I have never seen CSD included in the specs of drive units and my understanding is the drivers mfg listed specs include ALL critical data
Last edited:
Critical data would include distortion sweeps but hardly anyone does that. CSD as far as I am concerned is completely pointless on an unfiltered driver though.
Good I thought I was not utilizing all of the data as I should.
am not as experienced as most of you, but not a Newb either .
I will continue to focus on the data that can help me build better speakers
Thanks for your input
am not as experienced as most of you, but not a Newb either .
I will continue to focus on the data that can help me build better speakers
Thanks for your input
Last edited:
Yup, strength of harmonics, and even intermod distortion affect the sound.
But all in all, I'd like a smooth (poly) but detailed not soggy (paper), and no +5db peaks.
I'd been looking at the scanspeak 18w/8545 7" scanspeak, but that is not a full range driver. "heavily damped and carbon fibre impregnated paper cone". Carbon fiber/kevlar usually has 2 resonance, not sure about the 800hz hump (that has been verified).
The Madisound Speaker Store
Then again, the wild burro "betsy" drivers left carbon fiber/kevlar to a reported better sounding paper cone...................
Norman
But all in all, I'd like a smooth (poly) but detailed not soggy (paper), and no +5db peaks.
I'd been looking at the scanspeak 18w/8545 7" scanspeak, but that is not a full range driver. "heavily damped and carbon fibre impregnated paper cone". Carbon fiber/kevlar usually has 2 resonance, not sure about the 800hz hump (that has been verified).
The Madisound Speaker Store
Then again, the wild burro "betsy" drivers left carbon fiber/kevlar to a reported better sounding paper cone...................
Norman
I've seen some measurements of the recent 18w/8545, or at least one of it's close relatives and it had some peaking in the distortion at mid frequencies that you'd rather avoid.
http://www.scan-speak.dk/datasheet/reviews/18W-8545-01-VoiceCoil_2012-10.pdf
Yes, it's really only the second harmonic that peaks, but it's pretty severe.
http://www.scan-speak.dk/datasheet/reviews/18W-8545-01-VoiceCoil_2012-10.pdf
Yes, it's really only the second harmonic that peaks, but it's pretty severe.
You record the impulse response at different Hertz rates, for example 1000 different Hertz rates and then you construct a 3D representation of the 1000 different data points.
Was there a point within your question?
What do you mean by an unfiltered driver?
Just because no one or hardly anyone is interpreting CSD correctly does not make the variable "pointless", unless you are a flat frequency response freak and don't care about anything else or have tin-ears.
The point of the question is the CSD, like the FR is derived from the impulse response via FFT. It's the same data displayed differently. Indeed, applying the Hilbert transform to the FR data gives the minimum phase plot. Everything is contained within the impulse response.
Bob
The driver 18w/8545 was used in the meadowlark, odyssey, merlin, and proac I think.
Dang, people seemed to love it too.............
Oh well, thanks for saving me $165 each plus boxes.
It is neat seeing a waterfall "ring" longer on the peaks seen on the freq response.
Dang, people seemed to love it too.............
Oh well, thanks for saving me $165 each plus boxes.
It is neat seeing a waterfall "ring" longer on the peaks seen on the freq response.
Everything is contained within the impulse response, but not everything within the impulse response is contained within the frequency response.
Yes?
IR = everything
FR = not everything
Thus, we can not return to IR from FR.
Likewise, we can not take some kind leap violating all the laws of mathematics from FR to CSD.
"Frequency Response" needs qualification. When you have a *complete* FR (magnitude and phase) it is equivalent and identical to the corresponding IR, just another way to present the same data. CSD is yet another way to present it, etc.
Sorry no, that's not the way it is done. You generate an IR (many ways to do that, sine sweep and convolution seems to be the best, most robust method), then you step a short-term FFT with specific windowing (yielding a FR magnitude "slice") through the IR and plot the magnitude data slices.
Yes, if we have 1000 data slices of FR recorded over time, for example at 50 nanosecond intervals x 1000, then it is identical to IR.
If we have a single data slice of FR, we can not return to the IR.
Are we on the same page now?
I've noticed many people in this forum which believe they can predict CSD with a single data slice of FR, via SPL tracing and inverse FFT or whatever.
None of them could exemplify this.
No one can accurately predict CSD with a single FR slice, are we in concordance?
One that hasn't been passed through a coffee machine.
What do you think I mean? One that is running without a filter, be it active, passive, putting a sock in the way, one without any kind of filtering.
That's right, flat frequency response is a joke, lets all listen to loudspeakers with gross amounts of linear distortion.
I don't think anyone is interpreting the CSD incorrectly either, it's pretty obvious what it's telling you.
Way to go with that old chestnut.
Yes, if we have 1000 data slices of FR recorded over time, for example at 50 nanosecond intervals x 1000, then it is identical to IR.
If we have a single data slice of FR, we can not return to the IR.
Are we on the same page now?
I've noticed many people in this forum which believe they can predict CSD with a single data slice of FR, via SPL tracing and inverse FFT or whatever.
None of them could exemplify this.
No one can accurately predict CSD with a single FR slice, are we in concordance?
No.
The frequency response is not derived like this. Quite how it is derived mathematically I do not know, but when you are processing the impulse response, the gate makes a tremendous amount of difference to the FR itself.
If the FR were only displaying one snapshot then this would not be the case. If you increase the gate length to include more of the tail end of the impulse response then the FR changes, if you include a reflection or two then the whole things turns into a disaster. It is not simply displaying one snapshot in time, it is displaying all the data within the gated impulse response. Likewise if you do not gate the response it will display all the data that is present within it.
Everyone is, except for me. Just kidding.
What does this impulse response sound like ?! Can you tell me ?!
http://krashan.ppa.pl/articles/u1synth/
- Status
- Not open for further replies.