advertisement


Do NOS DACs fry your tweeters?

Dougal

pfm Member
Okay... it's a bit of a dramatic headline. But I'm wondering, if you run an unfiltered NOS DAC (putting out lots of high frequency noise about 20kHz) into a 'wide range' (20kHz and up) amp and speakers, could this ultimately damage the tweeters?

If not, how come so few mainstream manufacturers make a DAC of this type?

Please pardon my ignorance, I really want to learn the truth about this.
 
If they did might there not be a load of irate blown tweeter owners?

Good point! :)

I guess it depends on how the tweeters might be damaged and over what period of time.

I wasn't thinking of a catastrophic failure, more the effects of excessive 'excursion' over months or years.
 
It depends on how much HF hash actually comes out of the DAC. NOS DACs originally had heavy analogue filtration to avoid just this issue.

S.
 
Modern NOS DACs make a USP out of being filterless, so they tend to be ... filterless.

At first sight this indeed is a serious threat to your tweeters, but as so often first sight is wrong.

There are two reasons:

1) the spectral distribution of the vast bulk of music rolls off towards the treble. Most of the activity (and power) is below 3kHz.

2) the so-called zero-order-hold transfer function of a NOS DAC has a corresponding frequency response that is shaped like a Sinc(f) function. This function is about -3dB at Fs/2 (i.e. 22kHz for CD), and reaches a hard zero (0) at 44.1kHz.

This means that Sinc attenuates a bit the 22-30kHz images (which originate in the music's treble, which is not very active) and attenuates a lot the 40-44kHz images (which originate in the music's very active bass and mid regions).

Bottom line: no problem.
 
Elegant point werner

I still wonder whether it has something to do with the fact that most speakers (and probably many amps, including those on the output stages of many nos dacs) aren't really wide range either.

One thing seems clear enough though- if you really could hear anything over 20Khz, it would be somewhat annoying.
 
most speakers (and probably many amps, including those on the output stages of many nos dacs) aren't really wide range either.

Speakers, probably. Amplifiers I have to disagree with. It would take a serious effort to make an amplifier with hifi pretensions and yet with an appreciable rolloff below 40kHz, let alone 30kHz.

One thing seems clear enough though- if you really could hear anything over 20Khz, it would be somewhat annoying.

More elegance, of nature this time. High frequencies are detected at the entrance to the cochlea, where the basilar membrane is narrowest and tuned highest. The membrane is divided into frequency bands along its length, called ERBs (used to be Barks, but this has been refined) (*). Frequency discrimination AFAIK (#) is effected by sound triggering several neighbouring ERBs simultaneously (they overlap), allowing the auditory system to triangulate the original pitch.

The highest-HF ERB, some 3-4 kHz wide, is unique in that it lacks a neighbour near the entrance, and thus loses triangulation capabilities with progressive higher frequencies entering the ear.

What does this mean?

1) from 12kHz on (give or take) most adults cannot distinguish pitch. 13kHz and 15kHz sound exactly the same.

2) in digital audio: mild cases of an-harmonics due to aliasing or imaging (if audible at all) are not perceived as anharmonic (due to above pitch-deafness) but only as an overall brightening of treble, which is innocuous. Yes, aliasing is mostly harmless when confined to one ERB!

3) and does it surprise that ultrasonic detection by bone conduction (which is proven up to at least 100kHz) results in a perception identical to sound received on the upper ERB? I.o.w. bone-injected 100kHz sound is not distinguished from, say, 13kHz sound. Which makes a lot of sense if you look again at the physical structure of skull and cochlea.


(* I am not sure if the division into bands is static and defined by physical structure, or dynamic and defined by the lowest-frequency signals received, i.e. the lowest, deepest ERB that is triggered defines where the actual divisions start, back to the cochlear entrance. Which would be funky as sound needs 3-4 ms to travel down the membrane and reach that ERB. Can someone enlighten me?)

(# And if I am wrong then the underlying mechanism is certainly functionally similar to what I describe, in which case I am functionally right ;-)
 
Modern NOS DACs make a UPS out of being filterless, so they tend to be ... filterless.

Just goes to show how putrid some modern audio engineering has become. People who either don't know the basics or pretend that somehow doing things completely wrong is the new right.

Madness, and one reason why I keep away from all boutique products.

S.
 
Madness,.

If some people happen to better enjoy their music through such a DAC then it cannot be 'madness'.

But it is madness when this mode of replay is tauted to be the more 'correct' one. But that is just 'words' and one can ignore these words at will.

And it is madness when a DAC containing no more than 16 dirt-cheap and crappy TDA1543 chips, a receiver, and the supporting power supply (which can't be much as there is not much), and not even an output stage ... madness when such a product of low inherent value is sold for thousands and thousands of pounds. But of course no-one is forced to buy it.

Also madness when a DAC contains 16 PCM1704 chips, as this depletes the world's stock of these wonderful convertors ;-)
 
Many thanks for your comments, Werner. Very interesting! I'll be up at the Bristol show next week so will have to chat with Peter Qvortrup about this.
 
Many thanks for your comments, Werner. Very interesting! I'll be up at the Bristol show next week so will have to chat with Peter Qvortrup about this.

