Generating driver FRD setup questions for passive xovers

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Starting out, and getting confused by the many different setups to generating FRD files for drivers (for passive crossover design). Probably a two way (MTM) at first.

I'd like to generate my own FRD files for individual drivers, probably with REW, then use something like Xsim. Because my house is small and not empty, seems like I will need to use near-field for low end and some other technique for above that- still reading about that.

It's not clear to me when a driver should be in a baffle or a box or in neither for generating the FRD files. In audiojudgement's useful articles the driver is always shown in a box, regardless of measurement method, for instance. While Dickason says use a 2x2ft baffle (if I remember correctly). I see others using a baffle approx to the size they think the finished box will have.

So..? do I understand correctly that most people follow this procedure:

1) Obtain drivers based on what you want to accomplish, budget, etc.
2) Obtain T/S parameter measurements to establish size and type of enclosure.
3) Build box, install drivers.
4) Generate FRD files with drivers inside box, one driver at a time (I will use a UMIK-1).
5) Design passive crossover from the above FRD file.
6) Use REW (or similar) to test room response to tweak xovers and placement, etc.

Since I'm new to this, I'd really like someone to just confirm that is the general procedure, and that it's best to make the FRD measurements with the driver mounted inside the final box.

Thanks if you can help clarify,
Keith
 
I suggest you....

1. Build cabinet (complete, not just front baffle) and mount drivers

2. Perform measurements:
2A. Choose a far-field measurement position: on or off axis within the horizontal plane of the T in the far field (e.g. 1m away from T)
2B. First measure each driver's response at this position, one at a time
2C. Then measure the response of all drivers, all operating together with no filters used, at the same position. DO NOT MOVE THE MIC to a different position for measurements 2B and 2C
2D. Go back and measure a woofer nearfield (e.g. 1cm away from dustcap). This is just one measurement, done at the end.

3. Process your measurements: use the program FRDBlender (I am a co-author) to merge the nearfield and far-field responses for the woofer, and to generate minimum phase responses for woofer and tweeter

4. Crossover Modeling: use the program PCD (by Jeff Bagby) to:
4A. First you use the individual driver measurements PLUS the multiple driver measurement at a given position to determine the offset of the acoustic centers for each driver WRT each other. This gives you a model for the loudspeaker as seen by the mic at this position. Do it for the on-axis position first. You should repeat this for each off axis position as well, since the relationship between acoustic centers will change slightly.
4B. Model the response at the on-axis position. Use PCD to design your passive crossover.
4C. Use the crossover you have designed in 5B but now use the FRD data and acoustic offset info for your off axis position(s). This is a check to see how the system response looks off axis (it will be different). Continue to tweak crossover until both on and off axis responses look good.

The procedure above can also be used when designing a DSP crossover. Just use my ACD (A=active) program instead of PCD (P=passive) to do the crossover design. The general procedure remains the same. In ACD you can view multiple on and off axis responses simultaneously.
 
Last edited:
Thank you CharlieLaub. Under 2C, how does one "measure the response of all drivers, all operating together with no filters used,"?
You feed the same signal to all drivers. You can do this by connecting them in parallel, or connecting each to a separate amplifier channel and providing the same input to the amp inputs. You can do all drivers together, or pairs of drivers. Since there is symmetry about the T of an MTM if you keep the mic at the same elevation as the tweeter you can measure just the T and one of the M's. The other M will be mirrored about the T in terms of its position, and the distance from each M to the mic will be the same. For a beginner, just measure driver pairs (T and upper M, T and lower M).

Do you mean connect woofer and tweeter together with no xover?
Yes. You are using the measurements to probe the response of the drivers in the cabinet without the effect of the crossover. This is the information that will be captured in the FRD file of each driver. The crossover design program adds the effect of the crossover to each driver. The result is a model of the system response (drivers+crossover) at the microphone position and without any room effects.

Also, how would your suggested procedure differ for an MTM configuration?
It doesnt!
 
Thanks again, this is really helpful. I have a MTM box and drivers that I want to test, so that's a situation I'd like to be more clear about.

This has me confused: "For a beginner, just measure driver pairs (T and upper M, T and lower M)", because I'm not seeing how to integrate that with steps 2C and 2D. Can you clarify?

Do I hook all three drivers together in parallel (that could be quite a load!) in step 2C?

For 2D I still need only measure one of the two woofers?

Or are you saying for an MTM there are actually two different steps, one for "T and upper M" and a second for "T and lower M" for 2C?

