Alpha15a distortion measurements log

A year or two ago I built the H-frame open baffle 'enclosures' using the Eminence Alpha15a drivers, as designed by MJK. More info here. I've been very happy with them, but as 'progress' marches on, I've been examining other drivers, enclosures, etc. There were a few other options that really impressed me, like the XLS drivers in the Orion, or some of the Acoustic Energy drivers. Monte Kay has published a nice set of data, particularly woofers.

So I've become interested in taking a more in depth look at the Alpha15, and though I would share some of the findings, and perhaps get some feedback. First off, I know that people dispute the audibility of non-linear distortion. So lets not go there - we'll just talk about measurements here.

And in simulation, a lot of things look pretty good about this 'lowly' $60 woofer. In simulations, it seems to be more efficient than most other drivers. In an H frame, it has a particularly nice frequency response, requiring little to no EQ boost in the low end, if you can tolerate a f-3 of about 30Hz.

My only qualm with it is its smallish xmax - 3.8mm peak to peak (I think). It only takes about 8W to get there around 35Hz - but that isn't actually too bad, considering that you get about 96dB SPL out (at one meter), and minimal thermal problems compared to pumping 250W through some of the 'big boys'. But you can add 6dB for another driver (the other channel); and I just purchased another set (so I'll have two per side), so I should be easily able to hit 108dB SPL before running into big linearity problems.

Speaking of linearity, I had assumed that because I was already getting a lot out of a cheap driver, non-linear distortion would not be all that great. So I started doing some measurement today, and thought I would share. I use Soundeasy and a calibrated, DIY Linkwitz-style WM61 mic into a EMU 0404 USB. The mic was place at the center of the plane made by the opening of the H frame.

[IMGHTTPDEAD]http://img211.imageshack.us/img211/985/alpha1a0021.jpg[/IMGHTTPDEAD]

Well, things actually look pretty good. First, a look at the basic, unequalized frequency response and test conditions:
[IMGHTTPDEAD]http://img15.imageshack.us/img15/871/freqresp.jpg[/IMGHTTPDEAD]
Not bad, rolls off a little higher than predicted (28Hz), that may be due to age or being set on carpet.

So I did a set of single tone distortion tests, at 30, 40, 50, 60, 70, 100, 150 and 250 Hz, and a warble tone test at 60Hz, just to see how they compare. All were done with a 10Vrms (or 14Vpp on the scope) signal at the driver terminals. All graphs were normalized to 0dB, so just read the legend on the left for the relative levels of the harmonics.

First - 30 Hz - not spectacular, not terrible. A good bit of 3rd:
[IMGHTTPDEAD]http://img193.imageshack.us/img193/5465/30hzo.jpg[/IMGHTTPDEAD]

40Hz - getting better, still some 3rd:
[IMGHTTPDEAD]http://img188.imageshack.us/img188/6868/40hz.jpg[/IMGHTTPDEAD]

At 50Hz, things start to clean up their act:
[IMGHTTPDEAD]http://img189.imageshack.us/img189/7961/50hz.jpg[/IMGHTTPDEAD]

And they just keep getting better:

[IMGHTTPDEAD]http://img181.imageshack.us/img181/6256/60hz.jpg[/IMGHTTPDEAD]

[IMGHTTPDEAD]http://img375.imageshack.us/img375/4589/70hz.jpg[/IMGHTTPDEAD]

[IMGHTTPDEAD]http://img375.imageshack.us/img375/6710/100hz.jpg[/IMGHTTPDEAD]

[IMGHTTPDEAD]http://img218.imageshack.us/img218/9371/150hz.jpg[/IMGHTTPDEAD]

[IMGHTTPDEAD]http://img193.imageshack.us/img193/4327/250hz.jpg[/IMGHTTPDEAD]


Now, I would like to draw you attention to the data Monte Kay took from the Peerless 12" XLS, particularly F4 and F5 - they aren't terribly different.
[IMGHTTPDEAD]http://www.mfk-projects.com/images/DPL40.gif[/IMGHTTPDEAD]

Above 70Hz, the Eminence is its equal:
[IMGHTTPDEAD]http://www.mfk-projects.com/images/XLS70.gif[/IMGHTTPDEAD]

This may not be a fair comparison for a number of reasons, the first probably being that Monte does not use any baffle when he tests his drivers (there is no write up on that, but there is a picture of it on one of his pages). But both of our data should be valid, I would say a general comparison should be allowed - my conclusion is that the Alpha15 has surprisingly low non-linear distortion, particularly given its price and the number of other perks it has. Also consider how much less power it takes to drive the Alpha15 or and how much more volume displacement it has for a given voice coil displacement. And with 2 per side, they should be 'loafing'. I'd say thats not bad at all.

Now, how loud was I testing them at? Well, my mic/preamp has been calibrated, and is supposed to put out 199.2mV/Pa. Playing a 60Hz tone, the scope showed 7.3Vpp, or 5.16Vrms. I calculated this as being 122dB SPL (!). Remember this is right at the opening of the H frame, with no back wave cancellation - at a meter or two (where I listen) this would be a bit less. I hope someone checks my calculations. There are some very handy calculators here and here.

