I basically pretend that the capacitor is not in the circuit and start building my xover. I have attempted to add the capacitor's value, say 33uF, to the xover schematic in XSim, but this does not work. When you add a capacitor in series with a tweeter, the cap becomes an imbedded part of the tweeter. If you add another cap on the xover schematic, this is double counting the cap. Sort of like have two capacitors of the same value in series with the tweeter.
The capacitor will mess up the tweeters phase response a little bit, reducing the accuracy of your model. I generally use the raw phase data, including time of flight, straight from OmniMic, when I build my xovers. Bill
I generally use the raw phase data, including time of flight, straight from OmniMic, when I build my xovers. Bill
Bill, just an FYI. IIRC OmniMic sets t = 0 to the peak of the impulse response. This results with what Jeff B. calls "quasi-minimum phase". HolmImpulse (what I typically use) does the same thing. I still like to add the high/low tails and then create a truer minimum phase version, but the phase differences usually aren't very large between the two.
Ed is right, single channel measurement systems will not
have any "time of flight" included, since there is no reference of the
point in time where the signal left the speaker. For time of flight you
need a 2 channel measurement. I think the reason Jeff calls it "quasi minimum phase" is because t=0 is determined as the peak of the impulse, where I believe it truly should be the start of the impulse for minimum-phase. Holm will only do a single channel measurement, but you can apply an offset to shift the phase a bit if you need to.
To remove the cap from the measurement you can can add the cap in your simulation, then subtract the difference from the original measurement to provide a "reverse transfer function". If the cap is of a large value to provide a cutoff around 500Hz it will have minimal impact on the response anyway so you may be fine to not worry about its effect at all.
Thanks, Ed & dcibel. I stand corrected. I like your idea of creating a reverse transfer function to remove the cap from the measurement. I did not know that this could be done. I will give this a try.
OmniMic has "bass removed.frd" and "bass & midrange removed.frd" files that can be used in conjunction with "bass removed.wav" and "bass & midrange removed.wav" playback files to test tweeters without a series protection capacitor. But when you "normalize" your measurement to put the bass & midrange back in, the system also adds a bunch of low frequency noise into the normalized file.
Thanks, Ed & dcibel. I stand corrected. I like your idea of creating a reverse transfer function to remove the cap from the measurement. I did not know that this could be done. I will give this a try.
OmniMic has "bass removed.frd" and "bass & midrange removed.frd" files that can be used in conjunction with "bass removed.wav" and "bass & midrange removed.wav" playback files to test tweeters without a series protection capacitor. But when you "normalize" your measurement to put the bass & midrange back in, the system also adds a bunch of low frequency noise into the normalized file.
Yes, the procedure in Omnimic for the built-in "bass removed" sweeps does exactly that, when you normalize using that provided FRD you are subtracting the transfer function of the high pass filter from the measurement, but of course you can't put back information that wasn't there in the first place, so the noise floor is raised as a result. What it does correct is any errors in the driver response that the high pass imposed, and you can then apply a HBT to tail the response and remove the noise floor from the equation, resulting in a correct response as if there were no filter in place at all.
When using a physical cap, you want to do the same operation, but with the specific transfer function that the cap imposes on the driver impedance.
Seems like you are making a mess of the phase. I imagine you can adjust the phase on a single driver but how are you going to do that on the summed response when calculating offsets?
Same with normalizing the bass removed FRD. Why bother (normalizing) if you think you have meaningful content below 300 hz, then that's a tweeter than can take a short sweep. So no protection cap.
Using as measured phase is suppose to be easy, if you start manipulating the files it's easier to use minimum phase.
So If I'm worried about a tweeter (almost never) I do not use a cap and run the Base removed sweeps, trim tails, blend the woofer and extract min phase.
Yes, using a high pass in the measurement (whether Omnimic sweep or series cap) does mess with the response and the resulting phase. For design use I am always going to use minimum phase, if the cap or high passed sweep affects the response in any way, it is affecting the phase as well, and normalizing the response is a valid method of correcting this.
