From: uhclemLOSE.dec06@nemesis.lonestar.org   
      
   Kevin Michael Vail  wrote:   
   : I also remember the work it took to do the different notes from just a   
   : few sound samples, interpolating things on the fly.  That was pretty   
   : challenging as well.   
      
   I ended up with the ball on the various music/sound programs when Kevin   
   went to pastures greener in the summer of 1988 just prior to the launch   
   of the SL/TL systems...  (Or maybe he really got tired of listening to   
   a couple of those songs over and over and over and over - I certainly did!)   
      
   At least the PSSJ DAC could play (but not record) audio on pitch,   
   which was impossible for the TI 3-voice, as implemented on the   
   IBM PCjr, and then copied right down to this same mistake in the 1000s.   
   It would have required an additional crystal oscillator running at one   
   of the frequencies specified by TI to have the TI chip produce sounds   
   on-pitch.  Instead, hardware designers used an already-present clock   
   that was "close", probably not expecting anyone to use the audio for   
   anything other than zap-bang sounds and tinny music in games.   The   
   limited choice of audio waveforms (square, sawtooth and I forget the   
   other but it wasn't sine) the TI chip produced limited the produced   
   sound quality as well.   
      
      
   On the digital waveform audio side...   
      
   Getting plain recorded audio recorded on the 1000 to play back at the   
   exact same speed was also impossible in models before the 1000 TL/3.   
   This was due to yet-another bit of corner cutting in the gate-array,   
   a problem I had to explain and re-explain to management perhaps a hundred   
   times.  This was something that could not be consistently concealed by   
   any amount of software trickery.   
      
   Of course, what we had was about 35 cents of parts above what it   
   took to provide the joystick capability to provide sound, and   
   certain VP management types now expected the 1000s to produce grand   
   piano sound quality, AND be dead on pitch, but not change the   
   hardware to do it.  Worse, the entire point of having the ADC   
   was to cost-reduce the joystick hardware and eliminate the tedious   
   calibration that was done on each machine at the factory, so   
   "sound quality" wasn't even on the requirement list.   
      
   In the original PSSJ chip, there was only a single byte register   
   (the sampling latch), so a new sample could not start to be acquired   
   (done by successive approximation) until the previous one had been read   
   by the CPU or DMA, and how long that took depended entirely on what the   
   bus, CPU and DMA chip happened to be doing at that exact instant the   
   latest sample became available for reading and what speed the   
   processor was and what the bus speed of the machhine was.  The   
   Intel processors of the day would not release the bus to the DMA during   
   the execution of an instruction (or a sequence of instructions if LOCK   
   was used) and since different instructions and operands take different   
   amounts of time to execute, this made the timing even more unpredictable.   
      
   Only after that sound byte was finally read by the CPU/DMA would the next   
   audio sampling begin, a task that took 10 ticks of the clock driving   
   the approximator.  (The record clock in essence was 10X the playback   
   clock to the same device.)  This delay on making each sample meant   
   that recorded samples would be done at irregular intervals and the   
   effect was much as you would get with making a recording on tape at   
   a slightly slower speed than the eventual playback speed, PLUS having   
   some debris on the pinch roller only during the recording process   
   to also make the recording speed irregular.   
      
   After almost two years of persisting in wasting my time on various   
   ordered but doomed-to-be-less-than-desired attempts to conceal this   
   and other hardware omissions with more and more convoluted software,   
   Tandy hardware management finally allowed a new turn of the PSSJ   
   chip which added an entire second byte of buffer to the A-D sampler,   
   or about 20 to 30 gates.  (I think the real reason I got the buffer   
   was that management wanted changes in the printer section of the   
   chip to fix bi-directional printer port support and so were touching   
   the chip design anyway.)   
      
   These flaws in digital recording are also the main reason for the   
   alteration of the format of the .SND files, because the previous file   
   header format had no unused fields or even bits anywhere, and I needed   
   some information to indicate that the recording was made on old or new   
   chip and if not the newer chip, what older computer made the recording,   
   so a bunch of routines could activate and make efforts to conceal the   
   recording errors from listeners ears.  I left lots of room for future   
   expansion in the new file format once that was approved, but development   
   started to ramp down after the Deskmake sound/music programs and the   
   Sound Toolkit were made to work again for the TL/3 and 2500 platforms.   
      
   I remember waaaaay too much of that project...   
      
      
   Frank Durda IV - send mail to this address and remove the "LOSE":   
       http://nemesis.lonestar.org   
      "I'll huff and I'll puff, and I'll get promoted."   
       - Old Tandy management saying   
   Copyright 2006, ask before reprinting.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)