Test method for acquiring intermodulation distortion values

[Joseph Crowe]

I would like to share my latest blog post. I compare two woofers and propose a test method for acquiring intermodulation distortion values.

https://josephcrowe.com/blogs/news/eton-versus-dayton-12-woofer-comparison

 

[amirm]

Good work. I too experimented with different tones but gave up earlier than you did. I like the way you showed paired tones as a continuous graph. What I struggled with was believing the results I was getting. You did a decent try with listening tests but we need more conclusive evidence there.

I find that in low frequencies different types of distortion come and go and sometimes act together. This makes it hard to run with just one measurement.

 

[AnalogSteph]

It may be interesting to do some cross-checking with the kinds of frequencies that Purifi use for their drivers... for example, their 6.5" ones are spec'd with IMD spectra of 30 Hz + 255 Hz or 50 Hz + 425 Hz, all at 80 dB SPL.
https://purifi-audio.com/ptt6-5x08-nfa-01/

In this case the IMD testing confirmed that the Eton struggled more at larger excursion, but the substantial electric nonlinearity would not have been nearly as obvious without it. That's clearly the shorting ring making a difference there. (Not necessarily at the frequencies relevant for a sub though. A midwoofer would be a different story, but neither part lends itself to that application.) The Dayton would seem to be the much more modern construction in general. The only application that would favor the Eton is PA, as maximum midbass output still is 3 dB higher.

 

[Savi]

Really interesting article and measurements ! Thx !
When looking for a woofer, I often ask how to determine its maximum xo point (cone breakup, beaming, thd, imd ?). This kind of intermodulation measures could be a part of the answer.
Hificompass, Erin and french blog justdiyit.com also do these kind of measurements. You all have a different way to test it and present it: my favourite are yours and Erin. But in your example, I would have tried a larger F1 & F2 range. For instance 70Hz + 296Hz & 70Hz + 480Hz ?
For instance, based on your measurements, what would be the highest xo for the Dayton with LR4 ?

 

[oivavoi]

I would like to share my latest blog post. I compare two woofers and propose a test method for acquiring intermodulation distortion values.

https://josephcrowe.com/blogs/news/eton-versus-dayton-12-woofer-comparison

Interesting! Very cool test.

This is not central to your point, but I was wondering about this sentence: "If I attempted to match the frequency response between the two drivers I would end up with new variables such as increased group delay on the driver receiving additional EQ". This is probably a complete noob question on my part, but can you explain this? Why does eq increase group delay? Does this always happen when equalizing a driver?

 

[KSTR]

Interesting! Very cool test.

This is not central to your point, but I was wondering about this sentence: "If I attempted to match the frequency response between the two drivers I would end up with new variables such as increased group delay on the driver receiving additional EQ". This is probably a complete noob question on my part, but can you explain this? Why does eq increase group delay? Does this always happen when equalizing a driver?

Good points.

I will share a method that has proven to work and gives reproducible results:
- both drivers must be EQ'd to the excact same frequency response on a test baffle (of both magnitude and phase/group-delay), and levels matched. That response should be close to a realistic response one would choose in the acual application, so clearly band-limited and clean. Best done with FIR, from one of the DRC/XO packages like AudioVero's Acourate. As drivers are minumum phase and the target is (selected as) mimimum phase, the correction will also be minimum phase. When the drivers are of the same cone area, the cone movement will now be identical, and thus same operating conditions for each driver, at least to first order (neglecting the distortion)
- then record each speaker's response to piece of music playback, test signal, whatever. No change in driver or microphone position allowed .. this also applies for the initial measurements. Close miking (< 1m) and large damped room if feasible.
- listen to the recordings with headphones in a A/B-test (or ABX)

The obtained data (the measured impulse responses) can also be directly be used to "extract" the distortion and other ill effects in subtractive analysis:
- convolve the source signal offline with the filtered IR, that obtains the linear part of the response to the signals in a test file
- compare that to the original recording (subtraction, using DeltaWave), which will isolate or at least magnify the distortion when listening to the residual.