OK so I decided that to properly do the active crossover for my Sub to MTM's I needed some decent measurements of the "SUB's". I've put subs in quotes because they are actually my old three-ways with the wiring wired direct to the woofer and crossover detached.
I did nearfiled measurements summing the woofer and port response, performance was MUCH better than I expected. Bit of a problem with port resonance at around 700Hz, but since I'll be building new boxes (eventually) I'll address that then. Note that the nearfield measurement is the raw measurement, no smoothing or gating applied.
I also did a "farfiled" measurement at 1.5M Speaker around 1M off the ground and well away from walls at the rear. Was on the balcony and positioned to ensure the reflections went off into free air.
Matching up the nearfield and farfield responses seems to correlate nicely between around 300 and 500 Hz. so I spliced at 400Hz (which is also where I set the gating for the farfield) with 100Hz overlap.
The thing that is bugging me is that the spliced response looks too bass heavy (no signs of baffle step at all) tapering off from around 100Hz down to around 1500Hz (about 5db drop, which is the exact opposite to what I would have expected) I'm not sure if my methodology was flawed or not (should probably re-read D'appolito).
I've attached the individual measurements the overlapped ones and the final spliced one.
It does look like the response drops as frequency increases above 300Hz. Howver F3 Baffle step freq should theoretically be at 239Hz (480 mm baffle width) so I'm a little concerned that what I've got is not valid for putting into my crossover model for working out the active filter slopes.
Anyway here are the measurements: any feedback would be appreciated
Images in order, nearfield, 1.5M overlayed (levels adjusted to fit at 400Hz splice point) and finally the spiced results.
Tony.
I did nearfiled measurements summing the woofer and port response, performance was MUCH better than I expected. Bit of a problem with port resonance at around 700Hz, but since I'll be building new boxes (eventually) I'll address that then. Note that the nearfield measurement is the raw measurement, no smoothing or gating applied.
I also did a "farfiled" measurement at 1.5M Speaker around 1M off the ground and well away from walls at the rear. Was on the balcony and positioned to ensure the reflections went off into free air.
Matching up the nearfield and farfield responses seems to correlate nicely between around 300 and 500 Hz. so I spliced at 400Hz (which is also where I set the gating for the farfield) with 100Hz overlap.
The thing that is bugging me is that the spliced response looks too bass heavy (no signs of baffle step at all) tapering off from around 100Hz down to around 1500Hz (about 5db drop, which is the exact opposite to what I would have expected) I'm not sure if my methodology was flawed or not (should probably re-read D'appolito).
I've attached the individual measurements the overlapped ones and the final spliced one.
It does look like the response drops as frequency increases above 300Hz. Howver F3 Baffle step freq should theoretically be at 239Hz (480 mm baffle width) so I'm a little concerned that what I've got is not valid for putting into my crossover model for working out the active filter slopes.
Anyway here are the measurements: any feedback would be appreciated
Images in order, nearfield, 1.5M overlayed (levels adjusted to fit at 400Hz splice point) and finally the spiced results.
Tony.
Attachments
Last edited:
Hi,
AFAIK baffle step will not show up on nearfield measurements.
After that the issues get tricky, see PM's articles on measurement at Stereophile.
Also see :
FRD Consortium tools guide
http://www.rjbaudio.com/Audiofiles/FRDtools example2.html
rgds, sreten.
AFAIK baffle step will not show up on nearfield measurements.
After that the issues get tricky, see PM's articles on measurement at Stereophile.
Also see :
FRD Consortium tools guide
http://www.rjbaudio.com/Audiofiles/FRDtools example2.html
rgds, sreten.
Curve splicing is only an approximate technique and there is always some arbitrariness to the result.
Remember what you have: a nearfield curve represents the 2 Pi response. If you flush mount the system in an infinite baffle, then the nearfield and the farfield curves would be the same. Keele showed this in his original paper on the subject. But with a conventional system then the nearfield and farfield curves are different and the difference is the baffle effect or the 4pi to 2pi conversion curve (4pi at LF and 2pi at higher Fs).
