Test Setup and Methods

Test Setup

Measurement Conditions

All my testing is done in my garage.  It is not an ideal environment but I have taken steps to ensure the data is as accurate as it can be.  This means spot treating the room with acoustic panels, and testing when the temperatures are approximately equivalent to ensure accuracy and consistency within my own data set.

Test Baffle Discussion

The test baffle I use was initially a scaled IEC baffle.  However, once I used that for some time I realized that it was terribly impractical at over 8 feet tall and nearly 6 feet wide.  It literally consumed my bonus room.  After a period of using this baffle I decided to go with something more space friendly and ultimately scaled down the original baffle to a size of 1256mm wide, 1508mm tall.  I added large caster wheels, providing me the ability to roll the setup in and out of place as needed.  The drawback, like any other measurement baffle, is a baffle influence in the measurement.  Since all testing has this same baffle influence the results within my data are comparable.  For the added benefit of full disclosure, I have simulated the baffle size/shape in Tolvan’s EDGE software and provided farfield and nearfield response simulations due to the baffle itself and included them below.

Baffle Dimensions

FF NF Sim

My frequency response testing is done in a manner that merges nearfield and farfield response at a frequency that will mitigate the wild lower frequency shifts you see above.  When performing off-axis measurements, the baffle also has an influence as well, dependent upon mic proximity relative to the baffle.  This is not simulated but you should have enough data above to do this if you desire to see those results.

 

Measurement Hardware (Please See Updated Information Further Down)

All my testing is done using the industry reknowned Klippel Distortion Analyzer R&D unit and a Klippel-supplied MI-17 Microphone along with supplied measurement cables and ICP power supply (for the microphone).  With this hardware and accompanying software modules I can obtain measurements that even some manufacturers are not capable of attaining themselves.  Thus giving the viewer a clearer understanding of the electromechanical properties of the driver tested and insight in to the source of various distortion components.

IMG_0693 IMG_0694 IMG_0700 IMG_0701 IMG_0703 IMG_0706

Update 01/11/15:

I no longer will be using the Klippel products for all my testing (no drama related; just had to simplify my gear for financial reasons).  I am now using the combination of Dayton’s DATs module for basic Thiele-Small Parameters, and either Dayton’s OmniMic or Room EQ Wizard plus a Dayton UMM-6 Microphone for FR/HD/IMD testing.

 

Test Methods

Thiele-Small Parameters

All drivers are sufficiently broken in prior to testing.  I have measured differences and I have indeed found that break in can affect measured results to some degree.  Not just in T/S results, but distortion results as well.  Whether these results are audible is for debate but it is indeed a measured phenomena so all drivers are broken in before test results are posted.

As stated above, my measurements are done in my garage.  The environment does affect measurements and I have taken steps to ensure this is not a problem.  However, this absolutely means that my results will almost certainly never match a manufacturer’s results to a T.  Standard measurements such as Fs, Qts, Re, and Le are not influenced by the environmental conditions.  Other parameters such as Vas are, though.  Keep this in mind when viewing results and use caution when comparing directly to others’ results.  Not only does environment affect measurements but the quality control of a manufacturer does as well.  You may find that many manufacturers specify a tolerance on their spec sheets because of this.

To those wanting to extract T/S parameters to design an enclosure, I recommend buying something like Dayton’s DATS system.  It is extremely easy to use and will provide you with measurements under your conditions in less than a couple seconds.  For the DIY’r, this is a must have.  I use this for some in-car testing on the fly.  It takes the guesswork out of building and modeling enclosures.

Frequency Response Measurements and Associated Harmonic Distortion Results

All Frequency Response (FR) and Harmonic Distortion (HD) results are given at 2.83v/1m.  This is to give you a baseline to compare other tests with so you can have a good idea of the measured sensitivity.  Linear Distortion is a term used to describe a driver’s frequency response.  Frequency Response does not change with volume (if you discount compression effects) and therefore, this is called the “linear” distortion parameter.  Non-Linear Distortion, aka: Harmonic Distortion, Intermodulated Distortion, etc, is used to describe a driver’s actual distortion components as the volume is raised (hence: non-linear, because this type of distortion changes with volume).

