Driver distortion testing... strange results?

[Undefinition]

I have recently started getting into distortion testing of drivers (specifically, harmonic distortion)... not so much for just "consumer research" (I think Zaph and Mark K have that area pretty well covered), but more so I can find "problem spots" in a driver's passband.

I have been using Steps (part of ARTA) because, from my understanding, using stepped sine waves is the simplest way to get reliable harmonic distortion data. However, this is my first time using stepped sines; up until now JustMLS has served me very well; super-reliable.

Anyway, things SEEM to be set up correctly, but for some reason, my distortion measurements are coming out a bit weird. After testing a few drivers and getting some results, I decided to use a "control" driver (eg: one that has been distortion tested a lot): a Seas 27TBFC/G. Now I'm starting to wonder if I've screwed something up.

Anyone have any idea why the odd-order distortion is showing up so much louder than the even order distortion? Running sweeps of just the amp itself shows pretty normal results... 2nd order distortion about 80 dB down, everything else well below 100 dB, so I don't think it's the amp. I suppose it could be the Behringer mini mixer I use for a mic pre; but it's pretty much the same old UB802 everyone uses (oh, and the mic is an ECM8000).

So anyone with any experience doing distortion testing I would really appreciate any time or thought or input you could offer.

Thanks!
-Paul

[mazurek]

Try playing with mic preamp gain, I bought a m-audio audio buddy a while ago, and found that past a certain gain level it is full of distortion. Also, make sure that 'monitor' is unchecked when you are doing measurements (i.e. you don't want to play back what you are recording). Also, make sure the sample rate in ARTA matches the sample rate that your sound card is set at. One more thing, I've found that some equipment (such M-audio mobile-pre) can buzz if you hot plug things, on occasion I've found I just have to turn off the computer and stuff and restart and everything is magically better.

[Jed]

Paul, Do a FR sweep in ARTA for me and see if the results look ok. What jig are you using?

[Davey]

I have a 27TBFC/G tweeter sitting here loose I can run an ARTA distortion test on for comparison. I'll post the results when I'm through.

Here you go.

My microphone was about 1" from the mesh screen. I ran a test with 250mV and 1.0 volt drive level. I don't normally use ARTA for distortion measurements but these results seem reasonable based previous testing I've done with the tweeter.

Hope that helps.

Dave.

[Undefinition]

Paul, Do a FR sweep in ARTA for me and see if the results look ok. What jig are you using?

Jed, Here is a FR sweep I did in ARTA (using Pink Noise) at 50 cm. It looks like a 27TBFC/G should. By the way, the "jig" I use to do ARTA are just the probes I made for JustMLS. (Is that a problem? )

Davey, Thanks for doing those tests. It's interesting that you did the test at 1". I'm still unsure what is the normal distance for distortion testing. I was trying 50 cm because I think I saw that's what JKrutke uses, and it seemed like a reasonable starting point. Just for kicks, I tried mic'ing the tweeter at 1" like you did, and my measurements ended up coming up a lot more like yours.

Is there a "de facto" distance to measure driver distortion?

So that's where I'm at--still confused. Also, I've tried running distortion sweeps on the mic preamp, and its distortion is at about -100 dB, so I don't think the problem is there, either.

[Davey]

I think the closer the better for distortion testing since you're not really interested in baffle effects....and since you're using a non-MLS technique with ARTA you can swamp out pretty much everything but what you're interested in.

I believe John Krutke does his distortion testing with the mike about 4" from the driver. (Don't quote me on that.)

Cheers,

Dave.

[jkrutke]

I believe John Krutke does his distortion testing with the mike about 4" from the driver. (Don't quote me on that.)

Quote away, it's true, for dome tweeters on my infinite baffle at least. As the driver gets bigger I have to add a couple inches so the top end remains accurate. Otherwise nulls/droops start to occur as the difference from the mic to the cone center and cone edge sum. A usable distance basically depends on the driver radiating surface diameter.

