I'm building a 3-way speaker based on a Hypex FA253 plate amp with an integrated DSP.
All three elements are flush mounted in the same plane. I use the time delay feature in the DSP to get the timing in synch (due to different "sound pressure distances" and different distances to the listening position).
The crossover filters are standard LR4 filters both between low-mid and mid-high (~390Hz and ~2500Hz).
I have worked a lot with the getting the phases in synch in the crossover section and ended up with a rather large "separation" between the High pass and Low pass frequencies in each separation.
Example: The mid is set at 370Hz and the low is at 409Hz and with this configuration I get "clean" tones close to the crossover section (based on listening to music, not test tones).
Is it normal to get this wide separation of frequencies, or have I missed something important elsewhere?
Another observation is that I hear a difference with single Hz steps (409 is better than 408 or 410).
Is this normal?
Please share your experiences.
All three elements are flush mounted in the same plane. I use the time delay feature in the DSP to get the timing in synch (due to different "sound pressure distances" and different distances to the listening position).
The crossover filters are standard LR4 filters both between low-mid and mid-high (~390Hz and ~2500Hz).
I have worked a lot with the getting the phases in synch in the crossover section and ended up with a rather large "separation" between the High pass and Low pass frequencies in each separation.
Example: The mid is set at 370Hz and the low is at 409Hz and with this configuration I get "clean" tones close to the crossover section (based on listening to music, not test tones).
Is it normal to get this wide separation of frequencies, or have I missed something important elsewhere?
Another observation is that I hear a difference with single Hz steps (409 is better than 408 or 410).
Is this normal?
Please share your experiences.
My understanding is that LR4 filters should give you a result where both the drivers in phase - but the driver with the HPF (i.e. the low and mid) will be one cycle behind. You can work this out as the (1/frequency of crossover point) seconds and apply that as the delay to the driver with the hpf. along with this you can apply the delay as per driver centre distances.
ideally both lpf and hpf for LR crossovers should the same frequency to give you a flat frequency response - but I feel the best way to work for both amplitude and time delay response would be to test the speakers with a mic and something like REW and apply those corrections to the filters.
ideally both lpf and hpf for LR crossovers should the same frequency to give you a flat frequency response - but I feel the best way to work for both amplitude and time delay response would be to test the speakers with a mic and something like REW and apply those corrections to the filters.
When working with my Hypex FA253 [ New active 3-Way, Hypex and SB ] I found that the lower 200 Hz crossover was much more challenging to get dialed in than the upper 2 kHz crossover. The upper crossover was simple because I was working with gated quasi-anechoic measurements which had no room effect. The lower 200 Hz crossover was harder because I could never be certain how much my magnitude and phase measurements were being affected by room boundaries.
I also found I was fairly sensitive to phase and timing variations in the 200 Hz regions. A 200 Hz wavelength is 68 inches, but I found I could hear differences of as small as 4 inches of delay between the woofer and the mid.
So I am not surprised by what you report. I would iteratively make a change, listen and record my observations in writing, then measure and see if I could correlate what i heard with what was measured... repeat...repeat...
https://www.diyaudio.com/forums/class-d/306815-hypex-fusion-plate-amps-219.html#post6175835
What kind of measurements have you made so far?
I also found I was fairly sensitive to phase and timing variations in the 200 Hz regions. A 200 Hz wavelength is 68 inches, but I found I could hear differences of as small as 4 inches of delay between the woofer and the mid.
So I am not surprised by what you report. I would iteratively make a change, listen and record my observations in writing, then measure and see if I could correlate what i heard with what was measured... repeat...repeat...
https://www.diyaudio.com/forums/class-d/306815-hypex-fusion-plate-amps-219.html#post6175835
What kind of measurements have you made so far?
Both filters should have the same crossover frequency.
To adjust the delay in one of them, you reverse the polarity of the driver and you measure the notch in the frquency response to be as deep as possible while adjusting the delay.
