First couple of measurements
Here's my first two measurements worth uploading
Blue is a 3 way AS classic speaker.
Red is a 2 way AS classic speaker - different brand. Red one'w woofer runs full range w/no series coil. Wonder why all the phase inversions.
Blue one was measured on the floor. Red on was measured a couple of feet off the floor.
Here's my first two measurements worth uploading
Blue is a 3 way AS classic speaker.
Red is a 2 way AS classic speaker - different brand. Red one'w woofer runs full range w/no series coil. Wonder why all the phase inversions.
Blue one was measured on the floor. Red on was measured a couple of feet off the floor.
Attachments
Same tab Device and Signal under output device is PCM amplitude. Might want to lower it. If that doesn't work lower the Art MP. Basically if you are clipping the inputs you have to turn down whats feeding it or else all you are doing is turning down a clipped signal.
Sorry if you already understand this but I just see this question pop up frequently with people recording from vinyl expecting to be able to turn it down within the computer not realizing they are clipping the analog stage of the soundcards inputs.
Mr. Holm. What does the dash line in the impulse response represent? It curves down from the top of the graph and slopes down to the right.
And, why can I take one dashed line and move it to match the shape of another dashed line if I have taken two measurements?
There's more than one window in the impulse measure?
I was referring to the default impulse measure window.
I am now back into practice and need a few more details about the above method for time alignment.
If I look at the phase of each driver separately, I can find an area (for the twitter position, relatively to the woofer) where those two phase plot seem to agree.
My problem is that by looking at this agreement, I cannot find a very precise position for the twitter.
I would think that in a range of 1-2 cm, the 2 curves are kind of agreeing.
I am not sure how to do this in HOLM Impulse.
Could you, please, develop that part?
Thanks
I haven't gotten back into this much - just other more pressing experiments taking up my time - and I haven't come up with a simple way of doing it yet. But basically I think what we need - provided we have a reference mic and ADC/DAC - is a way to take an impulse reponse and do an inverse covilution that acts only on the phase response and disregards the FR. So how would we get rid of the FR from an impulse reponse leaving just the phase? Manually editing the file somehow?
Maybe someone can help me with these measurements. I am trying to test 32-bit float converted to 24-bits and truncation vs dither at these bit-depths. At first I was just trying to import and compare the 2 files but it didn't seem to show me much. I came up with this instead and am not sure if the measurements are useful or I am interpreting them correctly.
I took 10 secs of a master of one of my songs rendered at 32-bit float. I then male a copy of the file convert it to 24-bit and then convert it back to 32-bit float. I take the original and invert mix it with the altered file which I would think only leaves the difference between the two files - noise and distortions. I then take that difference file and compare the original to it using Holm. I get these results.
Green = Dithered
Red = Dithered
Blue = Truncated
Am I wrong in assuming that the truncated example has less noise (for the most part) than the dithered examples?
I took 10 secs of a master of one of my songs rendered at 32-bit float. I then male a copy of the file convert it to 24-bit and then convert it back to 32-bit float. I take the original and invert mix it with the altered file which I would think only leaves the difference between the two files - noise and distortions. I then take that difference file and compare the original to it using Holm. I get these results.
An externally hosted image should be here but it was not working when we last tested it.
Green = Dithered
Red = Dithered
Blue = Truncated
Am I wrong in assuming that the truncated example has less noise (for the most part) than the dithered examples?
Perhaps you want to try the 2-channel correlation?
Options > 'Import response from file' > 'Signal & Recording'
Signal wave-file should be your original wave file
Recorded wave file should be your modified wave file
Then also try the dB-option for the impulse response
Like I did here with mp3-files:http://www.diyaudio.com/forums/showthread.php?p=1913539
Last edited:
That is how I did it. It's just that I did the invert cancellation prior to the correlation. The correlation on it's own was a little hard to interpret without the inverse cancellation prior to correlation. Basically if I don't do the invert isolation all I get is a perfectly flat FR and PR.
Last edited:
I do remember zooming in and seeing spikes before but I guess it just wasn't looking as easy to interpret as my other method. I am not sure what you are saying about the dB setting. You mean the one under options after you take a measurement?
Anyway here is what it looks like if I correlate without doing the cancellation first.
Anyway here is what it looks like if I correlate without doing the cancellation first.
An externally hosted image should be here but it was not working when we last tested it.
Thanks for the graphs!
I mean the little dB Checkbox in the lower right corner of the IR
I believe what we see here is the noise floor of the file- this is how much the new file correlates to the original.
If you correlate the original with the original we should not get any noise at all
What is the latest situation on Vista64 (apart from that its dead). Should I use the normal WIndows (2000, XP, Vista, 7) version or the "x64" version?
The 32 bit version works on 2000, XP, Vista, 7
The 64 bit version works on Vista, 7 64 bit