I would do measurements and calculations before designing the XO / EQ network.
For my OB speakers (Woofers SS26W8861T00, 160Hz-, midwoofers SS18W8531G00 160Hz - 2.3kHz) I have managed to have two opamps only for the midwoofers and 2-3 (not decided yet) opamps for the woofers and achieved all required EQ, including a buffering opamp, shelving LP filters, and BW3 18dB/oct acoustic roll-offs, and a "notch" at 600Hz for the woofers. In the LR4 4th order XO for the midwoofers and the tweeter (Dynaudio T330D), I use passive XO. Only two capacitors are used in the passive network to achieve a near LR4 response with phase perfectly aligned. The frequency response as measured is quite flat. Using a 11ms gate, i.e. including the ceiling and floor responses, from 100Hz to 20kHz the response is +/-3dB variations of narrow band.
I did extensive simulations and calculations before I even designed the cabinet. Some simulations were proved to be accurate and some were not. If I relied on simulations only it would end up as a failure. If I did not do my design heavily based on simulations there was no way I could achieve my goal, including the simplest possible EQ networks and the least component counts, as I believe any capacitor in the networks adds a veil to the sound.
The dipole peaks and nulls are rather unpredictable due to driver directivity and other factors so no text book formulae can be used to derive the EQ / XO network. Such network can only be designed and built based on the measured data.
My 2 cents.
Regards,
Bill
For my OB speakers (Woofers SS26W8861T00, 160Hz-, midwoofers SS18W8531G00 160Hz - 2.3kHz) I have managed to have two opamps only for the midwoofers and 2-3 (not decided yet) opamps for the woofers and achieved all required EQ, including a buffering opamp, shelving LP filters, and BW3 18dB/oct acoustic roll-offs, and a "notch" at 600Hz for the woofers. In the LR4 4th order XO for the midwoofers and the tweeter (Dynaudio T330D), I use passive XO. Only two capacitors are used in the passive network to achieve a near LR4 response with phase perfectly aligned. The frequency response as measured is quite flat. Using a 11ms gate, i.e. including the ceiling and floor responses, from 100Hz to 20kHz the response is +/-3dB variations of narrow band.
I did extensive simulations and calculations before I even designed the cabinet. Some simulations were proved to be accurate and some were not. If I relied on simulations only it would end up as a failure. If I did not do my design heavily based on simulations there was no way I could achieve my goal, including the simplest possible EQ networks and the least component counts, as I believe any capacitor in the networks adds a veil to the sound.
The dipole peaks and nulls are rather unpredictable due to driver directivity and other factors so no text book formulae can be used to derive the EQ / XO network. Such network can only be designed and built based on the measured data.
My 2 cents.
Regards,
Bill
Thanks Bill, do you have a thread/page which describes your system and how it was designed and implemented? Obviously I am more than interested to reduce the component count.
I think there are few schools of thought when designing OB system. Such example is MJK and John K where they do heavy simulation, and the other is SL and Fikus where their approach is build-measure-listen and then iterate the design. I think either way would work.
I practically just follow SL's steps here:
http://linkwitzlab.com/faq.htm#Q24
Very important thing he pointed out is to use live performances as reference, not other speakers.
Prototype:
I managed to complete the XO part while waiting for the paint to dry.
The baffle is based on 1200mm x 450mm x 18mm particle board (I can't stand MDF dust). Nothing too fancy, the only critical construction is the opening for P13WH midrange.
I think it will be only a few hours away before hearing them the first time. This anticipation is always the most fun.
I think there are few schools of thought when designing OB system. Such example is MJK and John K where they do heavy simulation, and the other is SL and Fikus where their approach is build-measure-listen and then iterate the design. I think either way would work.
I practically just follow SL's steps here:
http://linkwitzlab.com/faq.htm#Q24
Very important thing he pointed out is to use live performances as reference, not other speakers.
Prototype:
I managed to complete the XO part while waiting for the paint to dry.
The baffle is based on 1200mm x 450mm x 18mm particle board (I can't stand MDF dust). Nothing too fancy, the only critical construction is the opening for P13WH midrange.
I think it will be only a few hours away before hearing them the first time. This anticipation is always the most fun.
From my own experience with chipboard, you may need just a little other material, such as 40 * 19 hardwood between the mid and woofer section for strength and to damp the vibration.
Are you using 90 * 40 for uprights?? because that gives you a small U-frame
Are you using 90 * 40 for uprights?? because that gives you a small U-frame
Except for the base 18mm particle board looks stiff enough, but the final design will have thick pine/oak edging like this anyway so hopefully they will be better. I had unsuccessful attempts with u-frames and steering away from them. Always had the impression that the sound is shifted but it may be just placebo effect too. I would sacrifice the few Hz for flat baffle (which can be equalised electronically anyway).