This means that in the crossover region both drivers have the same phase when the polarity is back to normal.
Usually around -20 dBs is enough.
To adjust the delay in one of them, you reverse the polarity of the driver and you measure the notch in the frquency response to be as deep as possible while adjusting the delay.
This means that in the crossover region both drivers have the same phase when the polarity is back to normal.
Usually around -20 dBs is enough.
Interesting replies.
The measurement methods so far has been in-room REW/UMIK-1 measurements, both close to elements for a flat SPL curve, as well as listening position measurements to find which room modes to cancel out in the DSP (<200Hz).
This has been combined with listening tests (mainly using a set of "reference songs" with emphasis on different aspects).
The listening tests suggest that correct phase alignment in the crossover sections are more important than a flat SPL curve which is "destroyed" anyway by the in-room reflections.
The room is a standard living room without special acoustic treatment, not terrible but far from perfect. The impression so far is that the speakers significantly outperform the room (confirmed to some degree by nearfield listening), and additional acoustic treatment is currently not an option.
My approach so far was to first correct timing between the elements without filters (75 microseconds for 4" mid, 160 micro seconds for 1" dome tweeter, compared to 8" woofer), and then in a second step adjust for the phase shift by adjusting the crossover frequencies, ending up in a rather large gap.
If I understand you correctly, you suggest the other way around, to avoid a peak in the crossover region(?)
The measurement methods so far has been in-room REW/UMIK-1 measurements, both close to elements for a flat SPL curve, as well as listening position measurements to find which room modes to cancel out in the DSP (<200Hz).
This has been combined with listening tests (mainly using a set of "reference songs" with emphasis on different aspects).
The listening tests suggest that correct phase alignment in the crossover sections are more important than a flat SPL curve which is "destroyed" anyway by the in-room reflections.
The room is a standard living room without special acoustic treatment, not terrible but far from perfect. The impression so far is that the speakers significantly outperform the room (confirmed to some degree by nearfield listening), and additional acoustic treatment is currently not an option.
My approach so far was to first correct timing between the elements without filters (75 microseconds for 4" mid, 160 micro seconds for 1" dome tweeter, compared to 8" woofer), and then in a second step adjust for the phase shift by adjusting the crossover frequencies, ending up in a rather large gap.
If I understand you correctly, you suggest the other way around, to avoid a peak in the crossover region(?)
If you want to be technically accurate, think I'd get a friend to help with blind testing.
BTW, nice to see you experiment with non-identical crossover choices. A lot of people think you must have the same frequency and the same slope for the higher and the lower range because it saves wear on your slide rule. By mic and by ear, best settings may not be identical.
B.
BTW, nice to see you experiment with non-identical crossover choices. A lot of people think you must have the same frequency and the same slope for the higher and the lower range because it saves wear on your slide rule. By mic and by ear, best settings may not be identical.
B.
Last edited:
for the LR filters to operate correctly, it is necessary that the two drivers radiate from the same acoustical plane. This can be achieved mechanicaly by appropiate mounting or electrically using delays.
See "conclusion" of the original article here:
JAES papers
Again, many thanks for all replies and suggestions.
I have now tried the "other way around", i.e. same crossover frequency and aligning phase using delays.
Im my case delays are easy since I use the Hypex plate amp DSP for this. Digital sampling frequency is slightly above 90 kHz, so the alignment can be adjusted in steps of approx. 11 microseconds.
The Hypex SW simulation suggests the following delays for best phase alignment at the (two) crossover frequencies:
Woofer: 267 (microseconds)
Mid: 43
Tweeter: 0
Previous distance based alignment measurement gives a delay for mid 75 and for tweeter 160 microseconds, ending up in the following Hypex configuration:
Woofer: 149 (microseconds)
Mid: 0
Tweeter: 53
Early listening tests with live recordings with a lot of room information indicates that the new "delay aligned" setup not reaches the same level as the earlier "frequency aligned" setup when it comes to the for me very critical PRaT (Phase Ryhm and Timing) aspect.
