EMUSIC-L Digest Volume 39, Issue 13 This issue's topics: Sampling Samplers around $1000 sampling & playback (4 messages) Your EMUSIC-L Digest moderator is Joe McMahon . You may subscribe to EMUSIC-L by sending mail to listserv@american.edu with the line "SUB EMUSIC-L your name" as the text. The EMUSIC-L archive is a service of SunSite (sunsite.unc.edu) at the University of North Carolina. ------------------------------------------------------------------------ Date: Thu, 30 Apr 1992 13:38:11 EDT From: "D. Yuen" Subject: Re: Samplers around $1000 I was thinking about putting an Frontal Lobe PCM Channel on my M1, but before I go too far - what is on the market in terms of Samplers fro around $1000.00? Any one using any Frontal Lobe stuff out there? What do you think of it? Andy Rooney question of the day...what is the attraction of Akai samplers? Whenever I read about some *rich rock and roll star* with keyboards they allways have some 10 Gigabite Akai with 4-wheel drive and dual outboards in their rig. Are the Akai's just sonicly cleaner, better, more resolution? Better Filters, programability or what? Dan ------------------------------ Date: Thu, 9 Apr 1992 11:39:49 MDT From: Julius Pierce Subject: sampling & playback I have two questions about digitizing & playback: 1. I assume that oversampling is an output technique where recorded samples are emitted at a faster rate (by duplication of the samples) and these samples are then digitally low-pass filtered. What kind of low-pass filter algorithm is used? Is it just a linear interpolation or something more sophisticated? 2. I know what "flash conversion" and "successive approximation" == "binary search" methods of A/D conversion are. What is "1 bit delta- sigma" conversion? julius@math.utah.edu ------------------------------ Date: Thu, 9 Apr 1992 14:53:00 EDT From: RAY BROHINSKY Subject: Re: sampling & playback Julius, I can't answer 1, because I just don't know. 2, however, I can explain, but only in terms of the only implementation I've ever seen. This consisted of a GAP-1 (which was a pair of op-amps connected to a switch in a single package) which was wired so that one of the amps would charge a cap, and the other would discharge it. (oh, yeah, the package also had a comparator that drove the switch!) If the cap voltage was above the input, the comparator drove the switch to connect the discharge amp, and the cap voltage would come down to the input level. The opposite held if the cap voltage was below the input. The comparator was run at a 1Mhz rate, so that the input was sampled at that rate, and corrections were in quanta of charge or discharge that lasted for 1 microsecond. Once the cap voltage is matched to the input, it will tend to follow it with a hunting-like pattern, charging above and below it alternately, with more quanta devoting to charging if the input is rising, more to discharging if the input is falling. Once the comparator drive signal is determined in this way, it can be transmitted to a similar device through a single line, but in the similar unit, instead of the comparator sampling the difference between an input and the local cap, it is simply driven as a slave from the master switch drive signal. I have also seen a similar system that determined how far off the cap was from the input, and adjusted the slew rate of the amps. To keep the system to a single bit of information, however, the converters worked on the basis that if so many consecutive bits were positive, then the slew rate would be raised. (actually, if so many consecutive bits were positive or were negative). If the bits alternated, the slew rate would be lowered. In any case, because of the high sample rate, and the fact that there was no storage involved in the cases I'd seen, the parts count was very low, and CMOS drivers and receivers for the one bit digital line made the system very resistant to noise that was generated outside the system. I can't say a lot about noise generated in the system, although the write-up seemed to think that having a clock freq that was about 50 times above the highest freq of interest would allow very simple low-pass filtering. Translating all that into digital language is left to the student 8^) raybro *** Nieder mit den Tueten! Freiheit fuer Gummibaerchen! *** ------------------------------ Date: Thu, 9 Apr 1992 16:33:35 EDT From: Brian Adamson Subject: Re: sampling & playback > > 1. I assume that oversampling is an output technique where recorded > samples are emitted at a faster rate (by duplication of the samples) > and these samples are then digitally low-pass filtered. What kind > of low-pass filter algorithm is used? Is it just a linear interpolation > or something more sophisticated? There are a couple of techniques used. Th linear interpolation can be used with what you mentioned, i.e. for 4X oversampling make 4 samples w/ 3 samples interpolating between 2 samples read off of the CD. Or use a higher order polynomial approximation for better interpolation. I think the more popular technique is to excite a low pass digital filter with impulses where the impulses are your original samples with zeroes in between. That is, for samples S1, S2, S3 ... to get 4X oversampling you would excite a digital low pass filter with S1, 0, 0, 0, S2, 0, 0, 0, S3, 0, 0, 0, ... If your original sample rate was 8 kHz, your low pass filter would be designed for 32 kHz sample rate with whatever desired cutoff freq (e.g. 4 kHz) and whatever order and shape of low pass filter you like (n-tap FIR, 2nd order Butterworth IIR). -- Brian Adamson NRL Code 5523 adamson@itd.nrl.navy.mil ------------------------------ Date: Thu, 9 Apr 1992 22:06:47 MDT From: Julius Pierce Subject: Re: sampling & playback Thanks Brian. That's certainly the way I'd prefer to increase the sampling rate. I just didn't know how sophisticated the li'l Si chippies were getting. julius@math.utah.edu ------------------------------ End of the EMUSIC-L Digest ******************************