Initial listening:
Albeit wires strewn on my living room I managed to get them ready for initial listening. Until the quiet morning hours and wide range of materials from known jazz audiophile classics to linkin park.
In short I like them.
They sound clean, even the mediocre tweeter which are now crossed at 1.5kHz sounds ok. I had to listen "hard" for the previous shortcoming of P13WH (cuppy sound). Not as pronounced as I remembered it. I'll experiment with notch filters to see whether they're actually needed or not. There is no apparent lack of resolution from the P13s. It was very enjoyable I listed to the full album of Peter Hicks, which in testing point of view - a waste of time 😀. It was 2.30 in the morning when I walked to bed.
My worry about tweeter height was largely unfounded. The tweeters are about 10-15cm above ear level and it's fine. This may be due to high dispersion of 19mm tweeters. Changing them to 25mm such as DX25 may change this.
I walked around the room and no longer hear any beaming out of the alphas so 1/2 wavelength rule is proven in this case. This makes me wonder whether listening sweetspot for CS2 would have been very narrow.
Don't likes:
- I really do not like the look of the baffles. They appear like very tall monolith with empty spaces below (reserved for 2nd pair of woofers). If I later decide to not use them they will be cut and the ratio will be good.
- As with other OB iteration, serious lack of punch. My gripe with any OB so far and somewhat accepted the fact. I do really miss the chest-thumping percussion of Hossam Ramzy.
I wonder whether making the second pair of alphas as closed box to serve 20-35Hz will help without affecting dipole bass sweetness...
Next I will do measurements and investigate whether delay correction filters are worth the extra component count.
Initial listening:
Albeit wires strewn on my living room I managed to get them ready for initial listening. Until the quiet morning hours and wide range of materials from known jazz audiophile classics to linkin park.
In short I like them.
They sound clean, even the mediocre tweeter which are now crossed at 1.5kHz sounds ok. I had to listen "hard" for the previous shortcoming of P13WH (cuppy sound). Not as pronounced as I remembered it. I'll experiment with notch filters to see whether they're actually needed or not. There is no apparent lack of resolution from the P13s. It was very enjoyable I listed to the full album of Peter Hicks, which in testing point of view - a waste of time 😀. It was 2.30 in the morning when I walked to bed.
My worry about tweeter height was largely unfounded. The tweeters are about 10-15cm above ear level and it's fine. This may be due to high dispersion of 19mm tweeters. Changing them to 25mm such as DX25 may change this.
I walked around the room and no longer hear any beaming out of the alphas so 1/2 wavelength rule is proven in this case. This makes me wonder whether listening sweetspot for CS2 would have been very narrow.
Don't likes:
- I really do not like the look of the baffles. They appear like very tall monolith with empty spaces below (reserved for 2nd pair of woofers). If I later decide to not use them they will be cut and the ratio will be good.
- As with other OB iteration, serious lack of punch. My gripe with any OB so far and somewhat accepted the fact. I do really miss the chest-thumping percussion of Hossam Ramzy.
I wonder whether making the second pair of alphas as closed box to serve 20-35Hz will help without affecting dipole bass sweetness...
Next I will do measurements and investigate whether delay correction filters are worth the extra component count.
Midrange investigation:
Measurement reveals that while the front radiates perfectly (as in datasheet-perfect), the rear is not. I can only guess the huge magnet blocks and create diffractions.
The notch is 600Hz and -1db (!). Yes this minute peak from the rear is audible and manifest in "cuppy" sound. Most prominent on diana krall recordings. I feel that this justifies the the additional op-amp.
Tweeter:
Further listening with wider material reveals that they are strained as thought previously.
Nothing to report on the woofers. Contrary to many popular belief the low-end extension is not the most difficult to remediate in OB. Simply "just add more". I think midrange is the most tedious to get right.
It's time to relax the ears for few days and then see whether they still sound "correct". It's 3.30am and I'm really sleepy (love this hobby!). I hope to find a live unamplified performance around or otherwise I just take my ninja to the mountains this weekend -- different kind of music this beast 😀.
Measurement reveals that while the front radiates perfectly (as in datasheet-perfect), the rear is not. I can only guess the huge magnet blocks and create diffractions.
The notch is 600Hz and -1db (!). Yes this minute peak from the rear is audible and manifest in "cuppy" sound. Most prominent on diana krall recordings. I feel that this justifies the the additional op-amp.
