Damn! That is so NOT what I wanted you to get!
I was hoping for flat amplitude, of course. The adjusted values you tried in the second plot I quoted (Q=0.5) presumably give a -6db dip in amplitude? Or a complete deep notch? Can't quite remember what Linkwitz-Riley does when wired up wrong.
Still, there may be a way out of this with high source impedance like a zero-feedback amp...
Last edited:
Hi,
Note, the conditions under which it worked.
1) Midrange (fullrange) with basically flat response 150Hz-3KHz capable of handling a fair bit of LF (excursion) and a mild rolloff above 3KHz, crossover at 3KHz
2) Woofer with basically flat response up to around 1KHz, no major resonances, crossover at 300Hz.
3) Resistive tweeter with a modest HF Rolloff towards high frequencies easily equalised flat.
4) Enclosure width, places the mid bass dip in the region of the LF crossover.
As said, it took 15+ Years and many pairs of speakers to find one combination where it works. In most situations series crossovers do not deliver.
No dog ate my laptop, a thief broke in and stole it. Now my external FW card no longer works (the replacement Laptop lacks Firewire) and I need to get something else to measure, too much hassle.
Well, my work on Speakers (not my little series X-over pet project) got reviewed (and measured) by Paul Messenger, you can look at that to see if I know what I'm doing or not.
Ciao T
Note, the conditions under which it worked.
1) Midrange (fullrange) with basically flat response 150Hz-3KHz capable of handling a fair bit of LF (excursion) and a mild rolloff above 3KHz, crossover at 3KHz
2) Woofer with basically flat response up to around 1KHz, no major resonances, crossover at 300Hz.
3) Resistive tweeter with a modest HF Rolloff towards high frequencies easily equalised flat.
4) Enclosure width, places the mid bass dip in the region of the LF crossover.
As said, it took 15+ Years and many pairs of speakers to find one combination where it works. In most situations series crossovers do not deliver.
No dog ate my laptop, a thief broke in and stole it. Now my external FW card no longer works (the replacement Laptop lacks Firewire) and I need to get something else to measure, too much hassle.
Well, my work on Speakers (not my little series X-over pet project) got reviewed (and measured) by Paul Messenger, you can look at that to see if I know what I'm doing or not.
Ciao T
Hi,
I don't think of it as "step" at all. It is a change of radiation angle. In the power response of the speaker there is no step.
With respect all these "Baffle Step compensated" speakers are just speakers with loudness build in and undefeatable...
If systems had a correctly implemented and calibrated physioligically corrected volume control these Speakers would show their overblown bass (they also do so when played loud).
At best there is a modest discontinuity, in room between where the sounds bends around the baffle and where room gain restores it to flat or even boosted.
Ciao T
I don't think of it as "step" at all. It is a change of radiation angle. In the power response of the speaker there is no step.
With respect all these "Baffle Step compensated" speakers are just speakers with loudness build in and undefeatable...
If systems had a correctly implemented and calibrated physioligically corrected volume control these Speakers would show their overblown bass (they also do so when played loud).
At best there is a modest discontinuity, in room between where the sounds bends around the baffle and where room gain restores it to flat or even boosted.
Ciao T
So, it took 15 years to get a combination that "worked", nice
"The dog ate my laptop" was meant like a student saying "The dog ate my homework" doesn't matter how you lost the data, it's gone now and it's all "say so"
If it's too much hassle to measure then......well........whatever.
So I can also "look it up" since you can't be bothered showing a link to your work........nice again.
Bye bye
To get back to System7 I have built many crossovers outside of the box connecting components with clips. Using a volt meter testing the voltage has taught me a lot.
I have done a lot of experimenting and they sound more natural with mid bass woofer Zobel and tweeter R2. Zobel not used in subwoofer.
I have done a lot of experimenting and they sound more natural with mid bass woofer Zobel and tweeter R2. Zobel not used in subwoofer.
Hi,
One of us sure is.
Please comment. What is the effect of the baffle step on the power response of a given speaker? I am talking about a real speaker designed for use in an enclosed space, not simulations.
Ciao T
One of us sure is.
Please comment. What is the effect of the baffle step on the power response of a given speaker? I am talking about a real speaker designed for use in an enclosed space, not simulations.
Ciao T
What you ask for has already been done in another series thread. Check out post #85 here: http://www.diyaudio.com/forums/multi-way/27969-series-crossover-2.html#post2711135
There are a few active series xo threads here. Do a search on 'series' and you'll get many hits. Of course, not all will be about xo's if just the word 'series' us used.
I've expressed some scepticism and also explained why series connected networks weren't for me. At the same time I've said at least twice that they can probably be made to work, and nicely, under some circumstances.
Carry on with your analysis and if you show that good results can be had, then we'll all shut up.
I would be interested in:
Do series connected networks need particular driver curves to work?
