Take an on axis measurement. Move the mic a known distance away from the first measurement point along the axis and take another measurement. Look at the difference in time it took the signal to arrive between the two measurements. Divide the distance between measurements by the time difference. Now you know the speed of sound, although the accuracy of your measurement depends on your previous measurements. Now look at the time between zero and the signal start at the point you're interested in and divide the speed of sound you measured by this time. I would not use holm for this - I'd use a 2 channel system where I could take the reference channel to be the voltage at the driver terminals or something like that. Then you could compare the time between the same points on the time signal (if they looked recognizably the same, which is a big if).
However I'm not sure I'd ever find a use for this in speaker design. Everything is relative to everything else - as long as you keep a constant time origin and measure your drivers in their final positions at one mic position, it's not necessary to known any absolute distances. In addition, the acoustic center of a driver will typically change with frequency, so depending on what you're doing, you might want to modify the above technique to use a band-limited signal.
However I'm not sure I'd ever find a use for this in speaker design. Everything is relative to everything else - as long as you keep a constant time origin and measure your drivers in their final positions at one mic position, it's not necessary to known any absolute distances. In addition, the acoustic center of a driver will typically change with frequency, so depending on what you're doing, you might want to modify the above technique to use a band-limited signal.
Thanks! It's close to what I had in mind, but just could not wrap my head around a few things here. I'll try it this way.
Of course relative measurements is where it's (mostly) at in design; but if I can get somewhat consistent results, even with HolmImpulse for this, it could be quite educational for me.
IG
Of course relative measurements is where it's (mostly) at in design; but if I can get somewhat consistent results, even with HolmImpulse for this, it could be quite educational for me.
IG
There is no way to get the precise point, you can only get close. As you pointed out, the only thing that actually matters is relative offset.
I don't recall if this has been mentioned in the thread. One way to do it is shown here. You have to be sure to have the same absolute setting for the window start time marker for each measurement. The system latency will be the same, so that isn't an issue for determining relative offset.
Dave
An FRD is a text file. What you may need to do is open the file with a plain text editor and delete any leading comment lines, leaving only the data.
Dave
Let me correct myself. If you use the method shown at the link, the start time marker isn't necessary other than what is required to get a good window for the FR. You'll be using the measurement output from HOLM to import into design software (since HOLM doesn't do that). The method shown does not require measured phase, you create the driver models and generate phase for each driver in your design software, such as the PCD. The summed response of these models is compared against the measured summed response to get the proper offset.
Dave
Yes, but I'm not trying to find relative distance, but absolute, within reasonable accuracy (only as good as I can take physical measurement and system latency I suppose). This is mostly for educational purpose, not for system design.
So is the start time marker the first numbered marker or a point on the dashed line?
IG
Last edited:
The window is the whole dashed line. You can use the Measurement Options window to manually set the marker (not sure how it actually relates to the dashed lines) after each measurement to the same sample point. Set the window/sample point initially, then re-set it to this same point each time.
I also don't know exactly how you would relate this to the absolute center, partly because that can't be determined precisely and because I'm less familiar with the windowing used in HOLM. The systems with which I'm most familiar use half-windows, not available in HOLM.
You can get an idea of this from the screen capture I posted some time ago, linked here.
Use the Impulse Time-Window setting and experiment a bit. You'll have to unlock it each time to manually set it.
Dave
I can now see how you can display the measurement signal, which is spot on time zero naturally, followed by the IR of my speaker, but it places it at a ridiculously small distance compared to the actual distance my mic is located.
Example: my microphone is roughly 50cm away from speaker, the measurement signal is on sample zero, speaker IR 1st +peak follows at 0.37 sample. For 44.1kHz, this is roughly 8.39µs. This is ~2.88mm distance.
IG
Example: my microphone is roughly 50cm away from speaker, the measurement signal is on sample zero, speaker IR 1st +peak follows at 0.37 sample. For 44.1kHz, this is roughly 8.39µs. This is ~2.88mm distance.
IG
Last edited:
This is always a large number, say ~30k samples, which at 44.1kHz means my distance would be insanely large, or is there something I don't look at right?
I've pretty much nailed the relative distance procedure, but the measurement signal peak VS the 1st peak of the IR is always extremely close, I don't know why.
IG
HOLMAcoustics is a single-channel measurement system. You cannot get any absolute time reference with it. Even a dual-channel system that uses a feedback probe from the speaker input terminals cannot accurately determine an absolute acoustic center. I don't know why I had not thought to point this out earlier. You'd be better off putting your efforts into other areas.
Dave
I'm talking about the "grey" curve, which is the measurement signal impulse centered at time zero. The impulse response of the speaker follows this, but at a distance/time that is ridiculously small. If anything, hardware/software delays should make that longer.
I'm able to get correct relative distances/times when locking time-zero after a first measurement, but this take away the possibility of seeing any "absolute" values.
IG
I thought it still had to know travel time from source for the purpose of relative measurements, but I guess that locking zero-time takes away all unknown hardware/software delay from the equation. I guess I can see what you mean and it makes sense that what I want to do might not be possible then.
IG
This is something I have not tried, because it doesn't matter much to me.
Hmmmm... if you get consistent offsets with the stream always on, can you subtract a known distance and arrive at a constant offset? Will it hold if you stop and start the stream?
If it does, you might then know the latency of the measurement circuit. Offset-known distance = device latency.
Hmmmm... if you get consistent offsets with the stream always on, can you subtract a known distance and arrive at a constant offset? Will it hold if you stop and start the stream?
If it does, you might then know the latency of the measurement circuit. Offset-known distance = device latency.
If anything can be kludged to arrive at a rough absolute measure, it could still be worth trying. I have not worked enough with HolmImpulse or similar software to know how to do this. Might need somewhat of a step-by-step for dummies.
IG