Tweeter:
Further listening with wider material reveals that they are strained as thought previously.
Nothing to report on the woofers. Contrary to many popular belief the low-end extension is not the most difficult to remediate in OB. Simply "just add more". I think midrange is the most tedious to get right.
It's time to relax the ears for few days and then see whether they still sound "correct". It's 3.30am and I'm really sleepy (love this hobby!). I hope to find a live unamplified performance around or otherwise I just take my ninja to the mountains this weekend -- different kind of music this beast 😀.
Quick update.
After few quiet days I am found that the dipole peak is more serious than previously thought. This is due to driver resonance at the back (air trapped between the chassis etc.). I have few variants of notch filter and not really happy with the measured results. So more investigation is needed. Unfortunately I can only do one or two changes in a day as extended listening makes my ears "used" to the sound, only to come back to them 2 days later and they don't sound real.
Aside from this midrange notch, the XO, shelving, and delay correction seems to be correct. There is an absence of artefacts and measure well. My impression of the delay correction is coherence of the drivers but I really don't know how to measure them so placebo effects is a real possibility too.
Very interesting experience: I got 'lazy' and played around with notch filter depth and felt it was getting 'right' and quite enjoy their sound improvements.... until they were measured and no difference was noticed on the graph!. Quick investigation revealed that the filters were wrongly implemented (!).
This convince me more that the ears are not to be trusted and how easy it is to perceive a difference of sound when one "wanted" or "expect it".
After few quiet days I am found that the dipole peak is more serious than previously thought. This is due to driver resonance at the back (air trapped between the chassis etc.). I have few variants of notch filter and not really happy with the measured results. So more investigation is needed. Unfortunately I can only do one or two changes in a day as extended listening makes my ears "used" to the sound, only to come back to them 2 days later and they don't sound real.
Aside from this midrange notch, the XO, shelving, and delay correction seems to be correct. There is an absence of artefacts and measure well. My impression of the delay correction is coherence of the drivers but I really don't know how to measure them so placebo effects is a real possibility too.
Very interesting experience: I got 'lazy' and played around with notch filter depth and felt it was getting 'right' and quite enjoy their sound improvements.... until they were measured and no difference was noticed on the graph!. Quick investigation revealed that the filters were wrongly implemented (!).
This convince me more that the ears are not to be trusted and how easy it is to perceive a difference of sound when one "wanted" or "expect it".
gainphile said:
Now, ya see, if ya did the right thing and used a shallow U-frame, it would do 2 things.....
1. brace the baffle to remove any baffle resonances
2. allow you to install 2" thickness of wadding behind the mids.
ps.. glad you are having so much fun ! 😀 😀
Gainphile,
I like your S6B configuration, it should give you excellent sound, however with your small room, the sound may be overwhelming. This config will give a huge sound stage I am afraid your room maybe too small.
The mid, P13 has a low high freq cut-off, at 1300Khz, hence difficult to match tweeter. The Peerless HDS 810921 is very good at AUD$125 each is very expensive.
If money is no problem I will pick a high SPL mid-range around 95-96dB eg Audax, B&C ,PHY or AE, if you are interested I will provide details.
Cheers.
I like your S6B configuration, it should give you excellent sound, however with your small room, the sound may be overwhelming. This config will give a huge sound stage I am afraid your room maybe too small.
The mid, P13 has a low high freq cut-off, at 1300Khz, hence difficult to match tweeter. The Peerless HDS 810921 is very good at AUD$125 each is very expensive.
If money is no problem I will pick a high SPL mid-range around 95-96dB eg Audax, B&C ,PHY or AE, if you are interested I will provide details.
Cheers.
If you have measurements, it is not difficult at all using LTSpice and Speaker Workshop to design your active circuit that tailors the response to a flat response.
It is true that even OB can have too much bass. Where it differs from box speakers are these overload are even and easily can be trimmed using level/volume control. I am not going to preach the quality and articulation of dipole bass. In developing this prototype I am using 3 integrated amps so level matching is readily accessible.
I worked again the peaks last night. While those peaks are easily identified using measurement tools, the remedy (notch filters) are not easily done and they do not always relate directly. My measurement is from listening position and room/baffle influence are at their fullest effect. I also use software parametric EQ during intial assessment of the peak notch frequency, depth and Q. Yet, constructing 6db notch, for example do not directly remove a 6db notch. This is where subjective listening and measurement iteration is necessary.
