As for the term "apodizing", it's borrowed from optical jargon where it means simply windowing i.e. the image of an "impulse" (star) is related to the Fourier transform of the aperature. You can fatten the central spot and attenuate the Airy disk by windowing the aperature, literally darkening from the edges to clear at the center. The math is exactly the same.
Using minimum phase reconstruction rather than linear phase is not really the same thing. I can't find any examples of actual music with/without only impulses, I expect the pictures would be far less dramatic. BTW Wolfson does mention early roll off in their implementation.
Any of these filters could be implemented by recording their impulse response and deriving coefficients.
Using minimum phase reconstruction rather than linear phase is not really the same thing. I can't find any examples of actual music with/without only impulses, I expect the pictures would be far less dramatic. BTW Wolfson does mention early roll off in their implementation.
Any of these filters could be implemented by recording their impulse response and deriving coefficients.
the Ayre "slow roll off" plot suggests ~ 0.4 % image frequency content near 30 kHz
may well be inaudible - but but you would think that much "inverted frequency", "non-harmonic" error would be worse than hypothesized "subtle phase distortions" that John and Charles are so dead certain ruins high feedback amp's sound
or are we now accepting that there is a audile frequency limit?
may well be inaudible - but but you would think that much "inverted frequency", "non-harmonic" error would be worse than hypothesized "subtle phase distortions" that John and Charles are so dead certain ruins high feedback amp's sound
or are we now accepting that there is a audile frequency limit?
Last edited:
I think making it switchable satisfies both needs. I have a Forsell Air Reference CD player since 1993 and it still sounds satisfying. As far as i know it has a conventional reconstruction filter but i should measure that before i argue. Peter got the sound by being extremely picky about component choices. I heard the final shoot out in his system between 8 converters on his modified Beveridges. He claimed that he had only swapped between 10 makers of resistors in 130 positions but with equal value. They sounded different. What made him broke was that in production the manufacturer did not use his exact choices. Believe it or not, i do not care any more.
I could imagine 10 DSP programmers each contributing their 130 favorite impulse responses to an open source library and a CD player with a USB upload feature.
Listening to different filters all the time with the same piece of music. I suggest "Jazz at the Pawnshop".
I think i wrongly included a 44.1 sinc response to get the image feedthru number estimate - if done without the zero order hold and with with their 8x upsampling you can pretty much just read the attenuation off the Ayre slow roll off graph
of course that makes the image frequencies near 22.05 kHz much bigger - I would think any of our Golden Ears claiming 24 KHz hearing would run screaming from the room - maybe the testers should include more prepubescent girls?
(I shudder to think of those test subject’s preferred music)
and SET or other "no feedback" amplifiers with "harmless" few % 2nd order distortion will cause difference products from the image frequency content to fold down into the conventionally audible range
of course that makes the image frequencies near 22.05 kHz much bigger - I would think any of our Golden Ears claiming 24 KHz hearing would run screaming from the room - maybe the testers should include more prepubescent girls?
(I shudder to think of those test subject’s preferred music)
and SET or other "no feedback" amplifiers with "harmless" few % 2nd order distortion will cause difference products from the image frequency content to fold down into the conventionally audible range
Calculating transient response
Not exactly about digital filtering, but I couldn't resist:
I found a vintage paper that I'm sure you guys will be interested in. About transient response, ringing, stability and the role of feedback. From a Mr. Roddam. From 1952!
Top left reference here: Historical Library
jan
Not exactly about digital filtering, but I couldn't resist:
I found a vintage paper that I'm sure you guys will be interested in. About transient response, ringing, stability and the role of feedback. From a Mr. Roddam. From 1952!
Top left reference here: Historical Library
jan
Another question, if I actually record an "impulse" on my sound card the data that would be written onto a CD is the impulse response of my A/D's anti-aliasing filter. The pre-ringing is already there, so how does the anti-imaging filter in the DAC remove this? Speaking rhetoricly here.
A single sample impulse or step can not be on a CD in the course of properly recording an analog input signal, unless a piece of software put it there.
@Jan: Interestingly, that paper proposes the idea that feedback occurs post hoc ("Delayed Feedback").
We all know feedback goes round and round.
And the painted ponies go up and down.
I could imagine that when an analog low pass filter is put before the AD converter the output of a linear phase DA converter would show no pre ringing.
I have found when I drive anticipating the curves in the road works better than driving by feel.
Feedback should occur in units of time measured by phase. Of course there always is feed forward available.
I think you need a few more bricks in the wall for that to work. Why not just go the NOS DAC route, that certainly makes a large group of folks happy?
The Roddam paper is interesting. However, I suspect that the analysis he presents is only true for a minimum-phase system, as only then would the 'real part' transfer function contain all the necessary information. Incidentally, he is using one of the wonderful things about analytic functions in the complex plane: if you are told something about the function you can probably deduce all the rest.
His talk about delay in feedback is rather misleading, as he assumes an average group delay across the whole audio spectrum when in fact almost all of the phase shift takes place in a narrow HF region. He admits that his treatment is incomplete, but that seems a bit weak having just misled readers. It is true that in order to get a flat total response at the output there may be peaks at intermediate nodes, and some designers forget that. To get delays you need storage, and that means resonances (e.g. HF in OPT).
His talk about delay in feedback is rather misleading, as he assumes an average group delay across the whole audio spectrum when in fact almost all of the phase shift takes place in a narrow HF region. He admits that his treatment is incomplete, but that seems a bit weak having just misled readers. It is true that in order to get a flat total response at the output there may be peaks at intermediate nodes, and some designers forget that. To get delays you need storage, and that means resonances (e.g. HF in OPT).
- Status
- Not open for further replies.