NOS DAC's without any digital filtering?

How are these DAC's able to perform as well or better than DACS that use filtering to diminish aliasing effects? I understand that there are some who believe that the best sounding DAC's in the world are NOS/non-filtering. How is is this possible?
Robertsong

Robersong, when you talk about aliasing effects, you are usually talking about a filter that comes BEFORE the converter (be it D/A or A/D). Such an anti-aliasing filter is almost always used when we employ A-D conversion. The anti-alias filter band limits the incoming signal between 0Hz & Fs/2 which is the fold-over frequency. Fs is the A-D converter's clock frequency. The purpose of the anti-alias filter is to prevent frequencies between Fs/2-Fs to fold into the A-D conversion band & create distortion/aliasing.
In a D-A converter you might not need to use an anti-alias filter since the incoming signal is a digital bit stream. In many D-A converters there is a PLL type IC used that locks onto the incoming bit stream, extracts the clock from this bit stream & by PLL action cleans up this extracted clock. Cleaning up the clock would mean reducing the jitter (or in analog terms reducing the phase noise of the extracted clock). If the incoming bit stream is sync'd to this clean clock, one can directly feed this bit stream into the D-A converter.
I *think* you mean to write "reconstruction filter" which is the analog filter after the D-A converter. Is that right?
While it is true that many non oversampling (NOS) DACs do not use explicit analog filters, they still filter the DAC signal. There are many tubed output stage NOS DAC where the tube itself acts like an analog low-pass filter that has a -3dB bandwidth of 22-25KHz. In other cases where the output stage is solid-state, many NOS DACs use a SInc filter where the sinc filter has a 0 amplitude at, say, 60KHz. This would mean that lower frequencies (such as 30KHz, 40KHz, 50KHz) are heavily attenuated (but not zero amplitude) such that any clock energy in this frequencies is also heavily attenuated.
So, contrary to what meets the casual eye, there is analog filtering taking place - either using the natural roll-off of the output tube or employing a higher freq sinc filter.
I think one of the key reasons that NOS DACs sound as good as they do is that they do not use up sampling or oversampling which is a digital filter to interpolate the incoming music signal as the incoming signal's data rate is increased from 44.1KHz to something higher. This digital interpolation filter is essentially software devised/created/invented by the manuf & it is a sophisticated algorithm to interpolate while keeping the distortion to a minimum as defined by the manufacturer. If you agree with the manuf then you will like his/her interpolation hence upsampling/oversampling DAC. If you don't agree, you'll be back in the market hunting for another DAC.
When you buy an up/oversampling DAC you are essentially subscribing to the manuf interpretation of guessing at the music signal as the data rate is increased. That's why there are so many different up/oversampling DACs - each manuf has his/her own interpretation of this process & nobody is right or wrong; they are just different. The other thing is that since they are up/oversampling i.e. DSPing the incoming signal, the original music signal has been changed much more than if one were to use a NOS DAC. Any time you meddle with the original signal you add distortion, no matter how little. And, this can detract from the listening experience a little or a lot.
In a NOS DAC if one keeps the electronics to as precision circuits as possible then this sort of DAC does the least required to convert the incoming music signal to an analog equivalent. OTOH, an up/oversampling DAC does the most processing to covert the incoming music signal to an analog equivalent.
Of course, as in all of audio, NOS DACs are not always the best - there are many excellent up/oversampling designs.

I immediately thought of Ebm as well when I read Roxy54's post. Some of his snidy posts make me laugh.
good to know he's joking because sometimes I cannot be sure - he writes with such a straight-face.
let's see if he posts here now or not...

09-24-15: Robertsong
Bombaywalla, I think the confusion was....

when I mentioned "aliasing", I actually meant the "pre-ringing" associated with steeper filter slopes from lower sampling rates (ie. 44.1khz). I thought this was same thing. I have heard hi-res tracks from several different companies and the vast majority sound subjectivly better to me, and I assume that this perceived difference has to do with "pre-ringing".

