Digital, but not by the numbers

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The differences have always been trivially obvious to me, even when listening over mediocre PC speakers. The same track, resampled to higher and higher resolutions always steadily improved, and this is specifically in the region of high frequency accuracy. The fine shimmer of a ride cymbal is an easy marker -- it starts off sounding like some vague hissing noise at lo res, and at the other end of the spectrum becomes a proper instrument ...

Frank
 
Curious to hear what you think is the cause of the improvement, considering the PC speakers probably don't go above 15 kHz...
Agree about the frequency thing, to me it's all about the ability of the usually nondescript S-D D/A to do its job properly, in the face of heavy interference from surrounding electronics. With a lo res version of the track, so much on the fly processing has to be done, on the PC side and then in the chip itself, that the quality of the D/A conversion suffers. With a high res signal being fed in, minimal extra processing is required and quality improves.

I've done this exercise several times, and provided the chip and speaker electronics were thoroughly "warmed up" it was easy to pick the variation.

Frank
 
I can always pick a difference with regard to a 44.1khz track played back at different sample rates. And with regard to higher frequencies, I tend to agree that they sound better at higher rates. But where I disagree is that overall the sound is "better".
I find that up-sampling is a trade-off. I have found that ( when listening to the same track) I perceive bass / mid tonality and dynamics to sound better at 44.1 or 48khz.
In other words, with up-sampling highs sound better, but everything else sounds worse!

But I do find that a native 96khz recording played back at 96khz can sound perfectly fine.
 
But where I disagree is that overall the sound is "better".
I find that up-sampling is a trade-off. I have found that ( when listening to the same track) I perceive bass / mid tonality and dynamics to sound better at 44.1 or 48khz.
In other words, with up-sampling highs sound better, but everything else sounds worse!

But I do find that a native 96khz recording played back at 96khz can sound perfectly fine.
Very interesting. At least someone else has noted that upsampling makes a difference, and that the highs, which is usually where the problems are, are improved. That the dynamics, etc, are lessened I find very surprising, doesn't make sense to me. Perhaps different ears, different equipment, different resampling software -- I use Audacity to do the job.

Have you tried downsampling that native 96kHz to 44.1, then back up again to 96, and comparing original to new?

Frank
 
It does not make much sense but this is what I heard.

well, I think we have actually been talking about different things.
I think when you say upsampling you are talking about converting a file to a different sample rate, then saving the file, then playing it back.

I have been talking about using the player, either foobar or cPlay to playback a file in real time, at a different sample rate.

I think your way might add dithering, I think my way does not.
Can we discuss?
 
Listener fatigue is the unknown factor, in listening tests.
This is why I prefer AB (two points of comparison) tests rather than ABCDEFG tests. Anyone can become muddled when listening to too many tracks, or equipment. But it is usually easy to declare a preference when comparing only one thing to another.



And this is why I am taking a whole week to analyze and verify mods done to my tda1543 russian dac.
 
Last edited:
On second thought cPlay uses selectable SOX or SRC to upsample.
And now I remember that I did prefer one over the other, but that both of them created a weird sound that irritated me, and so now I just play back at either 44.1 or 48khz. Even though this has a downside, I find this one the less irritating than the other.

Neither way is correct... back to the drawing board....
:)
 
Very interesting. At least someone else has noted that upsampling makes a difference, and that the highs, which is usually where the problems are, are improved. That the dynamics, etc, are lessened I find very surprising, doesn't make sense to me. Perhaps different ears, different equipment, different resampling software -- I use Audacity to do the job.

Have you tried downsampling that native 96kHz to 44.1, then back up again to 96, and comparing original to new?

Frank

Ive also noticed a difference when using my sony's dac and 48kHz instead of 44.1kHz

And yes, the highs are less harsh and I prefer this sound on the sony's mystery dac when its being run at 48kHz, but I always assumed that there was some kind of DSP screwing around with the sound to make some kind of profile for it at different modes.

But since building the tda1543 with a cs8412 I haven't noticed much if any change between 44.1 and 48kHz, it might need more time for me to listen in.
 
Last edited:
well, I think we have actually been talking about different things.
I think when you say upsampling you are talking about converting a file to a different sample rate, then saving the file, then playing it back.

I have been talking about using the player, either foobar or cPlay to playback a file in real time, at a different sample rate.

I think your way might add dithering, I think my way does not.
Can we discuss?
Indeed we have. The crucial thing is that any conversion is done offline, otherwise the processing of the player software also joins the party, in terms of creating electrical interference.

