Pretty much. A conventional spectrogram is produced using a fixed window width, which gives it the same time resolution at all frequencies. The frequency resolution is similarly a fixed number of Hz, 10 Hz if a 100 ms window was used, for example. At low frequencies that is a big octave fraction (1/1.4 octaves at 20 Hz), at high frequencies a very, very small octave fraction (1/1386 octaves at 20 kHz). For a time-frequency plot it would be handier if the tradeoff between time and frequency resolution varied with frequency, with higher time resolution at high frequencies and lower at low frequencies. A wavelet transform can achieve that, specifically a constant Q wavelet transform. A constant Q wavelet transform is mathematically equivalent to using a frequency-dependent window to produce the spectrogram (which is what REW does). For reference, this is how a perfect impulse looks with the same settings.
1) how does this compare to a s(tockwell) transform? (superficially they seem like quite similar things)
2) what does the dotted line really mean? it is described as the "peak energy curve" but I'm not sure what that actually means?
3) should one normalise or not?
ST and CWT with Morlet wavelet are very similar when looking at amplitude. ST retains absolute phase information, but that's not relevant for the uses these plots are put to.
The dotted line shows where the maximum level occurs at each frequency.
If the main interest is relative timing normalisation is helpful in seeing where the peak occurs vs frequency, but of course it very much distorts the amplitude since it lifts the content at each frequency up to achieve the same peak value.
I was giving this yet another try recently, this time focusing on phase aligning my subwoofer to my midbass speakers. See the attached MDAT...
Trace 1 is the Subwoofer by itself, with no time delay.
Trace 2 is the Midbass speakers playing together without the Sub.
I compared those traces in the Overlays Window using the Phase tab. I then adjusted delay on my sub and took additional measurements until I got the phase to align at the crossover range. 80Hz is where the sub and midbass are crossed over.
Trace 3 is the Sub with 1.59ms delay. This was the delay I found that aligns the phase well at the crossover. At 80 Hz, the sub and MB are showing 18 degrees phase, and the phase plot looks very stable through the crossover.
Trace 4 is the Subs w/ 1.59ms delay and the Midbass speakers playing together. I did this to observe the SPL support through the crossover, which to me looks like it was good.
For further confirmation, I applied Filtered IR @ 80Hz for the Sub w/ 1.59ms delay and the MB speakers. The filtered IR peaks line up very well in the Overlays Window for Impulse.
Let me know if this looks like I'm on the right track?
subterFUSE, I was reading over these old posts as driver to driver alignment thru crossover is an area of great research and learning for me. I perform car audio system tunings at my work, therefore I have the rather unique perspective of having worked with a very wide range of different drivers, cars, processors, etc. I see that you said you inverted the subwoofer by 180 degrees, but your IR chart clearly shows otherwise. The subwoofer polarity hasn't changed from one measurement to the next. Were you by any chance using the Helix DSP when these measurements were taken? I also experienced the same thing (when using the helix) where as most processors allow a simple polarity inversion of each channel, the helix is different.
Sorry.... you are correct. I did not invert polarity. I adjusted the phase control to 180 degrees. The phase control on the Helix DSP Pro is realized by a 2nd Order Allpass filter. Edit: Also, I think the images you linked are pretty old.... when i had even less of a clue what I was doing than I do now. LOL I'm busy tuning my car for INAC Finals this weekend, so I'll try to post some new measurements here that might be more informative.
awesome, that should be fun. I'm currently working on a way to determine how the different crossover selection, different timing, etc. affect the transient response of a system. I have come up with a way to measure, in real time, the step response, and filter the signal coming from the mic to just the range of frequencies shared by the drivers that are sharing a crossover, so, in other words, the crossover region. So far, using this method has proved fruitful. I also have begun making recordings of live instruments, (drums so far) using the same measurement mics that I use to tune systems. This way, the mics own freq. response and IR characteristics are nullified, plus, since I was the one who made the recordings, I know what the live event sounded like. So much fun and an incredible learning experience.
I have always used REW with a calibrated ECM8000 mic and a Steinberg UR22MkII sending the output from the UR22 to the analog input on my AVR for the speaker being measured. That setup used Input 1 and Output 1 as loopback. Input 2 connected to the mic and output 2 connected to the analog AVR...
I had bought a U-Mik several years ago and ran some measurements but never fully understood what REW was all about.. I sold the U-Mik as I did not really understand it's worth. Now some years later I have a better understanding of REW and why we need to take measurements..
My Old room...
So it's been a couple of years since I've played with REW and it's showing. New computer and I can't seem to get a good sound card calibration.
I have a simple stereo headphone to stereo headphone jack going from the line output to the line input on the computer.
Here is the soundcard...
I have used REW over many years, but not for the last 6-8 years.
I want to re-calibrate my sub (four x 15" in an infinite baffle).
I have downloaded the latest REW, but am using old equipment - EMM6 mic into a XENYX803 pre-amp, and then into a Sound Blaster Creative Digital Music LX connected to...
Does any one works with Mathaudio Room EQ or could make an comparison with other Roomcorrection software. Found it quite simpel to work with incomparison with REW The results are for me excelent. I work with Mathaudio in combination with Foobar2000 no cost involve.