Attached is a smoothed FFT plot done by Speaker Workshop. It's the response of a stereo pair of Locust I's at about two meters, in my office. The subwoofer is muted.
It looks reasonable down to 30 Hz or so. It starts rolling off around 70 Hz, as Martin's worksheet says it should. It does not roll of as quickly as predicted, being down 6 dB around 50 Hz rather than 63 or whatever. Maybe that's because of room lift in the small office. Otherwise the rolloff curve looks plausible. But what's going on below 30 Hz? I feel pretty sure those little speakers are not actually functioning at those frequencies.
I don't know how to set the FFT parameters in Speaker Workshop. For this plot, I checked the "Use entire range" button, and used uniform windowing.
It looks reasonable down to 30 Hz or so. It starts rolling off around 70 Hz, as Martin's worksheet says it should. It does not roll of as quickly as predicted, being down 6 dB around 50 Hz rather than 63 or whatever. Maybe that's because of room lift in the small office. Otherwise the rolloff curve looks plausible. But what's going on below 30 Hz? I feel pretty sure those little speakers are not actually functioning at those frequencies.
I don't know how to set the FFT parameters in Speaker Workshop. For this plot, I checked the "Use entire range" button, and used uniform windowing.
Attachments
Ideas
I've got a few ideas for possible sources of error;
First, (collecting data) What microphone are you using? A lot of microphones don't work right for extreme low frequencies.
Second, (processing) What's the sampling rate/bit depth? I've noticed with some software I have used that the results of a FFT in the low frequencies changes based on sampling rate and/or bit depth. If this is a software problem, it would be best to post on the forum for Speaker Workshop. It looks like they have a good forum over there.
Third, what if the data is right?
How large is your office? Perhaps the speakers are setting up some sort of resonance; or perhaps its a subsonic rumble from passing trucks. If there are low frequencies present and the software is accurately processing them, then the next thing to check would be the environment.
Those are just my guesses. Good luck!
Joe Carrow
Dave Jones said:
I've got a few ideas for possible sources of error;
First, (collecting data) What microphone are you using? A lot of microphones don't work right for extreme low frequencies.
Second, (processing) What's the sampling rate/bit depth? I've noticed with some software I have used that the results of a FFT in the low frequencies changes based on sampling rate and/or bit depth. If this is a software problem, it would be best to post on the forum for Speaker Workshop. It looks like they have a good forum over there.
Third, what if the data is right?
How large is your office? Perhaps the speakers are setting up some sort of resonance; or perhaps its a subsonic rumble from passing trucks. If there are low frequencies present and the software is accurately processing them, then the next thing to check would be the environment.Those are just my guesses. Good luck!
Joe Carrow
I agree. the information below 50hz seems erroneous
Search the local forums for SW help
www.audioasylum.com has some,use its powerful search utility
Take them outside and compare( if possible )
Search the local forums for SW help
www.audioasylum.com has some,use its powerful search utility
Take them outside and compare( if possible )
It's spurious errors caused by the inaccuracy of your soundcard. You will notice with most soundcards below ~20Hz or so, depending on how soon they start to roll off, that there are a bunch of spikes in measurements.
Ignore them, or get a better soundcard. Ithink SBlive will frequently get down to 10Hz without big errors like this, but all soundcards I have used (onboard or cheapie PCI) have had problems below 20-30Hz.
Ignore them, or get a better soundcard. Ithink SBlive will frequently get down to 10Hz without big errors like this, but all soundcards I have used (onboard or cheapie PCI) have had problems below 20-30Hz.
I don't use Speaker Workshop, but I have noticed some peculiarities when measuring. The microphone output from low frequency wind noise can be much greater than the signal amplitude. Look at the mic preamp output with a 'scope and you'll see the signal waveform bounce off screen. Move your arm towards the keyboard and it bounces again. I built a mic preamp for the electret insert and included a 7Hz 4th order high pass filter, which removed the bounce.
Soundcard measuring systems have a similar lf response, but even if the fft filters the low frequencies, the amplitude may be enough to clip the inputs. When the mic output is sampled, how does the fft know if the voltage is the signal picked up from the speaker, or the draught created by your body movements?
When I've auto re-measured the response, at 1 to 2 second intervals, the higher frequency response doesn't change, but the lf response changes with each aquisition.
So maybe you should keep still and out of draughts when measuring.
Hope this helps.
Soundcard measuring systems have a similar lf response, but even if the fft filters the low frequencies, the amplitude may be enough to clip the inputs. When the mic output is sampled, how does the fft know if the voltage is the signal picked up from the speaker, or the draught created by your body movements?
When I've auto re-measured the response, at 1 to 2 second intervals, the higher frequency response doesn't change, but the lf response changes with each aquisition.
So maybe you should keep still and out of draughts when measuring.
Hope this helps.
Do your impedance measurements show spurious peaks at low frequencies as well?
IF not, it may not be the probem I describe. My measurements often show peaks at 5Hz that are higher than the resonance peak.
Perhaps it is aliasing? Try the fastest sample rate if you haven't already, or try several sample rates. I know the SB Live worked best at 48kHz because it resampled all other sample rates to that rate internally.....
IF not, it may not be the probem I describe. My measurements often show peaks at 5Hz that are higher than the resonance peak.
Perhaps it is aliasing? Try the fastest sample rate if you haven't already, or try several sample rates. I know the SB Live worked best at 48kHz because it resampled all other sample rates to that rate internally.....
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.