-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
DacMagicPlus Filter Mirror Images - Short Study
- Thread starter pma
- Start date
If a signal is band-limited to 20 kHz and then sampled at 44.1 kHz, there will be no aliases. Had there been content at 24.1 kHz, that would alias into the baseband, sure, but the premise was that the signal had been limited to 20 kHz.
- Thread Starter
- #43
This is all trivial. If the signal is sampled at 44.1kHz and then reproduced, everything depends on anti-alias filter in the ADC and interpolation filter in the DAC. Properly anti-aliased signal coming into the ADC has no frequencies above 22.05kHz. It may have frequencies <=22kHz. If this properly antia-aliased and sampled signal is then reproduced by the DAC, everything depends on DAC interpolation filters, in case of S-D oversampling DAC, or on analog output filter in case of NOS DAC. If the output filter has leaks (not steep enough) above 22.05kHz, then mirror images (not aliases) are created at Fs - Fsig frequency. The mirror image is at Fs - Fsig frequency. Fsig and the mirror image create the beat frequency. This is bullet proof and cannot be doubted.
Fsig = 21.5kHz and its mirror image at 44.1 - 21.5 = 22.6kHz
- Thread Starter
- #45
This is trivial and somehow you completely missed his point.
My point is that you never know and the DAC must properly handle the signals <=Fs/2 (22.05kHz here). It must not create mirror frequencies and there is no excuse, otherwise you violate conditions of proper signal reconstruction. Go and learn how to design a proper filter and add it to Topping DAC filter menu, that would be the solution.
Not completely on topic but somewhat relevant, one should never judge the bitrate or sound quality of these codecs by eyes. Someone in HA made this claim and of course, violated the famous TOS8. I played a game with that member, the results are hilarious. A very good example of expectation bias.
https://hydrogenaud.io/index.php?topic=111736.msg927918#msg927918
MP3 files can contain frequencies up to fs/2 even at low bitrates, and they can look very beautiful on spectrograms. Psychovisual plays tricks. The whole thread only has three pages, so read from start to end won't cost a lot of time.
solderdude
Grand Contributor
Ah.. I was looking at spectrum plots from MP3 created from RBCD and then converted back to WAV.
The files I played with all topped out sharply around 18kHz-19kHz (LAME enc, 320kbs, CBR, high Quality) where CD versions topped out at 20kHz - 20.5kHz.
Of course that will also depend on the content and perhaps VBR differs as well. Not all MP3 are created equal...
The files I played with all topped out sharply around 18kHz-19kHz (LAME enc, 320kbs, CBR, high Quality) where CD versions topped out at 20kHz - 20.5kHz.
Of course that will also depend on the content and perhaps VBR differs as well. Not all MP3 are created equal...
Last edited:
Most (decent) encoders will low-pass filter the audio at 16-18 kHz since what little spectral content exists above is barely audible (if at all). Removing the highest frequencies entirely allows all the bits to be used for the lower part of the spectrum. The resulting improvement in that range almost always gives a perceptibly better result. There is, however, nothing in the spec that mandates this behaviour.
Thanks for the nice writeup even i can understand it 
Question can we lay to the rest this "preringing" fad of lately ? or is that another tread .
My layman understanding .
Some class of filters does not exibit preringing when shown in a graph wit a specific step test signal .
This is a special step test signal made to test filters for various purposes (here i don't know why or the history of it ) .
But this kind of signal is "forbidden" by proper bandwith limiting so it will not exist in any properly made music you can buy ? The ADC would not let trough any step (or corresponding digital filters and effects in your DAW or synthesizer ) Is this correct ?
Therefore your music will ofcourse not exhibit "ringing" even if the filter can be made to do that with a purpose made test signal !!
Arguments i read all the time in audio press sales literature and forums is about how preringing is unnatural and not what naturally occurring sounds will do. This is of course correct, but the great misunderstanding so it seems is that this happens with music and not only with a step test signal that can not occur normally ?
Or is reality somewhat more murky , can signals that do exist in our music files and CD provoke this to any audible level ? I realize that bandwidth limiting may not be perfect on all recordings but does it really matters in practice ?
It's seems that most implementations with apodizing or preringing free filters comes with some obvious penalties .
My layman understanding that I want you guys to correct if possible is that by removing this ringing "problem" you actually get other more real problems that where already solved , arguably it may not be audible anyway ? I have a feeling ( I could be wrong ) that this a myth similar to the "digital staircase" we where to believe in the early days of digital ?
Question can we lay to the rest this "preringing" fad of lately ? or is that another tread .
My layman understanding .
Some class of filters does not exibit preringing when shown in a graph wit a specific step test signal .
This is a special step test signal made to test filters for various purposes (here i don't know why or the history of it ) .
But this kind of signal is "forbidden" by proper bandwith limiting so it will not exist in any properly made music you can buy ? The ADC would not let trough any step (or corresponding digital filters and effects in your DAW or synthesizer ) Is this correct ?
Therefore your music will ofcourse not exhibit "ringing" even if the filter can be made to do that with a purpose made test signal !!
Arguments i read all the time in audio press sales literature and forums is about how preringing is unnatural and not what naturally occurring sounds will do. This is of course correct, but the great misunderstanding so it seems is that this happens with music and not only with a step test signal that can not occur normally ?
Or is reality somewhat more murky , can signals that do exist in our music files and CD provoke this to any audible level ? I realize that bandwidth limiting may not be perfect on all recordings but does it really matters in practice ?
It's seems that most implementations with apodizing or preringing free filters comes with some obvious penalties .
My layman understanding that I want you guys to correct if possible
is that by removing this ringing "problem" you actually get other more real problems that where already solved , arguably it may not be audible anyway ? I have a feeling ( I could be wrong ) that this a myth similar to the "digital staircase" we where to believe in the early days of digital ?If I interpret this correctly (I am not a native English speaker), did you mean ADCs are supposed to be perfectly bandlimited? The answer is no, and in some cases even more severe than DAC filters.
https://www.audiosciencereview.com/...u-m4-audio-interface-review.15757/post-504008
Another interesting read:
https://archimago.blogspot.com/2019/03/measurements-look-at-audio-ultra-high.html
Basically DAC and ADC makers violate the rule quite often, especially on higher samples rates, as non-artificial signals are often much weaker at higher frequencies. Datasheets are widely available from different chip makers, better download several of the popular ones and have a read.
Here's my attempt at explaining it: https://troll-audio.com/articles/filter-ringing/
- Thread Starter
- #53
It is not forbidden. You create perfectly band-limited square and see the rising edge of the step.
A very good article indeed .
It leaves as small gap
when this small margin does not exist , but as bennetng says non artificial signals are weak at high frequencies .And its an issue for the music producer not the DAC as you say . Thankyou both .
Sorry for my bad English this is band limited and of course "allowed" the single sample pulse exemplified mansr nice article is not something that's on your CD's or files and it can not get trough your ADC, bandlimited squares are ok I understand . But these filters are often demonstrated with single sample pulses , that are not bandwidth limited.
- Thread Starter
- #56
But it is also OK. You want to know the filter impulse response. Impulse response is nothing but a derivative of the step response. You can easily let pass the impulse or square through ADC, with input anti alias filter and limited to <=Fs/2, record it and play through DAC and you will still get pre-ringing if you do not use the minimum phase filter. So it goes.
Just for fun, attached is a properly generated square at 44.1kHz Fs. Then it is upsampled to 96kHz and as you can see, nothing is above original Fs/2, except for some possible low level quantization noise.
Just a reminder - this thread topic is not about pre-ringing!
Last edited:
I do understand there is a technical need to know the filter impulse response . I wont do more off topic then (mod can move this if he feels the need) .
- Thread Starter
- #58
No it is fine. The impulse response is the basic transfer characteristics of the system and you will find it in every digital filter design textbook. The resulting signal is a convolution of the signal with impulse responses of individual samples.
I made a a mistake here above, now corrected - the impulse response is a derivative of the step response.
The statement doesn't make sense. Several parts of it. It would be best to show an example maybe.
Pavel, I have a deep respect for your knowledge, but before you attack him you should understand that CS43198 does not even allow for an external digital filter (lacks proper 8x input mode) and it is unclear if you can get the proper response with the number of taps allowed via the custom coefficients feature.
Similar threads
- Poll