Good luck with that. If you're a believer you will no doubt find him charismatic. He looked in a while ago on a thread on hifi wigwam. Can't say I found his contributions illuminating, but he is an inspirational figure to many.
 
Speakers, probably. Amplifiers I have to disagree with. It would take a serious effort to make an amplifier with hifi pretensions and yet with an appreciable rolloff below 40kHz, let alone 30kHz.



More elegance, of nature this time. High frequencies are detected at the entrance to the cochlea, where the basilar membrane is narrowest and tuned highest. The membrane is divided into frequency bands along its length, called ERBs (used to be Barks, but this has been refined) (*). Frequency discrimination AFAIK (#) is effected by sound triggering several neighbouring ERBs simultaneously (they overlap), allowing the auditory system to triangulate the original pitch.

The highest-HF ERB, some 3-4 kHz wide, is unique in that it lacks a neighbour near the entrance, and thus loses triangulation capabilities with progressive higher frequencies entering the ear.

What does this mean?

1) from 12kHz on (give or take) most adults cannot distinguish pitch. 13kHz and 15kHz sound exactly the same.

2) in digital audio: mild cases of an-harmonics due to aliasing or imaging (if audible at all) are not perceived as anharmonic (due to above pitch-deafness) but only as an overall brightening of treble, which is innocuous. Yes, aliasing is mostly harmless when confined to one ERB!

3) and does it surprise that ultrasonic detection by bone conduction (which is proven up to at least 100kHz) results in a perception identical to sound received on the upper ERB? I.o.w. bone-injected 100kHz sound is not distinguished from, say, 13kHz sound. Which makes a lot of sense if you look again at the physical structure of skull and cochlea.


(* I am not sure if the division into bands is static and defined by physical structure, or dynamic and defined by the lowest-frequency signals received, i.e. the lowest, deepest ERB that is triggered defines where the actual divisions start, back to the cochlear entrance. Which would be funky as sound needs 3-4 ms to travel down the membrane and reach that ERB. Can someone enlighten me?)

(# And if I am wrong then the underlying mechanism is certainly functionally similar to what I describe, in which case I am functionally right ;-)
Very interesting. The amp thing reminds me of someone's recent anecdote on another thread involving an amp which rolled off at 27 kHz.
I was under the impression that some output transformers rolled off quite low too. I also understand there is/was often a rapid roll off with class d amps - see for example the bel canto ref 1000m which cost $6000 each http://www.stereophile.com/content/bel-canto-ref1000m-monoblock-power-amplifier-measurements

About the frequency discrimination.-
Just to clarify if everything over 12 kHz sounds the same then is the distinctive sound of the nos DAC due to the noise floor (which always looks really high in the stereophile measurements) or does the sinc function frequency response cause audible departures from linearity > 12 kHz, or is it just the amount of 12 kHz energy or ?

Presumably the lack of fine frequency discrimination may also help explain why jitter is not audible until it reaches quite high levels (at least jitter sidebands causing spuria over 12 kHz which I guess means low frequency jitter on hf signals or maybe hf jitter on lf signals assuming other hf content)
 
The typical 1" or 3/4" silk dome does not respond to signal above 22 kHz.
The ultrasonics from NOS DACs (not really high level either, just higher than a normal DAC) are more likely to cause intermods in the amplifier, especially SET valved
 
often a rapid roll off with class d amps - see for example the bel canto ref 1000m which cost $6000

555Belfig1.jpg


Amplifiers may roll off not too much above 20kHz, but the slope is not steep enough to yield a lot of attenuation below 40kHz or so. Or even 100kHz. Have a look at the vertical scale above: it looks steep, but it only covers 3dB. A decent AA or AI filter does 100dB or more in that frequency span.

Just to clarify if everything over 12 kHz sounds the same then is the distinctive sound of the nos DAC due to the noise floor (which always looks really high in the stereophile

Measuring a NOS DAC without precautions is ill-advised, because reference signals tend to be high level, much higher in the treble than real music, and so they trigger a host of intermodulation problems that are less in effect with real music.

This said, plenty of NOS DACs are comprimised in the sense that they use cheap and crappy DAC chips originally intended for budget gear and/or combine this with inept or intendedly colouring output stages. So you do get indeed more noise and intermodulation than with a clean and correct DAC. But then vinyl too gives a lot of noise and intermod ...

measurements) or does the sinc function frequency response cause audible departures from linearity > 12 kHz,

Unless explicitly compensated for it does indeed. A NOS DAC operated at 44.1kHz is mathematically equivalent to the treble knob turned down a notch or two.

Bruno Putzeys claims that if you equalise a normal DAC with a similar roll-off it sounds 'entirely' like a NOS DAC. I don't know. I never tried, am not inclined to try, and I only have a TDA1543-based DAC lying around anyhow. (I do have a stash of unused TDA1541s, though: I will never use them, interested parties may drop me a PM.)


Presumably the lack of fine frequency discrimination may also help explain why jitter

I wouldn't go that far. But then I have zero academic interest in jitter. Or in anything with tubes or transformers, for the matter.

But don't forget that the loss of frequency discrimination only applies to the upper ERB. All the lower ones are perfectly fine, thank you.
 
Thanks for the clarification.
Any particular reason for the lack of interest in jitter. Is it because you don't think it has any major impact?
 


advertisement


Back
Top