversion without low pass filter (reconstruction filter).

The scale is 50 us/div, so full scale is 500 us. At 44.1 kHz (or ksps) that would mean 22.05 samples. Counting the number of visible steps indeed reveals 22 distinguishable jumps. That square wave like distortion would account for a lot of the harmonics.

Am I right in thinking that most of the measured distortion isn't IM related at all, but are the harmonics of this square wave like distortion of the sine?

[SoNic_real_one]

Well... yes it is a lot of harmonics due to the square shapes. It will not be IM since in that case you play just one frequency.
But as you notice, without interpolation and oversampling, the original "16 bit" are worthless at high frequency (closer of 1/2 SR). At 1/4SR you have actually only 8 levels there because of large sampling time - that is only a 3 bit resolution.

PS: This observation was the basis of sigma-delta theory anyway...