SF10's Spectrum Analysys tool broken...? -Sound Forge - PC

Subject:SF10's Spectrum Analysys tool broken...? Posted by: shanksworthy Date:8/4/2010 11:19:52 AM I frequently use the Spectrum Analysys tool in Sound Forge 8 with swept sines and frequency impulse train input signals. Recently I had a need to use the multi-channel features of SF10, so I downloaded the demo version for evaluation purposes. However, in SF10 there appears to be a bug; when using the Spectrum Analysis tool to measure a stereo impulse train signal with identical left and right waveforms, the left-channel signal shows up in the frequency domain as being 6dB louder than the right-channel signal, even though the left and right channels are bit-identical in the time domain. Also, the left signal rolls off at about 2k, whereas the right signal is flat accross the entire frequency span. For example: 1. My input file: Stereo impulse, identical left/right content... [IMG=http://i138.photobucket.com/albums/q242/Shanksworthy/sound%20forge/waveform.jpg] 2. The above waveform graphed in the spectrum analysys tool. Note how the sa tool incorrectly shows a left/right imbalance, and also note the rolloff on the left channel only... [[IMG=http://i138.photobucket.com/albums/q242/Shanksworthy/sound%20forge/spectrum_analysis.jpg] 3. My settings (FWIW)... [IMG=http://i138.photobucket.com/albums/q242/Shanksworthy/sound%20forge/settings.jpg] Notes: - This doesn't appear to be a problem when using real-time monitoring; it only happens with static measuremts. - The behavior isn't exhibited when measuring sine waves ... only impulse trains (i.e. a series of single sample spikes with "infinite" frequency content). - Using the exact same inputs with the exact same SA settings in SF8, the graph shows the left/right frequency responses overlapping perfectly, as expected. Does anyone have any idea what's going on here, and how to fix it? Message last edited on8/4/2010 1:59:56 PM byshanksworthy.
Subject:RE: SF10's Spectrum Analysys tool broken...? Reply by: Geoff_Wood Date:8/4/2010 4:12:47 PM Will try and duplicate your results - more soon ... I assume your SR is 48K ? geoff Message last edited on8/4/2010 4:15:32 PM byGeoff_Wood.
Subject:RE: SF10's Spectrum Analysys tool broken...? Reply by: shanksworthy Date:8/9/2010 11:11:01 AM Thanks! Yes, my inputs in this case were 48k -- I can't remember if I tried with different sample rates.
Subject:RE: SF10's Spectrum Analysys tool broken...? Reply by: Geoff_Wood Date:8/9/2010 8:20:43 PM Yeah - I got something similar, athough the 'base' level before rolloff was at -104dB, which is a little odd for a 16 bit file which should max out at -96dB ! I ALSO got an error "Sound Forge has stopped working" and submitted it. Something went wonky in forgek.dll . (XP-Pro 32 SP3) geoff Message last edited on8/9/2010 8:23:12 PM byGeoff_Wood.

I frequently use the Spectrum Analysys tool in Sound Forge 8 with swept sines and frequency impulse train input signals. Recently I had a need to use the multi-channel features of SF10, so I downloaded the demo version for evaluation purposes.

However, in SF10 there appears to be a bug; when using the Spectrum Analysis tool to measure a stereo impulse train signal with identical left and right waveforms, the left-channel signal shows up in the frequency domain as being 6dB louder than the right-channel signal, even though the left and right channels are bit-identical in the time domain. Also, the left signal rolls off at about 2k, whereas the right signal is flat accross the entire frequency span.

For example:

1. My input file: Stereo impulse, identical left/right content...
[IMG=http://i138.photobucket.com/albums/q242/Shanksworthy/sound%20forge/waveform.jpg]

2. The above waveform graphed in the spectrum analysys tool. Note how the sa tool incorrectly shows a left/right imbalance, and also note the rolloff on the left channel only...
[[IMG=http://i138.photobucket.com/albums/q242/Shanksworthy/sound%20forge/spectrum_analysis.jpg]

3. My settings (FWIW)...
[IMG=http://i138.photobucket.com/albums/q242/Shanksworthy/sound%20forge/settings.jpg]

Notes:
- This doesn't appear to be a problem when using real-time monitoring; it only happens with static measuremts.
- The behavior isn't exhibited when measuring sine waves ... only impulse trains (i.e. a series of single sample spikes with "infinite" frequency content).
- Using the exact same inputs with the exact same SA settings in SF8, the graph shows the left/right frequency responses overlapping perfectly, as expected.

Does anyone have any idea what's going on here, and how to fix it?