Most interesting piezo test

[LelandCrooks]

Today I set up for the first time to test piezos using my new software. Always before it had been by spl and by ear. It's pocket pro rta running on my pda, and the pc, sigjenny for a sweep generator.

Now afore you start questioning my test procedures, understand that its only referenced to itself. So even if the test system is out of scale, it's out of scale only to other systems, not itself, so variations within it are consistent. My deq2496 went back to the band before I got to calibrate the meter in the pda. It does however calibrate itself to the internal mic in the pda.

First spl with pink noise. 24 piezos within 1db of each other from 2k to 12k. +- 2-3 db above 12k. Almost ruler flat to 12k.

Here's where it gets interesting. Sweep from 2k to 15k. 9 out of the 24 exhibited an odd characteristic. You would hear a secondary and tertiary harmonic somewhere between 8k and 15k, below the sweep, in the 4k and 2k region. Pretty loud on some, faint on others. It happened too fast for me to capture it, so I ran just 8k, 10, 12k, and 15k tones. I saw the expected bumps in response at 4k and 16k with the 8k tone, but no evidence of what I was hearing. Not there. I didn't want to spend all day looking for just the right freq, so I just ran sweeps from 8k to 18k. A couple more exhibited it at 18k, but they were of high enough frequency to be neglible.

The worst ones in the 8k region popped somewhere around 2-3k, which would be a tertiary harmonic. It was a very stepped sound. I'm betting this is the source of the some of the nastiness that piezos can exhibit. We end up swapping out elements because this shows up. I know I've built arrays, isolated each one with a tube and there are distinct differences in some of the tweeters. The array smooths it all out if it's not too bad.

The good ones were smooth as silk.

One nice thing I found is I can still hear to 15-16k. Pretty damn good for almost 50.

Thoughts? Suggestions for further testing?

[Bill Fitzmaurice]

I'd be looking at vibration of the unit. Were the horns glued up?
In my last batch I tested with pink one had an odd response, so I pulled the element. It looked fine. I reassembled it and it worked fine.
It would be interesting to view the signal on a scope as well.

[LelandCrooks]

Individuals before I build the array. I thought it might be something with the gear, but it was consistent. I tested the good ones two or three times, and the odd ones two or three times. Same results every time. I believe that rules out the gear, but maybe not. Granted my techniques aren't very sophisticated, but the fact that I could replicate it over and over I feel means I'm on to something.

I'll pull the elements on the odd ones and a good one and see if I can find a difference. I have a scope capability on the software, I'll just have to figure out how to use and capture it.

[Bill Fitzmaurice]

Individuals before I build the array. .

I meant whether the horn and throat plug had been glued up, if not there could be a vibration issue.

[Sydney]

I have a scope capability on the software, I'll just have to figure out how to use and capture it.

If the scope is PC based - Do you mean something other than a screen snap to jpg?

[LelandCrooks]

I was mistaken it's not a scope. I can do linear, log, various octaves, spectrograph and spl. And I did find the snapshot to jpg utility.

I've seen software scopes somewhere, I'll go looking again.

[LelandCrooks]

Having now done this on well over 100 tweeters a pattern begins to emerge. It's second order harmonics, below the fundamental. On some it's very pronounced, others very faint. Predominately in the 15-18k region, which puts it at 7.5-9k. Some will exhibit it as low a 12k. The ones below 15 are always much louder, drowning the fundamental, but that's probably my hearing. I have not disassembled any yet to determine a factor.

Because so many exhibit it above 15k, it's not possible to sort them, you'd only get 3 or 4 out of 50. But it is possible to find many with it as faint as possibe. Any below 15k I removed from contention.

[Bill Fitzmaurice]

Since it occurs an octave down it's not a harmonic, more likely its caused by some sort of non-linearity. If the signal triggering it is at 15kz and up it's not much of a concern in pro-sound, where there's very little content above 12kHz anyway. But those that exhibit it at 12kHz I'd can. Before doing so I'd remove the element, then reattach and retest to be sure it wasn't a misalignment issue.

[gdougherty]

Having now done this on well over 100 tweeters a pattern begins to emerge. It's second order harmonics, below the fundamental. On some it's very pronounced, others very faint. Predominately in the 15-18k region, which puts it at 7.5-9k. Some will exhibit it as low a 12k. The ones below 15 are always much louder, drowning the fundamental, but that's probably my hearing. I have not disassembled any yet to determine a factor.

Because so many exhibit it above 15k, it's not possible to sort them, you'd only get 3 or 4 out of 50. But it is possible to find many with it as faint as possibe. Any below 15k I removed from contention.

I'm wondering if a steep low pass at 14-15K would help tame the harmonics generated by the upper frequencies and improve the sound of the speakers at higher volumes. Thanks for the test methodology Leeland. I'm sorting through about 144 right now. Very helpful.

[LelandCrooks]

You're welcome. I've done this on over a thousand by now. It's very consistent. Under 15k it's pretty rare, above it, about 10%. I think a filter is a very good idea. It could be a very simple 12db per octave. I don't think any steeper is necessary.

[gdougherty]

You're welcome. I've done this on over a thousand by now. It's very consistent. Under 15k it's pretty rare, above it, about 10%. I think a filter is a very good idea. It could be a very simple 12db per octave. I don't think any steeper is necessary.

Maybe I need to slow down the sweep I'm using. It's 2.5K-16K over 10 seconds. I'm not seeing or really hearing the lower harmonic you're talking about using the RTA on the DEQ. By the time it sweeps up to 16K the RTA is reading a large blob from 10-16K and I'm not seeing anything an octave down from any of that range that's within 30db of the signal. What I am often getting is a resonance one octave above 2.5-3K. I have a bunch of tweeters I'd already pulled from the horns and some still mounted. The pulled ones I'm hand-holding tight against the horn and Ive noticed on some of them I can get it to disappear by pressing down even tighter. Mount those on a horn though and I'll often hear and see harmonics I didn't see when hand holding.
My plan is to take those and test them with damping material wedged between the cone and ceramic element plus a thin layer of silicone caulking (dried) on the element edge to seal it to the horn. It's possible the seal alone will eliminate some or all of it if pressing harder helped. I pulled at least 20% of the ones still mounted but only a handful out of about 80 that were off the horns. I'm thinking this might be the harshness I'd been hearing before that exhibited itself on piano and vocals with signal in that upper mid-range.

[Mark Coward]

I reported this before, using Pink Noise and RTA at least 10% of the 1016's exhibited a steep notch around 2.5K - 3k. The "good" ones had a fairly smooth rollof down past 2k. After a while I could hear the difference without looking at the RTA.

[LelandCrooks]

I reported this before, using Pink Noise and RTA at least 10% of the 1016's exhibited a steep notch around 2.5K - 3k. The "good" ones had a fairly smooth rollof down past 2k. After a while I could hear the difference without looking at the RTA.

Yep. No rta after you've done it a while

[gdougherty]

I reported this before, using Pink Noise and RTA at least 10% of the 1016's exhibited a steep notch around 2.5K - 3k. The "good" ones had a fairly smooth rollof down past 2k. After a while I could hear the difference without looking at the RTA.

It's not a notch I'm seeing, it's a resonance spike at 5-6K during the 2.5-3K signal period. After the sweep passes 3K the spike disappears. It's not severe, but it's ~20db down from signal level.

[Bill Fitzmaurice]

it's ~20db down from signal level.

If you can't hear it forget about it. Down 20dB is insignificant.