doug@homxc.UUCP (D.SULPY) (12/27/88)
I read here last week that one of the sampling programs (I forget the name) was able to digest several minutes worth of information. Is it worth dreaming of a system, then, which could sample a poor quality recording (say, a twenty year old radio performance), and use the abilities of the Amiga to clean it up? Or eliminate reverb (assuming the waves for reverb are simply smaller clones of the parent wave), or isolate an instrument or voice, or remove pops, clicks, and scratches from a record? I know some very expensive systems will do this kind of thing (generally using the Mac).
rsilvers@hawk.ulowell.edu (Robert Silvers) (12/29/88)
In article <4732@homxc.UUCP> you write: >I read here last week that one of the sampling programs (I forget >the name) was able to digest several minutes worth of information. >Is it worth dreaming of a system, then, which could sample a poor >quality recording (say, a twenty year old radio performance), and >use the abilities of the Amiga to clean it up? Or eliminate The main problem with this idea is that the Amiga is limited to around 30,000 Hertz 8 bit samples. This is not exactly Hi-Fi. I would describe it as "micro-cassette" quality. Basically we are talking about sound processing, as opposed to image processing. You could do it on a theoretical level with an Amiga, but not on the professional level. The NeXT would be a good system to use. It has stereo 16 bit sound output, and a high-speed signal processing chip. DAK sells a device that removes clicks and pops from records. I have seen other boxes that remove hiss. A system like the NeXT could be programmed to do all of this and more. That would be pretty neat. As for the Amiga, well, we will have to see if the 3000 will offer true compact- disc-quality sound. --Rob. Robert Silvers. rsilvers@hawk.ulowell.edu Box #1003 University of Lowell. Lowell Ma, 01854 (508) 452-5000 ex 2233
dleigh@hplabsz.HPL.HP.COM (Darren Leigh) (01/04/89)
In article <10955@swan.ulowell.edu> rsilvers@hawk.ulowell.edu (Robert Silvers) writes: >In article <4732@homxc.UUCP> you write: >>I read here last week that one of the sampling programs (I forget >>the name) was able to digest several minutes worth of information. >>Is it worth dreaming of a system, then, which could sample a poor >>quality recording (say, a twenty year old radio performance), and >>use the abilities of the Amiga to clean it up? Or eliminate > > The main problem with this idea is that the Amiga is limited to around >30,000 Hertz 8 bit samples. This is not exactly Hi-Fi. I would describe >it as "micro-cassette" quality. Basically we are talking about sound >processing, as opposed to image processing. You could do it on a theoretical >level with an Amiga, but not on the professional level. The NeXT would be a >good system to use. It has stereo 16 bit sound output, and a high-speed >signal processing chip. > > DAK sells a device that removes clicks and pops from records. I have >seen other boxes that remove hiss. A system like the NeXT could be >programmed to do all of this and more. That would be pretty neat. As for >the Amiga, well, we will have to see if the 3000 will offer true compact- >disc-quality sound. Be nice, Rob, and give the guy some real information! Never discourage someone who wants to hack. Of course the current Amiga sound hardware is limited in both speed and resolution but there is still plenty of kick there. Of course it won't be studio quality, but it's still worth playing around with. An excellent quality phone connection uses a 64 Kbps digital data stream, i.e. 8000 samples/second and 8-bit samples. It can, therefore, convey an analog signal with a maximum frequency of 4000 Hz and a 48 dB signal to noise ratio. The amiga sound outputs are eight bits wide and the supplied reconstruction filter cuts off around 7000 Hz; we can easily beat a phone connection bandwidth-wise. I understand you can disable the filter on the 500/2000 and use an external one to take advantage of the Amiga's 30,000 samples/sec capability. Granted this would sound horrible if you amplified it much (lots of quantization noise ==> hiss), but is quite tolerable to just listen too. As far as processing goes, the original poster wanted to get rid of hiss, clicks, pops and reverberation. Hiss can be taken care of to a certain extent with filters, but you always lose some information by getting rid of the hiss since they occupy the same portion of the frequency spectrum. For example, a high-pass filter will get rid of much of the annoying stuff, but will also remove the high frequencies that are part of the music. Remember, though, that you will never be able to do better than the 48 dB S/N ratio. Clicks and pops can be handled in the time domain (they're easy to see, you could edit them out by hand if you wanted). Reverberation is a little harder to get at. One technique I saw involves complex cepstrums (don't ask, it's something hairy that has to do with homomorphic signal processing) and is computationally expensive and only works for simple cases (one echo). Remember the Carruso (sp?) recordings that were cleaned up digitally and received so much press? They used a technique called blind deconvolution (more complex cepstrums) to get rid of some distortion. Other things that could be played with are FFTs: mess around, do voice prints, speed up or slow down a recording without modifying the pitch. There are all kinds of fun things to do that are easily within the reach of the Amiga if you can live with the bandwidth, S/N ratio and not being able to do things in real time. (And I maintain that these are easy to live with if you're not worried about professional quality). To get started you'll need an Amiga, a sampler and a good book on digital signal processing (I have Oppenheim and Schafer). Armed with these and some knowledge of calculus and complex numbers you should be up and hacking in no time. If I had a decent sampler I'd probably do something myself. ======== Darren Leigh Internet: dleigh@hplabs.hp.com UUCP: hplabs!dleigh