Do they need a particular impedance curve, or conjugation to work?
Do they give different transfer functions than similar ordered parallel netowrks?
Can they be made to give the same transfer functions, if desired?
How does one optimize the network?
David S
Carry on with your analysis and if you show that good results can be had, then we'll all shut up.
I would be interested in:
Do series connected networks need particular driver curves to work?
Do they need a particular impedance curve, or conjugation to work?
Do they give different transfer functions than similar ordered parallel netowrks?
Can they be made to give the same transfer functions, if desired?
How does one optimize the network?
David S
Hi,
Not really, however if you want to use very simple series networks, yes, they do, but the same holds true for parallel crossovers of equal simplicity.
Again, it depends. In the cases I looked at where "series works" it works in ways equally simple parallel does not work, it is because it to a degree ignores impedance rises beyond the crossover point (shunt element determines crossovers, as opposed to series element).
If we use no conjugates (impedance compensation) and if we do not need the crossover to equalise transfer functions of drivers designed with a non-flat response, yes, they offer different transfer functions.
If we do not limit component count we can browbeat any network to any response we want.
Same as you would optimise any other network.
Personally it means measuring, simulation, building, measuring, listening, changing iteratively.
Others have different views.
Ciao T
Not really, however if you want to use very simple series networks, yes, they do, but the same holds true for parallel crossovers of equal simplicity.
Again, it depends. In the cases I looked at where "series works" it works in ways equally simple parallel does not work, it is because it to a degree ignores impedance rises beyond the crossover point (shunt element determines crossovers, as opposed to series element).
If we use no conjugates (impedance compensation) and if we do not need the crossover to equalise transfer functions of drivers designed with a non-flat response, yes, they offer different transfer functions.
If we do not limit component count we can browbeat any network to any response we want.
Same as you would optimise any other network.
Personally it means measuring, simulation, building, measuring, listening, changing iteratively.
Others have different views.
Ciao T
Hi,
You are late to the party and all the above pretty much has already
been described. Note that the above is not true for 2nd order series
but is to a degree true for for 2nd order parallel.
Generally 1st order parallel, 2nd order series, 3rd order parallel etc.
are more likely to need / benefit from zobels / impedance peak
compensation than 1st order series, 2nd order parallel, 3rd
order series etc. for the reason given above.
2nd order electrical series doesn't appeal much compared to parallel.
rgds, sreten.
I don't want to get into a flat earth debate about baffle step.
It happens. Of course the total power response has no step,
but that is just side stepping the issues with the phenomenon.
Last edited:
Hi,
Yes, with a correctly arranged 2nd order parallel type similar effects may be observed.
I might agree here, but then to me 2nd order does not appeal at all.
This is not a "flat earth" debate. It is a debate based on how the human hearing works. If you make speakers that are intended to be listened to in enclosed spaces you must take note of these issues and correctly deal with them.
No, on the contrary, this is precisely the issue. Because what you hear as tonal balance is determined more by the power response than the on axis anechoic response.
I am more than familiar with speakers which, based on anechoic on-axis response one might be tempted to completely dismiss, yet they offer an excellent power response and in room balance.
It boils down to "You can have one of these two":
1) A speaker with a flat anechoic response (Baffle Step equalised) which will have a bass boost in the power response and thus boosted bass in the "in room" response.
2) A speaker with a "bass shy" anechoic response (Baffle Step not equalised) which will have a flat power response and thus flat in-room response.
One option is "wrong" in the actual listening situation at realistic SPL's, but has build in bass boost that, while not matching a physiological correction goes into the general direction and seems "right" when measured using pseudo anechoic methods.
The other option is "wrong" when measured using pseudo anechoic methods, but "right" in the actual listening situation at realistic SPL's but lacks any artifical bass boost (and hence will sound "thin" at low SPL's.
As a general side note, if program material has been "mixed" at a reference level of (for example) 94dB SPL and it is listened to at 74dB SPL, then a system with a flat response will make the resulting sound subjectively thin and bass-shy.
Physiological correction needs to be employed, it would need around 6dB at 50Hz with return to flat at around 500Hz for this example.
This should NOT be confused with the common "Loudness" Button which does not provide physiological correction at all.
As a further side-note, in German Studio systems in the 80's the Monitor Level controls where physiologically equalised and for certain types of recordings (gain riding for classical recordings) we also employed physiologically corrected faders.
The resulting recordings always translated well onto system that where also physiologically corrected. And especially the classical recordings avoided this problem of crescendo's sounding thin and pianissimo's sounding fat due to this physiological correction when levels where altered to reduce actual dynamic range for domestic use.
Ciao T
Yes, with a correctly arranged 2nd order parallel type similar effects may be observed.
I might agree here, but then to me 2nd order does not appeal at all.