In the picture below (ignore the SPL) it's clearly seen that the 600Hz has been tamed. Compare it with the perfect near-field response below it which looks pretty and good for xo confirmation - but somewhat useless sonically in real world. As female voices are one of my testing tools, at the current state I would evaluate the system as a good "mid-fi" and not a reference quality. More work is needed.
I worked again the peaks last night. While those peaks are easily identified using measurement tools, the remedy (notch filters) are not easily done and they do not always relate directly. My measurement is from listening position and room/baffle influence are at their fullest effect. I also use software parametric EQ during intial assessment of the peak notch frequency, depth and Q. Yet, constructing 6db notch, for example do not directly remove a 6db notch. This is where subjective listening and measurement iteration is necessary.
In the picture below (ignore the SPL) it's clearly seen that the 600Hz has been tamed. Compare it with the perfect near-field response below it which looks pretty and good for xo confirmation - but somewhat useless sonically in real world. As female voices are one of my testing tools, at the current state I would evaluate the system as a good "mid-fi" and not a reference quality. More work is needed.
What measurements do you use? Do you use MLS?
If you use gated measurements you can take the room out of the equation above about 200-300Hz. I would EQ it as flat as possible.
However, if the measurements are not gated, they are rather useless above the modal region.
If you use gated measurements you can take the room out of the equation above about 200-300Hz. I would EQ it as flat as possible.
However, if the measurements are not gated, they are rather useless above the modal region.
HiFiNutNut said:
I'm not entirely sure that you WANT to take the room out of the equation for open back speakers. The room and positioning relative to the wall behind the speaker make a big difference to the sound. They are very much part of the equation.
I'm using ARTA and a diy panasonic mic. I have yet to really calibrate them but they are very useful to investigate FR.
Rear Tweeters:
I installed the reverse-firing rear tweeters last night and what a difference that made. I have been using them before on previous iteration, but not while investigating midrange response. I initially thought this can be left for later once midrange sounds correct, but that apparently is the wrong way to go.
I don't know how to describe it, but switching back with/without rear tweeter the overall "body" sound hollow when they're not used and very prominent on midrange (not the highs surprisingly). I am not talking about "frequency-response hollow" but rather a "space-time hollow". The additional tweeters give a "full" sound. Surprisingly too, I can then use less aggressive notch due to this. The most probable explanation given by SL to this "effect" is that human brain has evolved to expect a component of sound reflection, and upon hearing partial reflection it's interpreted differently. I found too that talking in a heavily-damped rooms (we have a dedicated video-conferencing room in the office) just sounds unnatural and disturbing.
Early morning usually is a good time to listen to speakers as the ears have been rested. I found them sounding "correct" today albeit the limitations of cheap tweeters. My initial thought is I could live with this but I don't want to rush to conclusions before few weeks. There is a hint of warmness on female voice and piano which I may decide later whether to equalise or not. I do feel that to be a reference-quality the speakers should not contribute their own sound. Not noticeably at least.
Rear Tweeters:
I installed the reverse-firing rear tweeters last night and what a difference that made. I have been using them before on previous iteration, but not while investigating midrange response. I initially thought this can be left for later once midrange sounds correct, but that apparently is the wrong way to go.
I don't know how to describe it, but switching back with/without rear tweeter the overall "body" sound hollow when they're not used and very prominent on midrange (not the highs surprisingly). I am not talking about "frequency-response hollow" but rather a "space-time hollow". The additional tweeters give a "full" sound. Surprisingly too, I can then use less aggressive notch due to this. The most probable explanation given by SL to this "effect" is that human brain has evolved to expect a component of sound reflection, and upon hearing partial reflection it's interpreted differently. I found too that talking in a heavily-damped rooms (we have a dedicated video-conferencing room in the office) just sounds unnatural and disturbing.
Early morning usually is a good time to listen to speakers as the ears have been rested. I found them sounding "correct" today albeit the limitations of cheap tweeters. My initial thought is I could live with this but I don't want to rush to conclusions before few weeks. There is a hint of warmness on female voice and piano which I may decide later whether to equalise or not. I do feel that to be a reference-quality the speakers should not contribute their own sound. Not noticeably at least.
I agree with that. But that is a room problem, not a speaker response problem. I would not apply EQ to rectify the problem. Room response is as important as the speaker response. However, room problems require room treatments and speaker placements.
Room reflections are mostly not minimum phase therefore trying to EQ the speakers based on room-included-measurements to counter the room effects would not necessarily work.
Above 200Hz, I would definitely EQ the speakers to a flat response using gated measurements that take out the room completely. Any abnormalties (room problems) would be dealt with room treatments.