The difference is rater subtile, but not very difficult to spot.
One explanation could be that the alignment outside the (rather narrow 24dB/octave) crossover zones is more accurate in the "frequency aligned" setup.
Please comment.
I have now tried the "other way around", i.e. same crossover frequency and aligning phase using delays.
Im my case delays are easy since I use the Hypex plate amp DSP for this. Digital sampling frequency is slightly above 90 kHz, so the alignment can be adjusted in steps of approx. 11 microseconds.
The Hypex SW simulation suggests the following delays for best phase alignment at the (two) crossover frequencies:
Woofer: 267 (microseconds)
Mid: 43
Tweeter: 0
Previous distance based alignment measurement gives a delay for mid 75 and for tweeter 160 microseconds, ending up in the following Hypex configuration:
Woofer: 149 (microseconds)
Mid: 0
Tweeter: 53
Early listening tests with live recordings with a lot of room information indicates that the new "delay aligned" setup not reaches the same level as the earlier "frequency aligned" setup when it comes to the for me very critical PRaT (Phase Ryhm and Timing) aspect.
The difference is rater subtile, but not very difficult to spot.
One explanation could be that the alignment outside the (rather narrow 24dB/octave) crossover zones is more accurate in the "frequency aligned" setup.
Please comment.
Thanks for wise and courteous reply. Right. I'd only add, implemented by DSP and then tested by mic and ear.
So maybe those (perhaps needlessly) fussing on these matters need to add to Linkwitz' always good math is the empirical measurement of the drivers. Because without nailing down the driver behaviour, math can't be right.
And even then, you need some dental impression goo so as to fix your head immovably in a bite-board.
Raise your hand if you know the planes of your drivers. (Hint: I don't think it is a matter for a tape measure. But glad to learn how to do it properly, please.)
B.
Last edited:
The following elements are used:
Visaton TIW 200 XS
Visaton TI100
ScanSpeak D2608/913000
I have no information about the "acustical planes". Listening based delays from impulse based sounds ("metronome" sound from a test record) suggest 75 microseconds delay for the mid and 160 for the tweeter at a listening position approx. 190cm from the speaker (at tweeter level).
Visaton TIW 200 XS
Visaton TI100
ScanSpeak D2608/913000
I have no information about the "acustical planes". Listening based delays from impulse based sounds ("metronome" sound from a test record) suggest 75 microseconds delay for the mid and 160 for the tweeter at a listening position approx. 190cm from the speaker (at tweeter level).
I have done two speakesrs with Minidsp and one pair with Hypex FA123. I use REW and UMIK-1 for measurements.
Johan, you must set baffle step compensation for the woofer and perhaps or the mid too - nearfield measurement misses baffle diffractions! Delay settings should be measured at 1 - 1,5m at listening axis. Reflections and room modes will make analysis difficult, but gating the measurement for 4ms per mid/tweeter and 8-12ms for woofer-mid works quite well.
In all of my speakers delay for woofer is 0, around 40us for mid and 42us for tweeter. LR4 xo must have same corner frequency and responses equalized straight at least one octave past xo (half/double Hz)
Start delay setup by playing only mid and tweeter, with same polarity. Look at spl response and step response (Impulse window, toggle Step) ideal step response shows peaks of tweeter and mid impulses combining snoothly. Then add delay to tweeter and take a new sweep, then compare to previous and take notice on changes. Then take third measurement with different delay and compare to previous - this way you learn to see and analyze changes and find the right delay. Verify and fine-tune dealay with new measurements with reversed polarity o the tweeter 8there should be a sharp dip of spl at xo. The dip is off and smoother from xo if delay is full cycle wrong.
Then play only woofer and mid and repaet the previous procedures. This is harder because of reflections messing responses.
