I have been getting FRD and ZMA files from Holm and LIMP, respectively, and am wondering about accuracy. For testing purposes, is it better to test on a large open baffle or, with them mounted in the enclosure that they will be used in? If it's the latter, will an open area, or situated where they will be used, be the better option?
I'm downsizing from a active system (multiple amps and Xover) to a simpler single amp and passive Xovers so I need to design an acceptable passive Xover. Mostly working with PSD-lite.
Thanks for any help. 😉
I'm downsizing from a active system (multiple amps and Xover) to a simpler single amp and passive Xovers so I need to design an acceptable passive Xover. Mostly working with PSD-lite.
Thanks for any help. 😉
You'll probably get the most relevant results with your FRD file made from a collective of samples representative of what your speaker is putting in to the room. Going this way they should be in their boxes but should not include room information, although at the lowest frequencies where the room will be clearly involved you might want to do something from the listening position.
For the same reasons it's hard to get room free measurements down low to begin with so you might, for example, want FRD files focussed toward each particular cross point where you can choose the level of detail and room involvement you can tolerate or use.
For the same reasons it's hard to get room free measurements down low to begin with so you might, for example, want FRD files focussed toward each particular cross point where you can choose the level of detail and room involvement you can tolerate or use.
If you have access to an infinite baffle setup you could measure the drivers in such a way, so you can use the data and simulate whatever enclosure you want before physically build it. See here for this kind of setup: Zaph|Audio
The alternate method is to measure the drivers in the final enclosure. The only problem with this approach is that if the enclosure adds some effect in the same way as the driver, i.e. a bump in the response in both the "natural" FR of the driver and the enclosure. And of course you'll find this effect only after you built the enclosure.
Either way you should always use the quasi anechoic method with a gate, in order to leave out the reflected sound. The length of the gate is dependent of the space around, what is important is the difference in path from the straight distance from driver to mic and the reflected one. Looking at the impulse response it is obvious where the first reflection kicks in. My setup (mic some 80cm to the driver) limit the valid range to above 200-250Hz, under that frequency I have to measure near field, and merge the two measures.
Ralf
The alternate method is to measure the drivers in the final enclosure. The only problem with this approach is that if the enclosure adds some effect in the same way as the driver, i.e. a bump in the response in both the "natural" FR of the driver and the enclosure. And of course you'll find this effect only after you built the enclosure.
Either way you should always use the quasi anechoic method with a gate, in order to leave out the reflected sound. The length of the gate is dependent of the space around, what is important is the difference in path from the straight distance from driver to mic and the reflected one. Looking at the impulse response it is obvious where the first reflection kicks in. My setup (mic some 80cm to the driver) limit the valid range to above 200-250Hz, under that frequency I have to measure near field, and merge the two measures.
Ralf
Hi Thump Lump, what I suggest as a "test" of whether your measurements are ok or not is to do the following:
Position your speaker as far from any objects you can preferably at an angle to any parallel walls (even better do it outside) Try for mid way between the floor and ceiling.
Place your mic about 1M away on axis with the tweeter.
Take a measurement of the tweeter with Holm and check the impulse, if it looks upside down (main peak going down rather than up) toggle the "Invert recorded Signal" tickbox on the first tab.
Go back to the measurement tab and click options for the measurement. Change the Time Zero auto detect to highest positive peak. click auto detect, click it again. Close the options window.
Go to the second tab and select time zero locked. then click on the Use button beside last detected.
Go back to the measurements tab and measure the woofer (don't move anything).
Now connect a simple crossover (it may be one you have been working on)
Do a third measurement (again don't move anything).
Do your measurements of the impedance of each driver in the box at a fair distance from any other objects (you might as well do them in the same spot you had the speaker for the FR measurements). You would be surprised what interference will show up in the impedance plot.
Load your frd and impedance measurements into PCD and put in the details of the crossover components you used for the third measurement (make sure you put in coil dcr). Don't put in any Z offset information.
Export the simulated crossover and import it back into holm impulse. Compare it to the actual measured result. (I actually use speaker workshop for the simulation and the comparisons, but the above should be fine).
If your simulated and measured results are very similar in spl and phase then your measurements can be trusted for the crossover design. It won't tell you if your mic is accurate (ie flat FR) but it will tell you whether you have measurements that you can trust for getting good results with the simulation.
If you are only doing a two way then the above should be "good enough" I get usable measurements down to about 250 Hz off my verandah pointing into the yard. I've found it very difficult to get meaningful measurements below that in room. if you need accuracy below about 300Hz then simulated response, or appropriately corrected nearfield measurements spliced into the "farfield" measurements will need to be used.
Tony.
Position your speaker as far from any objects you can preferably at an angle to any parallel walls (even better do it outside) Try for mid way between the floor and ceiling.
Place your mic about 1M away on axis with the tweeter.
Take a measurement of the tweeter with Holm and check the impulse, if it looks upside down (main peak going down rather than up) toggle the "Invert recorded Signal" tickbox on the first tab.
Go back to the measurement tab and click options for the measurement. Change the Time Zero auto detect to highest positive peak. click auto detect, click it again. Close the options window.
Go to the second tab and select time zero locked. then click on the Use button beside last detected.
Go back to the measurements tab and measure the woofer (don't move anything).
Now connect a simple crossover (it may be one you have been working on)
Do a third measurement (again don't move anything).
Do your measurements of the impedance of each driver in the box at a fair distance from any other objects (you might as well do them in the same spot you had the speaker for the FR measurements). You would be surprised what interference will show up in the impedance plot.
