Measuring Driver Frequency Response

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I have started measuring driver parameters with the DATS v2, very easy to use. I also picked up a UMIK-1 microphone with the intention of using REW to generate some frequency response curves for vintage drivers I have laying around. I would like to know what would be a good way to measure the FR, I see this arrangement from another member (in pic), is this good enough? I was planning on hanging the driver off a ladder and using a boom mic, any thoughts on that? Seems like it would be easier, not having to mount the driver. Perhaps a recommendation on a mic stand that would work well with the mic I have? If there is a standard setup people are using for this, I would appreciate a photo or some sketches, I have come across some theory on this, but surprisingly, no photos or videos on actual setups.


My goal here is to come up with replacement drivers for drivers that tend to fail in some of the vintage speakers I have. For example, the Energy Pro 22 tweeter is failure prone and there are some suggestions for replacement here and there, but none that are backed by measurements. The Thiel CS3.5 mid range, the 13m8521 Scanspeak, is an utter disaster of a driver, I have 5 failed units on hand and just discovered one on my main rig has gone south.
 

Attachments

  • baffle_front_s.jpg
    baffle_front_s.jpg
    9.2 KB · Views: 342
Last edited:
That is one way, but it is only representative of what the FR would be when the final installation has a front baffle of that size. Also the bass response would not be indicative of the final result once in a specific alignment. Me personally I use test enclosures of roughly the same size and internal volume as the final actual project.
 
As big a baffle as possible is best if the aim is to obtain a 'reference' measurement in order to compare drivers of different sizes. Ideally the baffle should be infinite in size, as should the room.
Another approach would be to use a very large sphere (or perhaps hemisphere with open back) - but this presents more of a construction challenge.

For obtaining a measurement to design a speaker crossover, you'll want to measure it mounted in the completed box to account for baffle edge diffraction effects:
Diffraction from baffle edges
Some people obtain a 'reference' measurement such as that in a drivers datasheet and then use their simulation software to add a simulated edge diffraction - I don't recommend this. If you have the means to accurately measure the driver mounted in the real box, that measurement is best.
If you wanted to investigate edge diffraction effects from changing the baffle dimensions or driver placement on the baffle, you could also measure it on an open baffle the same dimensions as the intended front baffle of a box. The edge diffraction effects will just be twice as severe compared to a closed box.

Finally, edge diffraction effects and room effects can be minimised by placing the microphone closer to the driver. On my 1.2x0.9m baffle shown above, when the mic is placed at around 10cm away from the driver the baffle edge diffraction effects and room effects are minimal. However, there is a limit as to how close you can place the mic to certain size speakers before the high frequency response becomes inaccurate because different parts of the cone can radiate out of phase. For accurate high frequency response I'd recommend spacing the microphone back at least twice the diameter of the cone. When designing a crossover I'd measure no closer than twice the height of the array of drivers.
 
Last edited:
Just another Moderator
Joined 2003
Paid Member
The other thing that you can do if you use a large baffle such as TMM's (or even better an IEC baffle if you have the space) is that you can actually subtract the simulated baffle response from the measurement and then add back in the simulated baffle response of your proposed cabinet. This allows you to do one meter measurements and not have to worry about the baffle diffraction issues (which on an IEC baffle are fairly minimal).

I also suggest that for crossover design you should use measurements in the final box, as TMM suggests, but for initial box design using a baffle difraction simulator can be very helpful in working out what will work well and maybe not so well with a particular driver.

Another thing I have been playing with lately (from manufacturer traced curves) is subtracting the box alignment from the FR curve and then adding in the intended box alignment to see how the response looks with a different alignment. If you also measure the T/S params, and have a known volume box (say 20L) you can model the driver in 20L export the response curve and subtract that from the actual measurement (do a division).. You then model the wanted alignment (say a reflex) and add that curve back to the one you subtracted the sealed alignment from (do a multiplication).

I've not done any real world tests of this yet, but it seems to give valid results with drivers (such as peerless) where the information about the measurement setup is available (something you have when you are doing the measurements yourself).

Both of these things give you a range of options for trying different designs before you even cut any wood (other than your standard test baffle and sealed box).

Tony.
 
So if I am looking to find a replacement tweeter for a 2 way, disconnect the woofer and measure the tweeter mounted in the box nearfield? Next step would be looking for a tweeter of similar response by comparing manufacturing FR curves with the measured one? Purchase what might look like a good match and then test it inbox?