If the tweeter is mounted in an enclosure or small dipole, and you're only looking for the distortion profile of that driver without baffle effects, I'd go even a little closer, maybe 1-2 inches.

If you're looking for a system distortion profile, you've pretty much got to back up the mic to get proper summing, and accept the room artifacts. That's why various system HD sweeps on my site have a big disclaimer on them.

But to the OP's query, damn that is weird and I don't have any ideas about it at the moment.

[Jed]

Jed, Here is a FR sweep I did in ARTA (using Pink Noise) at 50 cm. It looks like a 27TBFC/G should. By the way, the "jig" I use to do ARTA are just the probes I made for JustMLS. (Is that a problem? )

[/IMG]

I use the JustMLS jig as well, so I don't think that should be an issue. Odd that ARTA FR is ok, but HD testing is so far off. I'll keep thinking what could be the culprit. Maybe a bad mixer board setting? Maybe you have something coming off the jig incorrectly, and I'm assuming you restarted your computer and the program a few times to reset everything? One time I had bad HD results when I shouldn't have and all it took was a reboot of the computer.

I also measure HD at 4" from the tweeter. You have to turn the input volume on your mixer down so it doesn't clip. Sorry if all this is obvious, just writing whatever comes to my head.

[Dennis H]

Paul, what kind of mic are you using? I saw a test where the Behringers start distorting pretty heavily above 100dB which isn't really very loud if the mic is close to the driver. Maybe you're measuring mic distortion rather than driver distortion.

[seattle_ice]

Paul, what kind of mic are you using? I saw a test where the Behringers start distorting pretty heavily above 100dB which isn't really very loud if the mic is close to the driver. Maybe you're measuring mic distortion rather than driver distortion.

That is a very good point. I remember trying for days a few years back to get clean sound out of our speakers from a mic'd drumset, only to finally realize it wasn't the electronics or the speakers, it was a couple of the microphones. Someone should run some of the tests with different microphones to see if and at what levels the microphones themselves are distorting the sounds.

[dlr]

That is a very good point. I remember trying for days a few years back to get clean sound out of our speakers from a mic'd drumset, only to finally realize it wasn't the electronics or the speakers, it was a couple of the microphones. Someone should run some of the tests with different microphones to see if and at what levels the microphones themselves are distorting the sounds.

You may want to read the section at Linkwitz's site on distortion and microphones. IIRC he shows a modification to make to a Panasonic electret to allow for higher SPL measurements that may be encountered with close mics used for distortion.

Dave

[Jed]

Or try a lower harmonic distortion SPL test... say around 85db 1 meter.

[mazurek]

If you are doing two channel testing (i.e. with reference), try switching to one channel testing and see if that helps. I believe simple is better when troubleshooting, and I've seen odd results like that if I've got two channel testing enables and the reference is bad.

[JonP]

Paul, what kind of mic are you using? I saw a test where the Behringers start distorting pretty heavily above 100dB which isn't really very loud if the mic is close to the driver. Maybe you're measuring mic distortion rather than driver distortion.

I was going to mention that too. Could also be the mic preamp hitting it's wall as well.

Doing the lower SPL test should tell you that. If so, it might be good to work up, and find out what volume you shouldn't exceed when you're measuring.

Another thing, you were doing this with the sweeps (in Steps?), and the FR was done in Arta with a MLS? Might be that between the two animals, the levels/overload levels are different.

Could be the behavior is different as well. I've only done a little swept tone testing in SoundEasy, but I seem to recall a trickyness to level setting, You start out low at a fairly good level, then the gain of the driver goes up with frequency, and next thing you know you're clipping due to the rise...

[Jed]

I think he has a bad mini jack cable. Paul, awhile back I noted that your MLS measurements had excess phase rotation/spikes in the low end as well. It could be there is something more going on with your equipment that needs further investigation. It's more than likely not: the mic, the mixer, clipping, etc etc etc, because the fundamental level is where it should be.

ARTA/STEPS/ JustMLS is very sensitive to cables, and if there is the slightest possiblity that the cables are not grounded properly or there is a slight short things like this can happen very easily. I've had a few bad cables from RadioShack and ended up making my own with better wire.

