rfg@hound.UUCP (R.GRANTGES) (01/18/85)
[?] Sounds like someone who knows what he is talking about, so perhaps you can comment on a concern I have: Assuming there <is> something to the argument that brick-wall filters mess up the sound audibly so that oversampling is beneficial, what about the brick-wall filter the signal is put through in the recording (encoding) process? a) I assume there is one. b) I assume it's brick-wall. c) Doesn't it mess up the phase in the same way that similar filters are alleged to in the decoding process? d) Is there anyway to undo those effects? e) is there any alternative? e.g., can oversampling or something be used at the encoding end with similar alleged benefits (still retaining the 44.1 rate on the disc)? Breathlessly awaiting enlightenment. -- "It's the thought, if any, that counts!" Dick Grantges hound!rfg
karn@petrus.UUCP (01/19/85)
> ... Assuming there <is> something to the argument > that brick-wall filters mess up the sound audibly so that oversampling > is beneficial, what about the brick-wall filter the signal is put > through in the recording (encoding) process? a) I assume there is one. Yes, there is. Its purpose is to filter out all frequencies above half the sampling rate. This is a much more critical function than the low pass filter on the player, since allowing a high frequency through the filter would produce unwanted aliased signals in the audio band. There would be no way to remove these components once they had reached the digital domain. > b) I assume it's brick-wall. c) Doesn't it mess up the phase in the same > way that similar filters are alleged to in the decoding process? d) Is > there anyway to undo those effects? e) is there any alternative? e.g., > can oversampling or something be used at the encoding end with similar > alleged benefits (still retaining the 44.1 rate on the disc)? Ideally, it would also be brick wall. But brick wall filters cannot be built so some practical approximation must be used. I don't know exactly what is in the Soundstream recorders, but you can make very good analog filters if you are willing to pay enough. In particular, if you make a Bessel (maximally linear phase) filter you need LOTS of sections to get a quick cutoff slope. Perhaps in commercial recorders the extra money available makes it feasible to build them this way. I suspect that they simply use Butterworth filters, compromising between the Bessel and the elliptical (Cauer) filters which are common in CD players. Digital low pass filtering similar to that used in the Phillips players could be used here, but is considerably less practical since it requires faster A/D converters which are considerably more expensive than faster D/A converters. If you knew exactly what kind of filter was used in the recording process, yes, all you would have to do is construct a filter with a compensating phase response to make the concatenation of the two "linear phase". In modem design, this is known as "equalizing the channel". But this is easier said than done (it requires a lot of analog allpass sections or, more easily, transversal digital filtering), and it all hardly matters anyway. Just sit back and enjoy the music and quit worrying about all the red herrings. Only my oscilloscope can tell the difference. Phil
shauns@vice.UUCP (Shaun Simpkins) (01/21/85)
Well, Phil said in greater detail than I would what I was going to say about input filtering, so I won't say it. However, I can offer some interesting evidence to his contention that one should just sit back and listen, not worry - a study has been brought to my attention (I am trying to get a copy of it now) that examines the psychoacoustic effects of brick-wall filters. High-order filters similar to those used in Sony products were cascaded and auditioned. Two cutoff frequencies were employed - 5KHz and 20KHz. Result - differences could be heard between the single 5KHz section and a cascade of 4, while no differences could be heard with the 20KHz cutoff filters, again in a cascade of 4. I do not know whether these filters were compared with Bessel filters of equal order. I suspect they were, and will report to the net. By the way, the bandwidth shrinkage that you would assume in such an experiment (final bandwidth = sqrt N where N is the number of identical cascaded stages) is not correct. Since these filters' responses aren't Gaussian the shrinkage is much less. Therefore, the researchers claim that any audible differences are primarily caused by phase aberrations. Whoopee. Check out the latest issue of TAS. They auditioned the CD-X1, the NEC and the KYOCERA player. RADICAL differences between them. DISGUSTINGLY inferior to the digital master tape. I didn't heard much difference at all when I listened - I might believe that the CD is less pristine than the master. I'm still waiting for Rounder Records to issue their first CD... The wandering squash, -- Shaun Simpkins uucp: {ucbvax,decvax,chico,pur-ee,cbosg,ihnss}!teklabs!tekcad!vice!shauns CSnet: shauns@tek ARPAnet:shauns.tek@rand-relay