Eight inch woofer testing, report

[DaveFred]

Hello Everyone,

I am looking to build a speaker with an 8" woofer and a 4" tweeter in 1.5 cubic feet or less. These are the constraints, please don't make suggestions on what to try outside these guidelines.

Because I do not know which are the best drivers to choose, I am going test the following,

Woofers,

- SB Acoustics SB23NRXS45-8
- Human Speakers PRO 005
- Ciare HWB200
- Dayton Audio RS255-8
- HiVi LK8A
- Seas U22REX/P-SL
- SB Acoustics SB23NACS45-8
- Wavecor SW223BD02
- Dayton Audio RSS210HF-4

Passive radiators,

- Scan Speak 26W/0-00-00
- Dayton Audio DSA270-PR
- Dayton Audio RSS265-PR

Tweeters,

- SB Acoustics SB26STWGC-4
- Seas 27TBCD/GBDXT
- Human Speakers PRO 002 M2
- SB Acoustics SB29RDAC-C000-4
- BoZhen CQ66



Tests will be carried out on a 1.5 cubic foot poorly braced test box 20"x13.5"x15". The test box has replaceable 10"x10" panels so woofer of different sizes can be swapped in and out, same with 6"x10" panels for tweeters and side panels for passive radiators.



Testing equipment will be an Earthworks M23 XLR microphone on a home made height adjustable stand for far field and a Umik-2 for near field. Amplifier is a Neurochrome LM3886 DONE RIGHT built by me, along with a Motu M4 audio interface. Arta will be used for dual channel farfield measurements with loopback. Automated turntable is home made under control of Arta.





I strive for as accurate as measurements as possible as well as repeatability. Microphone distance from baffle will be 1m, stand is fixed to turntable with a backstop so test box can be returned to the same position again and again. Baffle has been aligned to center of turntable rotation and mic centred vertically and horizontally. Mic stand allows for one axis of movement, up and down only.






First up will be the 8" woofers.

Thanks for watching.

David.

[Evil Twin]

I'm sure a lot of folks will be interested to see how this works out and how the parts look- obviously, besides SPL measurements, HD2 and HD3 would be of interest, along with impedance tests free air and in your test enclosure.

How big is your test room? I would give consideration to pulling the whole setup out from the corner by a couple of meters/and or setting up some acoustic batting to dampen the rear wall reflections somewhat. And heavy batting or even just thick moving blankets folded up on the floor in front of the DUT cabinet at the obvious "reflection point" between the cabinet and microphone could be helpful. But don't let me be a "damper" on your efforts!

🤣


As a check on your "in room results", consider some quasi-nearfield checks at say, 1/2 meter, as they will show much less influence from room reflections, and as narrow as your test enclosure is, should still be fairly useful.

BTW, what is your proposed methodology for testing the PRs?


I think you're going to have fun doing this, (I hope!) and we'll certainly enjoy seeing your results.

[svenarajala]

I have to admire the laser level! That was a nice touch! Looking forward to see how your measurements come out!

[technodanvan]

Beautiful turntable! Do you have a build thread for that somewhere?

[Efalegalo]

I've seen Dave's post on other websites. I commend him on willingness to perform such time-consuming tests, and then further share results with the DIY community.

A very big Thanks You, Dave.

[DaveFred]

I'm sure a lot of folks will be interested to see how this works out and how the parts look- obviously, besides SPL measurements, HD2 and HD3 would be of interest, along with impedance tests free air and in your test enclosure.

How big is your test room? I would give consideration to pulling the whole setup out from the corner by a couple of meters/and or setting up some acoustic batting to dampen the rear wall reflections somewhat. And heavy batting or even just thick moving blankets folded up on the floor in front of the DUT cabinet at the obvious "reflection point" between the cabinet and microphone could be helpful. But don't let me be a "damper" on your efforts!

🤣


As a check on your "in room results", consider some quasi-nearfield checks at say, 1/2 meter, as they will show much less influence from room reflections, and as narrow as your test enclosure is, should still be fairly useful.

BTW, what is your proposed methodology for testing the PRs?


I think you're going to have fun doing this, (I hope!) and we'll certainly enjoy seeing your results.

My limiting factor is ceiling height, which is only 2 meters,



Which means my gate for windowed measurements will only be ~3.5ms



Does this not remove the floor/ceiling/backwall reflections? The floor is the closest thing to the mic/speaker, everything else is >1m away.

I will be doing distortion tests, free air and in box impedance and others. I will try to make one post about each driver with all the points.