[Undefinition]

Or try a lower harmonic distortion SPL test... say around 85db 1 meter.

Yeah, I think this is where the culprit lies. I have been sick the last few days, so I haven't had the energy to futz with measurement stuff.

But I will get out my SPL meter next time. It seems to me the sweeps were REALLY loud... much more painful than MLS Pink noise. In that case, the mic element itself could very well be distorting.

Thanks to everyone so far who has offered their $.02 for help. I'll report back with new results when I'm feeling good enough to get off the couch.

[Jed]

Happy new year Paul and get well soon! Sounds like you'll get this worked out in time.

[Dennis H]

Get better soon, Paul!

To test for mic distortion (not exactly how much but just to see if it's a big deal), start with a pretty loud sweep with the mic right by the cone. Then keep the signal the same and back the mic away in steps. Driver distortion should stay the same but SPL at the mic will get lower as you move away and you can see if the distortion curves change. Naturally, you have to apply a bit of Kentucky windage to decide if any changes are due to room effects or mic distortion.

[Undefinition]

A-HA! Culprit found!

Well, I pretty much drove my wife and dog crazy last night running stepped sines for several hours, trying to get to the bottom of this.

Finally I had to resolve to myself: "Maybe it's time to try another amplifier." So I grabbed a T-amp (newer version)... mostly because it was much smaller and lighter than any other amp I had on hand. The sweeps immediately came out much more normal!

I decided to go back and run a stepped sine sweep on just my testing amp (an Alesis RA-100) itself, using the "probes" as an input, and lo and behold, the odd-order distortion was louder than the even order distortion! My powers of deduction lead me to believe the prominent odd-order harmonic distortion was the amp's fault. It will now be severely punished and defenestrated (well, that or I just sell it, or relegate it to simpler duties). I think I'll build one of Bob Elliott's designs or some sort of Gainclone amp to replace it.

So here's the sweep of my (now former) testing amp: an Alesis RA-100. Now that I objectively look at the FR, it's evident why I liked listening to it; the bass is just slightly rolled off, as are the very high frequencies... midrange sort of takes center stage. Reminds me of one of my favorite microphones, the AKG-414.

[JonP]

Wow... I'm surprised that you were even able to use a T-amp. When I got one years back, one use I was thinking of was a nice, compact test amp.

Turned out, since it's output has got both sides floating at 1/2 the DC supply voltage, (BTL) I always would short the output when I connected probes. Thinking back, I think my setup had the shield grounds connected to the amp minus output, which gave a path back to the amp ground on the inputs.

If yours didn't do that, it just would be 6VDC into the line in, not a good thing, but not a dead short.

Do you use a resistor divider, or a series resistor in your line in probes?
Maybe that's what was going on with you... the DC bias affecting the input, causing it to clip early, but not blowing it out.

Any rate... good to hear you got it figured out with the new amp. Nice low distortion floor there! Next phase is setting up a quiet room, and testing to see what SPL your mic/preamp can handle without adding anything.

Another note... if you have some large value caps, you could put one on BOTH T-amp outputs and retest it. That would block the DC, and maybe you'd see a clean output. The T-amps are supposed to be very good, and that would verify it was a DC bias effect on the line inputs. I'd still use the other amp, though...

[Undefinition]

Wow... I'm surprised that you were even able to use a T-amp. When I got one years back, one use I was thinking of was a nice, compact test amp.

Turned out, since it's output has got both sides floating at 1/2 the DC supply voltage, (BTL) I always would short the output when I connected probes. Thinking back, I think my setup had the shield grounds connected to the amp minus output, which gave a path back to the amp ground on the inputs.

If yours didn't do that, it just would be 6VDC into the line in, not a good thing, but not a dead short.

Do you use a resistor divider, or a series resistor in your line in probes?
Maybe that's what was going on with you... the DC bias affecting the input, causing it to clip early, but not blowing it out.

