Driver balancing.

[Guest David H]

I have recieved several requests for information about driver balancing, and I want to build a set guidelines for this purpose.. So far everyone I ask has a different way of testing. I am not sure which way is currently the best so I am reaching out to the community for sugested proceedures.

This is a rough draft of how to balance drivers using nothing but an SPL meter and noise source. I will also build a set using an FFT/RTA for the same purposes.

Comments and corrections welcome. When I have a comprehensive set of directions I will post them on my site for download.

Hopefully I can also make pink and white noise available for download as well, if not I will link to someone who does.

[biglaz]

Can you explain the use of the resistors?

[Guest David H]

Using the reisitors in place of the driver is to simulate the drivers load to the crossover while eliminating acustial output. "dummy load"

Another viable option is to completely cover the driver, in a horn you can simply place a rag in the throat, a woofer is a bit more difficult.

Dave

[biglaz]

OK. Thanks

[babadono]

Dave,

This looks good and is probably more than most would take the trouble to go through. An RTA with a calibrated microphone would be the way to go instead of a SPL meter although I trust some of those are quite good (and more common). Being meticulous (scientific) about placement of the microphone (i.e. a tripod or microphone stand) when changing from one speaker to another would be in order. Maybe this is off topic but it seems once you have the drivers balanced and an RTA handy the next step is to EQ the entire system from the listening position.

Eric

[jwc]

Dave.

I use a TrueRTA and an old Audigy sound card. Behringer mic. Calibration is easy for that setup as the calibration files can be downloaded.

In all honesty......if I'm trying to balance drivers connected to a crossover........I put the mic about ten feet away and measure outside. However, lately....for big speakers....I don't do that as I'm tired of straining my back. I've done raw curves this way too and just rune same tests back to beck with different drivers. However.....the XO messes with stuff sometime and like to measure eachoutput of the horn/driver behing a crossover.

The other way......with a crossover......is to leave a driver disconnected and measure with the mic right up to the horn/woofer. save the curve and immediately run it the other way.

If you have a stand alone signal generator.....I have one.....and want to use an SPL meter.....that will work one frequency at a time.....but I dont balance drivers that way. That could help tune a cabinet but I general use the setup described for that too.

I confess to measuring inside for purposes such as balancing......or similating my own EQ. I know that isn't perfect...for for preliminary starts....I do.

[jwc]

BTW....I got started with my setup by reading a blurb from the AVSforum about TrueRTA years ago.

[Guest David H]

JWC, thanks for your input. I also use True RTA and a Behringer mic. I dont like my Phantom power or sound card so I think I am going to pick up the Tascam unit.

Dave

[mikebse2a3]

Dave

IMO if someone is using pink noise for a test signal then I would suggest using bandwidth limited pink noise that is close to the crossover region of interest to minimize interference from frequencies outside the crossover region. This should allow one to get closer to the best balance before any fine tuning begins.

Your suggestion to listen and tweek the balance of the drivers is valid due to driver/horns polar response variations shifting the perceived tonal balance of the system in real rooms and the somewhat unpredictable nature of how a listener will perceive these variations. I've found fine tuning of less than +/- 2db is sufficient if the above method is done correctly for most systems I've balanced this way.

mike tn

[WMcD]

I agree with Mike. What is measured by a microphone close to the horn mouth or direct driver is probably going to be a different than what is experienced out in the room. So some tweeking is probably going to be necessary.

There is also some technical discussion around indicating that flat response as measured by a microphone at the seating position does not equate to what is perceived as balanced sound. The Belgian review of the Cornwall discussed this and it is probably the reason for the X-curve in theater set ups. Without this caveat, some fellow may be puzzled to find his perfectly balanced set up is harsh sounding.

The other issue is running a crossover filter without a load. One of our moderators years ago pointed out that a second order crossover (and I'll add,probably higher order crossover) has an unanticipated effect. The L and C become a series resonant circuit which presents a short circuit to ground at resonance frequency. This, in my memory this dip in impedance occurs at close to what is otherwise the crossover frequency.

Of course a burned out driver is the same as no driver. I've wondered whether this does not explain some of the train wrecks you read about. The driver is overdriven and fails. Then the amp is pushing full power into the L and C (at some freq). Then the amp may fail or go into protection mode. Or the L or C fails.

In the post mordem, the amp is blamed, or the crossover is blamed.

So a dummy load is a good idea if you disconnect the driver.

WMcD

[Guest David H]

The other issue is running a crossover filter without a load.