Not every driver will be tested with a PR, just the ones that look promising. I have had success in merging nearfield PR and nearfield woofer response in REW. Generally the sims show the roll off tails crossing at 20Hz, so that is what I lined up when combining the responses.



I have lots to learn, so suggestions and corrections welcome, just be nice when I am wrong please .

Thank you, David.

[DaveFred]

Beautiful turntable! Do you have a build thread for that somewhere?

I don't know if this link will work for you, but give it a try,

https://www.facebook.com/groups/462464294109413/search?q=arta%20turntable

Thank you, David.

[technodanvan]

Really cool stuff, thanks for the link!

[JonMarsh]

You're really geared up nicely, and it looks like you're already getting some good experience with the tools. Not knowing anything about your acoustics or engineering background, I'm going to give a few hints that may help you out in your situation, and if I'm covering stuff you've already thought about and have plans for, then please accept my apologies. Some of this is going to be a bit of philosophy about purpose behind measurements, and some will be hints and ideas for coping with your circumstances (2 meter ceiling!!) and how to get the most useful driver information under those conditions...


🔊 Speaker Measurement "Philosophy"

I look at speaker measurement as two similar and related tasks, but different techniques to produce optimized results for the needs at hand...

• Driver characterization, to understand its performance capabilities and quirks in a setting similar to what is used for loudspeaker data sheets. This includes collecting...
  • SPL amplitude response with little or no impact from baffle step, and managing measurement technique and windowing technique to get the data needed with minimal impact from the measurement environment. A flat near IB is preferred to eliminate diffraction issues in SPL response.
  • Driver impedance, which gives hints for issues with driver performance (including resonances) and crossover challenges
  • Polar behavior, to estimate or calculate a specific forward power response for a specific angle or for total output power response
  • Distortion- which adds an extra degree of difficulty, because if you want to measure HD3, you need to have a system that is good to go for 3x the highest frequency of serious interest. That means targeting high sample rates and using wide band measuring devices.
• Application measurement- for a specific design concept in a specific cabinet with specific radiation goals.
  • Using a cabinet or test baffle that matches the application requirement or initial plan for physical construction
  • Measurements may be focused more on the SPL characteristics for the intended frequency range- but polar characterization is needed, if only to verify that some range (like up to 45 degrees off axis) is consistent and can be assumed equivalent to a specific axis angle measurement
  • It may be desired that the measurement environment is adjusted to replicate the anticipated playback environment, as regards boundaries and locations, and their impact on both LF behavior and overall behavior, as well as cabinet diffraction behavior at different angles.

✅ Hints for extracting useful measurements


Optimizing the measurement bandwidth and effective "signal to noise" ratio are key goals often at odds with each other.

You mention using a gating window of 3.5msec due to first order boundary reflections, but that window duration for getting a correlated impulse response from a sine chirp will have very limited low frequency data, and it really only suitable for tweeter evaluations, and even then, lacking LF data for full characterization tweeters I often use.

The necessary goal for in room measurements is literally to optimized the "signal to noise" ratio of the impulse response data. That means, quite basically, reducing the level of reflected wavefronts versus primary measurement wavefronts. This can be done both by physical means and by choice of measurement windows for the correlated impulse response.

I'm going to talk about measurement windowing first, as I use it in one of my two main tools, Fuzzmeausre. The other measurement tool I use is Pocket CLIO- it is very useful in some regards, but also has significant limitations. Fuzzmeasure does many kinds of acoustical measurements, one happening to be a mode suitable for loudspeaker characterization.

Note, these hints and methods are what I'd call my GTD techniques- "Getting Things Done"- you may like or reject any as you're so inclined, but they work for me within the context of my overall design conceptual process. I'm going to use a single tweeter measurement as an example, taken from the Isiris update project, all the following screens are just one measurement capture, but showing how differently things work out and look for creating a measurement that I'll export FRD files from, and further process in VituixCAD (scaling and minimum phase conversion, sometimes additional smoothing). This is mounted on the cabinet tested in my living room. Ceiling is around 12 ft. Cabinet is moved out further into the room for tests.





Note: because any single measurement in Fuzzmeasure can be used for a variety of analysis types, including distortion, the way I setup things is to optimize the dynamic range and signal to noise ratio for the DAC processor in the Motu or whatever outboard processor I use (used to be TC Konnekt). So, I don't worry about displaying in a calibrated level, because that is easy to check once to validate factory specs, and then to scale to for FRD files used for crossover design. Instead, I'm working with a dB scale referred to 0dBFS and below, and normally setup for the max signal level to be around -10dBFS in order to optimize both distortion and dynamic range. Note, I also use a digital storage scope to monitor actually speaker driver levels at all time, using a sweep speed that allows me to capture the whole sweep; I have a small Teledyne scope on my Pocket CLIO cart, and a Siglent scope on the Fuzzmeaure cart. On the Fuzzmeaure cart, a Schitt Freya+ is used as the preamp (in SS gain mode, not passive) and a Behringer A500 power amp, chosen for it's low noise and low distortion at low to moderate power levels.


