advertisement


A bit of filter theory

philiphifi

pfm Member
hi - perhaps a qualified engineer could help with this question. Many (many) years ago, i had to learn electronics design and it was gospel that one needs filter (forget the order of the filter for now) for anything above 20kHz to avoid higher freq reflection into the audio freq range. Now i read that An and others are now using filterless DACs. How do they work ? If it is the case that the noises are non audio-able then what have they been teaching students!
 
… Many (many) years ago, i had to learn electronics design and it was gospel that one needs filter (forget the order of the filter for now) for anything above 20kHz to avoid higher freq reflection into the audio freq range.
Yes at the Analogue-to-Digital (strictly the sampling) stage.
… Now i read that An and others are now using filterless DACs. How do they work ?
That's at a different stage - the Digital-to-Analogue (strictly the reconstruction) stage. Reconstruction creates new aliases of the signal, which should be removed to recover the exact (original, filtered) signal. You can omit the reconstruction filter and the human ear's limitations can perform that filtering. However the aliases might cause problems in amplifiers not designed to handle ultrasonic signals.
… If it is the case that the noises are non audio-able then what have they been teaching students!
The teaching is correct technically, but does not necessarily address non-technical aspects such as product positioning in a market via differentiation and other customer preference issues.
 
On the DAC side, IC designer can use upsampling and internal digital filtering to avoid external filters
 
Can you give an example link to a DAC where this is referred to? The problem is that the language used may not be clear. e.g. 'filterless' *might* mean 'no analogue filter'. Or it might mean one of the designs that do no upsampling or filtering of any kind. Or something else.
 
IIRC from a review I read some years ago it simply has no filtering and feeds all sorts of ultrasonic crap at pretty high levels to the amplifier...
 
IIRC from a review I read some years ago it simply has no filtering and feeds all sorts of ultrasonic crap at pretty high levels to the amplifier...
Yes that's my issue with non oversampling conversions to CD players. Your ears might not be able to directly hear the ultrasonic noise in the signal....but your amp can.
 
I don't think I can hear ~50KHz but my amplifier can amplify it. I think the speakers I am using only go up to ~25KHz though.
 
..so why bake the tweeters with the random IMD [unheard or unperceived?] screaming of a power amp subject to unwanted/incoherent/ [non-compliant with the Nyquist-Shannon Samplng theorem, of which a proper Reconstruction filter is a part] ..noise?

My objection to the 16/44.1 non-oversampling cant isn't only just that it sounds poor, but also it's just so ...inelegant/wilfully-ignorant, mathematically. It is predicated on a total misunderstanding.

Yes, you can get over the first objection by bumping the sample rate of your CD source (but hey, that's identical to oversampling!) but not the second.


[And all of this is for individual pleasure, and we are all entitled to our own choices in that regard: and I am definitely not taking prisoners here ;) ]
 
ETA: obvious afterthought - 16/44.1 NOS dac used with a wilful/ancient design of valve amp is no problem: simply becasue the amplifier doesn't have the gain-bandwidth to run into such problems...
 
An example of this is Audio Note DACs I believe, NOS and no reconstruction filter. Within the context of an Audio Note system this may work fine, it does sound fine to my ears, but what happens in other systems??? Perhaps they are only really used in AN systems and it is all good ...
 
Except with all this conceptualising, no one has mentioned the voltage size of this noise. Are we talking 60dB down on the audio?

A bit of crossover dither might make your power amp better. Most speakers are a high impedance above 20kHz
 
[And all of this is for individual pleasure, and we are all entitled to our own choices in that regard: and I am definitely not taking prisoners here ;) ]
Personal preference notwithstanding it seems to me there are two identifiable poles of thinking about reconstruction filters in the market:
  • The no-filter approach (several adherents and more providing the option)
  • The perfect-filter approach (Chord mainly I think)
I believe the nature of human hearing is that the perfect-filter approach to reconstruction is good but unnecessary, and that the nature of our electronics means that the no-filter approach is risky. To me the real requirement of the reconstruction filter seems to be not the mathematical re-construction of the waveform but:
  • to not audibly damage the wanted signal (either linearly in amplitude and phase, or non-linearly)
  • to remove the unwanted aliases sufficiently to prevent subsequent electronics from damaging the wanted signal.
Most DACs seem to take this fit-for-purpose approach. Any other approach seems to be primarily marketing-led.
 
NOS DACs score in two areas.

1. Relying on "natural filtering" post DAC leads to post-ringing rather than pre-ringing.
2. No problems with digital headroom for inter-sample overs (assuming analogue headroom is rarely an issue).

Both of these can also be achieved digitally, but often they're not. As mentioned, against NOS are tweeter or amp distortion products leaking down into audible range.
 
To me the real requirement of the reconstruction filter seems to be not the mathematical re-construction of the waveform.

It is a requirement. Mathematics is not a question of opinion or preference. A reconstruction filter is an essential part of the DAC process. A NOS no-filter DAC just leaves that function to be done (in an uncontrolled manner) by your amp, speakers and ears.
 
The average dome tweeter oil drum modes will convert 44.1 ksps unfiltered artifacts into audible mush all too well.
This is perceived my many as more "detail" or "analogue"
 
Results with such a DAC are certainly likely to be very amplifier dependent! Many amps have a passive LPF at the input which could remove much of the ultrasonics before the amp sees it but the turnover point of this filter varies widely...
Then there is the linearity at these frequencies to take into account. Some may be non linear enough at ultrasonic frequencies to generate audible noises from the difference frequencies of the various ultrasonic artefacts.
Depending on the level of the supersonics etc there could even be the risk of slew rate limiting.
 


advertisement


Back
Top