General debate thread about audio measurements

[Calexico]

Where do you see a sine and harmonics ?
I see noise bands about 20khz wide around a 353kHz band (8x oversampling) and a smaller one at 16x oversampling.

How loud do you think your tweeter will reproduce frequencies around 350kHz ?
Will the tweeters react at all ?
Will they produce IM ?

Depending on how much your amplifier distorts and produces IM which, when it were to fold back, will again be lower in amplitude, certainly when using tube amps, will it 'harm' the sound ?

If those ultrasonic noise measurements test can validate the design of the analog low pass filter and shows if it's calculated for high res or dsd or cd quality that's enough to justify it.

 

[gene_stl]

That means it's better to stay cd quality.

I have not yet listed to enough high res music to disagree with the above statement but I will say that even though CD audio is considered to be just below the limit of "high rez" is certainly can be darn good. But it seems high rez ought to be better. Enough so that I am spending a not so small fortune, in money and even more valuable time,, to find out.

 

[Calexico]

@gene_stl that's why i ask for a test to validate that dacs perform as good in high res as in cd quality and that they do it thru the whole frequency range offered by high res

If tests show no better measurements with high res samplerate then high res is pointless

Before asking the format we should ask if the dac exploit it all

 

[solderdude]

If those ultrasonic noise measurements test can validate the design of the analog low pass filter and shows if it's calculated for high res or dsd or cd quality that's enough to justify it.

You still don't get why the analog post filter is there and that said filter is never laid out for CD bandwidth.
You really need to understand HOW it works and not mix up the various filters.
If it weren't designed for the highest sample rate it would drop off considerable at 50kHz which it doesn't. It suppresses the higher frequencies so does its job. Regardless if you understand or not.

 

[solderdude]

@gene_stl that's why i ask for a test to validate that dacs perform as good in high res as in cd quality and that they do it thru the whole frequency range offered by high res

If tests show no better measurements with high res samplerate then high res is pointless

Before asking the format we should ask if the dac exploit it all

You show VERY poor understanding of DAC's and their frequency response and how that relates to sample frequencies.
VERY poor... you really need to learn about it.
Measurements often show 44.1 response only because there is where a DAC performs the poorest. It only improves for higher sample rates.
As most music is still 44.1 it makes sense to show those results.
One can see from impulse response how much 'garbage' can be expected above Nyquist.
The fact that you can't doesn't mean others can't either.
There really is NO DAC at all that supports a bitrate above 48kHz that does not extend well above 22kHz.
You just think those exist and MUST be measured so YOU can be sure.

 

[Calexico]

You show VERY poor understanding of DAC's and their frequency response and how that relates to sample frequencies.
VERY poor... you really need to learn about it.
Measurements often show 44.1 response only because there is where a DAC performs the poorest. It only improves for higher sample rates.
As most music is still 44.1 it makes sense to show those results.
One can see from impulse response how much 'garbage' can be expected above Nyquist.
The fact that you can't doesn't mean others can't either.
There really is NO DAC at all that supports a bitrate above 48kHz that does not extend well above 22kHz.
You just think those exist and MUST be measured so YOU can be sure.

Delta sigma dac have more noise at high frequencies that's why i d like to see how they perform with high res

 

[Calexico]

You show VERY poor understanding of DAC's and their frequency response and how that relates to sample frequencies.
VERY poor... you really need to learn about it.
Measurements often show 44.1 response only because there is where a DAC performs the poorest. It only improves for higher sample rates.
As most music is still 44.1 it makes sense to show those results.
One can see from impulse response how much 'garbage' can be expected above Nyquist.
The fact that you can't doesn't mean others can't either.
There really is NO DAC at all that supports a bitrate above 48kHz that does not extend well above 22kHz.
You just think those exist and MUST be measured so YOU can be sure.

That's not your greatest post. It's only confusing without any argument.
Picture attached your see that the digital filter at different samplerates has no the same cut off freq
On the datasheet they explain how to design the output analog low pass filter.
Then we see that for dsd they recommend other values for filtering
Nothing show that designers respects those values and if they chose to optimize the lpf for dsd or not .