In keeping with TLR, I'm showing first the measurement processing result used to create FRD export, which includes SPL and the windowed impulse response, and also the distortion plots for HD2 and HD3 for that measurement exported to a separate graphic file. I prefer distortion plotted in dB, as percentage, like Accuton uses, compresses and obscures information in the bottom end of the scale.











Sampling rate is 88.2 kH, bit depth is 24bits, and the measurement mic for the Fuzzmeasure cart is currently an Earthworks M30; I have two ACCO Pacific microphones misplaced somewhere in one of my two local storage units; haven't had time to ransack them. Would have bought another one, but they seem to be out of business. They have an online quote form which they did not respond to, and the online ordering which used to be there is no longer on the site. Emails were not responded to.

Note, the default impulse correlation length is 200msec, which can be then edited with window modifications.


Fuzzmeasure is more flexible than CLIO as regards windowing configuration, having rectangular, Hanning, Half Hanning, Hamming, Half Hamming, Bingham, and Half Bingham, and the ability to graphically edit the window and see the results with a pop up window.


Here's what you'd be looking at with default settings (rectangular windowing) and a nice big room with few early reflections:






Something to note here, as a hint- looking at the impulse response, you can see that the early ones are at a much higher level than what comes later. This is just partly due to the square law effect, where if you double the distance to a source, the intensity drops by 6dB. This can be used to advantage to work against early reflections at times, by reducing the primary distance to the DUT. So, for example, if you move the microphone to 1/2 meter, the direct sound goes up by 6 dB, but the reflections remain pretty constant. My own rule of thumb for quasi near field measurements varies depending on what you want to know- if you need to keep baffle step in mind, I suggest not going close than 2X the baffle width. If you're just interested more in characterizing the driver, then you might choose to measure at a distance equivalent to 2x the radiating width of the driver- for an 8" woofer, that is about 6", so you could approach 12" and have fairly representative data, with a better ratio of direct to reflected sound data.


OK, this is just a tweeter we're measuring here, so I think I'd be comfortable looking more at a 6ms window gate duration.







Well, that looks better, but it still doesn't exactly ring my chimes, to use a music metaphor...

Why are all those other window functions in this tool? Some can be pretty helpful, depending on the the type of acoustic measurements you're doing. I prefer the Half Hamming window, with the origin set just before the start of the impulse response...





This provides a shaped gain function relative to time, literally a Half Hamming window, and the components later in time are reduced in level but not eliminated.


What happens if we take the window duration down to 3msec? This can be useful to do as a sanity check, sometimes even with a Bingham or Rectangular function. Then you know you're just getting the direct sound.




But we can readily see that we're losing so much lower frequency data that I don't really have much confidence below 2kHz. And that won't do for developing a good data plot for crossover phase below the nominal crossover point.

These measurements were done at 1 meter, on axis, and if I thought my data from the 6msec window with Half-Hamming gate wasn't clean enough, then I'd go back and move to a 1/2 meter DUT to mic distance and try again. Remember, the data above the baffle step point can be used for level calibration in post processing.


Last point, in tight quarters with a low ceiling, do everything you can to keep things clean acoustically. I mentioned moving blankets on the floor previously because I've got so many of them, having had to move once a year since 2020. A more permanent solution I could recommend would be using ATS Denim insulation sheets to make some basic acoustic baffles, one for the floor in front of your DUT, and one for the ceiling. The Denim material works quite well, even for lining the interior of speaker cabinets, and it's probably a lot less problematic from an air borne contamination view point than the more common acoustic insulation materials like fiberglass and rock wool.

Hope these hints help, of course, others should chime in with their experiences and tips.

[Reet]

Rectangular windows, data no good below 2kHz, what is going on here?

But we can readily see that we're losing so much lower frequency data that I don't really have much confidence below 2kHz. And that won't do for developing a good data plot for crossover phase below the nominal crossover point.

