[hpsdr] Question about I and Q position in protocol

[John Marvin]

  So I've been trying to play around with direct access to the Hermes 
data, but I wasn't getting the data I expected.  I spent some time 
trying to figure out what was going on, but eventually decided to create 
a test file with samples in HPSDR protocol order and then inject them 
into PowerSDR to help visualize what was going on (I just opened the 
file and substituted the samples coming from Hermes with the samples 
from my file in a loop).  I created the test file by taking 4 songs, 
band limiting them to 6 Khz mono and putting them in a -48 to 48 Khz 
range using libfftw. If everything worked as I expected I would have got 
each song in order from left to right, each in a 6Khz range with a 6 Khz 
blank range between them. Instead what I got was everything backwards.

Luckily it was easy to verify that, since LSB reads everything 
backwards. I must admit I spent close to an hour just playing around in 
PowerSDR with the sub receiver feature and listening to my four songs.  
Part of that was just listening to the four songs I happened to choose, 
since they were old favorites I hadn't listened to in a long time; 
however, there was something strangely fascinating about using PowerSDR 
to switch between four (low quality) songs by changing VFO frequencies. 
Here's a screen capture for those who are curious:

http://www.themarvins.org/hpsdr/foursongs.jpg

So, the answer appeared simple. Somewhere I must have switched I and Q.  
But I couldn't find where I had gone wrong.  So I eventually tried 
tracking the source in PowerSDR, from where samples were received from 
Hermes until they were made into complex samples.  I finally wound up in 
process_samples() in DttSP/sdr.c, where I found the following code:

    CXBimag (rx[thread][k].buf.i, i) =
                 bufl[i], CXBreal (rx[thread][k].buf.i, i) = bufr[i];

I have carefully followed the path and believe that bufl was the "left" 
sample from Hermes and bufr was the "right" sample from Hermes, and here 
the code is assigning the left sample to the imaginary component and the 
right sample to the real component.  The protocol document indicates 
that left is the I (in phase or real) sample and right is the Q 
(quadrature out of phase or imaginary) sample. So, am I missing 
something, or should the I and Q positions be switched in the protocol 
document?

Thanks,

John