Below 200Hz, I don't mind taking the first floor reflection for the woofers (if mounted low and XOed low) into the EQ of the woofers, and I don't mind removing one or two substantial peaks in the room response with EQ.
Update:
After a week or so listening while notches are identified and sound is good I still feel something missing from this pair of speakers. As the ears again cannot be trusted I took them outside and do a proper no-reflection measurement, and gated to to eliminate neighbour fences.
What I found was surprising. The midrange area have major peak-dip which are caused by dipole summation. This most definitely is related to unoptimised baffle size. Quick study from the internet I found that P13WH which is is about 11cm in effective cone diameter is simply too small for baffle size of 45cm width.
The rule of thumb of optimum baffle width is 2 * cone diameter. When appropriately sized, the beaming effect of drivers sums nicely with dipole peaks and yields to flat response. Unfortunately I was not aware of this, well I guess there is no better learning than experience.
This then leads to ..... what is the optimum baffle size? Just my luck while reading the topics John from MusicAndDesign actually sells such software so I bought and played around with it. Even more surprising when I do simulation of my baffle they practically matched with my measurement 😱 . This simulator is really great and accurate. At $20 aussies, this tool is so invaluable, just compare the blue lines below with above measurement result!
So now.. back to the drawing board, still using same drivers but hopefully this time more equipped with important information on how to make good dipole baffles. And more understanding why the orions, nao, etc. looks like what they are.
After a week or so listening while notches are identified and sound is good I still feel something missing from this pair of speakers. As the ears again cannot be trusted I took them outside and do a proper no-reflection measurement, and gated to to eliminate neighbour fences.
What I found was surprising. The midrange area have major peak-dip which are caused by dipole summation. This most definitely is related to unoptimised baffle size. Quick study from the internet I found that P13WH which is is about 11cm in effective cone diameter is simply too small for baffle size of 45cm width.
The rule of thumb of optimum baffle width is 2 * cone diameter. When appropriately sized, the beaming effect of drivers sums nicely with dipole peaks and yields to flat response. Unfortunately I was not aware of this, well I guess there is no better learning than experience.
This then leads to ..... what is the optimum baffle size? Just my luck while reading the topics John from MusicAndDesign actually sells such software so I bought and played around with it. Even more surprising when I do simulation of my baffle they practically matched with my measurement 😱 . This simulator is really great and accurate. At $20 aussies, this tool is so invaluable, just compare the blue lines below with above measurement result!
So now.. back to the drawing board, still using same drivers but hopefully this time more equipped with important information on how to make good dipole baffles. And more understanding why the orions, nao, etc. looks like what they are.
Are your P-13s centered om the baffle or offset?
If they are offset by what distance,???
thanks for doing the research as I still want to use these mid-range drivers myself.
If I remember you are using and modeling single P-13s, what happens if you model MTM mid-ranges??
If they are offset by what distance,???
thanks for doing the research as I still want to use these mid-range drivers myself.
If I remember you are using and modeling single P-13s, what happens if you model MTM mid-ranges??
The P13 is single and in the center (no offset). From my further reading, horizontal offset can smoothen but at the expense of poor off-axis response to one side.
The tool do not simulate MTM arrangement. There is the popular xlBaffle which does this but when I tried to simulate my measurement on it there is no significant difference by having MTM arrangements.
The tool do not simulate MTM arrangement. There is the popular xlBaffle which does this but when I tried to simulate my measurement on it there is no significant difference by having MTM arrangements.
If you have a standard flat baffle or standard U-frame baffle for the woofers, the ABC dipole from John K is very accurate up to the first dipole peak. John K told me this a couple of times when we exchanged emails.
Indeed I found that is the case. I have measurement of my U-frame woofer and it matches exactly with John K's modelling software. Beyond the first dipole peak and dip things are rather unpredictable due to driver directivity and other factors.
The software also has a notch filter calculator and using LTSpice you can model very accurately U-frame woofer response without measurements.
That is a great relief for the DIYer because it is very difficult to obtain meaningful measurements at low frequencies with OB woofers while the software can be relied on to do a great job.
Regards,
Bill
Indeed I found that is the case. I have measurement of my U-frame woofer and it matches exactly with John K's modelling software. Beyond the first dipole peak and dip things are rather unpredictable due to driver directivity and other factors.
The software also has a notch filter calculator and using LTSpice you can model very accurately U-frame woofer response without measurements.
That is a great relief for the DIYer because it is very difficult to obtain meaningful measurements at low frequencies with OB woofers while the software can be relied on to do a great job.
Regards,
Bill
- Status
- Not open for further replies.