I think the HM-602 has high frequency roll-off in the software or firmware.
In post #35 I said it does not have a filter, alright, perhaps it has a low-pass filter.
It is not upsampling, it's a non-upsampling device, hence the 1 kHz sine.
The 1 kHz sine still looks like a blocky staircase, which is the issue here.
I always thought the ultrasonic content in NOS DAC's was completely inaudible, are you guys saying it actually folds down and severely messes up the normal 20 - 20 area?
So when I upsample 4x, I move all the ultrasonic content above 176 kHz, thus improving the 20 - 20 sound?
In post #35 I said it does not have a filter, alright, perhaps it has a low-pass filter.
It is not upsampling, it's a non-upsampling device, hence the 1 kHz sine.
The 1 kHz sine still looks like a blocky staircase, which is the issue here.
I always thought the ultrasonic content in NOS DAC's was completely inaudible, are you guys saying it actually folds down and severely messes up the normal 20 - 20 area?
So when I upsample 4x, I move all the ultrasonic content above 176 kHz, thus improving the 20 - 20 sound?
Last edited:
Here is the frequency response of the device measured, no UHF content, I think?
Look at the cyan line, HM-602.
http://sonove.angry.jp/RMAA/iPodClassic6_HP-P1_HM-602_x1060.htm
You can't see anything above 20kHz with a soundcard at 44.1 or 48kHz sampling, so you don't know what is or is not present above 20k.
Even soundcards that sample at 96 or 192 often do not show anything above audio.
The QA400 for example can sample at 192 but the audio bw is around 45kHz.
Jan
Of course it is there because it is produced by 48kHz DAC! What is your point?
Surely you are not suggesting that the 'blocky' view produces 'blocky sound'?
Jan
Most of that 'blockiness' would disappear at a higher sampling rate. OTOH, increasing the bit depth without changing the sampling rate would make a negligible difference. In that picture, there's only about 44 samples from left to right, but the vertical resolution is in the 10s of thousands.
BTW, I'll disagree with Jan and say that the steps are centered around 44KHz.
O.k., hold on......
------ question ------
Let's say we are using a non-upsampling 24-bit R2R DAC, with all the IMD and ultrasonic content removed, in any way you like, it's just not there.
I play a 24-bit / 48 kHz music file.
After that, I play the same file reduced to 16-bit / 48 kHz.
------ question ------
There I edited it, is the question acceptable now?
Yes, upscaling SD to HD video has more detail, upscaling images show more detail, upscaling colour bit-depth has more detail, upsampling audio does not have more detail. Is that your thesis?
Last edited:
This is NOT the output of a 16 soundcard. A 16 bit DAC output has an LSB of less than 0.15 Millivolt at 10V Peak-Peak.
Anyway a working reconstruction filter would filter the steps out...
The steps in the picture seem to be at least 7 millivolt high....
My tipp: Resonances by a misused output opamp (cable too long?)...
Bye some quality audio analyzer first.
Anyway a working reconstruction filter would filter the steps out...
The steps in the picture seem to be at least 7 millivolt high....
My tipp: Resonances by a misused output opamp (cable too long?)...
Bye some quality audio analyzer first.
Last edited:
Ok so this is all very interesting but I have to clarify the initial part of this thread.
Audio bit-depth is limited to 22-bit in ideal conditions, if we include the noise floor.
If we don't include the noise floor, what is the limit of our bit-depth hearing?
I used the R2R DAC as an example, it didn't work very well!
How about this!
There is currently a discrete 24-bit R2R DAC project on diyaudio made in Denmark, it includes upsampling.
If human hearing is limited to 22-bit, then why would a 24-bit R2R DAC need upsampling?
Another example, Neutron player for Android, it has 32-bit rendering. Why?
Audio bit-depth is limited to 22-bit in ideal conditions, if we include the noise floor.
If we don't include the noise floor, what is the limit of our bit-depth hearing?
I used the R2R DAC as an example, it didn't work very well!
How about this!
There is currently a discrete 24-bit R2R DAC project on diyaudio made in Denmark, it includes upsampling.
If human hearing is limited to 22-bit, then why would a 24-bit R2R DAC need upsampling?
Another example, Neutron player for Android, it has 32-bit rendering. Why?
Last edited:
O.k., so when I 4x upsample my desktop NOS DAC, it removes the blockiness, via shifting the ultrasonic content into the area above 176.4 kHz, this is why you suppose it sounds better, with upsampling?
If I play a 24-bit / 48 kHz song and then the same song in 16-bit / 48 kHz on a 24-bit R2R, it shouldn't make much difference?