I do have the DATS v2 which measures all the T-S parameters and VAS, is that data sufficient to find a reasonable replacement (and then test it in place) or must I also look at the FR curves?

The drivers I am looking to replace are vintage with no published measurements.
 
The Thiel CS3.5 mid range, the 13m8521 Scanspeak, is an utter disaster of a driver, I have 5 failed units on hand and just discovered one on my main rig has gone south.
I don't think the problem is with the Scanspeak 13m8521. The Thiel CS3.5 is 1st order network if I'm not mistaken. Playing too loud is usually the cause of damaged drivers.
 
I don't think the problem is with the Scanspeak 13m8521. The Thiel CS3.5 is 1st order network if I'm not mistaken. Playing too loud is usually the cause of damaged drivers.

This is a speaker that must be played loud, sounds too good not to. The idea is to model and take measurements of the speaker with the proper drivers and crossover, and look for a substitute for the midrange, modify the crossover if need be, purchase said substitute driver and do some measurements to confirm the selection and crossover mods. Scanspeak has some similar 4" drivers, but in different ohms. Crossover I have recently rebuilt with new caps and resistors, the electrolytic on the midrange was literally blown to pieces and others have experienced same. Tweeter is the Dynaudio D28AF (the D28/2 was also used without crossover modification) and the woofer seems robust and frequently come up on eBay. Again, weakpoint here is the midrange and the reason why so many of these get parted out, in addition to difficulty of shipping (70 lbs). I figure it would be a good project to get my feet wet in speaker building and do some good.
 
Last edited:
Are you planning to use a replacement tweeter with the existing crossover?

Well I just found out that the tweeters can be rebuilt for a reasonable $160CAN by ABI Tech, at least for now (rebuilder disappeared for a few years), so maybe finding a replacement (which would also require a crossover mod) is not as pressing, but the 2 way aspect of this speaker might make for an easier project than the CS3.5.

Someone was considering the Audax TW034X0, since it goes down low enough and is a drop in replacement, but would certainly need a crossover mod.
 
Just to echo what others have said, if you are trying to measure the "raw" response of drivers to compare one driver to another, and look at the drivers response in isolation then a large "infinite" baffle where the driver can be properly flush mounted is best.

Just beware of the effects of testing on an open baffle - unless it's very large (a couple of metres) so that it can be windowed out, the rear wave diffracting around the edge of the open baffle and interfering with the front wave will put serious ripples in the measured response below approx 1Khz that aren't the fault of the driver and won't exist on any completed speaker. (Unless it was also open baffle)

This could be avoided by putting a closed box behind the driver that attaches to the measurement baffle to contain the backwave, but then of course you change the drivers low frequency response so it's not really a "raw" response anymore unless you compensate for that in your measurements by applying a correction.

Better to just use as large a baffle as you can reasonably make and use in a room, then use time windowing to exclude the now sufficiently delayed rear wave from the measurement.

For measurements that you will be using in a crossover simulator / design software such as Vituixcad to design and optimise your crossovers you absolutely want to use measurements that are taken using the drivers in the completed cabinets.

This accounts not only for baffle diffraction and baffle step transition effects but also for effects like the closed box on the back of a midrange driver, how much stuffing is in the box etc...anything that will affect the drivers raw response is taken into account automatically by measuring in the final box. Critically, it also takes into account phase changes that result from any driver acoustic centre offset errors, which may need to be compensated for in the crossover design.

Some might try to take the raw "infinite" baffle measurements perhaps published by the manufacturer and then transform them using some baffle diffraction simulation software, however I'd advice against it as it is simply not realistic or accurate enough, especially relative phase of different drivers.

It is probably OK for a "first pass" estimate of whether a design or driver choice is feasible or not and to give you an idea of what kind of crossover topologies might work, but I would not base an actual crossover design on it certainly not to the component value level.

I would definitely measure the drivers in the final box from a single microphone position and taking care to measure phase correctly and then use those measurements to base the crossover design on - forget about the infinite baffle measurements completely, they don't matter for this purpose.

In fact I would say that there isn't a lot of need to do your own infinite baffle measurements unless you are doing a carefully controlled comparison of many drivers - good manufacturers measurements should be in the "ballpark" enough to choose a driver, then just take the in-box measurements when you're ready to design the crossover.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.