Techniques to accurately measure drivers with confidence?

[devils_trill]

Hello all,

It’s common that many “Pro” DIY builders measure their drivers before the formal design process.

If one’s own measurements match those of the driver manufacturer fairly closely, then the measurement efforts are a good confidence check.

Not surprisingly, the measurements made by the DIY builder are often different than those of the driver manufacturer.

However, how can we be confident that our measurements are actually better than that of the manufacturer? Before measuring the drivers for a project, wouldn’t we need to calibrate our equipment against some kind of ‘benchmark’, such as a driver whose parameters are known with great certainty? More precisely, the DIY’er would measure a driver whose parameters are already known with certainty. Next, these measurements would be compared to the known correct values to verify that the measurement procedure is being performed correctly and the measuring equipment is functioning properly.

Can anyone here offer some insight into how a typical DIY builder can overcome such issues and measure with confidence?

Thanks for all responses!

[romanbednarek]

Re: Techniques to accurately measure drivers with confidence?

I'll try to tackle this one based on my personal experience but I'm not going to talk about measuring T/S parameters, only frequency and impedance response measurements.

The problem with most manufacturer's data is that it is only as reliable as the manufacturer and in some cases there is a lot of variation from batch to batch and even from driver to driver in some cases. Also, manufacturer's data is usually based on measurements taken on a large, "infinite", baffle while project measurements should be taken with the drivers in the enclosure but this is less of an issue because baffle diffraction simulation tools are pretty good at predicting the differences.

As far as measurement accuracy goes, most of the time all that is required is that you get a calibration curve made for your microphone because "usually" all other parts of the chain are accurate and most of the times when errors occur in those parts the measurements are off by quite a bit which makes it pretty obvious to pick them out. Some other advantages of having a measurement system is that you can take accurate off axis measurements which are more difficult to simulate due to differences in drivers which act differently in terms of transition to pistonic behavior (where the produced sound projects only forward but not to the sides). So basically you can get a better idea of what is going on overall because off axis response does contribute to the overall sound in most rooms.

So to summarize, the advantages of taking your own measurements are to get a more accurate idea of how "your" driver really performs which tends to be more important if the manufacturer consistency is in question. It also allows you to get a more accurate model of each driver's performance in your project cabinet. With a calibrated microphone, measurement accuracy is usually very good as long as you know how to use the software and set up everything properly and most of the time if there are issues with the system they will lead to obvious errors.

I hope that helps you get a better picture of why a lot of DIY'ers like to take measurements. I've taken measurements of both raw drivers in an enclosure as well as those of the same driver with the crossover installed and I would have to say that most crossover simulators are accurate enough that the simulated crossover response derived from raw driver measurements almost always match the measurements of the speaker with the crossover installed. Some measurement tools, like SoundEasy, allow you to take measurements even further with non-linear distortion measurements. Also you can learn more about the influence your room has on the performance by taking ungated measurements of the speakers from your listening position. Basically having a measurement system really opens the door to so many more options when it comes to optimizing a crossover but you should always trust your ears when it comes to tweaking, using the measurements as a guide.

I'm sure others will have additional comments and I hope that they can contribute to this discussion.

[djarchow]

Re: Techniques to accurately measure drivers with confidence?

Roman's explanation was very good. I would like to add though that measurments are only as good as the person taking them. Measurements can be impacted by lots of things, the environment, your equipment, how you measure etc. So to get good measurements, you have to understand all the variables that can impact your measurements. This isn't meant to imply that taking good repeatable measurements is difficult, because it isn't; but understanding those measurements including those things that can impact them does take some experience.

If you want to learn more about measuring, I would strongly reccomend, Testing Loudspeakers by Joe D'Appolito.

Regards,

Dennis

[jonpike]

Re: Techniques to accurately measure drivers with confidence?

Roman and Dennis pretty much nailed it.

Many different folks calculate crossovers for their self teste drivers, and have the predicted results match the actual very well. Dr D'Appolito's book is probably the only serious single volume on the subject, and his professional cred goes without saying. It is a great reference on the subject.

It's worth searching in the forums and websites for comments from the "senior" designers... lots of nuggets are out there to be found, for practical tips. I like to save them when I run across them.

Its easy to cover the basics of good setup, quiet background, and ways to test your rig (loopback tests, measuring known caps and inductors, etc) for accuracy... and thereby be rather secure with your results.

[toddshore]

Re: Techniques to accurately measure drivers with confidence?

Concerning measuring T/S parameters, I currently have a good example of why you should do it yourself.

I am building a set of in-wall MTMs using two Tangband W5-704D woofers and the Vifa D27TG-35 tweeter.

Let's start with the tweeter. This driver model has been around for many years. Or has it? Mine is labelled Peerless and not Vifa. Not only is it a different batch from any driver whose specs have been published (at least that I can find), it is likely made in a different factory and a different country. PE has its Fs as 650Hz, both on the web page and in the PDF spec sheet. On the Typhany website, they have Fs at 880Hz. I measured 1004Hz and 960Hz on the two tweeters I have. Which do you think I am going to trust? I know I'm not going to trust a SPL trace.