Sorry, it's probably easier done then said...
 
I will try to claify. I think the confusion is with 2C: measure the response of all drivers, all operating together with no filters used...

Here's more info about the measurements. Now that measurements software and calibrated mics are available to the DIYer it is possible to make accurate measurements in your home. The problem is that the space in your home has a finite size, that is there are room boundaries (floor, ceiling, and walls) acoustically nearby. When you make a measurements (typically an impulse measurement) the driver will emit the impulse. It travels through space in all directions. The shortest path is directly between the driver and the mic. Slightly longer paths are from driver-->floor--> mic (the floor reflection) or reflections off of the other surfaces. These will come later, a few milliseconds later in fact, and add to the signal of the "direct" sound. They essentially "contaminate" the measurement and if/when you include the reflected sounds the measurement will start to have large peaks and dips from them. This will in general be different depending on where in the room everything is located: speaker, mic, other stuff in the room, etc. What you want to know is the frequency response of the speaker itself. What can be done is that the time record of the microphone's signal can be inspected. It's typically relatively easy to see where the reflected sounds kick in. You just throw out all the data starting from that point on. What you are left with is a few milliseconds of the initial time record. Then this data is converted from a time record into a frequency response. The problem is that in order to get accurate low frequency information you need a long time record, but because you had to truncate your data at a few milliseconds you don't have that for frequencies less than, e.g. 200Hz.

But you also made a nearfield measurement. With a nearfield measurement, the "loudness" of the sound from the driver is much greater than that from the room reflections of that sound, such that the room reflections become so much less in magnitude that they can be ignored - you can use the full time record. This means that you can also get very accurate low frequency information. But the nearfield measurement cannot capture effects like the baffle response and other frequency response changes that are seen farther away from the speaker itself.

The far field frequency response valid down to 200Hz captures these plus most of the baffle step. What you can do is then take the nearfield response, adjust it to account for the baffle step using a model (it's semi-accurate) and then combine the far-field and near field responses together. This used to be done by splicing them at a single point, however, I can up with a way to "blend" them together over a range of frequencies (thus the name of the program FRDBlender). The FRD Blender allows the DIYer to properly combine responses, account for the baffle step, and then will generate the "minimum phase" response for the combined curve. You can think of the minimum phase response as the frequency response of the driver at the microphone position, but without the additional phase delay caused by the time it takes the sound to travel thru the air. The process, as I have explained it up to this point, is repeated for each driver to yield that driver's wideband (e.g. 20Hz-20kHz) minimum phase response. For the tweeter you do not need to make a nearfield response measurement, only for woofers and other drivers that have <200Hz sound within their passband.

Strictly speaking, you should do separate measurements on each driver. In some systems like an MTM when there are symmetries about the tweeter (including the cabinet) each of the M's will have the same minimum phase frequency response. If your MTM cabinet is e.g. a floorstander so that one M is near an edge and the other is kind of more central in the cabinet, then you should measure each M separately.

Now that you have processed your measurements into one minphase response per driver, the next step is to create the model of the loudspeaker in the crossover design program. The missing information is the relative location of the acoustic centers of each driver in space. Remember the minimum phase measurement doesn't have the phase delay resulting from the time it takes the driver's output to travel between the acoustic center (where the sound "starts") to the microphone. But there is a clever way to determine it, or make it possible to figure it out: the multiple driver measurement. When two or more drivers are operating together the sound at the microphone is a mixture of all of them. Because of the different pathlengths from each driver to mic, the phase lag will be different and each driver will have its own frequency response due to cabinet diffraction and the off-axis angle. You import e.g. the minphase FRD data for two drivers AND the data taken when both were operating together into the crossover design program. Then you vary the parameters (e.g. X,Y,Z location) of one of the drivers until the crossover design program's frequency response prediction matches the MEASUREMENT of the two drivers operating together. At that point you have correctly determined the relative XYZ location between driver 1 and driver 2. Then you move on to the third driver, replacing one of the FRD datasets (e.g. driver 2) with that of driver 3 and importing the two-driver measurement (e.g. of driver1+driver3). AGain you vary the XYZ coordinates for driver 3 until the prediction by the crossover program matches the driver1+driver3 measurement. At that point you have enough information to know the relative positions of drivers 1,2, and 3, which means you have a good model of the output of the loudspeaker at that microphone position.

Once you have an accurate model of the wideband output from the drivers and their relative positions in space you are ready to move on to designing the crossover.
 