Load your frd and impedance measurements into PCD and put in the details of the crossover components you used for the third measurement (make sure you put in coil dcr). Don't put in any Z offset information.
Export the simulated crossover and import it back into holm impulse. Compare it to the actual measured result. (I actually use speaker workshop for the simulation and the comparisons, but the above should be fine).
If your simulated and measured results are very similar in spl and phase then your measurements can be trusted for the crossover design. It won't tell you if your mic is accurate (ie flat FR) but it will tell you whether you have measurements that you can trust for getting good results with the simulation.
If you are only doing a two way then the above should be "good enough" I get usable measurements down to about 250 Hz off my verandah pointing into the yard. I've found it very difficult to get meaningful measurements below that in room. if you need accuracy below about 300Hz then simulated response, or appropriately corrected nearfield measurements spliced into the "farfield" measurements will need to be used.
Tony.
Last edited:
Thanks for the replies. In the particular case I am now into, the speakers are used for near field monitors used for a small home recording studio. Listening distance is about 18" or .5m. The speakers are sitting on a shelf behind the computer monitor as far away from the back wall as possible and about 33" apart, angled in to the seating position. Not exactly the same as in the living room, 10 feet away.
Given this, I have measured both the tweet and the woofer on axis from the listening position. Measuring each a few times (they come out almost identical). The ZMA files were created with the drivers in the enclosure. I felt with the singular listening position, it would be better to measure with that in mind, in this particular instance.
I loaded up the FRD and ZMA files into PSDlite and filled in the current baffle dimensions and driver placement. Ignoring the lower freq., I can get a Xover to simulate a 2+ dB total variation between the high and low points in the SPL curve. Honestly, it look a good deal flatter compared to a full range test in Holm with the active system I am currently using.
If anyone is interested in playing with this and think they can do a better Xover than what I have come up with, let me know and I can post the ZMA and FRD files.
Attached is the sim graph from PSDlite as well as the Xovers
Edit.....If you notice, the Xovers do not show the inductors DCR. I had put those in and it made no real difference. In the process of changing components, I didn't bother looking up the DCR of various inductors that I was trying.
Given this, I have measured both the tweet and the woofer on axis from the listening position. Measuring each a few times (they come out almost identical). The ZMA files were created with the drivers in the enclosure. I felt with the singular listening position, it would be better to measure with that in mind, in this particular instance.
I loaded up the FRD and ZMA files into PSDlite and filled in the current baffle dimensions and driver placement. Ignoring the lower freq., I can get a Xover to simulate a 2+ dB total variation between the high and low points in the SPL curve. Honestly, it look a good deal flatter compared to a full range test in Holm with the active system I am currently using.
If anyone is interested in playing with this and think they can do a better Xover than what I have come up with, let me know and I can post the ZMA and FRD files.
Attached is the sim graph from PSDlite as well as the Xovers
Edit.....If you notice, the Xovers do not show the inductors DCR. I had put those in and it made no real difference. In the process of changing components, I didn't bother looking up the DCR of various inductors that I was trying.
Attachments
Last edited:
Hi Thump Lump could you post original driver frd plots with phase and also the simulation with phase? I have found that it is possible to get quite flat FR whilst having not so good phase. The shelf in the tweeter and woofer response between about 2.5Khz and 5Khz concerns me 🙂 also reverse the tweeter and post the response. That will tell you a lot about how well the phase between the drivers is tracking.
Tony.
Tony.
Hi,
Baffle dimensions and and driver placement should not matter
with those files, if they do something is not being done right.
rgds, sreten.
FWIW I've never ever seen a x/o with that sort of overlap
2KHz to 5KHz, what does the reverse null look like ?
Last edited:
I went back and tried to duplicate the Xover plot above and discovered that I didn't check 4 boxes (2 for each driver). Those being to compensate for driver offset and baffle step. With the FRD files generated with the drivers in their enclosure, offset and baffle step will already be there and no further compensation needed. I am going to go back and do new FRD files since I moved the mic vertically to be on axis with each driver and this time, split the difference between driver centers and record both from a fixed mic. position.
Edit.... Given the center to center distance is only 5.5", do you really think it will make a significant difference? That would only make it 2.25" off center for each driver.
Edit.... Given the center to center distance is only 5.5", do you really think it will make a significant difference? That would only make it 2.25" off center for each driver.
Last edited:
Seems like I'm going 1 step forward and 3 steps back. I decided to try again with the FRD files I have and with the aforementioned boxes checked. Short version....I simply didn't like what I was coming up with or driver response. So, I decide to check the active system I have been using. I kept coming up with the same hole in response at around 8K-9K. To high to be a Xover issue. Simply a tweet response issue. So lets take a look at Vifa's data sheet. It shows the same on axis dip. Looking a bit closer, at 30 degrees off axis, it disappears for the most part. Time to retest the active system and see what happens with the enclosure tilted back. I tried 20,25 and 30 degrees of tilt and found that 25 was the best. The system response smoothed out considerably from just below the crossover point on up. This must come down to driver design as well as a closer to zero driver offset. This was all hidden in my first try because of the previously mentioned un-checked boxes concerning baffle step and driver offset.
I guess it's back into the shop to build another pair of enclosures. I think I will try to leave the woofer vertical and tilt the tweeter back 25 degrees. I need to look closer at the woofer's response to the tilt and go from there. They could stand a slight increase in volume as well.
I guess it's back into the shop to build another pair of enclosures. I think I will try to leave the woofer vertical and tilt the tweeter back 25 degrees. I need to look closer at the woofer's response to the tilt and go from there. They could stand a slight increase in volume as well.
- Status
- Not open for further replies.