Due to the reflective nature of the room, all far field (standard 1 meter distance) testing is gated to no more than 12ms.  This is the time before the first reflection in my environment causes irregularities in the measured response and allows me to essentially remove them from the measurements.  This is known as a quasi-anechoic measurement.  Because of this window, farfield measurements only have resolution down to ~120hz, or 300hz in the cases where I window out >3ms.  However, information below this can easily be derived from high output harmonic distortion testing (see below) which is done in the nearfield (see below).

Each FR/HD measurement provides on and off-axis measurements.  This is to give the user a good idea of the response in varying axes.  While some believe this isn’t particularly important for midwoofers due to the typical crossing point below the point of beaming (ie: where a driver’s off axis response separates from on-axis response), it does give a clear indication of modal issues often due to the cone.  These are spotted more easily when off-axis results are given because these modes will appear in all axes.  If an on-axis plot shows a bump in response at 5khz and the other off-axis measurements do not mimic this, then it’s likely a non-resonant characteristic of response.  However, if the off-axis measurements do show this, then it is likely a resonant mode in the driver itself and can give an indication of high(er) frequency performance and provide a better idea of where a low pass filter and slope should be used.

Harmonic distortion is always relative to the on-axis measurement and therefore the PHD (a distortion limit based on likelihood of audibility) is given based on this.  You’ll see in each of my results I provide a fundamental mean, which is simply an average SPL.  The PHD line is given at -40dB from this average line, where -40dB is equal to about 1%.  This is done to give the user a quick feel for where the higher and more (potentially) audible distortion components are.  A quick glance can tell you where any issues may lie.  Whether or not they are audible to you depends on much more than I am even knowledgeable on.  Suffice it to say, 40dB from the fundamental is a low audibility value and, it seems to be common knowledge that even order distortion is less ‘harsh’ than odd order distortion.  So, if you see 2nd order distortion rise above the PHD line it likely isn’t as offensive as 3rd order distortion (or any odd order distortion component) above the PHD line would be.

For what it’s worth, audibility of distortion can very on many different things such as individual listener thresholds but also the type of music being played.  I have found through my own individual testing that 3% is a good average number, which is equal to 30dB down from the fundamental.  Above this value, distortion audibility increases.  I provide the PHD limit at 1%, but keep in mind there is wiggle room in this regard.

 

High Output Harmonic Distortion Testing

This testing is always done directly on-axis in the nearfield; at approximately 4.75″ distance from the driver being tested.  This is done in order to mitigate the room’s effect on measured response and overcome any ambient noise floor which could potentially influence the results.  This means more accuracy for lower frequency content as well.  Pair the frequency response measurement here with the one provided in the farfield and you can have a good idea of where the driver rolloff is and performance on the low end.

All testing is done at an equivalent to 96dB at 1 meter.  In some cases I could push the driver under test harder but I think this is a very good number to use as a measurement point because 96dB for one speaker is actually quite loud.

The reason for this additional HD test is to give the user a better feel for real-world use and potential crossover choices when it is driven harder.

Intermodulated Distortion (IMD) Testing

Like the high output Harmonic Distortion tests, the IMD testing is done in the nearfield at 4.75″ and referenced to 96dB at 1 meter.

The IMD testing consists of two components:

  1. Voice Sweep
  2. Bass Sweep

The purpose here is to sweep a driver in a desired passband while also driving it with a single tone.  This better emulates real world use where a driver is playing multiple tones at one time.

1. Voice Sweeping is performed in different ways depending on the type of driver.

  • Tweeters:  Fixed bass tone is 80hz.  Swept frequency range is from 800hz to 10khz.
  • Midrange:  Fixed bass tone is 40hz.  Swept frequency range is 600 to 6khz.
  • Midwoofer:  Fixed bass tone is 20hz.  Swept frequency range is 400 to 6khz.

 

2. Bass Sweeping is performed for mids and midwoofers in the following manner:

  • Midrange:  Fixed voice tone is 800hz.  Swept bass range is 40hz to 300hz.
  • Midwoofer:  Fixed voice tone is 2khz.  Swept bass range is 20hz to 120hz.

 

Parting Thoughts

Should I alter my test method in any way, I will note it in the individual results.  If you have any questions, feel free to contact me by leaving a message.

Leave a Reply

Your email address will not be published. Required fields are marked *