Difficulty interpreting pulse and frequency response data

[ByronInLawrence]

Greetings,

I am having difficulty interpreting my pulse data and frequency response data done with Speaker Workshop and a Wallin Jig. I went to great lengths to record without reflections. I removed about a 7ft section of my balcony railing, and put the speaker on the edge of the balcony facing away from my house. Then I put the microphone on a tripod duct taped to a ladder. the speaker and microphone were 10ft off the ground. The speaker was 13ft away from the house (width of my balcony). The microphone was 1 meter away from the speaker. The only reflective surfaces were the narrow balcony railings I left up, 3.5ft on either side of the speaker.

The second graph shows the pulse response with (black line) and without (blue line) the tweeter hooked up. The first pulse comes at 9.7ms, which is 6.8ms later than it should have been, but I'll accept that the sound card or my amp may have some delay. But what is this second pulse at 12.6ms? It is as big as is the first pulse and represents 1.1meters of sound travel, so it's not likely a reflection. When I disconnect the tweeter, the big spike is gone, but the smaller wiggles are still there. Is this second spike my tweeter? If so, why is it delayed by 2.9msec? Is it a crossover thing? I'm using 2nd order centered at 2.5KHz with an L-pad in series with the tweeter. More importantly, what effect does it have on my frequency response measurement?

The graph on the left is my frequency-response with the reflection time set at 2.9ms (black line), 100ms (red line) and way out at 700ms (green line). All are smoothed to 1/4 octave. You can see the reflections I told Speaker Workshop to ignore don't have much of an effect on the plot except at low frequencies, but why do they look so awful? I'm sure the speaker doesn't sound that bad. The port needs tuning, but otherwise, it sounds pretty good. Is this plot affected by that mysterious second peak in the pulse response?

Picture on the right is the speaker with cat for scale.

Any advice from more experienced builders would be greatly appreciated!

-Byron

[wintermute]

That looks very odd.. I'd try setting your start gate at the begining of the second pulse (pretty much where the black line is) and your end gate somewhere out past 18 ms assuming it is still as clean as it is showing at a18 ms and see what the frequency response looks like then.

It looks like you are getting two impulses played 2ms apart. The reason the big spike disappears when you disconnect the tweeter is that the high frequency energy is in the very first bit of the impulse and the lower frequencies in the tail. Without the tweeter hooked up there is not enough information to register at the begining of the pulse

If it were purely two impulses being played 2ms appart though I'd expect them to look similar, but perhaps not as the speaker may not have recovered when the second one starts playing so it may give a different response.

I'd try doing some impulse response measurements in SW wiith just a loopback (no mic) and see whether you are getting something similar. If you have a creative sound card make sure that the record without monitoring option is set! You may find that what you are seeing is the soundcard playing the recorded sound back out again which would certainly account for the 2ms delay.

Tony.

[DBMandrake]

I'll bet dollars to donuts that the problem is local record-playback pass-through in the sound card is not turned off...

Just before making a measurement tap on the microphone and if you hear it through your speaker, that's your problem.

You say the first pulse comes at 9.7ms, 6.8ms earlier than you're expected, eg you're expecting 2.9ms, which is about right for 1 metre. A 6.8ms (and somewhat randomly varying with each measurement) delay is typical of many sound cards.

The first peak will be the ~6.8ms delay of the sound card, plus the 2.9ms delay of the sound travelling through the air 1 metre, eg 9.7ms.

When it is received by the microphone it will be immediately passed through from the mic/line input of the sound card to the speaker output without delay (analog pass-through is used in most sound cards) which will send a second copy of the impulse to the speaker, which 2.9ms later will arrive at the microphone again, hence the 12.6ms impulse.

This second copy of the impulse is not just delayed, it is what the microphone recorded the first time around not the original impulse, so the second impulse will have been through the speaker/microphone chain twice, causing any deviation in response to compound on itself

This will royally screw up your measurements and there is no satisfactory way to gate it out and still maintain an acceptable low frequency response.

As wintermute suggests, check your sound card recording settings for any settings that relate to "monitor while recording", "record without monitoring" etc.

On most sound cards you can go into the playback settings, mute the microphone/line input, but leave it unmuted in the record settings - this is what you want.

Some of the creative cards like my Audigy 2ZS are unusual in that the setting is "Record without monitoring" in the recording advanced settings, and the pass-through is only muted whilst an application is actually recording. (Very annoying actually, thanks Creative...)