I mean no disrespect but I strongly disagree with this statement. Window length determines the frequency resolution of the measurement, window type affects how the dots are connected between each measurement sample. kimmosto has shown that for minimal overshoot/undershoot of low resolution data, the best options are either Cosine window or Tukey 0.75. I'm not sure why anyone would purposefully use a rectangular window for a speaker measurement, as it provides maximum error in the low resolution portion of the data, but I'm sure there's a use case for it somewhere, IMO this isn't it.

For comparison, I traced your measurement data above and overlaid it in everyone's favourite software. Here you can see very good agreement in your own measurements down to 250Hz, only difference is that one includes a bunch of room interaction from the long windowing.



If you really wanted to make an argument on phase data for crossover development, you should have shown the measured phase for comparison. Of course, it's easily done with our own measured data. Simply load into VituixCAD IR tool and compare with reflection free window, vs long window that includes the room. Below the Green lines show correct windowing at 4.2ms length, and default setting of Tukey 0.5. You can see very good agreement of measured phase down to 200Hz, apart from all the junk added by the room reflections.


Going the other direction, decreasing the window down to 3ms, phase is still exactly the same as what we get at 4.2ms window, and I would say not much data is lost for evaluation of this tweeter. Problems only start to rear their heads at 2.5ms where the low resolution start to affect the response slope in the 500Hz range. With tweeters in mind, if 3ms is all you can muster, you can get by just fine for crossover development with that measurement data.


As far as using damping material to increase effective window length, I think this may be a problematic solution. Considering the goal in Dave's case would be to extend the low end frequency resolution for better merging of nearfield and far field measurements, the main problem with damping material is that it's effectiveness is frequency dependent. It will damp high frequency energy well, but not so much the low frequency energy that we're most interested in. It effectively puts a low pass filter on the reflection, making it harder to discern visually on the impulse response, however the extended window may still be contaminated at low frequencies.

For evaluation of this, I would recommend first to gather a good reflection free measurement of the speaker in a large open space, use a friends house with a bigger space or go out in the driveway and annoy the neighbors for a few minutes, but you need a baseline of the good expected response to evaluate the effectiveness of the damping material to absorb reflections in the frequency range you're interested in. high absorption in 200-300Hz range could be a challenge though, even anechoec chambers have low frequency limitations.

[kimmosto]

Tukey 0.50 is still initial window function in Convert IR to FR. It is a bit more accurate at high mid...low treble than Tukey 0.75, and it's available in other FFT tools such as ARTA and REW. Steeper windows are okay for tweeters which do not produce LF content -> significant IR data ends before window function ends. In addition, applicable range of the driver is above possible errors due to steep-ending window functions. This story changes with LF radiators producing LF data and wish/requirement to reach as low frequency as possible with as short window as possible and as accurate as possible. Window functions landing smoothly to zero seem to be better options. Overall response is more accurate because window function does not generate extra content to FRD by ending rapidly where IR level can be still far from zero. Narrow band deviations are clearly smoothed, but that is not so bad as wrong overall level which can lead to incorrect balance when merging NF to FF. Unfortunately this equation has quite many variables in a small measurement space, and with different radiator types so result is case dependent. "Best for everything" does not exists.

One of my test materials was not minimum phase. Some IR content was too early giving a bit optimistic results with shallow functions such as Hanning and Tukey 0.75. Measuring complete speaker with steep IIR XO having some excess group delay has same kind of consequences. Some LF data could drop out of time window. For example "bad old" Genelec 8260 having massive excess GD at LF due to steep IIR XO, time window of 10 ms is barely adequate for capturing frequency response right below lower XO frequency. Steeper windows may look tempting, but the result is probably still wrong because window ends sharply in the middle of woofer's effective data.

[tktran]

Let’s not let perfection be the enemy of good.

Look forward to your results.

[kimmosto]

Results of window function test with minimum phase IR

[AJINFLA]

If that is your intended baffle shape, I would recommend a tweeter that can cross very low (1k), but still have good dispersion >10k. The SB Acoustics SB26ADC-C000-4 comes to mind. Directivity matching a 8>1 on a flat baffle like that will be difficult, if that is a criteria for you.
cheers,

AJ

[DaveFred]

First Driver Report, SB Acoustics SB20PFCR30-8, approximately $54.00 USD each,

[DaveFred]

First Driver Report, SB Acoustics SB20PFCR30-8,


Free Air impedance,



In Box inpedance,



Distortion 2.83 volts, 315mm,







Raw response 2.83v, 1m,




Normalized,




Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.










This is obviously the first one of these I have posted, how does the presented data look (not the driver, the usefulness of the data)?

I have data for a number of drivers, just want to get opinions on how this first one is posted before I go on with the rest.