For splicing to be accurate you need to splice at frequencies below the transition range. The difficulty is that, if you could get nice curves accurate at those frequencies, you wouldn't need splicing! You can see though, that lower splicing frequencies would get the combined curve a little bit more downshelved at LF, closer to what you suspect the true response is.
Be careful that your impulse truncation doesn't warp your 1.5m response (try smoothing rather than truncation) and splice lower, and I think you will have a more accurate response.
David S.
Remember what you have: a nearfield curve represents the 2 Pi response. If you flush mount the system in an infinite baffle, then the nearfield and the farfield curves would be the same. Keele showed this in his original paper on the subject. But with a conventional system then the nearfield and farfield curves are different and the difference is the baffle effect or the 4pi to 2pi conversion curve (4pi at LF and 2pi at higher Fs).
For splicing to be accurate you need to splice at frequencies below the transition range. The difficulty is that, if you could get nice curves accurate at those frequencies, you wouldn't need splicing! You can see though, that lower splicing frequencies would get the combined curve a little bit more downshelved at LF, closer to what you suspect the true response is.
Be careful that your impulse truncation doesn't warp your 1.5m response (try smoothing rather than truncation) and splice lower, and I think you will have a more accurate response.
David S.
Thanks Sreten and David I'll check out the FRD tools a bit later, just posting before heading for work.
I tried different splicing just now.
3rd octave smoothed looks pretty ordinary (and lesser smoothing a lot worse). I gated at 200Hz as well which is clearly including reflections but I think it was better. the three measurements attached.
Tony.
I'll check out the FRD tools a bit later, just posting before heading for work. I tried different splicing just now.
3rd octave smoothed looks pretty ordinary (and lesser smoothing a lot worse). I gated at 200Hz as well which is clearly including reflections but I think it was better. the three measurements attached.
Tony.
Attachments
See David Ralph's two-way build here for an in depth study about your problem.
SB Acoustics 2-Way Title Page
SB Acoustics 2-Way Title Page
Good Idea Shaun! I've got BDS (and it is on the FRD site that Sreten linked to as well) Brain not engaging obviously!! It's a Vifa M26WR-09-08. I actually downloaded a new BDS type spreadsheet from Jeff Bagby only a couple of days ago as well... Loudspeaker Design Software
Thanks for the Link David I will try and digest that tonight! Very comprehensive write up I must say!!
[temporarily suspend reality] I was thinking this morning on the way to work, I need to take the measurement outside with the speaker at the top of a steep hill (sloping away at an angle that the reflections cant come back up to the mic) with nothing around at all, and a boom going up and out and back down again to position the mic.... also would need to be quiet, no wind and fine wheather.. not too much to ask is it Of course would also need a power source (perhaps a 12V battery and an inverter)... [restore reality]
edit: I've just done a baffle sim in Jeff's new sim (it is a bit simpler than BDS to get something done!) and attached here. It does look like the 1.5M measurement is following the curve expected in the sim. The port resonance problem at 700Hz is showing as a deviation but other than that, it looks to be fairly close.
Tony.
Thanks for the Link David I will try and digest that tonight! Very comprehensive write up I must say!!
[temporarily suspend reality] I was thinking this morning on the way to work, I need to take the measurement outside with the speaker at the top of a steep hill (sloping away at an angle that the reflections cant come back up to the mic) with nothing around at all, and a boom going up and out and back down again to position the mic.... also would need to be quiet, no wind and fine wheather.. not too much to ask is it
Of course would also need a power source (perhaps a 12V battery and an inverter)... [restore reality] edit: I've just done a baffle sim in Jeff's new sim (it is a bit simpler than BDS to get something done!) and attached here. It does look like the 1.5M measurement is following the curve expected in the sim. The port resonance problem at 700Hz is showing as a deviation but other than that, it looks to be fairly close.
Tony.
Attachments
Last edited:
Hi Terry, I have to admit I just followed my nose!
I did the measurements with Holmimpulse
Woofer about 5mm from the centre of the dustcap, thought OMG that looks terrible...
I then measured each of the two ports with the mic about 5mm from the plane of the baffle in the centre of the port. This was an area I wasn't sure about as I was uncertain whether there may be phase issues since the port exit is further forward than the position of the mic for the woofer measurement.