Here' some information from Jeff Bagby that should help. If you don't have Excel for PCD you can use WinPCD by David Ralph. He basically duplicated the PCD functions with a few enhancements. Since it's a Widows app you don't need Excel.
 

Attachments

  • Bagby White Paper - Accurate In-Room Frequency Response to 10Hz.pdf
    754.9 KB · Views: 96
If the speaker is not large and the port and driver cone are right next to each other, I think you can get away with positioning the mic in between the two for the nearfield measurement. It is not as good as being right next to the driver's dustcap (the valid upper limit of the nearfield data will be reduced somewhat), but it should work out fine on a small bookshelf monitor type speaker.

If the port is on the rear of the enclosure, you have to measure the port and driver outputs separately and must scale the port output by the ratio of the areas of port and cone and then account for the relatively longer distance from port to listener. There is an online tech note about nearfield measurements of ported system by Klippel here:
Nearfield Measurement with multiple Drivers and Port
 
For information:

Quick manual for measurements and data preparation
Includes recommended measurement gear, method and data processing.
Dual channel measurement gear and mode is recommended to support simulation of off-axis responses, power response and DI.

Home of VituixCAD
Feature lists of VituixCAD and tools.
Links to youtube video lessons.
Download of test projects.
Download of setup, versions 1.1 and 2.0.
 
@CharlieLaub- in reading Bagby's white paper, which describes using FRDBlender, he always says he's using a baffle, not a completed box. I want to confirm with you that is OK, since I have some drivers which I don't yet know what type or size box to put them in. My thought is that as long as I use a baffle that's very close to the final size (to get the step response) that I don't need to have the drivers in a completed box. Correct?
 
DO NOT USE ONLY A BAFFLE. caps intended for emphasis! Really.

Using only the front baffle will result in a dipole unless the baffle is very large (e.g. 1x1m or larger). Even then, the response will be quite different than when the driver is put in an enclosure eg. "box".

The whole point of measuring the driver in its final enclosure is to "measure" the exact diffraction and baffle step response that will occur for that driver, in that enclosure, at that location in the baffle. When you change those things, the response will also change.

If you are "not sure" about your enclosure, then I suggest doing some modeling of the low end from TS parameters (which you should also measure), modeling of the baffle step as a function of the baffle shape, and then look at the manufacturer's SPL plots to get an idea about the driver's response in the breakup region (at higher frequencies). I use all of these approaches to figure out approximately where I will cross over a driver, and what I might expect when I put it in the box. At some point you just have to jump in and try it. You can make up a test box out of particleboard if you want to experiment with different sizes/volumes and shapes.

Also, regarding Jeff mentioning the baffle: it's essentially the dimensions of the front baffle, and the location of the driver in that baffle, that determine the "baffle step" effect on the driver's response. The "depth" of the box does not really contribute. That is why Jeff talks about the baffle (the front baffle) because that is what is influencing the response. It's not because he is only using the front baffle and the rest (sides and/or back) is open.
 
Last edited:
If the speaker is not large and the port and driver cone are right next to each other, I think you can get away with positioning the mic in between the two for the nearfield measurement. It is not as good as being right next to the driver's dustcap (the valid upper limit of the nearfield data will be reduced somewhat), but it should work out fine on a small bookshelf monitor type speaker.

If the port is on the rear of the enclosure, you have to measure the port and driver outputs separately and must scale the port output by the ratio of the areas of port and cone and then account for the relatively longer distance from port to listener. There is an online tech note about nearfield measurements of ported system by Klippel here:
Nearfield Measurement with multiple Drivers and Port

Thank you very much!
 
@Charlie or others,

After reading through the thread, am I right and assuming there are two nearfield and two far field measurement taking place?

  • A. I would also assume the first set if near field and far field you are describing are between the speaker driver and it's resonance within it's speaker cabinet?

  • B. Ed Simon had described the method to determine near field and far field by using and SPL meter and walking forward from the back of the room and when the SPL meter reads +3dB from the original reading, then at that point moving froward to the speakers I would be moving in the near field.
B method would be for speakers in a room/venue etc.​

So, A, the bagby method is for the drivers and the cabinet. And B, the speakers and the room.

If so or if not, then please clarify and add to the discussion.

Cheers,
 
Last edited:
I couldn't quite follow what you are saying RE "two measurements" etc.