Zaph has measured the Tangband W5-704D. I think. Maybe it was the 704, maybe the 704S. His T/S parameters are different than Tangbands and slightly different than the four drivers I measured. Was it the 704D? Mine all measure very close to each other in several different measurement sessions. His parameters give just enough of a different box tuning than mine that I think I'll go with mine since I am sure what was measured. Again, those differences don't make me comfortable tracing someone else's frequency response graphs.

[jkim]

Re: Techniques to accurately measure drivers with confidence?

Concerning measuring T/S parameters, I currently have a good example of why you should do it yourself.

I am building a set of in-wall MTMs using two Tangband W5-704D woofers and the Vifa D27TG-35 tweeter.

Let's start with the tweeter. This driver model has been around for many years. Or has it? Mine is labelled Peerless and not Vifa. Not only is it a different batch from any driver whose specs have been published (at least that I can find), it is likely made in a different factory and a different country. PE has its Fs as 650Hz, both on the web page and in the PDF spec sheet. On the Typhany website, they have Fs at 880Hz. I measured 1004Hz and 960Hz on the two tweeters I have. Which do you think I am going to trust? I know I'm not going to trust a SPL trace.

Zaph has measured the Tangband W5-704D. I think. Maybe it was the 704, maybe the 704S. His T/S parameters are different than Tangbands and slightly different than the four drivers I measured. Was it the 704D? Mine all measure very close to each other in several different measurement sessions. His parameters give just enough of a different box tuning than mine that I think I'll go with mine since I am sure what was measured. Again, those differences don't make me comfortable tracing someone else's frequency response graphs.

Hi Todd,

Inconsistency between manufacturer's specs and DIYers' measurements is usually much worse for T/S parameters than for the frequency response. I've modeled hundreds of crossovers using Zaph's measurements and also compared most of his measurements to manufacturers'. His FR measurements are very, very close to manufacturers' except for a bit of sensitivity mismatch in some cases. FYI, read Roman's comment on a similar issue:

http://techtalk.parts-express.com/showpost.php?p=1547302&postcount=32

In the case of the D27TG35 there has been a consistency issue as you already noted. If the Fs of your tweeters is high, then I suspect its low end response below 2 kHz may have a rise (i.e., a more pronounced high Q "knee"). You will need to change the notch filter values in the tweeter network of the LR2 crossover I designed for you. You may also need to lower the primary cap value in the tweeter section of the LR4 crossover.

In any case, however, you won't need to make major changes on those crossovers. I can even say that even if you use the crossovers as they aree, the speakers won't sound bad. Keep me posted on your progress.

BTW, what Zaph measured (currently posted at his website) is W5-704D. He posted the W5-704S measurement before but pulled it when it turned out no longer available at PE.

-jAy

[curt_c]

Re: Techniques to accurately measure drivers with confidence?

As I have some background in metrology, I can confer a couple of salient points when it comes to quantitative measurements as typically utilized in driver / speaker measurements.

Tolerances add up: If you measure a 1% resistor with a meter exhibiting a 1 percent accuracy, you can say the reading is accurate to +/-2%. More on this later…

The accuracy of any measurement has to be referenced to a known standard. Without comparison to a known standard, all measurements are relative, at best. That is, they can be compared to each other, but their absolute accuracy is unknown.

Accuracy vs. resolution: Your meter or measurement device (think computer soundcard / software may provide reading with 4 digits past the decimal point, but if your standard is only 1%, all those ‘extra’ digits are meaningless.

Known standards must also be periodically calibrated. Don’t count on that new DMM you just bought to be accurate unless you purchased it with a certificate of calibration. -It could be close, it could be way off: the bottom line is you just don’t know. Typically, equipment used for quantitative measurements are calibrated at least yearly.

So what’s all this mean for driver measurement? For impedance measurements you should run a calibration on the device making the quantitative measurement. For computer based measurements, this would normally entail 2 measurements. One would be to measure the impedance of a short, and set the software to ‘0’ at this point. The other would be to measure a known value resistor (the ‘standard’) to verify the scaling of the device is correct. This could be any reasonable value, but something around 10 to 100 ohms would be good choices as these would be values we’d typically expect to see when measuring drivers. If you use a 5% resistor, then the most accurate measurement the device can measure is 5%.

Finding an acceptable standard for response measurements is a bit tougher. A driver measured on another calibrated setup is one possibility. I use a Hiq tweeter, as Oskar sends these out with individual measurement plots. This can be a good ‘sanity check’ assuming your measuring conditions are similar. Fortunately for us, in most cases absolute SPL readings are unnecessary, only relative SPL between the drivers.

Other issues that may affect response measurements:
While the mic may come with a calibration file, there is no guarantee that it will ‘hold’ that calibration, especially if misused or dropped. This should be calibrated periodically.
While 2 channel measurements supposedly take and errors from the power amp and sound card out of the picture, the response and distortion products of the mic amplifier are not.

C