I very much agree. I posted the dummy load reccomendation in the original document.

Thanks for all the additional input.

Dave

[Guest David H]

IMO if someone is using pink noise for a test signal then I would suggest using bandwidth limited pink noise that is close to the crossover region of interest to minimize interference from frequencies outside the crossover region. This should allow one to get closer to the best balance before any fine tuning begins.

Would you use bandwidth limited pink noise if you were using a passive crossover, or was this consideration for use with active, or both?

Thanks, Dave

[Guest David H]

Double post.

[mark1101]

With an active crossover all you have to do is mute every channel except for the one you are measuring. Of course the crossover filters should be setup.

[mikebse2a3]

Would you use bandwidth limited pink noise if you were using a passive crossover, or was this consideration for use with active, or both?

For both types Dave.

Any given driver/horn will have variations in it's operating bandwidth often up to +/- 3db even for good drivers/horns so sending signals that have very little relationship to the crossover region itself just opens the door to misleading results. Your looking for a good summing of the drivers/horns ideally in this region. IMO if there are frequency balance issues of the loudspeaker system outside of the crossover regions then that should be addressed through further refinement of the passive crossover design or if Active DSP then through fine tuning of the program.

mike tn

[Panacea Engineering]

Dave,

I have been reading this thread with great interest. As I am getting ready to start some extensive testing on the Jub-Like Drones, it raises the following question for me:

Should I uise the two way passive with a dummy load for the HF or should I use one of my DBX DriveRack 260's where I can manipulate the upper cutoff range?

This would enter into the mix, a whole new set of parameters, to establish the upper end of the cabinet. Is this something we would want to do? I certainly do not want to sabatoge your thr3ead so I will as the same in the build thread. I look forward to responses in the build thread, if everyone doesn't mind.

W. C.

[DrWho]

IMO if someone is using pink noise for a test signal then I would suggest using bandwidth limited pink noise that is close to the crossover region of interest to minimize interference from frequencies outside the crossover region. This should allow one to get closer to the best balance before any fine tuning begins.

Would you use bandwidth limited pink noise if you were using a passive crossover, or was this consideration for use with active, or both?

Thanks, Dave

This approach does not work in practice because the SPL meter is an average power measurement device. As such, the bandwidth of the signal being measured affects the overall energy captured by the SPL meter.

For example, if you measured a tweeter that could play from 1kHz to 21kHz, then that's a total bandwidth of 20kHz. If instead you measured a tweeter that could play from 1kHz to 11kHz, then that's a total bandwidth of 10kHz. That will result in a 6dB difference. In practice, it's a bit more complicated because pink noise has a higher crest factor, which will come into account with the ballstics of the averaging built into the SPL meter.

With constant sine wave test tones, you have the problem of the +/-3dB from the driver itself, in addition to all of the room interaction. Just play a signal on your system and move the microphone around a few inches and you'll see what I'm talking about.

The only way to properly dial in a system is to use a time domain measurement - there are three ways that I know of that work exceedingly well, but I stick to the log-sweep because it's artifacts are well understood and you can do it for free with REW.

I've spent years playing around with the SPL meters and all sorts of test tones and approaches, but it will not yield results better than what you can do by ear.

[Arkytype]

Dave's quest to come up with a low-cost procedure for balancing loudspeaker drivers is laudable.

With the availability of low or no-cost acoustic software such as Room EQ Wizard, ARTA, True Audio (Level Two) and microphones such as the Dayton EMM-6, one can easily perform measurements that only a few years ago required spending tens of thousands of dollars in test gear for essentially the same results. In 1967, JPL scientist Dick Heyser revolutionized the acoustic measurement field with his Time Delay Spectrometry technique. The first implementation of TDS required tens of thousands of dollars in test equipment and constant tweaking to assure repeatable results. Now you can download a TDS-based system with a mouse click!

Testing your loudspeakers in the listening environment has potential pitfalls such as axial and tangential reflections from walls, floor and ceiling.

DrWho wrote: The only way to properly dial in a system is to use a time domain measurement - there are three ways that I know of that work exceedingly well, but I stick to the log-sweep because it's artifacts are well understood and you can do it for free with REW.

Agreed. Measurement software with "gating" will allow you to approach anechoic measurement conditions by ignoring all but the direct sound from the loudspeaker to the microphone.

DrWho also wrote: ....the SPL meter is an average power measurement device. As such, the bandwidth of the signal being measured affects the overall energy captured by the SPL meter.

