An MLS measurement question

[tcpip]

Guys,

I've been trying to take SPL measurements of an MTM box I'm trying to build. I use Speaker Workshop and a Creative USB sound module, place the mike (a two-dollar Panasonic mic capsule from Digikey) about 1.1 metres from the tweeter, and take readings. Since this is a USB thing, I experience long and variable latencies. To take care of the problem, I take impulse measurements, never SPL readings. I then visually move the start and end gates to appropriate positions and do an FFT to get SPL.

This time, I noticed something I've never seen before. I saw a small spike a few milliseconds before the main spike. Check here:



At this point, if I put the start gate at 0msec and the end at 21.87msec after the main spike, it looks like this:



The FFT from this, taking the data between the markers, is:



Clearly, I'm not getting a good clean reading. And I don't have a clue where the the first little spike is coming from. So I try to take the start marker close to the first spike: 16.59msec to 21.87msec:



The FFT now shows:



This was better than the 0msec start marker, but not very good. Lots of jagged spikes in the HF region, hinting at room reflections or something else. So I move the start marker a bit more to the right: 17.10msec to 21.87msec. The start marker is now just a tiny bit before the first small pulse.



The FFT now shows:



Moving the start marker still more to the right gives me 17.38ms to 21.87ms:



The FFT now shows:



This looks clean enough to compare with most SPL measurements I've seen.

My questions:

• Where is that small first spike coming from?
• Can one just move these markers around arbitrarily to get clean-looking SPL curves, and then just take the curve which is visually clean enough? Which of these SPL curves is "right"?

[SY]

If I were to guess, the first spike is electrical noise pickup from the stimulus. It is advanced just about the right amount. The "noise" in the top octave is unlikely room reflections, more likely reflections off nearby bits like mounting screws or other protrusions (including the mike's mounting hardware). It could also be from falling response in the reference spectrum.

That's actually a pretty decent SPL curve.

[tcpip]

If you want to know what I'm up to, here is a write-up of work in progress: the Asawari Mark 2

[tcpip]

Originally Posted by SY

If I were to guess, the first spike is electrical noise pickup from the stimulus. It is advanced just about the right amount. The "noise" in the top octave is unlikely room reflections, more likely reflections off nearby bits like mounting screws or other protrusions (including the mike's mounting hardware). It could also be from falling response in the reference spectrum.

That's actually a pretty decent SPL curve.

Hi SY, nice to hear from you. It feels strange, being back among friends after a gap of three years.

Which one is the "decent" one? The one for 0 to 21.87 msec?

What do you mean by "advanced just the right amount" -- care to elaborate?

If you had this impulse graph, what gating would you use to get the "right" SPL curve?

[tcpip]

Guys,

Anybody? Any suggestions/ideas/explanations?

[SY]

Originally Posted by tcpip

Hi SY, nice to hear from you. It feels strange, being back among friends after a gap of three years.

Which one is the "decent" one? The one for 0 to 21.87 msec?

What do you mean by "advanced just the right amount" -- care to elaborate?

If you had this impulse graph, what gating would you use to get the "right" SPL curve?

Nice to "see" you as well!

There's going to be a delay between the electrical stimulus of the speaker and when the acoustic output reaches the mike. The electrical path is for all intents and purposes instantaneous. So if you're picking up stray electrical signals, they should appear advanced from the acoustic pickup by the speaker-to-mike distance divided by speed of sound. That's about 3 ms/meter.

If someone doesn't beat me to it, I'll give you a gating recommendation later- I'm being called into a meeting...

[bwaslo]

It's from cross-talk in your soundcard. The signal is going right from line out to the (line or mic) input, usually through leakage capacitance. You can probably reduce it by turning up the volume on the power amplifier (assuming it has its own volume control and that you aren't just using the computer's volume control slider to adjust level).

[tcpip]

Originally Posted by SY

There's going to be a delay between the electrical stimulus of the speaker and when the acoustic output reaches the mike. The electrical path is for all intents and purposes instantaneous. So if you're picking up stray electrical signals, they should appear advanced from the acoustic pickup by the speaker-to-mike distance divided by speed of sound. That's about 3 ms/meter.

Originally Posted by bwaslo

It's from cross-talk in your soundcard. The signal is going right from line out to the (line or mic) input, usually through leakage capacitance.

Thanks a ton, SY and bwaslo. Both of you are saying essentially the same thing. I'll try turning up the power amp volume the next time I set up my rig for more measurements -- let's see if that helps.

So, the next question is: should I put the start gate just after this small spike? Seems logical, doesn't it? When does your meeting get over, SY?

[wintermute]

Hi tcpip long time no see I gave up trying to get good spl measurements with speaker workshop and started using HolmImpulse I still use SW for impedance measurements and also import the holm measurements to use the SW crossover modeling.

I was having a similar problem recently though it was having a much nastier effect on my measurements than what you are seeing. major ripple right throughout the frequency range.

I would definitely move the start gate after the blip and for the end gate, I'd say you need to work on your speaker and mic placement in your room... I normally display my impulse response with distance rather than ms so I'm having dificulty relating to your impusle response but it doesn't look like there is any point that I would say was a good place to put the second marker..

Attached is a plot which shows the problem I was having with the same sort of blip. Note how the impulse is quite flat for a while and the point where I have the gating is a bit before where the first big spike appears after that. The wiggle in the graph is from the pre-impulse, it completely goes away if the marker is moved before that.

If you have at least 1M from the speaker to any possibly reflective surfaces and the mic at 1M then you should be able to get a relatively clean impulse response like the one I have attached. Try angling the speaker and mic so that it isn't directly parallel (or at right angles) to any walls.

edit: On looking at that last implusle again, I'd say try 17 to 26ms and see how it looks... It might go all ragged but it might be ok, where you have it at the moment is not going to be giving you very much low frequency resolution.

Tony.

[tcpip]

You know it occurred to me that if this is indeed stray crosstalk from the output channel to the input channel, then maybe I can simply reduce the mic input gain on the computer, and increase the volume on the power amp. That way, the mic input will be less sensitive to any adjacent channel's signals, and maybe the (relative) amplitude of the crosstalk spike will reduce. Let me try it out and let you know.

BTW, I did keep the speaker and mike at an angle to the walls -- it was not all perpendicular and parallel. Also, the distance from tweeter to mic was about 1.1 metres.