When W/M is ok, change tweeter delay to match the M/T delay´determined earlier. Then measure the whole thing WMT response and also with M in reverse polarity (you should see two dips in response)
Learning and understanding wil take time, lots of time! Be patient. Save different dsp settings and respective measurements!
More info Refining a 4-way open-baffle speaker with the miniDSP 2x4
Johan, you must set baffle step compensation for the woofer and perhaps or the mid too - nearfield measurement misses baffle diffractions! Delay settings should be measured at 1 - 1,5m at listening axis. Reflections and room modes will make analysis difficult, but gating the measurement for 4ms per mid/tweeter and 8-12ms for woofer-mid works quite well.
In all of my speakers delay for woofer is 0, around 40us for mid and 42us for tweeter. LR4 xo must have same corner frequency and responses equalized straight at least one octave past xo (half/double Hz)
Start delay setup by playing only mid and tweeter, with same polarity. Look at spl response and step response (Impulse window, toggle Step) ideal step response shows peaks of tweeter and mid impulses combining snoothly. Then add delay to tweeter and take a new sweep, then compare to previous and take notice on changes. Then take third measurement with different delay and compare to previous - this way you learn to see and analyze changes and find the right delay. Verify and fine-tune dealay with new measurements with reversed polarity o the tweeter 8there should be a sharp dip of spl at xo. The dip is off and smoother from xo if delay is full cycle wrong.
Then play only woofer and mid and repaet the previous procedures. This is harder because of reflections messing responses.
When W/M is ok, change tweeter delay to match the M/T delay´determined earlier. Then measure the whole thing WMT response and also with M in reverse polarity (you should see two dips in response)
Learning and understanding wil take time, lots of time! Be patient. Save different dsp settings and respective measurements!
More info Refining a 4-way open-baffle speaker with the miniDSP 2x4
Attachments
Last edited:
https://www.diyaudio.com/forums/multi-way/324454-setting-lr4-active.html#post6222334
i could do my ex-speaker tweeter with both polarities.
Vituixcad simulation showed slight different on power response between these two
i could do my ex-speaker tweeter with both polarities.
Vituixcad simulation showed slight different on power response between these two
Many thanks for the detailed description of the process and for the links.
I will do my best to try the suggested method.
I do have a baffle step compensation in the mid at 630Hz (closed box) and a rolloff compensation/extension for the woofer at 46Hz (closed enclosure also here).
Any comments to the thoughts about the (decreased) timing precision for the frequencies outside of the crossover regions?
I will do my best to try the suggested method.
I do have a baffle step compensation in the mid at 630Hz (closed box) and a rolloff compensation/extension for the woofer at 46Hz (closed enclosure also here).
Any comments to the thoughts about the (decreased) timing precision for the frequencies outside of the crossover regions?
Your delays in post #10 looked odd to me. I was about to write a response and then saw Juhazi's post 14. I agree completely with his post.
In my active 3 way with Hypex FA253, the 12 inch woofer is in a separate cabinet which is further from the listening position than the mid and tweeter. However, if the woofer cabinet was brought forward so its baffle was in the same plane as the mid and tweeter, I would get the following delays:
12 inch woofer: 0 us = 0 mm
6 inch mid driver: 288 us = 100 mm
1 inch dome tweeter: 416 us = 143 mm
In my active 3 way with Hypex FA253, the 12 inch woofer is in a separate cabinet which is further from the listening position than the mid and tweeter. However, if the woofer cabinet was brought forward so its baffle was in the same plane as the mid and tweeter, I would get the following delays:
12 inch woofer: 0 us = 0 mm
6 inch mid driver: 288 us = 100 mm
1 inch dome tweeter: 416 us = 143 mm
The contribution of these variations itself varies. To put it simply, the compromise can be somewhere between the close up response and the in room, but a better situation is to focus only on the speaker when crossing, but treat the room. Better still is to design the speaker to deal with the room.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.