Next up, I'll be measuring the distortion levels at different power levels, and see were things crap out. I'll try and measure the in room response and SPL at the listening position. I would like to encourage people to share their perspective on this, as I'd like some more ideas of what to examine. In a recent post by Lynn Olson, he was saying that very low bass frequencies will cause substantial IM distortion of upper bass in OB configurations - so my thought was to high pass the OB somewhere around 40Hz and use a sub. I think the data above suggests this also.
 
Okay, here is some more data - harmonic distortion sweeps at different power levels. Each graph has 5dB higher drive signal, with the first being about 4.25Vpp, the second being 8Vpp, then 10Vrms, and then 17Vrms.

This first one is encouraging - low distortion, especially tall order. Its the 'first watt' that counts, right.

[IMGDEAD]http://img197.imageshack.us/img197/9541/425v10db.jpg[/IMGDEAD]

Distortion then gets progressively higher (ignore the two tall narrow spikes - the computer glitched here).

[IMGDEAD]http://img200.imageshack.us/img200/8131/8v5db.jpg[/IMGDEAD]

This graph should match the data in the first post - they have the same drive levels.

[IMGDEAD]http://img199.imageshack.us/img199/1834/14v0db.jpg[/IMGDEAD]

Distortions getting high here with a 24Vpp (17Vrms) drive signal. But read the notes about this below.

[IMGDEAD]http://img145.imageshack.us/img145/4762/24v5db.jpg[/IMGDEAD]
 
I measured/estimated/calculated the dB SPL of that last signal at my listening position about three meters from the drivers. In the 100's I was getting about 105dB - louder than I thought! But look at this - yes, the ugly truth:

[IMGDEAD]http://img354.imageshack.us/img354/7364/analogsweeplisteningpos.jpg[/IMGDEAD]

This is the frequency response of my listening position. Looks horrible huh? This was done with an ungated analog sweep, drive level 24Vpp. That broad dip from 50-100Hz concerns me, as does the overall large level drop below 100Hz. Where is my f3 of 30Hz? Who knows. I was getting about 95dB SPL through 60-100Hz. I only did a quick estimate while the signal was sweeping - I could measure individual tones, but right now I'm happy with an estimate. Although the frequency response is pretty bad, the SPL level is quite high, and as you can see in the last graph of the post above, tall order harmonic distortion was -60dB or better above 60Hz, which I think is okay. Add the speaker from the other side (+6dB), then the two more I'll add later (+6dB), and things should improve a bit - plenty of SPL capability, and lower distortion. I guess I should look for a better listening position though.

I also measured the harmonic distortion at the listening position with the same drive level, just to see what happens. In general the levels haven't changed, but just gotten a bit rougher.

[IMGDEAD]http://img150.imageshack.us/img150/4915/24v5dblisteningpos.jpg[/IMGDEAD]
 
Here is another interesting tidbit. I wanted to see what happened to the drivers distortion profile when tested without a baffle - which is how Monte Kay and Siegfried Linkwitz do it.

The two graphs are for 10Vrms drive signal, and then 5dB lower, 8Vpp (whatever that is rms, I didn't calculate it).

[IMGDEAD]http://img150.imageshack.us/img150/1834/14v0db.jpg[/IMGDEAD]

[IMGDEAD]http://img145.imageshack.us/img145/8131/8v5db.jpg[/IMGDEAD]

The main thing here is that distortion is considerably higher - about +15dB for the taller orders. What this implies is that the XLS driver has 25dB less tall order distortion than the Alpha15 (MK's graphs show its 10dB down from my woofers, which are in enclosures; then, when my woofers are taken out of the enclosures, they gain about 15dB distortion - so a total (estimated) difference of 25dB). I would guess this is a significant difference.

The other interesting thing is how much smoother the individual harmonics are as they sweep up when unbaffled. 3rd and 5th show increasing levels at higher frequencies, which the driver doesn't show when baffled.
 
Thanks for sharing your knowledge.

Looking at the picture and guessing at the depth of the H frame I am not surprised that the SPL is much lower below 100Hz.

I have found John K's ABC Dipole to be very accurate when used to calcuate driver response when mounted on U / H frame. I did on 3 different U frames with 3 different sets of drivers and my measurements corresponded to the simulations very well.

With my latest U-frame I only targeted -3dB at 55-65Hz, and will use the Linkwitz sub below that.

Regards,
Bill
 
Here is a question I could use help with - does one use the Vrms or Vpp of the mic output to calculate dB SPL (it makes a 3dB difference).

Here is an example - I play a 2500Hz tone through my tweeter - I connect the scope to the output of my mic/pre, calibrated to giv 199.2mV/Pa. The scope shows 1.75Vpp - this is equivalent to 1.24Vrms. Divide .1992V by both Voltages to convert to Pa, then compute [20log(Pa)+94] to get the dB SPL. There is a three dB difference between the two options, and I can't figure out which to use, the rms or pp Voltage...
 
