XXHighEnd

I don't understand why the Phasure DAC, or any other DAC without low pass filtering, does not damage tweeters.

It seems to me that the DAC would output a pulse train whose frequency equals the sample rate and whose amplitude is only a few dB lower than the amplitude of the audio signal.

Music normally has treble content that is a small fraction of the total music power, so tweeters do not have to be designed to handle as much power as woofers.  It seems that continuously outputting 192 or 384 KHz pulses at a 10 watt power level would risk burning out a tweeter.

Am I overlooking something?

On the other hand, it's possible that one of the reasons the Phasure DAC sounds good is that superimposing an ultrasonic signal on the music reduces tweeter distortion due to hysteresis near the zero crossing, sort of like the ultrasonic bias signal used in magnetic tape recording.

Hi Bob - and welcome here !

Who says that no low pass filtering is applied to the NOS1 ? It just does. BUT, it is a requirement that this happens in-software, which is per the explicit design. So, in the case of XXHighEnd this would be the Arc Prediction filtering and in other software it's the means of "upsampling" applied, which also implies the filtering.

I would not call this low pass filtering per se because to me that would be like filtering out high frequencies which -then- would be in the material. This is of course not what we do (in any software); it is to be maintained because it is supposed to be on purpose and good. But, this then would need the 2x higher sampling rate compared to the highest frequency in there (and we'd call that Hires of any form). So ...

Your low pass filtering for sure would be in the A/D process. For example (and to the extreme) when a recording (microphone etc.) can capure 25KHz, low pass filtering must make that 22.049 KHz for 16/44.1 Redbook.
This is where your post starts in the process ...

No too high frequencies are in there anymore, BUT, now we need to make the "cut out" samples into a legit stream (wave) again. And this can't be with 16/44.1 when left alone; huge distortion will be our part, and, higher frequencies will *emerge* now because the too squary waves the higher frequencies imply, will cause, say, infinite high frequencies. It will be that what you are talking about ...

Granted, you can easily call what's to apply now low pass filtering, but it merely will be about reconstruction of the wave. So, if first that "too square" is deformed into "sine", no low pass is needed anymore. This is what Arc Prediction is doing to it's best capabilities.
This is not what normal (brick wall) filtering does literally, because that really reconstructs (and changes) and you sure can call this low pass filtering I think. But let's keep in mind : this is the low pass filtering of the too high frequencies just created because of that too low sampled file (implying pure squares at 20KHz when the sample points would be at the proper edges).

Possibly I just told you the obvious, but the key point is that Arc Prediction is ahead of this all. So, there these too high frequencies do not emerge in the first place. And also keep in mind : this happens in-software before it can reach the DAC.


This does not happen. So, the fact that the NOS1 "upsamples" to 768 does not imply that there's a frequency continuously going on of 384KHz. There's nothing in the source material of that.

What remains for Arc Prediction is the last portion of the audio band which is and remains too square. Well, actually above the audio band. So, when a sound of ~20KHz would play a sort of continuously then there's fairly high output at the mirror side of the sampling frequency / 2 (fs/2), 24.1 KHz. This reoccurs at double of that (mirrors around 44.1) but already 6dB down. Same at 88.2, 176.4, 352.6 where it would be 24dB down. But remember, only when that high frequency plays, which it does not and when it would it would be down xxdB already because of the far less power in that frequency by nature, unless synthesizers create it explicitly (which "explicitly" would be a moot thing for people who can't hear that anyway and otherwise it won't do anything really -> no way I'd be creating such frequencies ever on purpose with my synths).

I hope this suffices for an explanation why the NOS1 can sound good instead of that it should sound bad from theories. The theories are just different ...
Besides that I never heard a single voice of really anyone who perceived problems one way or the other (could be amplifiers).

Best regards,
Peter

I apologize that my question was unclear.  I was not referring to the software-implemented low-pass filtering in the digital domain.  I was referring to the absence of any analog low-pass filter after the D/A conversion.

The analog signal produced by the D/A conversion stage will not be smoothly continuous, but will look like a staircase, with sudden steps between each sample.  Therefore, pulses at the sample rate (not the Nyquist rate which is half the sample rate) will be superimposed on the audio signal.  In other words, the analog signal will look sort of like a pulse train at the sample rate that is amplitude-modulated by the audio signal.

Okay, from that angle and apologies that I misunderstood :

Stepping now is a 705.6 KHz (for 44.1 based material) and in 24 bits resolution. This will show beyond 352.8KHz anyway (so not before that) and that too (IIRC) is 6dB down. So yes, a high bandwidth amplifier (like 1MHz which sure exists) will pass that on. And next ? well, next think about the numerous "genuine NOS/Filterless" lovers which play through that. There too this will happen, but way more excessive because also right beyond the 22.05KHz, on and on and on.

I agree that this has always been a question of many, including myself (many times put in doubt on this forum by me myself), but apparently it doesn't work out like devistating; could be the cables not having the bandwidth, the tweeter having its natural roll off, etc.

But also think of that analogue filter : which filter would not influence the audioband and - thus be far enough from it not to influence, and next has a useful lowpass ? one octave overhere is 192KHz of width you know ... (thinking simple 6dB filters which won't influence too much in the wrong direction but won't do much in the right direction in the mean time -> no escape).
So it's not in there on very purpose. And since one highest quality resistor already completely kills the sound ... (I'm serious - tried everything and all).

Regards,
Peter

Thanks, Peter.

Regarding even one resistor of the highest quality killing the sound, my power amplifier has a few resistors, so I guess the sound is close to death by the time it reaches the loudspeakers :)

Haha. Depends on where those resistors are. But maybe that's why I use GainClones with two of them in there only. I won't claim that GainClones are the best out there, but at least it adds nothing (FFT wise) to the output of the DAC (yea, the gain).

Less is more really applies ...