Radio Shack SPL Meter and justMLS - I'm getting odd measurements...

[fjhuerta]

Hi all!

I decided to take my speaker building project very slowly, and learning first about what makes a good speaker.

Anyway, I got lspCAD 6, and was happy to see justMLS included. I had never heard about it, so I began playing with it.

The instructions are not ver clear... at all. I'm using a laptop, a Sound Blaster MP3+ card, and the Radio Shack SPL meter as a mic.

I built the reference cable (the one that goes from the speaker to the Line In, with the voltage divider circuit). If I plug it into the sound card, a weird "tick-tick-tick" noise (loud!) is heard on the speaker.

But the weirdest thing is - although I can measure frequency response (it looks terrible, but the shape corresponds to that shown on the tweeter's literature), the phase is all over the place. It crosses 0 10 or 20 times and never settles down. I tried calibrating the sound card, and it's no use...

Is there a FAQ where I can read a bit more on justMLS? I have read the instructions twice, tried measuring using one and two channels, changed drivers, played with the distance, and nothing seems to help. Alternatively, do you think the Radio Shack SPL meter is playing serious tricks with phase?

Any help will be greatly appreciated!

[simon5]

I would look at the reference cable you built and also at your soundcard. Soundblaster cards are not that accurate over 10 KHz, the FR oscillates.

Try to keep the imput voltage (or output of the SPL meter) of your soundcard at around 1.5V I guess. Soundblaster cards have a bad habit of saturating their op-amps when you use the standard 2V level.

Something seems to bring the whole thing into oscillation, too much capacitors somewhere? Hehehe!

It could be the RatShack meter too...
Good luck!

[MBK]

I don't know enough of your setup and JustMLS but a few things to consider:

- usually such software calculates phase from the frequency response. It calculates the frequency response by doing an FFT from the MLS signal's time response.
- the FFT is really a discrete calculation. Fitting a discrete calculation onto a continuous spectrum means risking artefacts: Depending on the parameters set for the FFT you do get both "real" terrible looking frequency wiggles, and artefactual wiggles
- those real and artefactual wiggles = mathematically generated noise, will now be included in the calculation of phase... with the result that you get.

Possible solutions:

- more FFT points, though there is a point depending on length of MLS stimulus where more FFT points will not increase precision
- averaging of the frequency response and hoping that the software uses the averaged frequency response for phase calculation
- forget phase and use averaged frequency response for a first order mental approximation of phase (rule: FR goes down->phase becomes negative; FR goes up->phase becomes positive - this is why you get such a mess in phase, it comes from the up and downs of you FR)

Incidentally much of the terribly looking wiggles you get come from the room interaction (if not gated) and the baffle edges (here even gating won't help much).

If you want to test your microphone and soundcard measure the tweeter at 1/2 inch from the dome. This way you'll have essentially just the tweeter and it should look very smooth and flat. If it doesn't this indicates some peaks and dips of your microphone.

But first, test your soundcard. Measure distortion over your reference wire (just the wire, no mic and no speaker - output to input), using a single frequency sine wave signal instead of MLS.Play it continuously and increase input and output levels one at a time, while doing a real time FFT (watch screen while turning knob) until you see weird spikes appearing at the harmonics (multiples of the signal frequency ). This indicates clipping distortion onset. Remember that both outputs and inputs can clip: outputs from weakness, inputs from overload. For the input, overload can occur at the mic level, the mic preamp level, and the coundcard level.

If you want to look any of this kind of measurement to look good, just use octave averaging and a display scale of 100 dB on the y axis - that's what manufacturers do, hey, and you won't see any wiggles at all

Hope this helps

[simon5]

Sorry, I forgot that the Soundblaster cards are reconverting stuff to 48 kHz most of the time to apply some effects, the conversion engine is really crappy, so use 48 kHz sampling frequency on the tone software generator and the microphone line-in if you can to bypass this issue that might show false results. Check that CMSS is disabled and things like that.

Also, follow the advice of MBK because he knows more than me hehe!

[SY]

But the weirdest thing is - although I can measure frequency response (it looks terrible, but the shape corresponds to that shown on the tweeter's literature), the phase is all over the place. It crosses 0 10 or 20 times and never settles down. I tried calibrating the sound card, and it's no use...

I don't know your software so I can't tell you the fix. But I can tell you the problem- you haven't corrected for time of flight, i.e., the time it takes to get from the speaker to the mike. If there's any kind of distance variable there, that's what you need to play with.

For example, let's say there's a millisecond of delay because your mike is a foot (30cm) away. At 1kHz, that's a full cycle, so the dumb program thinks that, in addition to the real (minimum) phase, there's another 360° of phase there. At 2kHz, there's another 720°. Since the phase display "wraps" at +/-180°, what you get is a display that has a wildly varying phase curve.

[Jennice]

And how would one correct for this distance/frewquency/phase "file of flight" error?

I am following the thread, as I am considering this tool also. However, this phenomenon seems more of a hhysical phenomenon than a fault with the setup, am I correct?

Jennice

[SY]

Originally posted by Jennice
And how would one correct for this distance/frewquency/phase "file of flight" error?

I've never used that particular software, so I can't tell you where the menu buttons are to do that. If you want to understand the theory of unwrapping the phase curve, d'Appolito's book explains it very well with lots of illustrations.

[dhenryp]

I think SY is right. I have version 5 of Just MLS and it has an input that I think is called distance. It allows you to enter the distance between the microphone and the driver in either meters or miliseconds. You can start by measuring the distance with a tape measure and entering that. Then make a small adjustment - you will see the phase get worse (i.e. it will show more cycles or peaks) or better. Keep making corrections in the direction that reduces the number of peaks. Eventually you will come to a distance that minimizes the number. One warning - if you enter a value that goes too far in the "right" direction, you will see things getting worse again.

I can't help you with the clicks - I've not had that happen (yet). You are not alone in finding JustMLS to be tempermental. I'm getting better at it but I'm still having problems repeating measurements. I'm sure it's something I'm doing (as opposed to the software) but I will let the users' manual share the blame.

Regards,

Denis

[Jennice]

I agree,

I've downloaded the demo, and the documentation there seems short to say the least.

I wonder if it's better If I end up buying the real thing after following this thread.

Jennice

[MBK]

Didn't even think of that one too.

Again, I don't know JustMLS but say in Audiotester, you have a time domain diagram and you can tell the software when to start the FFT. That way you don't even need to measure, you see the time point where the impulse arrives, and you click to make this time zero.

JustMLS must have that feature too, because surely there will also be an echo point / gating feature, and for that you need the time domain graph as well.