Well, here is the next installment, I think some people might find interesting.

This is a continuation of in room measurements (although this set was done out doors to reduce reflections). When I did my first in room measurements, I was fairly surprised by how bad they looked - but I thought I noticed an unexpected trend - and here it is:

[IMGDEAD]http://img174.imageshack.us/img174/3503/a15deltadistance2.jpg[/IMGDEAD]

Here we see a successive series of measurements with only the mic distance changed; starting at the mouth of the H-frame (nearfield) then moving back to 1 foot, 3, 6 and 9 feet. Notice the trend to a 6dB/octave rolloff below the baffle peak (about 280Hz). Also, there are plenty of reflections in this data, particularly at longer distances, but the 6dB/octave trend should be obvious.

One of the main reasons for using the Alpha15a in an H-frame (as shown by MJK) was the high Qts of the driver leading to a knee in the bass frequencies, and subsequently, no need to give the driver an EQ boost to compensate for the OB, 6dB/octave roll off below the baffle peak. This is clearly seen in the nearfield measurement - but unfortunately that knee disappears at greater distances (aka, listening distances), something that I haven't seen demonstrated yet.

This may come as a big hit to the popularity of the Alpha15 and the "high Qts is better" crowd (me included)! I guess there are no free lunches. At this point, I want to mention John K's article, dipole woofer on axis response in relation to listener distance, which is a description of the same phenomenon, I think - just worded from a different point of view. There is a very valid point in there, wrt the back wave reflection having a different EQ than the direct wave.

I guess the main implication of the above measurements is that those of us with high Qts drivers may need to add some EQ!
 
Wow! Thanks for the charts and comments! Very much appreciated!

http://www.zillaaudio.com/speakers-tweeter-woofer-bib.htm

My Alpha 15's in H-frames sound great in my room, full, fast and dynamic bass! My kids have on occasion run from the room... i think i was playing the Happy Feet soundrack lol... I wanted to note I use my H-frames from lower than 100hz down, maybe more like 70hz down (i play with the variable crossover on the amp but settled on about 70hz) so the charts look very good to me from 100hz down. Personally, going higher than say 150hz the woofers begin to call attention to themselves in a negative way when used with my main speakers. Certainly different mains will need to be blended differently but using the H-frames in this manner works very well for me and the critical as well as casual listening i do from time to time.

Thanks again.
Godzilla
 
This is a continuation of in room measurements (although this set was done out doors to reduce reflections). When I did my first in room measurements, I was fairly surprised by how bad they looked - but I thought I noticed an unexpected trend

In a room and with the crossover would be a more interesting set of measurements. I don't know if your conclusion would be the same. Then maybe do the same test with a lower Qts driver that is Eq'ed to be flat down to 40 Hz and see if the results change. At that point an interesting comparison of low and high Qts driver responses could be formulated.
 
MJK said:


In a room and with the crossover would be a more interesting set of measurements. I don't know if your conclusion would be the same. Then maybe do the same test with a lower Qts driver that is Eq'ed to be flat down to 40 Hz and see if the results change. At that point an interesting comparison of low and high Qts driver responses could be formulated.


I just got in a 12" woofer with Qts of .33 (mfg. spec). I'll try what you suggest, it might take me a couple of weeks though - I'm still moving into my new place...
 
Here is another set of outdoor measurements, which I believe better show that between 30 - 100Hz, at 6' and 9', there is a 6dB/octave roll off.

[IMGHTTPDEAD]http://img194.imageshack.us/img194/3521/a15deltadistancemkiii2.jpg[/IMGHTTPDEAD]

I made more measurements in other locations, and feel confident that there is a steady/smooth 6dB/octave roll off in the bass at listening distances.
 
Maybe I am missing something, but, to me, your results are the expected ones ( btw, I have similar results from my Alphas in shallow U-baffles).

If you measure outdoors and nearfield - and "nearfield"at 50 Hz is pretty far ! - you will miss the dipole effect, which progressively turns on as you increase the distance. If you think of a dipole as of 2 separated sources, measuring from a distance that is comparable to the equivalend distance between the sources will have one of the sources dominating the measurement - it is only as the distance grows when you will start seeing the full 6dB/oct effect as you do in your measurements.

I always thought that EQ-ing a H/U dipole flat nearfield does not make much sense, you need to do this further away (I took the listening distance).

Of course, the room will add it's own to that: in my case, however, in room response was not much off compared to outdoors measurement (flat EQ from about 2.5 meters)

BUT: I have found that the EQ in the 30-100someting range is incredibly important, as seemingly the perceived sound changes dramatically by small changes (especially the overall balance between low and high end of the interval)

I have also found that using a Linkwitz-style 100-200 shelving highpass filter to account for the full/half space transition (my crossover being at around 135 Hz) simply get things right: I could never move back to the simple level adjustmend between bass and mids.