Indeed !
I would even go 1 step further... try to build psu's that can perform better then batteries (and batteries don't need regulation). I know that sounds like a paradox as a psu is not a battery and only by regulation it can be mimicked, somehow that is not required... It can be done with shunt regulation as well but here we introduce noise (regulation noise). So passive must be the answer.
Hey, that sounds familiar ..... don't we also like passive I/V convertion and filtering on the output of the dac?
Hans, interesting.
So, essentially:
1) up-sample an 1Fs rate signal to 4FS, which means it is passed through an FIR filter to remove the image-bands starting at 22KHz.
2) then, down-sample the now 4Fs rate signal back to 1FS, which mean that it is, once again, passed through an FIR filter. This time, to remove any frequencies above 22HKz which might be present, and would alias down
The results of the test would seem to show either of the following results:
A) Should the sound after return to the 1FS rate, still show the NOS character as it did before the up/down conversion, then, that particular FIR filter utilized could be considered free of producing typical OS sound artifacts?
B) Should the sound after return to the 1FS rate, NOT show the NOS character as it did before the up/down conversion, then, that particular FIR filter utilized is implicated as a subjective culprit.
Are those your expectations as well, or have I overlooked what you expect that the test may reveal?
Last edited:
Ken,
That’s exactly what I meant.
By using the same hardware in both cases, apples are compared to apples, as against comparing a NOS Dac to a OS Dac where everything differs.
As a matter of fact, depending on the internals of the used software, downsampling from 4Fs to 1Fs does not need a filter but can be achieved by removing 3 out of every 4 samples, but going from 192 to 44.1 is more complex than that.
That’s why I suggested to try both.
And don’t forget the dithering when going back to 1Fs.
Hans
That’s exactly what I meant.
By using the same hardware in both cases, apples are compared to apples, as against comparing a NOS Dac to a OS Dac where everything differs.
As a matter of fact, depending on the internals of the used software, downsampling from 4Fs to 1Fs does not need a filter but can be achieved by removing 3 out of every 4 samples, but going from 192 to 44.1 is more complex than that.
That’s why I suggested to try both.
And don’t forget the dithering when going back to 1Fs.
Hans
Last edited:
That's a good question. One thought is that the second filter might undo the sins of the first. While I'n positively certain, I think that this wouldn't happen, and the effect would rather be that the impulse-response of the first filter would be multiplied by the impulse-response of second filter. So, I would expect the net up/down response to be like that of a single FIR filter of twice the order. In other words, like a single brick-wall filter except twice as steep.
Agreed. Simple decimation can be utilized so long as there is no signal content above 22KHz. Which, there wouldn't be if the signal had been previously up-sampled.
Last edited:
That would be interesting to see indeed. On the other hand it will not help us, as then we would add even more unkown factors into the equation and we still do not know the real difference between NOS and OS, don't you think?
Hans, I like your suggestion. We would want to also know some details about the kind of FIR filter which was utilized (equiripple, windowed-SINC, etc.), so that the results may ideally be placed in to a two-dimensional matrix mapping the FIR filter type, to the resulting subjective sound character. With the obvious hope that a clear positive-correlation between the two emerges.
You can then also use something like Foobar ABX to rule out some psychological factors.
Lampie519 has a point, though, when the interpolation factor is an integer. When you use filters of which the output signal passes through the original samples, depending on which samples are chosen by the decimator, it could be that the decimated signal equals the original signal except for the effect of dither. I don't think that could happen with the 44.1 kHz to 192 kHz and back version.
Lampie519 has a point, though, when the interpolation factor is an integer. When you use filters of which the output signal passes through the original samples, depending on which samples are chosen by the decimator, it could be that the decimated signal equals the original signal except for the effect of dither. I don't think that could happen with the 44.1 kHz to 192 kHz and back version.
I think once the original file is processed by any software, it will be impossible to arrive back at the original file. I played with ProTools (Vers. 7.3.1) when 24/192 material was rarety - I was creating high-resolution files by upsampling native, EAC-ripped 16/41 to 24/96 and 24/192 to test my DAC's... The internal processing engine (Dynamic Plug-In) specifications were a big unknown to me. I was able to choose dither; that I do remember.
I do not think you can control these internal engine plugins to simply arrive back at the original file.
I do not know at what number (integer or not) the ProTools upsampling engine would arrive at, even If a whole number multiple is chosen to upsample 44.1 material.
I think what's required here is to decide on few things:
First of all: are the measurements important, or are we are just doing a listening test. This is crucial to determine.
Then, we could digress from here...
1. Use a true NOS DAC and then try NOS vs OS with a single resistor used as I/V (no analog filters of any kind). This would measure bad, but in reality, it actually sounds really good (transformers sound pretty good as well if chosen correctly). Only one variable is manipulated at anyone time.
2. As per the above, stick with NOS... and then try different analog (passive/active) filters. Again, one variable; the rest is kept constant.
3. As per 1. above, use OS... and then try different analog filters as in 2.
It just means I don't understand you.
So far this thread has been about trying to figure out why some people prefer a DAC without oversampling and without an analogue reconstruction filter. Do they just like the slight treble roll-off or is there more to it, for example imperfections of the oversampling filters?
You mentioned measurements, but I haven't a clue what measurements you mean and how they relate to the subject of this thread.
So far this thread has been about trying to figure out why some people prefer a DAC without oversampling and without an analogue reconstruction filter. Do they just like the slight treble roll-off or is there more to it, for example imperfections of the oversampling filters?
You mentioned measurements, but I haven't a clue what measurements you mean and how they relate to the subject of this thread.
Last edited: