Oh ! Somebody tried to design a brick wall filter in analog ?
(Thanks for the link, the most complete paper on this subject I had seen).
Last edited:
Many times, and with a certain amount of commercial sucess.
Analog filtering was the only way to go until digital filtering was generally accepted.
Digital filtering requires that the signal be in the digital domain. Getting signals into the digital domain was a very expensive proposition until digital recording was made practical.
There was a generation or more of digital recorders before the CD such as the PCM 1600 and the PCM-F1 and AFAIK they all depended on analog filtering. I seem to recall that there were third party brick wall filter modules to replace the stock ones, that claimed improved sound quality. AM radio sound quality could be improved with improved brick wall filtering, and there were several stereo receivers that incorporated this.
Digital filtering can be enhanced by oversampling, but this is not required.
Sharp cut-off analog filters had been around in audio for a decade or more before consumer digital audio, in AM receivers, FM stereo decoders and Dolby noise reduction processors. However, they were usually not as precise, and generally only had from 1-4 stages.
Last edited:
Thankyou Arny I and a whole bunch of people over many years contributed to that. Some dates from old Philbrick app notes from the mid 60's.
Here's a link to the 480 slide classroom Powerpoint version, for those interested.
Communication Acoustics – Pulkki and Karjalainen
interesting fact:
if you take all the harmonics (all of which are odd) of a square wave and remove all the ones below the stopband frequency,(filters with flat response and flat phase in passband and stopband), you'll be left with all the odd harmonics of the square wave that were above the cutoff, obvious.
The part below the cutoff, from a flat mag and phase LPF, will be a square wave with Gibbs ripples as weve seen. The output of the upper harmonics by themselves, in the time domain, will be an inverted version of the Gibbs ripples, but without the squarewave shape part, and with some spikes where the square wave edges were.
The HPF ripples and spikes consist of the higher odd harmonics of the square wave, by themselves.
The LPF output consists of the square wave but with those harmonics removed.
If you summed them together again, you get the original wideband squarewave.
if you take all the harmonics (all of which are odd) of a square wave and remove all the ones below the stopband frequency,(filters with flat response and flat phase in passband and stopband), you'll be left with all the odd harmonics of the square wave that were above the cutoff, obvious.
The part below the cutoff, from a flat mag and phase LPF, will be a square wave with Gibbs ripples as weve seen. The output of the upper harmonics by themselves, in the time domain, will be an inverted version of the Gibbs ripples, but without the squarewave shape part, and with some spikes where the square wave edges were.
The HPF ripples and spikes consist of the higher odd harmonics of the square wave, by themselves.
The LPF output consists of the square wave but with those harmonics removed.
If you summed them together again, you get the original wideband squarewave.
Last edited:
Arnyk, I was joking. I made my first multitrack digital recording in the 80s on a (fragile) Sony machine, then the Mitsubishi.
later, i helped Akai to design their recorder/virtual editor, the DD1500.
Instinctively, a brick wall seems very unnatural for me, and the only place where it should have a reason to be is before AD converters, to avoid aliasing.
Our technology allows to over sample, so we can make a soft analog filter after the audio band, then any brick wall will be far away from this limit, just to kill very little residuals noise...
Am-I wrong ?
Our technology allows to over sample, so we can make a soft analog filter after the audio band, then any brick wall will be far away from this limit, just to kill very little residuals noise...
Am-I wrong ?
Interesting.... that is my conclusion also. In fact, the LPF on the ADC is all that is needed..... then it goes to the DAC already band limited.
But for that HF junk shoved up higher still needs to be removed and in a nice way.
THx-RNMarsh
You need a brickwall LPF before Fs/2 (in a non-oversampling case) after the DAC even if the input is perfectly band-limited. Nyquist-Shannon does not work without a reconstruction filter.
The DAC output is going to be a series of steps, no matter what. That needs to be filtered to produce the original input.
The DAC output is going to be a series of steps, no matter what. That needs to be filtered to produce the original input.
Last edited:
That's where 8X oversampling helps. I mean real ones, calculating intermediate values. Then, an softer low pass filter with a trap at 8X Fs will be enough to smooth the curve.
I prefer CFA amps, because their high slew rate. With a 6dB low pass filter at their input, assuring the signal entering in the loop will never reach this slew-rate. That's for IM.
Now, I'm interested to see if Richard can find anything about the little residues at HF(noise and remaining s harmonics).
Richard are we on the same page ? (post-DAC filtering)
You are still under the impression that a LPF will remove/minimize the ripples from a square wave (and not that the LPF is the reason for generating the ripples)?
Arnold, thank you for the link ( oh these engineers!)
A valuable gift
George
Yes George I'm at a loss as to how to make this point. Richard, let me put it this way the scope photo (with "ringing") that you posted has 1kHz plus the first ten odd harmonics ONLY i.e. the 21kHz harmonic is it, nada, nothing above that in the waveform. You CAN NOT reduce the peaking without removing signal in the audible range or violating Nyquist.
Doing this at 96kHz sampling you can filter and allow leakage from 22.05k to 48k and yes it looks better. When you burn a Red Book CD a CORRECT sample rate conversion process will put the "ringing" back.
Last edited:
Yeah, Richard, much of the smooting you speak of is ironed out not by analog filters, but by organoleptic filters, when excited by quality ouzo. Or in my case, Imiglikos wine.Unlike music, it NEVER fails to bring a smile to my face, ear to ear.
Early here, but the Champagne and pinot noir are ready for dinner.
:
Early here, but the Champagne and pinot noir are ready for dinner.
And those would be...? For us, Cedric Bouchard (Champagne and Pinot Noir all at once), some older La Chapelle (not sure which one, have to count bottles), and a Banyuls Sec.
- Status
- Not open for further replies.