 

[LTig]

That's not your greatest post. It's only confusing without any argument.
Picture attached your see that the digital filter at different samplerates has no the same cut off freq

Of course not, and this is what @solderdude meant when he wrote "There really is NO DAC at all that supports a bitrate above 48kHz that does not extend well above 22kHz."

On the datasheet they explain how to design the output analog low pass filter.
Then we see that for dsd they recommend other values for filtering

For DSD 64 I would use a low pass filter a little above 20 kHz, to get rid of the ultrasonic noise, 'cause it starts to rise just above 20 kHz AFAIK.

Nothing show that designers respects those values and if they chose to optimize the lpf for dsd or not .

That's another story.

 

[solderdude]

T
On the datasheet they explain how to design the output analog low pass filter.

You are confusing different filters and seem to assume that the steep filters and the extra LPF are the same. They are not. The 'steep' filters are digital ones used in the upsampling process. The analog post filters are not steep at all and are there to remove HF garbage that still comes out BUT is at a very low level already.
The different filter settings are for DSD types which have lots more (essential) noise and can be made to 'better fit' a specific use case.
In DAC's where multiple formats are used (DSD and or PCM) the filter is usually set at around 80kHz.
That does not mean that the 'steep' filters aren't also there. They are. And user selectable.
When the owner makes a poor decision to select a filter by ear or by 'belief' that is not one that complies to the theorem but is supplied by the manufacturer because consumers want/ask/demand it then you can have serious garbage.
Choose the right type of filters and you are free (enough) of US garbage as that will be removed by the digital filter.
The analog post filter subsequently lowers the remaining noise (with a NOT steep filter) a bit more.
That filter is NOT there to remove what you think it removes nor does it shift/vary when different sample rates are selected.

This will be so with all wide sample rate DAC's and is well documented here and there. Amir does not feel the need to measure that too just to please your curiosity. Despite you commanding/telling him he MUST do this because it explains a weird theory you seem to adhere on based on incorrect 'testing methods'.

 

[Calexico]

If the nalo

You are confusing different filters and seem to assume that the steep filters and the extra LPF are the same. They are not. The 'steep' filters are digital ones used in the upsampling process. The analog post filters are not steep at all and are there to remove HF garbage that still comes out BUT is at a very low level already.
The different filter settings are for DSD types which have lots more (essential) noise and can be made to 'better fit' a specific use case.
In DAC's where multiple formats are used (DSD and or PCM) the filter is usually set at around 80kHz.
That does not mean that the 'steep' filters aren't also there. They are. And user selectable.
When the owner makes a poor decision to select a filter by ear or by 'belief' that is not one that complies to the theorem but is supplied by the manufacturer because consumers want/ask/demand it then you can have serious garbage.
Choose the right type of filters and you are free (enough) of US garbage as that will be removed by the digital filter.
The analog post filter subsequently lowers the remaining noise (with a NOT steep filter) a bit more.
That filter is NOT there to remove what you think it removes nor does it shift/vary when different sample rates are selected.

This will be so with all wide sample rate DAC's and is well documented here and there. Amir does not feel the need to measure that too just to please your curiosity. Despite you commanding/telling him he MUST do this because it explains a weird theory you seem to adhere on based on incorrect 'testing methods'.

If you listen or convert to dsd it's better to know if the dac analog lpf is designed for dsd.

 

[KSTR]

You are confusing different filters and seem to assume that the steep filters and the extra LPF are the same. They are not. The 'steep' filters are digital ones used in the upsampling process. The analog post filters are not steep at all and are there to remove HF garbage that still comes out BUT is at a very low level already

It is worth to note all those digital filters do *not* filter the out-of-band stuff, that is, they don't filter the intrinsic mirror frequencies of the zero-order hold function any DAC has by design (no other way). They filter the *input* signal to the DAC core after the basic upsampling process to remove the digital images from that process, not its output signal.
That means the analog images have the level as predicted by the sinc envelope (see reference I gave above) which is rather high. The upsampling process helps to place these images much higher in frequency than without and then the analog post filter requirements are much more relaxed.
Sidenote: With D/S PCM DACs we also have modulation noise (and with DSD as well, as this is just the recorded D/S-modulator output of the D/S ADC used to record the signal) which also only sees the attenuation of the analog post filter (which may partly reside in the chip itself, the voltage-mode AKM chips being an example).