Hmmm....the last time I checked my Ivie, Bruel & Kjaer and General Radio sound level meters, they all indicated they measured sound pressure level, not average power. Sound pressure level is a measurement of the effective sound pressure relative to a reference value. By knowing the sound pressure level, the distance from the source and the Q, one can compute the acoustic power. You cannot read it directly!

DrWho also wrote: For example, if you measured a tweeter that could play from 1kHz to 21kHz, then that's a total bandwidth of 20kHz. If instead you measured a tweeter that could play from 1kHz to 11kHz, then that's a total bandwidth of 10kHz. That will result in a 6dB difference.

Electronic circuit noise voltage increases or decreases by 3.01 dBv for each doubling or halving of the bandwidth. Should that "law" apply to acoustic measurements as well?

I set up a test using some Krohn-Hite, Bruel & Kjaer, General Radio & Audio Precision gear to test DrWho's contention that halving the acoustic bandwidth would result in a 6 dB lower SPL level.

A Krohn-Hite 3384 was programmed with two cascaded 48 dB/octave filters set as HP and two set as LP resulting in a passband from 1 kHz to 21 kHz with 96 dB/octave slopes at each end. Pink noise from a General Radio 1382 was fed to the input of the 3384 and the output was fed to a B & K 2610 and AP SYS-2722. This setup eliminates the measuring microphone, loudspeaker and room--so the resulting measurements should simulate a perfectly flat microphone and loudspeaker in a free field environment.

After taking an initial reference reading on the analog meter of the 2610 (set to slow response) and the AP (with a bargraph display of incoming voltage), the 3384 was set to a 1 kHz to 10kHz bandwidth. Rather than seeing a 6 dB drop in level, there was only about a 1.2 dB change. Next, the bandwidth was changed to 1 kHz (1 kHz to 2 kHz) and the result was about a -6.5 dB on both instruments. I tried this experiment at lower frequencies with essentially the same result--halving the bandwidth showed only about a 1.2 dB drop in level. Next week, I hope to do some actual acoustic measurements to see if there is something I am missing.

Thoughts anyone?

Lee

[John Warren]

With the availability of low or no-cost acoustic software such as Room EQ Wizard, ARTA, True Audio (Level Two) and microphones such as the Dayton EMM-6, one can easily perform measurements that only a few years ago required spending tens of thousands of dollars in test gear for essentially the same results.

Thoughts anyone?

Lee

So I'm clueless and I've got tons of measurement capability. Now what?

What do I measure? How to interpret? What do I change? What do I do next? Did I even do it right? Is there even a "right" for what I'm doing? What's my goal? Where do I want to be?

[DrWho]

DrWho also wrote: ....the SPL meter is an average power measurement device. As such, the bandwidth of the signal being measured affects the overall energy captured by the SPL meter.

Hmmm....the last time I checked my Ivie, Bruel & Kjaer and General Radio sound level meters, they all indicated they measured sound pressure level, not average power. Sound pressure level is a measurement of the effective sound pressure relative to a reference value. By knowing the sound pressure level, the distance from the source and the Q, one can compute the acoustic power. You cannot read it directly!

Just to clarify, I'm not referring to the toal power from the speaker....I'm talking about the power arriving at the microphone capsule of the measurement device.

DrWho also wrote: For example, if you measured a tweeter that could play from 1kHz to 21kHz, then that's a total bandwidth of 20kHz. If instead you measured a tweeter that could play from 1kHz to 11kHz, then that's a total bandwidth of 10kHz. That will result in a 6dB difference.

Electronic circuit noise voltage increases or decreases by 3.01 dBv for each doubling or halving of the bandwidth. Should that "law" apply to acoustic measurements as well?

Depends if the sources are correlated or uncorrelated. That said, I totally mispoke on this point (I'll blaim it on being interrupted at work) [A]

It's actually the square of the bandwidth. This is a power scenario so we're talking 10log(sqrt(BW1)/sqrt(BW2)), which should be 1.5dB difference for halving, and 6.5dB for the 1/20 bandwidth.

So in a real world example, pick a 15" two-way speaker crossed at 1kHz. The woofer will measure 6.5dB quieter than the tweeter. In a 3-way system crossed at 400Hz and 7kHz, the LF will measure 7.6dB quieter than the tweeter, and the MF will measure 1.5dB quieter. You can try to go into the measurement knowing this, but the non-flatness of the speaker makes it way more complicated - and there's no reason to manually compute integrals when there are other better methods that don't cost anything extra to do.