The goal is to build a "wideband" and quasi anechoic measurement of the driver-in-box. Wideband means for all frequencies, DC to infinity. Quasi-anechoic means without the influence of room reflections and room resonances.

You cannot make a wideband measurement directly. Instead you use a combination of measurements, modeling, and merging of data sets. The measurements will typically include farfield and nearfield types.

A nearfield measurement is done very close (e.g. 1cm away from dustcap) of a driver. At this distance the room reflections and resonances are much lower in intensity than the sound coming directly from the driver diaphragm. The advantage is that you can measure to a very low frequency (but still not DC!) e.g. maybe 1-10Hz. The disadvantage is that a nearfield measurement also does not capture any interactions of the sound wave with the cabinet, so the baffle step does not show up in the nearfield data. But because you can directly measure the low frequency response of a driver independent of the environment, it's very useful since few other techniques can do that (that you can do at home).

A far-field measurement, on the other hand, does capture the "baffle step" that is created from the diffraction from cabinet edges, etc. Remember that measurements are a time record of what the microphone picks up. At some point in time, the microphone will also start to gather sound that has reflected off of a room boundary first before reaching the mic. The closest boundary will show up first in the time record, then the next, then multiple reflections, etc. When you try to use all the data to generate the frequency response, you get a lot of interference and the frequency response coming only from the speaker is difficult to determine. The goal is to take a measurement that does NOT include these interferences so that you see only the sound produced by the microphone.

In the far field measurement, there are a few milliseconds of time after the sound first reaches the microphone that do not contain reflections - it's only the sound coming from the driver and the cabinet/diffraction. It's possible to use only this portion of the microphone's measurement, that is to say the first 5-10 milliseconds. The DIYer sets a "gate" after which all the microphone data is set to zero. Then this "gated" data is converted from the time domain into the frequency domain. The result is a frequency measurement where no room interaction is present, just like if the speaker could magically be suspended in free space far away from any boundary. But there is a penalty for setting the data points to zero: you cannot determine the response at "low" frequencies. How low depends on how much data you are able to keep after the microphone first picks up the sound and when the first reflection appears in the time record from the microphone. It turns out that the lowest frequency that you can practically resolve is about 200-300Hz indoors. Since the baffle step is not always completely finished by this frequency and may extend down to 100Hz or so, we don't have all that we need.

The current practice of many DIYers is to try and merge the nearfield and farfield measurements. But this can only be done if the nearfield measurement is "corrected" so that is includes the baffle step. Once that is done the DIYer can overlay the two frequency data sets and combine them. One convenient way to correct the nearfield measurement data for the baffle step is to add to it a modeled baffle step response that takes into account the dimensions of the front baffle (with and height) and the driver diameter and its location on the baffle. Once the nearfield has been corrected, you are able to merge a data set that has good low frequency info (the corrected nearfield) and good high frequency info (the gated farfield) to get a merged frequency response. To extend that to the "wideband" response you assume that the behavior at the low and high frequency extremes, above and below the merged frequency response respectively, has a constant slope. For instance a driver in a closed box has an ultimate rolloff of 12dB/oct. Determining the high frequency rolloff takes a little guesstimation but it's likely between 18 and 30dB/oct. These extra "constant slope" parts of the frequency response are known as "tails" because they tail off to infinitely high and infinitely low (towards DC) frequencies. Adding the tails gives you the wideband response.

NOTE: for a driver like a tweeter, the far-field data already captures the response down to 200Hz or so. This includes below-resonance where the response is asymptotically reaching it's ultimate rolloff rate of 12dB/oct. Most people do not bother to make a near-field measurement on a tweeter for this reason. You just don not really need that data. Likewise, a port is really only operating at low frequencies, so you do not typically make a far-field measurement of a port, and it would be mixed up with the woofer sound anyway. A nearfield measurement on a port is all you need. For other drivers like woofers and midranges, or any driver that has a response passband extending below about 500Hz, both a nearfield and farfield measurement are a good idea.
 
Last edited:
Hi Charlie,

I just read your response after I added to the thread the following questions. Thank's for clarifying.

Not to hijack Keiths thread, but when I read the link of the Kimmosto's "quick manual for measurement..." file I noticed the speaker tower shares similarities to the speakers I inherited and in process of having them live again. I have new drivers for them that will need to measure. Where would place the mic to measure the two side firing woofers for this cabinet? For the Driver,

and then for the listening position: In the near field. In the far field.

Where to place the mic for rear bass port?

I guess flush with panel in middle of port?

Cheers,
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.