[ByronInLawrence]

Thanks Wintermute and DB Mandrake. Unfortunately, it's a big pain in the a** to hook up all those cables and the jig and amp and microphone . . . So I may not get to trying your suggestions until later in the week. I did find an option within windows to adjust my sound card (Rocketfish 5.1 PCI) to adjust the volume for "playback" or "recording". The "playback" button was checked. Also, a had muted the microphone input, but not "line-in", which I think was necessary for impedance measurements (impedance works well). So I guess it was treating the microphone input as line-in. I really don't know what I'm doing, but I'll play around with it, and try the "loopback" experiment Wintermute suggested.

I don't want to be ungrateful to Eric Wallin, who described the Wallin jig, and detailed how to use it with Speaker Workshop. Without his hard work, which he gave away for free, I would be measuring nothing, unless I spent $100 on the Dayton toy which only measures impedance. However, Mr. Wallin used a sound card which he was able to modify to produce speaker-level outs up to 1 watt, and his entire tutorial uses that card. He also recommended getting impedance curves for T/S parameters with 2.83VAC. It turn out, unless you're using Eric Wallin's modified sound card, much of his tutorial is incorrect, which has caused me endless frustration. I'm thinking of writing a wiki for use of SW with a sound card with line-level outs and an external amplifier (like every other sound card on the planet) as soon as I get the hang of what I'm doing.

[wintermute]

If you haven't already got it, I'd highly recommend downloading the unofficial speaker workshop manual http://www.esnips.com/web/SpeakerW/

If you have gotten as far as you have without it you are doing well! make sure to at least read the sections on calibration.

I have been able to do impedance measurements directly off the sound card output on at least three different cards, a turtle beach montego II, Audigy II ZS and using the on-board sound on my asus M4A88TD motherboard, all without any modification. I use quite low levels no where near 2.83V for T/S params, most cards could not cope with 2.83V input without severe clipping.

When doing FR measurements I simply hook up the sound card to my amp, make sure the walin jig is in the -20DB position set the voltage to around 2V using my multimeter (make sure you remove it before doing any real measurements!) and start measuring I actually have stopped doing the FR measurements in speaker workshop and started doing them in holm impulse (in this case I still use the jig, but believe it isn't actually necessary). I had a lot of trouble getting consistent results with impulse responses in speakerworkshop due to the latency continuously varying. I don't know if that was a driver or a speaker workshop problem, but changing to holm impulse seems to have fixed it. I the export the results and import into speaker workshop to use the crossover modelling.

Tony.

[DBMandrake]

The line level output impedance of the Audigy 2 ZS is quite low - I measured mine at around 35 ohms, with a maximum output of 2.2v RMS, I find this plenty to drive a speaker directly through an impedance test jig (without an amplifier) for measuring impedance curves and T/S parameters.

I normally use ARTA, and my impedance "test jig" consists simply of a 34 ohm resistor from line level left output to speaker +, with left line in going to speaker + and right line in going to left line out.

With this I can easily drive an 8 or 12" woofer to 1-2mm excursion, more than enough for an impedance curve.

I also have another jig designed for impedance measurements when the speaker is driven via an amplifier (If I want to drive it harder) - the only difference is the left and right line inputs are sampled via a ~20dB L-Pad consisting of a 1k and 10k resistor each. I seldom use it though, I just use the single resistor straight off the sound card, but not all sound cards can do this.

As for microphone - I have an ECM8000 (balanced XLR) fed into a balanced to unbalanced step-up microphone transformer, (which also has two 9v batteries and two resistors for supplying phantom power) which I then connect to ether the mic or line inputs of the sound card.

Although it seems like using the mic input is the best idea, on the Audigy the mic input can't sample at 96Khz, and is also mono, making dual channel measurements impossible. In general most sound cards mic inputs are also quite noisy and non-flat in response so aren't well suited to accurate measurements.

Because of this I usually plug my mic (via step-up transformer) into the line input, and although there is a loss of about 20dB sensitivity there is still plenty of dynamic range available. One day I'll build an active mic preamp to replace the transformer, but it's nice having a purely passive device

I would also congratulate you on figuring out Speaker Workshop - I've tried it a few times and couldn't get very far with it before I got very frustrated and confused. I've tried Holm Impulse and it looks promising, but for me I find ARTA by far the easiest to use, but still extremely powerful.

Even though money is tight for me I went ahead and registered ARTA a couple of weeks ago and don't regret it. Being able to actually save results to file and examine them later makes all the difference...