I don't have a 24-bit R2R so I can't perform that experiment yet.
I see the commenting in this thread has slowed down, I think I'll come back tomorrow, thank you for the conversation.
Last edited:
My guess is that some people under some conditions would just be able to hear a difference, as the 16-bit is undithered.Kastor L said:
Do you mean oversample? It helps with filter design. A better reconstruction filter means a closer copy of the original signal.
Something you don't seem to have grasped is that if you take a sampled digitised dataset (whether audio or visual) and then increase the bit depth you have added no new detail from the original analogue signal. You may have added spurious detail, which may or may not mimic some of the detail omitted by the original sampling or digitisation.
The same thing is true of oversampling - going to a higher sampling rate (or more pixels for a picture). No new genuine detail, however much you might perceive it.
So if you start from 16-bit 44.1kHz then the best you can do is produce an accurate rendering of the information contained in that 16-bit 44.1kHz data stream. Converting to 24-bit 96kHz adds no new genuine information, but it may add some spurious information. It may also make it easier to produce an accurate reconstruction filter.
My understanding is that dithered 16-bit 44.1kHz material is indistinguishable from hi-res formats when peaking is not allowed. In fact, some 'hi-res recordings' are actually just processed 'lo-res recordings' - they contain no new information. This cannot be heard, but signal statistics and frequency response can be a giveaway.
Problem with your ears?
Ohh wait, we leave the realm of technical discussion now and go into personal, unverified, unreliable, psycho-acoustics-related, ad-hoc opinion now. Sorry, I take that back.
Jan
Last edited:
Where does this picture with the blocky sine wave come from?
Seems to me like your soundcard or HP-output is crappy or defect....
A decent 16 bit sine wave has a THD+N greater than 80 dB.
You cannot see any visibible blocks.....
Thank you.
Hmm.
The 24-bit file will have a slightly cleaner looking 1 kHz sine, which in turn is closer to a pure sine.
Does this mean the 24-bit file will result in lower THD?
Do all the digital to analogue systems reproduce equal voltage steps for each bit that is in the signal range.
eg.
a 16bit signal has 2^16 steps = 65536 steps.
The LSB to +1 bit for a 3Vpk signal would be from 0Vpk to 0.0458mVpk
at the MSB end the step from MSB to -1bit would be 3Vpk to 2999.9542mVpk
Does the same apply when working with lossy compressed signals? Is the LSB step the same all the way through the bit range?
eg.
a 16bit signal has 2^16 steps = 65536 steps.
The LSB to +1 bit for a 3Vpk signal would be from 0Vpk to 0.0458mVpk
at the MSB end the step from MSB to -1bit would be 3Vpk to 2999.9542mVpk
Does the same apply when working with lossy compressed signals? Is the LSB step the same all the way through the bit range?
Last edited:
http://sonove.angry.jp
I have heard that device he measured, so it's a useful measurement to me.
Here is another, I just looked it up now
Should I NOS-DAC ?
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
That looks like ~ 50steps judging by the size of the steps at the peaks.
6bit (2^6) is 64 steps.
That signal looks like approximately 6bit, not 16 bit.
If that signal was only using the 6bits at the LSB end of the signal capability, it would represent a signal of ~2mVac to 3mVac.
6bit (2^6) is 64 steps.
That signal looks like approximately 6bit, not 16 bit.
If that signal was only using the 6bits at the LSB end of the signal capability, it would represent a signal of ~2mVac to 3mVac.
Last edited:
@Kastor Have a view of these two informative videos on digital audio for your answers: Xiph.org: Video This information puts to rest the myth of stair cased or blocky sine waves.
From a practical perspective, if you want to listen for yourself to see at what bit depth you can or cannot hear at, try: Computer Audiophile - Fun With Digital Audio – Bit Perfect Audibility Testing Here you can download audio files at various bit-depth's and hear first hand where your audibility limit is.
More audio files to listen to and compare: Artifact Audibility Comparisons
From a practical perspective, if you want to listen for yourself to see at what bit depth you can or cannot hear at, try: Computer Audiophile - Fun With Digital Audio – Bit Perfect Audibility Testing Here you can download audio files at various bit-depth's and hear first hand where your audibility limit is.
More audio files to listen to and compare: Artifact Audibility Comparisons
Andrew what you see there is the sample rate which appears 44.1 kHz. You can't see the bit size as that would be encoded in the step amplitudes and the signal level in terms of a fraction of the full scale amplitude.
Jan
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.