Subject: Re: [linux-audio-dev] Gigasampler Clone
From: Juhana Sadeharju (kouhia_AT_nic.funet.fi)
Date: Sat Jul 01 2000 - 01:16:13 EEST
>From: Nolan <gte317h_AT_prism.gatech.edu>
>
>There is something called "willfull infringement" which is many times
>worse than normal patent infringement. Since you've just announced your
>intentions to violate (and your knowledge of) gigasampler's patent, you
>need to be damn sure you can win that case...
Good to know!
If we write the sampler playing samples directly from the disk, then
we have to find prior art on it.
The patent abstract broadly tells about getting low-latency file playing
by storing beginnings of each needed segments to memory. In sense of
plain file playing, I found prior art from Deja News on this low-latency
file playing.
We yet don't have any prior art for the sampler system using the above
low-latency file playing. (I designed and partially wrote such sampler
application, but at most I wrote one e-mail on it at 1992 or such.)
There indeed is one system done earlier. Music-dsp archives has following:
For the last 3 years I've spent my time developing a DSP audio
system, a high-end sound sampler with direct-from-disk sample playback.
The system is working now but is very expensive (about 50000 $),
so I won't get a chance to work with it as an artist.
Author Marcus Verwiebe <helium_AT_cs.tu-berlin.de> wrote me details.
I copy the mail here without permission:
The work we did was an exclusive development for an (mysterious)
electronic studio in switzerland. There had been plans for making it
available in the general market, but it never really finished as a usable
product. Also in this case, it took us three times longer than we planned
it, until it was, let's say, alpha released. It was a combined
hardware/software device, would run on ethernet with a silicon graphics
indy as host for user interface, sample editing and processing.
There was an synchonous bus for sample transfer, capable of 128 stereo
samples per frame (32, 44.1 or 49 kHz), our AUDIOBUS. There was a VME-Bus,
a Motorola 68040 Processor as a local controller, there was a
network-card providing TCP/IP socket-like connections, a
multi-channel MIDI-card with a dedicated 68k, a SCSI-controller card
with a dedicated 68020, and up to 4 DSP-cards, each equipped with 4
Motorola 65301 processors, 24kWords of RAM, and a pair of AES/EBU digital
audio interfaces.
The signals/instruments/orchestras were stored on a couple of
wide-differential SCSI drives, the SCSI controller card included 256 Mbyte
RAM for the first second of sample storage for up to 1000 different
signals. The ASIC logic on that board could handle sustain and release
loops, check for the need of sample block reload, and would shuffle
samples forth and back to the DSP cards for samplerate conversion,
envelope and LFO processing.
The immense DSP processing power was needed because of the high-end
approach, the interpolation filter in the samplerate conversion stage was
using
20-40 taps depending on the transposition, whereas the studio-standard
(patended) famous EMU filters (Emulator III ...) are using only 8 taps,
implemented in programmable logic hardware. We checked them for our
design, and it was not the frequency response that our customer requested.
I remember asking last if they made anything public but got no reply.
It is a very impressive system anyway.
Juhana
This archive was generated by hypermail 2b28 : Sat Jul 01 2000 - 01:42:28 EEST