I then simply put the woofer measurement and first port measurement into slots A AND B in holm-impulse and used the built in SUM function to sum to slot C. I then put that into slot A and the second port response into Slot B and then did the sum again, the result looked feasible (but better than I expected) so I figured it must have worked, as I didn't think that if I stuffed up that I would get something that resembled what the models showed.
The bit that surprised me was it being flat down to around 30Hz... I expected the -3db point to be around 36Hz so something could well be amiss. (I can't remember the exact tuning of this box, and I'm unsure if I have the original model, but I should be able to work it out based on the port lengths, and the enclosure volume (or alternatively an impedance plot).
I'll post the individual measurements when I get home, as I'm on the train on the way home at the moment and the measurements are on my desktop
Tony.
I did the measurements with Holmimpulse
Woofer about 5mm from the centre of the dustcap, thought OMG that looks terrible...
I then measured each of the two ports with the mic about 5mm from the plane of the baffle in the centre of the port. This was an area I wasn't sure about as I was uncertain whether there may be phase issues since the port exit is further forward than the position of the mic for the woofer measurement.
I then simply put the woofer measurement and first port measurement into slots A AND B in holm-impulse and used the built in SUM function to sum to slot C. I then put that into slot A and the second port response into Slot B and then did the sum again, the result looked feasible (but better than I expected) so I figured it must have worked, as I didn't think that if I stuffed up that I would get something that resembled what the models showed.
The bit that surprised me was it being flat down to around 30Hz... I expected the -3db point to be around 36Hz so something could well be amiss. (I can't remember the exact tuning of this box, and I'm unsure if I have the original model, but I should be able to work it out based on the port lengths, and the enclosure volume (or alternatively an impedance plot).
I'll post the individual measurements when I get home, as I'm on the train on the way home at the moment and the measurements are on my desktop
Tony.
Hi Terry, I have to admit I just followed my nose!
I did the measurements with Holmimpulse
Woofer about 5mm from the centre of the dustcap, thought OMG that looks terrible...
I then measured each of the two ports with the mic about 5mm from the plane of the baffle in the centre of the port. This was an area I wasn't sure about as I was uncertain whether there may be phase issues since the port exit is further forward than the position of the mic for the woofer measurement.
I then simply put the woofer measurement and first port measurement into slots A AND B in holm-impulse and used the built in SUM function to sum to slot C. I then put that into slot A and the second port response into Slot B and then did the sum again, the result looked feasible (but better than I expected) so I figured it must have worked, as I didn't think that if I stuffed up that I would get something that resembled what the models showed.
The bit that surprised me was it being flat down to around 30Hz... I expected the -3db point to be around 36Hz so something could well be amiss. (I can't remember the exact tuning of this box, and I'm unsure if I have the original model, but I should be able to work it out based on the port lengths, and the enclosure volume (or alternatively an impedance plot).
I'll post the individual measurements when I get home, as I'm on the train on the way home at the moment and the measurements are on my desktop
Tony.
Hi Tony,
The only reference I have left after the flood is the LMS manual. the formulae is
20Log(0.2820*SQRT of Sd) Sd is in square metres. Do this for the driver and the port. Just multiply port Sd by 2 for two ports. The answers are in relative dB. Then subtract the port answer from the driver answer and you have the difference that you need totake into account whenyou sum the two together.
You can't just add the two together because what you are trying to do is to add the volume of air displaced and Sd must come into it. That is why your graphs look so good. Sorry to disappoint.
I have never used holm-impulse but I would guess that it has somewhere to add the relative dB or to scale one of the curves.
Hope this helps. M26WRs are not quite that good
Terry
Aha! Thanks Terry so I was right to capitalise MUCH when I said it was MUCH better than expected
I'll check out speaker workshop Shaun, should save me having to do the calculations manually.... I will post back when I have something more in line with reality...
Here are the individual measurements on the same graph BTW.
Cheers,
Tony.
so I was right to capitalise MUCH when I said it was MUCH better than expected I'll check out speaker workshop Shaun, should save me having to do the calculations manually.... I will post back when I have something more in line with reality...