If the 300Hz high pass doesn't affect your measured response and roll-off, then go ahead and use it without any modification. If the high pass does (whether Omnimic sweep or series cap) , then you should correct it for accurate response and phase. Not doing so is just designing with inaccurate data.
FWIW , in the Omnimic software the "mid and bass removed" sweep is a 3kHz high pass which will affect most any driver response.
I generally use a full-range signal without protection, at ~85dB the driver receives <1W for a few seconds and they've all survived just fine. I don't use ribbons though, so protection is likely a greater concern for those fragile devices.
I'd have to ask Bill W. (unless you've tried this already?) but I think Omni mic. normalizing the frequency response is not going to make the phase equivalent to as measured. There's no way to blend or average the phase to make it continuous, it would have to be derived from the whole response.
I've never used the measured phase from Omnimic, I apply HBT to tail the response (remove the noise floor) and extract minimum phase always. It takes like 1-2 minutes to do, and the HBT will only make the phase more correct.
Comments
I basically pretend that the capacitor is not in the circuit and start building my xover. I have attempted to add the capacitor's value, say 33uF, to the xover schematic in XSim, but this does not work. When you add a capacitor in series with a tweeter, the cap becomes an imbedded part of the tweeter. If you add another cap on the xover schematic, this is double counting the cap. Sort of like have two capacitors of the same value in series with the tweeter.
The capacitor will mess up the tweeters phase response a little bit, reducing the accuracy of your model. I generally use the raw phase data, including time of flight, straight from OmniMic, when I build my xovers. Bill
Bill, just an FYI. IIRC OmniMic sets t = 0 to the peak of the impulse response. This results with what Jeff B. calls "quasi-minimum phase". HolmImpulse (what I typically use) does the same thing. I still like to add the high/low tails and then create a truer minimum phase version, but the phase differences usually aren't very large between the two.
To remove the cap from the measurement you can can add the cap in your simulation, then subtract the difference from the original measurement to provide a "reverse transfer function". If the cap is of a large value to provide a cutoff around 500Hz it will have minimal impact on the response anyway so you may be fine to not worry about its effect at all.
Thanks, Ed & dcibel. I stand corrected. I like your idea of creating a reverse transfer function to remove the cap from the measurement. I did not know that this could be done. I will give this a try.
OmniMic has "bass removed.frd" and "bass & midrange removed.frd" files that can be used in conjunction with "bass removed.wav" and "bass & midrange removed.wav" playback files to test tweeters without a series protection capacitor. But when you "normalize" your measurement to put the bass & midrange back in, the system also adds a bunch of low frequency noise into the normalized file.
When using a physical cap, you want to do the same operation, but with the specific transfer function that the cap imposes on the driver impedance.
Seems like you are making a mess of the phase. I imagine you can adjust the phase on a single driver but how are you going to do that on the summed response when calculating offsets?
Same with normalizing the bass removed FRD. Why bother (normalizing) if you think you have meaningful content below 300 hz, then that's a tweeter than can take a short sweep. So no protection cap.
Using as measured phase is suppose to be easy, if you start manipulating the files it's easier to use minimum phase.
So If I'm worried about a tweeter (almost never) I do not use a cap and run the Base removed sweeps, trim tails, blend the woofer and extract min phase.
If the 300Hz high pass doesn't affect your measured response and roll-off, then go ahead and use it without any modification. If the high pass does (whether Omnimic sweep or series cap) , then you should correct it for accurate response and phase. Not doing so is just designing with inaccurate data.
FWIW , in the Omnimic software the "mid and bass removed" sweep is a 3kHz high pass which will affect most any driver response.
I generally use a full-range signal without protection, at ~85dB the driver receives <1W for a few seconds and they've all survived just fine. I don't use ribbons though, so protection is likely a greater concern for those fragile devices.