Thank you,

David.

Please note, I know this data is all specific to my 1.5 cubit foot sealed box that is 20" high x 13.5" wide and has 1.5" radius side edges.

[Efalegalo]

Superb work, Dave! Any chance "Normalized" polars can be included?

[DaveFred]

Superb work, Dave! Any chance "Normalized" polars can be included?

Take a look again at the "raw response", I have added a normalized version right after. Is that what you were suggesting?

[Efalegalo]

Take a look again at the "raw response", I have added a normalized version right after. Is that what you were suggesting?

Yes - weird - the image was not loading previously and so I missed it. Yes, this is exactly what I was requesting. Thank you.

[DaveFred]

Ciare HWB200-4, approximately $145.00 USD each,

[DaveFred]

Free Air impedance,



In Box inpedance,



Distortion 2.83 volts, 315mm,






Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.

[DaveFred]

Dayton RS225-8, approximately $90.00 USD each,

[DaveFred]

Free Air impedance,



In Box impedance,



Distortion 2.83 volts, 315mm,





Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.

[Evil Twin]

You are making steady progress. My comment at this point is that the impedance curves of these drivers tend to be good predictors of inductivity related modulation. If HD climbs between 100Hz and 1kHz, this is a classic symptom of this issue. The new Dayton Epique E180HE-44 have the same issue.

[DaveFred]

Dayton Audio RSS210HF-4, approximately $140.00 USD each,

[DaveFred]

Free Air impedance,



In Box impedance,



Distortion 2.83 volts, 315mm,





Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.

[DaveFred]

Eton 8-412/C8/32 HEX Symphony II, approximately $200.00 USD each,

[DaveFred]

Free Air impedance,



In Box impedance,



Distortion 2.83 volts, 315mm,





Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.

[DaveFred]

HiVi LK8A, approximately $130.00 each,

[DaveFred]

Free Air impedance,



In Box impedance,



Distortion 2.83 volts, 315mm,





Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.

[DaveFred]

Human Pro 005 woofer, approximately $83.00 each,

[DaveFred]

Free Air impedance,



In Box impedance,



Distortion 2.83 volts, 315mm,





Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.

No XO and results to post with this sample.

This woofer was meant to run as is with the Human Pro 002 tweeter.

[DaveFred]

SB Acoustics SB23NRSX45-8, approximately $110.00 each,

[DaveFred]

Free Air impedance,



In Box impedance,



Distortion 2.83 volts, 315mm,





Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.

[DaveFred]

Seas U22REX-P-SL, approximately $165.00 USD each,

[tktran]

That RS225 really is a champion woofer from 40hz to 250Hz.... with the right crossover a good midwoofer up to past 1KHz too...

[DaveFred]

Free Air impedance,



In Box impedance,



Distortion 2.83 volts, 315mm,





Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.

[DaveFred]

Wavecor SW223BD02, approximately $300.00 each,

[DaveFred]

Free Air impedance,



In Box impedance,



Distortion 2.83 volts, 315mm,





Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.

[Evil Twin]

That RS225 really is a champion woofer from 40hz to 250Hz.... with the right crossover a good midwoofer up to past 1KHz too...

But do note how the RSS210HF is an even better choice, with regards to the distortion floor and overall smooth extension. And with its 9mm Xmax, I would designate it as a long throw mid woofer, not a subwoofer, though it's low frequency extension with the right design moves it into that realm also. It straddles those categories very impressively... and at not unreasonable cost.






Time spent examining the woofer impedance curves before testing can greatly winnow the "playing field", as it might be called...

On the application side, a series resonant network to address the primary breakup mode at 4500Hz will be desirable, along with a parallel shunt network before hand to address the 7 kHz area, if class leading performance is desired.

[Reet]

But do note how the RSS210HF is an even better choice, with regards to the distortion floor and overall smooth extension.

RSS210 is going to need a good 3dB higher output to really make apples to apples distortion comparison for a given SPL. Perhaps test at 4-5V instead of 2.83V would make sense for these low sensitivity subwoofers being considered.

[DaveFred]

Monacor SPH-200KE woofer, approximately $140.00,

[DaveFred]

Free Air impedance,



In Box impedance,



Distortion 2.83 volts, 315mm,






Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.

[DaveFred]

Peerless HDS-P830869, approximately $65.00 each,

[DaveFred]

Free Air impedance,



In Box impedance,



Distortion 2.83 volts, 315mm,






Raw response 2.83v, 1m,



Normalized,



Quick and dirty 1800 Hz LR4 on the woofer, something realistic to go with my tweeter requirement.