I might be wrong on parts of this explanation. If I am, I apologize, anyone is free to correct me.
If you have a signal coming into the DA7 on the analog mic/line input, you typically adjust the MIC/LINE input knob with the fader at 0dB so that you get a max. signal without clipping or distorting (same way you set the gain trim for most consoles). In some cases a certain amount of occasional clipping is tolerable if short enough duration. When you then push the fader above 0dB you start to distort the signal because the fader is adding level to something which is already peaking at 0dB (if you you have adjusted the channel trim correctly).
Same goes for digital tape returns from MDMs. You have recorded a signal to your ADAT and the signal peaks at 0dB, so you have the hottest possible level to tape. When that tape channel comes into the DA7 through lightpipe, you will have the identical level as recorded on your ADAT with the fader at 0dB (you can see this indicated on the meterbridge). If the fader is pushed above 0dB, you will add level to the signal and start to distort the signal.
What happens when you push the fader higher? The audio gets digitally scaled up. The details of the whole operation are probably a lot more complicated than I am explaining, but technically you are correct it does add more 1s and 0s to the bit stream in a way, it makes the value of the bit word higher as long as it does not pass max. value (I don't think it can add bits onto the bit word length).
Each sample has a value either above 0 for the positive half of the waveform or below 0 for the negative half of the waveform. With 16 bit signals for instance there are 65536 total values (2 to the 16th power), the max. positive value is 32767 (0x7fff) and the max. negative value is -32768 (0x8000), these are constant. When a signal is at "hottest" level, the maximum positive half of the waveform does not go above 32767 and the max. negative half of the waveform does not go below -32768.
A -3dB 16 bit digital signal peaks at a certain value below max. On the positive side of the waveform I think it's about 23197 (using the dB = 20log(value/32767) formula). When you add 6dB to the signal using the fader, it pushes that level above the 32767 max. positive limit of 16 bit and since there are no values allowed in 16-bit above 32767, unfortunately is clipped to the max. positive value of 32767. Same thing goes for the negative half of the waveform.
If you look at clipped signals in a waveform editor, it looks like you gave the waveform a crew cut.
Depending on how severe the clipping is, you may or may not hear it. Short instances of 10mS or so may not be noticeable. If the audio is pushed way over the threshold, you will definitely hear it and it will change the sound.
The short of it is, you should monitor each channel on the meter bridge to make sure you are not adding level to it and causing the channel to clip.
I believe scene memory 00 is filled with the current automation file start scene when you recall it. Save the automation files to guarantee each start scene is saved along with automation data.
Once again, sorry if any of this is incorrect. You should go to a book on digital audio for a more complete explanation.
[This message has been edited by gregk (edited 07-10-99).]