Wombat,
Thanks, I'm relatively new to this process. I do notice that using SOX to upsample to 176 kHz, with the parameters that I specified sounds very nice on my oppo-105. I'm using their minimum phase, high quality, bandwidth somewhat limited method. In the link I supplied, it is compared to the Meridian phase response.
Perhaps you could recommend a simple way to achieve even better results. In particular, is it possible to achieve results even closer to those achieved by Ayre and or Meridian? I'm interested in upsampling from 44 kHz to 176 kHz.
Sorry, can't help you with that because i don't know what Meridian or Ayre exactly archive besides having a patent to do some maketing with.
I am that kind of person that trusts in the DAC designer and i am not trying to voice sound with changing the input signal to much, especialy with upsamling.
Since you believe in it just enjoy to play around.
I am that kind of person that trusts in the DAC designer and i am not trying to voice sound with changing the input signal to much, especialy with upsamling.
Since you believe in it just enjoy to play around.
Even if you start with 16 bits?
"Obviously" is one of those words that usually mean the opposite.
Yes, you definitely need more precision temporarily when processing, but if it is 16 bits in, why do you need more than 16 bits out? Upsampling doesn't magically increase the SNR and dynamic range of the signal.
It may not increase anything on (your) paper, but it actually does in fact, perceptual. The improvements are obvious on an upsampled material, compared to a standard one.
We have now two different methods of upsampling. There are reported improvements in using both these methods.
The appropriate approach it may be in these cases to try to find explanations of these perceptual improvements, rather than only deny it. The theories (theoretical approaches) it may be a good thing when it supports the facts. When facts are denied to the sake of theories, then we face something else...
By the way, I have seen your plots (presentation) in another thread. It looks to me that these it were realized based on simulations. It could be nicer to have some real scope plots of real measurements of the stated behaviours...
As in "in the mind of the listener", yes, probably.
The plotting paper might be mine, but seems pretty much anyone who has actually studied digital signal processing agrees with the view that you can't recreate the extra bits that weren't there in the first case - the dynamic range and SNR is limited by the source resolution (16 bits). Not much you can do to change that afterwards. Yes, you can dither, distort and colour the signal to sound more pleasant, but it won't give a greater dynamic range or SNR.
Any ABX logs you'd be willing to share with us?
Not denying them at all. There are very simple explanations to those improvements. The key word is "perceptual". As in "perceived improvements".
Nobody is denying any facts here, as far as I can tell.
Feel free to do the simple simulations using an audio gererator and a scope. Which of the plots don't you think corresponds to a real situation?
The very straighforward explanation is that the DAC and following analogue circuitry behave audibly differently when presented with different types of data streams. The concept of 'digital' at some point has to cross over into the analogue world, and at this point all bets are off as regards varying distortion artifacts entering the equation ...
In effect, we are doing some of digital processing in advance on a PC therefore we need to keep the higher precision as we transfer information to our OPPO player at 176 kHz.
I will not say that your plots and the logic behind your presentation are not right. And I do not pretend at all that upsampling techniques mean introducing of extra (quality) bits.
There is not in discussion here the eventual effects applied to the original file. This is another discussion. Here is about only upsampling process, without no any other intervention on the original file. To be honest, I do not think that "colouring" the signal by some software tricks it bring some "pleasant" improvements... Just the opposite in my opinion.
The upsampling discussion is quite old on these forums, and the controversy is not ended yet about this subject (as we can observe in this thread...).
The strong argument of the theoreticians is that one do not introduce more informations to improve the quality of the resulting analogue signal, when process the files from 16 bit to 24. And this is right.
The answers of those who experienced improvements is that in such processing it may happen something with the smoothing of the resulting analogue signal by dividing that one (each) bit (from the 16 ones) in to 2 or multiple sample (with the same bit value), when increasing accordingly the sampling frequency. There is of course no any extra bit in all this process. There is the original analogue information coded in to 16 bit (44,1Khz), but that original information is now transcoded into 24 bit and to say, 174,6Khz sampling. There may be there some empty bits, or it may be some bits which repeat the same informations...
It is like when one see an 800 x 600 pixels image on a 800 x 600 display, and then see the same pixel dimensioned image on a HD display. One will of course not see more pixels of the original 800 x 600 pixels picture on the HD display, but the one will perceive an improvement.
So, while the theoretical stage is quite in place and it have some logical explanations, the practical mechanism it may not be very well explained yet. And it may differ from one hardware to another, as maybe some particular parameters, function the chip (with some unknown functions/functionalities inside)or components involved.
The upsampling process give a better perception of the instruments in a sound stage. This is quite obvious. It occur an increasing of the perceived definition of the sounds in the same sound stage. How is this possible without more added informations to the original digital file?
One can not hear more distinct the instruments in a sound stage, with increasing position precision, without heaving more sonic informations about that sound stage. Where it come that information from, if nothing change in the initial 16 bit coded file?
But here it may be mentioned (quite important) that the original recorded material it may not be at 16 bit, but at 24, 32, or even 64 bit. It may be mentioned also that the original analogue material it may not be processed at 44,1Khz, but at much higher sampling frequency, and then transcoded to 16bit/44,1 Khz, because the final physical media reasons...
So, it seems to be a little bit more to be explained here, than only showing some figures on a paper, or some simulated plots, or so...
Last edited:
You do the processing on the PC. For that specific processing (the sample rate conversion and the reinterpolation) you need temporarily added precision/headroom. Once the operation is done, and you transfer the data to the player, you don't need the headroom, as the transfer doesn't involve any non-linear processing.
Good.
I am glad we agree.
A 800 x 600 pixel image linearily zoomed onto a HD display looks just as coarse as on the original display. A 800 x 600 pixel image intelligently interpolated to HD resolution looks better.
Your brain?
But even if you had recorded the material in 2 MHz, 128 bit resolution, once you truncate and downsample it to 16 bits @ 44.1 kHz, it is 16 bits 44.1 kHz. The information is lost. Forever. There is nowhere for it to hide (maybe in the cracks between the bytes?).
I agree also that a DAC system it behave different when it is to process different types of data streaming... Here are quite many factors involved about the DAC chip itself construction, internal setup, or internal software logic/processing. The most of these factors remain unknown for the user (patent reasons, and so...)
Actually, most of those things are pretty well documented in the material for most DAC chips. There is very little hidden "magic sauce" there. It is just that it does take a bit of engineering knowledge to read that documentation - something that is often beyond the average DIYer.
Even brain`s appreciations are based on informations. It can not process something without informations. When it may happen, then is pathological, that is very wrong functioning... and that is another (medical) discussion...
BTW, when it happen that more different brains perceive similar in similar situations, then it may be some very solid facts involved...
If you will take f. ex. the ESS Sabre chips, then you may observe that not all the details are in the datasheet...
Nonsense. I've looked in vain for engineering detail on what's exactly inside these chips, and apart from useless abstract explanations there is nothing available. It's all about company IP, no-one's going to tell you the "important stuff" ...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.