Actually, the T-amp won't take "probes" as it causes some sort of bizarre "clicking" feedback. I think you just explained why. I only ran it single-channel (that is, no probes) to just get a sweep of the tweeter. Otherwise, I don't think it would make a good testing amp. (For just listening purposes, however, I have to admit the T-amp is quite nice)

Any rate... good to hear you got it figured out with the new amp. Nice low distortion floor there! .

To be clear, that is the distortion of the "faulty" amp (the Alesis). Yes, the overall distortion is pretty low, but that higher odd-order distortion was messing up my measurements. So as of now, I don't really have a testing amp--for distortion sweeps anyway. :cry:

[cjd]

I have some "spare" chip-amps around if you'd like to borrow one. Would need some connectors and a trafo. I'm thinking one of the "Rev-C" amps designed by Mario P and sold through TwistedPearAudio a while back. Just a monoblock should suffice, right?

[Curt C]

Actually, the T-amp won't take "probes" as it causes some sort of bizarre "clicking" feedback. I think you just explained why. I only ran it single-channel (that is, no probes) to just get a sweep of the tweeter. Otherwise, I don't think it would make a good testing amp. (For just listening purposes, however, I have to admit the T-amp is quite nice)

Hey Paul,

Why don't you modify the T-amp for a floating output? a 2200 uF cap in series with each of the outputs will allow the output to be referenced to ground and should provide flat response down to around 20 Hz. P/n 020-1122 will cost you .84 each. Nonpolarized caps won't be necessary here.

C

[Dennis H]

Paul, what's the max output voltage of your sound card? Is it possible you were overdriving the input of the Alesis when you measured the tweeter? The last sweep of the amp by itself doesn't look that bad.

[lhwidget]

Paul, I saw you mention maybe using one of Rod Elliott's amps as your test amp, and I'm thinking about doing the same thing.

I was wondering about chip amps & T amps also. The T amps (both the PE amps and the designs on the the net) appear to be based on the Tripath digital (PWM) amp chips, great performance for audio, but they require a fairly stout output filter to remove high frequency switching artifacts.

The chip-amps (LM3886) use extensive protection circuitry to guard against casual abuse (check Elliott's project for the LM3886). If the amp is driven hard, or is driving a difficult load, it may begin to trigger its protection circuitry.

Elliott's P3A amp is a plain-ol' (well documented & tested) class AB amp using discreet components. No protection circuits (a shorted output will instantly kill it). Frequency response is basically flat from a few Hz to ≈ 40 kHz, and it's a clean amp too.

Here's what I'm curious about:
1) How important is extended freq response? I realize it takes measurements to 50 kHz and a 96 kHz sampling rate for a full HD sweep to 10 kHz, but does extended freq response help when using MLS based systems?

2) Does anyone have freq response info on the LM3886 based chip-amps? I'm wondering if their high freq output is intentionally limited.

EDIT:
I realized after this post that adequate response to capture F5 during a signal sweep to 10 kHz requires the mic-preamp-sound card path to have a calibrated response to 50 kHz. The excitation signal stops @ 10 kHz. ops:

My question is: Is extended freq response on the signal side is of any benefit for MLS based measurements? (Assume the receiving side has extended freq response)?

[Undefinition]

Thanks everyone so far.

Chris,
That is a very generous offer. I'll get in touch with you about that, if for no other reason than to hear some of these "chip" amps.

Curt,
I like the idea. I'll try it next time I put an order in at PE.

Dennis,
I don't really know how to find the output voltage of my sound card. According to the "estimate" in ARTA (based on the generated 400 Hz sine, which I then measured w/ my multimeter), the max output voltage is 1788 mV. Times like this, I'm reminded of how much I wish I understood electronics better.

[Dennis H]

Paul, I guess the voltage numbers don't really matter too much. The idea is to turn the volume down low enough (-10dB maybe?) in the ARTA software that you aren't giving the amp too hot a signal for its input stage to handle. Then set it how loud you want with the amp's volume control. You'd have to experiment to see what's going on.

[Jed]

Paul, have you posted the result with the new amp yet?