What I'm getting at is...

does a redbook only NOS dac use a different method to reduce this ringing, or does it just compensate by providing [i]overall[/i] better sound somehow?

Of course it's the overall sound that actually matters, I just want a better understanding of WHY.

Thanks!

Robersong, pre-ringing & aliasing are not the same thing. Atleast I dont think of them as the same thing. Pre-filtering is an (bad) artifact of using a linear digital filter after up/oversampling. This pre-filtering creates hi frequency distortion & is often associated with "digititis" in digital playback. When people say "linear phase digital filter" I believe they mean to say an IIR (infinite impulse response) digital filter is used. IIR filters have 2 advantages: they can achieve a high attenuation factor (steep filter roll-off skirt) in a few number of poles (compared to a FIR filter) & 2nd, the delay thru an IIR is much less compared to an FIR. hence, if somebody wanted to monitor playback in real-time (such as in a studio setting) an IIR would be better suited. But, they have the down-side of pre-ringing that creates harshness in the sonics.
Aliasing is the effect of high frequency content, that is not low-pass filtered, folding back into the audio band.
When you up/oversample, you need a low-pass filter to band limit the resulting signal between 0 & Fs/2 & you can use a linear phase filter which, while low pass filtering, will create its own distortion in the form of pre-ringing. Maybe that's why you were thinking that aliasing & pre-ringing are the same thing. Pre-ringing has a similar effect as aliasing but the cause is different.

Yes, today a lot of the DACs have moved over to minimum phase filters which are FIR (finite impulse response) filters. these are purely digital filters with no analog equivalent & they are characterized with equal phase delay for all frequencies in the audio band. FIR filters simply cause a fixed delay in the playback path - when you press "play" it takes a short while before you hear the audio. IN non-real-time apps like home-listening it does not matter. And, yes, minimum phase filters get rid of the pre-ringing & have only post-ringing (which is considered natural).

In a NOS DAC that is playing at 44.1KHz there is no digital filter. There is no need for a digital filter because one is not up/oversampling at all. Since one is not passing the digital bit stream thru any digital filter, there is no signal processing on the music signal & the music signal remains untouched as it goes into the DAC. That is why in a NOS DAC playing at 16/44.1 the digital bit stream looks perfect in the time-domain i.e. there is no pre-ringing & no post-ringing (which are created by digital filters). On the output side, there is a low-pass analog reconstruction filter to smooth out the analog output of the DAC & to remove any high freq content that might be amplified by wider band preamplifiers &/or power amps. This low-pass filter could take the form of a tubed buffer output stage where the natural freq band limitation of the tube acts like a LPF. Or, in the case of my Scott Nixon Saru DAC+, there is a sinc (which is a sinx/x) filter with a zero amplitude at 65KHz. This means that lower freq like 30K, 40K, 50K are also attenuated (but not zero amplitude). And, it also might mean that the very high frequencies of 20Khz might also be attenuated a little.

In A NOS DAC there has to be some low-pass filtering to avoid the high freq hash from getting into the preamp & power amp which often have higher power bandwidths of 60KHz & even 100KHz. If this hi freq hash from the NOS DAC was not LPFed, the power amp will amplify it & it will degrade the overall sonics quite a bit.

All R2R NOS DACs are not necessarily 16/44.1, as Kijanki pointed out. Many are but there are several that operate at higher sampling freq. For example, my Saru DAC+ can accept upto a 24/96 input (which is the case in my setup - the digital gets fed into a Monarchy 24/96 DIP & its output feeds my NOS DAC. I believe the DIP is outputting a 24/88.2 signal). If that particular NOS DAC can accept a higher sampling rate such as 88.2K, 176K, 192K then you can upsample the digital bitstream on your computer (which can do a much better job than the processor in a DAC) to that higher sampling rate & feed it into your NOS DAC at the higher data rate without any digital filter.