No dithering occurs with the resampling, it's all about interpolation, working out the inbetween values. This is what happens anyway, one way or the other, when playback occurs ... my exercise was for that to happen during another time period ...

I went through a bizarre example where a 20k bytes MP3 file ended up the size of around 500Meg. And blow me down, that monstrous end result was far superior to the original, compressed version, on playback.

Frank
 
OK, well I do remember doing an upsampling conversion like this years ago.
I am confident this would be better than converting on-the-fly with the playback software. I'll try it again sometime soon.

What is the deal with this MP3 you are talking about?
what did you convert the file to?
I think it started as a 128k bit rate, from a download. Then exported as 16/44.1, was much better, then converted to 16/88.2, better again, got up to 24/352.8. Even tried next higher again, but all the players choked on that !!

Frank
 
I can always pick a difference with regard to a 44.1khz track played back at different sample rates. And with regard to higher frequencies, I tend to agree that they sound better at higher rates. But where I disagree is that overall the sound is "better".
I find that up-sampling is a trade-off. I have found that ( when listening to the same track) I perceive bass / mid tonality and dynamics to sound better at 44.1 or 48khz.
In other words, with up-sampling highs sound better, but everything else sounds worse!

But I do find that a native 96khz recording played back at 96khz can sound perfectly fine.

Likewise !
 
But, even if it sounds better I'll be buggered if I'm converting all my files...
Yeah, the file storage alone will kill you!

However, this is all about proving a point. That is, it's all about what goes on in the processing, in the PC and the DAC, having a real lot to do with with what the final sound is like. So, if there's something real going on here, then sharper, better hardware can make even pretty crappy digital files sound pretty damn good ...

Frank
 
"digital" is getting to the point where it is "analog". Dirty secret is it always has been, digital only exists in books, even our computers are nothing more than billions of analog signals switched on or off with transistors.

With the high rez DSD the computer is doing the "digital" to "analog" conversion. Won't be long and all we will need is an analog stage, thats how fast the "digital" signal is getting.

There was a rudimentary design where a guy fed a straight DSD signal to a tube stage no DAC and said he got decent sound. The new NAD-51 is basically doing this, but all the processing it does in "digital" can be moved to our computers I think Signalyst can do this. The speeds probably aren't fast enough for top end output yet, but look how fast SATAIII is, won't be long someone will plug in an analog stage to a computer and the DAC will be extinct.

Then there will always be the other camp where signal speed is the enemy (see building the ultimate NOS TDA1541 thread).

But no matter what you do unless you like spinning disc's there will always be some hashy high speed device connected to your audio chain.
 
"digital" is getting to the point where it is "analog". Dirty secret is it always has been, digital only exists in books, even our computers are nothing more than billions of analog signals switched on or off with transistors.

OK, I know this is elementary stuff that we all know, but...

Nothing new or "dirty" about it. The switching, storage etc, is "analog" on the fundamental physical level, but the beauty of digital is that the "analog" signal is recreated and resynchronized so that errors don't get propagated (and don't matter).

In a digital signal, the only thing that matter is the *logical* (as opposed to physical) data. As long as the error/noise/distortion is small enough not to cause a "0" or "1" to be misinterpreted, all that matters is delivering them with the correct timing to the DAC.

With the high rez DSD the computer is doing the "digital" to "analog" conversion.

Not sure what you are saying here. DSD is pulse-density encoding. It is just another way of encoding, that trades bit depth for bit timing.

There was a rudimentary design where a guy fed a straight DSD signal to a tube stage no DAC and said he got decent sound.

And my first DAC, 35 years ago, was a R-2R PCM converter that consisted of nothing more than a handful of resistors connected to the outputs of a parallel port. Industrial embedded applications are full of pulse width modulation control systems that do the DAC conversion simply by proper timing of a (digital) output pin, just like your "DSD signal straight to a tube stage". The missing part in your example is the low-pass filter required for a successful pulse density conversion - a tube amp is probably good for that purpose. :)

all the processing it does in "digital" can be moved to our computers I think Signalyst can do this.

The only processing that needs to happen is the low-pass filtering - and some products leave that to your amp and speakers.

The speeds probably aren't fast enough for top end output yet, but look how fast SATAIII is, won't be long someone will plug in an analog stage to a computer and the DAC will be extinct.

Even basic SATA I is more than fast enough.

And nothing new under the sun... There was the MUTRAN program/language that ran on an IBM 1620 in the 60's that produced music by precisely timed program loops that caused enough RF noise that you could pick it up on an AM radio placed close to the mainframe computer. Was that perhaps the first wireless computer audio? No DAC to be seen anywhere.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.