This is not a "flat earth" debate. It is a debate based on how the human hearing works. If you make speakers that are intended to be listened to in enclosed spaces you must take note of these issues and correctly deal with them.
No, on the contrary, this is precisely the issue. Because what you hear as tonal balance is determined more by the power response than the on axis anechoic response.
I am more than familiar with speakers which, based on anechoic on-axis response one might be tempted to completely dismiss, yet they offer an excellent power response and in room balance.
It boils down to "You can have one of these two":
1) A speaker with a flat anechoic response (Baffle Step equalised) which will have a bass boost in the power response and thus boosted bass in the "in room" response.
2) A speaker with a "bass shy" anechoic response (Baffle Step not equalised) which will have a flat power response and thus flat in-room response.
One option is "wrong" in the actual listening situation at realistic SPL's, but has build in bass boost that, while not matching a physiological correction goes into the general direction and seems "right" when measured using pseudo anechoic methods.
The other option is "wrong" when measured using pseudo anechoic methods, but "right" in the actual listening situation at realistic SPL's but lacks any artifical bass boost (and hence will sound "thin" at low SPL's.
As a general side note, if program material has been "mixed" at a reference level of (for example) 94dB SPL and it is listened to at 74dB SPL, then a system with a flat response will make the resulting sound subjectively thin and bass-shy.
Physiological correction needs to be employed, it would need around 6dB at 50Hz with return to flat at around 500Hz for this example.
This should NOT be confused with the common "Loudness" Button which does not provide physiological correction at all.
As a further side-note, in German Studio systems in the 80's the Monitor Level controls where physiologically equalised and for certain types of recordings (gain riding for classical recordings) we also employed physiologically corrected faders.
The resulting recordings always translated well onto system that where also physiologically corrected. And especially the classical recordings avoided this problem of crescendo's sounding thin and pianissimo's sounding fat due to this physiological correction when levels where altered to reduce actual dynamic range for domestic use.
Ciao T
Looking on as an observer, this comment seems to sum it all up. If the two types of filters have "similar effects" then what's the point!? They both work, and whichever one give you the ACOUSTIC result that you need (and they both seem to), then so be it. Why all the bantering about on such insignificant points?
Why not worry about things that actually make a difference? Or is that just too sensible?
Just say'in!
For the L/R //second order xover;L=2R divided by 2Pi.f.and C=0.5 divided by2Pi.f.R. As far as baffle step compensation goes,Thorston L is vindicated as it appears in the vast majority of actual measured responses as opposed to misapplied theory to be a non event in practical speakers.Just consider the measurements made in authoritive magazines such as "Audio" since Richard Heyser stepped up the ante in assessing speaker performance;no sign of it.Let us consider that we listen on axis and that if you realise that the sideways energy from the baffle at the transition from 2 Pi to 4Pi conditions(although it is also of minor consideration in a listening room as opposed to anechoic conditions) is time delayed,then it should be clear that we judge the speaker performance on the direct sound in the main and to a lesser extent by the resultant reverberant energy.All the same,it pays to make a baffle as wide as possible and always avoid over symmetrical placement of a driver(s) to keep diffraction to a minimum.
I'm not sure, it could give a good power response. Anyway, I was thinking from earlier of bringing up a third order series. I can't see the benefit myself but that's not the point.I'm a bit disappointed that I seem to have got the transfer function for second order constant resistance series crossovers wrong.
By the way designing one of these iteratively is a great way to kill time.
In the second plot I've used the above network against third-order Butterworth slopes (ideal conditions).
Attachments
You're too clever by half, Allen!
Let's get this right, you've designed a flat amplitude third order butterworth series there, starting with 1.273mH and 19.9uF as in Rod Elliott's first order. I think I'm right in recalling that a pair of 3rd order butterworths sums nicely to be flat amplitude even with parallel.
The constant resistance solution would be those values multiplied by 0.707 and 1.414 for a second order, and 0.5, 4/3 and 1.5 for a third order, no? I worked that out from the Edward Norton Constant resistance filter paper. Keeping Z1 X Z2 = R X R being the reason.
We now need to list the values at 1000Hz and 8 ohms for the various flat impedance and flat response series networks to be useful for design purposes, don't we?
Let's get this right, you've designed a flat amplitude third order butterworth series there, starting with 1.273mH and 19.9uF as in Rod Elliott's first order. I think I'm right in recalling that a pair of 3rd order butterworths sums nicely to be flat amplitude even with parallel.
The constant resistance solution would be those values multiplied by 0.707 and 1.414 for a second order, and 0.5, 4/3 and 1.5 for a third order, no? I worked that out from the Edward Norton Constant resistance filter paper. Keeping Z1 X Z2 = R X R being the reason.
We now need to list the values at 1000Hz and 8 ohms for the various flat impedance and flat response series networks to be useful for design purposes, don't we?
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.