So, in a NOS DAC, by avoiding any digital filtering we completely eliminate pre-ringing hence manage to keep a perfect time-domain response of the DAC. The frequency domain response of the NOS DAC doesnt look that great because there is a lot of hi freq content but we can deal with it by intelligently choosing a reconstruction filter (like the 2 examples I gave above) & containing the high freq hash to manageable levels. That's why some well-implented NOS DACs sound really good & give many up/oversampling DACs are true run for their money.
Hope this helps....

09-25-15: Robertsong
Okay, that makes complete sense now. Thanks!

So even with a NOS DAC I assume there is still benefit of using hi-res files? Why would somebody want to up-sample redbook in their software (foobar, JRMC, Amarra, etc.)? I haven't tried this in a few years but I recall that I preferred the sound w/o the upsampling (in JRMC).

Hmmm.

when you input 16/44.1 into the NOS DAC, the output contains not only the converted analog equivalent but also all the clock energy at high frequencies above 22.05KHz to 60Khz, 100KHz & beyond albeit in decreasing amplitude as the frequencies keep getting higher. By using LPF from a tube buffer or a sinc filter some of this noise is reduced but not eliminated. There could be sufficient energy in the higher freq for the power amp to amplify it & create distortion thru the speakers.
You can eliminate this by upsampling redbook by, say, 4X to 176.4KHz using software (like the examples you gave). Now the noise will contained in frequencies 88.2KHz (Fs/2) & above rather than 22.05KHz & above. By upsampling you just moved the problem up in frequency to a place where most power amps are likely to have very little power gain hence the probability of distortion is much, much reduced. Additionally you can employ an analog LPF with a -3dB cut-off at a much higher frequency, say, 44.1KHz (or even higher since the software upsampling has ensured there is no noise in the 0-88.2KHz region) such that you do not roll-off any of the high freq response of the DAC.
Since the digital signal is upsampled in software & given your NOS DAC can accept 176.4KHz you can feed this signal directly into the NOS DAC without any hardware digital filtering.
You reap all the benefits of the NOS DAC & you dont have to worry about high freq noise.
That's why one would want to upsample redbook in software. Hope this makes sense.....

09-26-15: Robertsong
Bombaywalla, your explanations have been super helpful for me.

thanks for your kind words. Glad i could be of assistance.
yeah I know what you mean about trolling the internet for all this info. And, there is a LOT of bad & incorrect info which would confuse & misguide a layman very quickly.

Is an upsampled 16/44.1 file just as good as as "hi-res" file of the same sample rate when using a NOS DAC? No advantage at using a "hi-res" track at all???
Robertsong

this is a difficult question for me to answer as it is so subjective. You & I could be listening to the same hi rez file on your system as one of us could like it & the other not. What then??

I personally think that redbook done well these days in HDCD, XCRD, XRCD2, SHM-CD, etc are really superb & leave very little to be desired.
Many others dont agree & think that hi-rez is the way to go. Each to their own.

As Zd542 pointed out a 16/44.1 when upsampled to a higher rate does not improve in fidelity. You have what you have at 16/44.1 & you can only degrade if you go to a higher rate & at best keep the fidelity the same as what is was at 16/44.1.
With a hi-rez file, like Zd542 wrote, you start off with a higher number of bits (20 or 24 or 32) & so you have a lot more information encoded in those bits. The sonics of the hi-res file will be different from that of the 16/44.1 file & it's really up to you whether you like it or not.
I would say that you should try it esp. if you have hi rez files on-hand &/or have a subscription to a hi-rez msuic website. Try the same track in 2 different formats & see which you like better.
It's really up to the individual...
Sorry for this non-commital reply - it's so subjective.

love your post Bcgator! :-D

09-29-15: Kijanki
There is a difference between upsampling and oversampling,

I completely agree but I don't think everyone gets it. You see people lump over & upsampling together & often use it interchangeably. There's a difference in the DSP that goes into upsampling vs. oversampling. Glad you pointed this out....