[SOLVED!] loud digital distortion on peaked clips -- is there a setting for this?

I feel like I'm going to fail at trying to describe this but I'll try anyway:
 
In my old setup (8.5/Vista) I could hit the red from time to time and no real harm done.  Yes you would hear some distorting if you went too loud but it sounded like typical analog distortion.  In my new setup (X3d/Win7) if I hit the red one time it causes massive LOUD distortion.
 
I only hear it on playback, when I'm recording I don't hear it but I can see it if I hit the red if I'm watching.  Here's an example, I set this so it just barely hit the red: http://www.koskenmaki.com/flute.wav (edit: oops I didn't realize I was printing the delay as well but you get the idea.)
 
Nothing has changed in my hardware.  This sound is triggered on an external computer which goes ADAT into my mixers (02R96) then ADAT into my 2408s and recorded in Sonar.  (I print flute samples all the time because I prefer to apply audio effects in Sonar.)
 
Clearly you don't want to hit the red but I don't want an occasional peak to ruin a take.  Is there a setting I'm missing?

I could not hear your example but what you describe would be digital clipping. To fix it in future be sure that you have the source at the right level. In Sonar be sure the record meters are at unity gain.  That is zero.  Read the peak level. Do not adjust those faders when you are recording. They will have no effect on the signal but will give you a false reading if you move them. You set your level at the source and at the inputs to your audio inputs. 

I'm not talking about mixing. I'm only talking about recording. One way to check that the recorded audio is clipped or not is to look at it. Clipped audio will look flat where it should have a smooth parabola looking top and bottom. 
 
Your levels for recording are dictated by the very loudest sound you record. If that is not clipping nothing else will.
I am not that worried about the signal going into the red. Its clipping that is of great concern. Clipping means it is louder then the A/D converter can handle and will simply cut off the top. Its that cutting off the top that can sound bad.
 
I did listen to your sample and it very well could be due to clipping but I'm not sure.
 
I get the feeling that you are not understanding what I am telling you about setting up for recording.
 
This is very important. If you set the record track fader at any value other then zero you will not get a true reading of what is streamed to disk.  Adjusting it while recording will not change the level going to disk only what you see and hear if you are monitoring. It will look like you are lowering the level when you are not. That has to be done at the signal source. 
 
 

Yes!  This is what changed!  Thanks a ton.  And thanks John also for your help much appreciated.
 
This was set to 16-bit I set it to 32-bit and now on recording I can *see* the clipping and I'm sure the sound is compressing but I don't hear the massive digital distortion.  Should this be set to 64-bit?  I never messed with it before I don't remember what my old system was set to.  Currently record and render are both set to 32-bit.
 
All the tweaking I've been doing to the audio settings I always look under "audio" never thought to look under "file".

The bit depth isn't going to do anything except give you greater dynamic range and a lower noise floor. Plus no A/D converter can record above 24 bits. 
 
You need to lower the output from the source. 

Matt
OK So Audio Driver Bit Depth is set to 24 that's what I should set record bit depth to?

Makes sense to me. The driver delivers 24bits so record at 24bits.

Audio within Sonar can be any bit depth and its rendered that way. From 16 bits to 64 bits.  Audio coming in to Sonar via an A/D converter is converted to a maximum of 24 bits no matter what the file setting is. 
 
You should not have the problem you describe by just going into the red, slightly. (Going into the red is not clipping. It has to go above zero to clip) Sonar as rule is working at 32 bits floating point or 64 bits floating point and should not clip while mixing unless you send audio out to the main outs that is too loud. Sonar wont clip internally. Something else is causing the problem and I suspect its overloading the outs of your audio device. It will clip.

Even with all that SONAR is communicating with the 2408 via a 24bit driver.

Sounds like also the signal is taking a very roundabout route to get from you source to the DAW. Are all your digital clocks setup so they are locked correctly. If they are not at any point you can get some strange sounds too.

I have an Apogee Big Ben providing clock to every digital device, it's rock solid.
 
The routing is not that complex I just do everything with an external mixer, not within Sonar like a lot of you.
 
I have 6 external computers that output 48 digital outs (8 each) into a pair of 02R96s.
I have 16 analog synths that get summed down to 8 outputs via a pair of Neve 5059s and then into the 02R96s.
I have 7 channels of ADAT audio (56 outputs) going out of the 2408s into the 02R96s.
 
Make sense so far?  112 channels meet and get mixed in external mixer.
 
Going back IN to Sonar via the 2408s I just have 1 channel of ADAT (8 inputs) so I can record in Sonar plus the stereo digital SPDIF signal which carries the final stereo mix which I record/bounce in Sonar.
 
Despite not being able to communicate my issue/problem, it is now "fixed" and I am very happy.
 

Two things. 
Why are you re recording a sample from another computer? Can you not just simply transfer the WAVE / FLAC or whatever file with a USB stick? Or at least use "what you hear"  as a input and record it internally. 
 
This does sound like a clocking issue as you are using a chain of digital devices. 
The clock would have to be driving the first computer, the RME interface, the 02R96, the ADAT card and the recording DAW....This is a job for a master clock source for sure. 

Please don't take this the wrong way, but you should go and read up on gain staging.
 
Getting your levels set correctly all the way through the chain is the easiest & most effective way to avoid these sort of problems

Based on what you've said, what solved the problem, and what I saw in the recording, I suspect there was some sort of "wrapping" of amplitude between peaks and troughs going on due to mismatched bit depths. I've seen some cheap audio interfaces do this when the input level exceeded the limit of the A/D converter. Instead of clipping it, the converter would "wrap" the sample value around to the other side of the waveform. Some sort of Two's Complement programming error.