 

[solderdude]

If the nalo
If you listen or convert to dsd it's better to know if the dac analog lpf is designed for dsd.

The analog post filter for a DAC that has multiple DSD rates will always be fixed for the highest DSD rate which means that DSD64 will always show higher noise just above the audible band compared to higher DSD rates.

You are expecting too much 'filter effect' from that extra low pass.

How much garbage there is above the pass band depends on the type of (user) selected filters for PCM. When one chooses 'slow filters' one can be certain there are unwanted signals above Nyquist, whether or not they MAY become audible depends on multiple factors outside of the DAC.
When using steep filters that comply to the theorem then there is little to worry about.

 

[Calexico]

If steep filter is a

The analog post filter for a DAC that has multiple DSD rates will always be fixed for the highest DSD rate which means that DSD64 will always show higher noise just above the audible band compared to higher DSD rates.

You are expecting too much 'filter effect' from that extra low pass.

How much garbage there is above the pass band depends on the type of (user) selected filters for PCM. When one chooses 'slow filters' one can be certain there are unwanted signals above Nyquist, whether or not they MAY become audible depends on multiple factors outside of the DAC.
When using steep filters that comply to the theorem then there is little to worry about.

If steep filter is at 96khz for a 192khz samplerate we must check thd + noise vs freq until 96khz to see if there is no ultrasonic noise ath those high frequencies. (Not due to imaging but to poorer performance for high freq by the reconstruction filter)

 

[solderdude]

why ?

You have already seen plots with BW of 100kHz ... seen any noise of concern in 192kHz files ?

Why would 90kHz be any 'worse' in a 192kHz file compared to 20kHz on a 44.1kHz file ?
Same filter, different frequency.
Do you intend to listen to 90kHz ?
If so with what speaker, what recording and what do you think you will be perceiving/hearing ?

Are you using digital amps ?
The ones with less well designed input filters ?
In that case just use 44.1 or max 96kHz files with steep filters and you are good to go.

 

[Calexico]

Measurements are meant to validate it. If no measurements i don't know the answer.
That's Why i ask for them.

 

[Blumlein 88]

Measurements are meant to validate it. If no measurements i don't know the answer.
That's Why i ask for them.

What are you asking for? And by this I mean what tests and what results of those tests would convince you? If you can't be convinced we are wasting our time. If you can be convinced, then maybe we can show you the way.

 

[Duke650]

 

[Calexico]

@Blumlein 88
Tests i ask are very clear:
For 192khz samplerate
- thd vs freq until cut off freq of digital filter (around 96khz)
- white noise test until freq as high as possible to see where is the ultrasonic garbage.

Why not making same test with 768khz samplerate to see if it's good or not to convert to this format.

No test do that then no test validate that reading high res on a dac performs good.

I don't ask to be convinced i just ask tests to validate things that are not validated.

 

[Blumlein 88]

@Blumlein 88
Tests i ask are very clear:
For 192khz samplerate
- thd vs freq until cut off freq of digital filter (around 96khz)
- white noise test until freq as high as possible to see where is the ultrasonic garbage.

Why not making same test with 768khz samplerate to see if it's good or not to convert to this format.

No test do that then no test validate that reading high res on a dac performs good.

I don't ask to be convinced i just ask tests to validate things that are not validated.

I'm not home so don't have the files handy. Thd is fine up there. Maybe .005 becomes .05 with some gear. Noise varies more but isn't an issue. As for 768 don't have the info and don't care.

 

[solderdude]

Why not making same test with 768khz samplerate to see if it's good or not to convert to this format.

Why would anyone want to convert to 768kHz ? Those files are HUGE and don't add any info.
Do you know many consumer DAC/ADC's that support 768kHz ?
Convert from what .. 44.1, 96, 192 or 384 ? upsample ? What would be the benefit ?
Why record at 768kHz when there are no microphones nor speakers that can reach even 100kHz.
Even when you are convinced you can 'detect' 100kHz (which isn't recorded anyway nor audible) why would one want to have a 300kHz audio bandwidth ?

What's your fascination with 'noise' outside the audible band ?
Do you own poorly designed digital power amplifiers ?