Here are the individual measurements on the same graph BTW.
Cheers,
Tony.
Attachments
OK so here is the nearfield when calculated correctly! Much more like what I would have expected (in otherwords a bit disapointing) which is to be expected considering the construction of the boxes... I thought that the bodgy sealing job I had done must have worked much better than I'd hoped. Just shows that if it looks too good to be true then it is
Blue trace is the one I calculated in speakerworkshop, red one is the original one I had just summed in HolmImpulse without adjustment for port area.
Just shows I should re-read the docco before doing any measurements! Thanks for picking that up Terry!!
Also shows that I should have room for improvement when I build some proper boxes, I was a bit perplexed as I was hoping for improvements when I built the new boxes (as the current ones really do have some issues) but when I saw the first incorrect result I couldn't see how I could possibly improve it apart from the port problem!!
Tony.
Blue trace is the one I calculated in speakerworkshop, red one is the original one I had just summed in HolmImpulse without adjustment for port area.
Just shows I should re-read the docco before doing any measurements! Thanks for picking that up Terry!!
Also shows that I should have room for improvement when I build some proper boxes, I was a bit perplexed as I was hoping for improvements when I built the new boxes (as the current ones really do have some issues) but when I saw the first incorrect result I couldn't see how I could possibly improve it apart from the port problem!!
Tony.
Attachments
Hi Terry, Here is the impedance plot warts and all (more evidence of need of improvement)
From the look of it the box is tuned to 30Hz. I thought it was 34Hz from memory but I'd have to simulate to find out. The box is nominally 70L. Just looked at the T/S params and there are some differences between the two drivers and I don't know which is which... will average them and do a model.
I made the new baffles (on existing boxes) for these drivers probably 10 years ago not really sure when it was.... was the first excercise I did in modelling and "real" speaker construction, so I'm not sure what was going through my head, but I do remember deviating from a maximally flat response to try and get an extended response, which doesn't look to have been successful
Tony.
From the look of it the box is tuned to 30Hz. I thought it was 34Hz from memory but I'd have to simulate to find out. The box is nominally 70L. Just looked at the T/S params and there are some differences between the two drivers and I don't know which is which... will average them and do a model.
I made the new baffles (on existing boxes) for these drivers probably 10 years ago not really sure when it was.... was the first excercise I did in modelling and "real" speaker construction, so I'm not sure what was going through my head, but I do remember deviating from a maximally flat response to try and get an extended response, which doesn't look to have been successful
Tony.
Attachments
Yes it is at 30Hz, sounds close to right. I am not sure if your ports are working as a pair should or they are too small for the M26. Some further modelling before you cut would be time well spent. I can't run any sims until I ressurect my sunk computers. I am hoping I can at least get my data back. Much of the older stuff was not on the backup hdd. Time will tell.
Terry
Terry
You can compare measured impedance trace versus simulated to check whether the physical implementation is correctly aligned.
You need to have both SPL and Imp plots to see what and more importantly where a problem is. Phase helps as well. As an example, the above SPL plot shows a problem at around 22 Hz. You can see the corresponding blip in the phase curve. That usually means it is driver rather than enclosure related although in this case it could be port as well. All these things can be inter-related.
Terry
Terry
Hi Tony,
The only reference I have left after the flood is the LMS manual. the formulae is
20Log(0.2820*SQRT of Sd) Sd is in square metres. Do this for the driver and the port. Just multiply port Sd by 2 for two ports. The answers are in relative dB. Then subtract the port answer from the driver answer and you have the difference that you need totake into account whenyou sum the two together.
You can't just add the two together because what you are trying to do is to add the volume of air displaced and Sd must come into it. That is why your graphs look so good. Sorry to disappoint.
Terry is correct about area scaling. Also note that you have to add the sections with phase being considered (complex summation). The port contribution will be in phase above resonance and out of phase below resonance. If you don't see the port addition reducing the response at LF then you haven't got it right yet.
Also note that there will be crosstalk between port and woofer so this technique isn't perfect. I've had slightly better luck with inside-the-box measurements.
David S.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.