bjorno said:
Thanks alot, I'll look for it. That article is hereby inquired, if anyone collects old Speaker Builder issues and have a scanner, those pages would be highly appreciated 🙂
Another thing relating to the crossover;
I chose the series filter topology mainly out of curiousity and for the excercise (never made a series crossover before).. I like the tuneability of the topology very much as I feel the response is very easily tailored which makes baffle step comphensation quite easy.. but if anyone would be able to sensible describe to me how "back-EMF" (electro-motive force of the cone movement "generating" signal induced in the tweeter??) can be a problem in this circuit topology, I'd be delighted to hear your sayings on the matter. So far, that has been a non-issue as I have *no* clue on how to calculate it's effects.. 😕
BTW I found a webshop which sells that Speaker Builder issue, so it's all good I guess. Thanks Björn.
I chose the series filter topology mainly out of curiousity and for the excercise (never made a series crossover before).. I like the tuneability of the topology very much as I feel the response is very easily tailored which makes baffle step comphensation quite easy.. but if anyone would be able to sensible describe to me how "back-EMF" (electro-motive force of the cone movement "generating" signal induced in the tweeter??) can be a problem in this circuit topology, I'd be delighted to hear your sayings on the matter. So far, that has been a non-issue as I have *no* clue on how to calculate it's effects.. 😕
BTW I found a webshop which sells that Speaker Builder issue, so it's all good I guess. Thanks Björn.
phase_accurate said:
Ouch...
Is that a filter design that you use? Now I don't see any frequency axis in your graph, but if it is a typical woofer-tweeter filter, the humps in the responses will IMO be clearly audible, regardless of that the sum straight. The reason that they sum straight is that the contributions are of nearly opposite phase. For some other directions than what you have optimised for, the cancellation will be less efficient and there will be a hump in the response. Even if you are not sitting in that direction, this sound will be reflected by the walls.
Rocky said:
Rocky,
PSpice will do it for you. Just use your own schematic xo.gif and short the input or add an estimated low value resistor for ohmic losses in cables and substitute for amplifier output impedance.
Then replace both drivers with a Back Emf generator (Voltage Source) in series with their nominal impedance values.
Now you can perform ac and/or transient analysis.
For example, feed or sweep at the woofer terminals with a signal covering the woofer resonance + - one octave and probe the tweeter.
In this case you have to short the tweeter voltage source and add a differential buffer with gain equal one to sort out the ‘grounding’ issue.
This will give you voltage readings to compare. If you want to compare impedances instead, change the voltage source to a current source by adding a resistor>= 1 k Ohm in series with the voltage source as you do when making swept speaker impedance plots.
B
Svante said:
For me it is correct to caculate with the phase shift cancelation, as long as both drivers are situated close enough together that both can be considered as one point compared to wave length.
But I am wondering since years about phase alignment in general.
Let's consider a perfect low midrange speaker.
A violine playing the tone a at 440Hz, and generating some 2nd + 3 harmonic as it's natural sound....
I am sitting in 3m distance of the speaker.
440Hz has a wave length of 0.773m, means there will be 2.588 waves between me and my speaker. 440Hz will reach my ears with a phase shift of 2 x 360 degrees plus 212 degrees.
880Hz has a wave length of 0.386m, means there will be 5.176 waves between me and my speaker. 880 Hz will reach my ears with a phase shift of 5 x 360 degrees plus 63.5 degrees.
1320Hz has a wave len........
Now we could argue:"...yes that's one reason why speakers always sound artificial."
NO!!! We have exactly the same behaviour already when we listen to the original violine in a live concert. The different frequencies will reach your ears at another phase angle than they were generated by the violine.... But still life concerts sound natural no matter if our distance 5m or 15m from the violine.
I just found this link:
http://sound.westhost.com/biamp-vs-passive.htm
Explains the issue better than I did.
The Crossover - a Different View
B
http://sound.westhost.com/biamp-vs-passive.htm
Explains the issue better than I did.
The Crossover - a Different View
B
Thanks Björn, I'll be doing some reading today. I also found a chapter on back-EMF in a school book from last year (a chapter that wasn't pensum) so I have several leads now..
I've also been sneaking around on the tech docs of Basta!, and noticed the speaker expressed in both electrical, mechanical, and acoustic impedances at http://www.tolvan.com/basta/Basta!TechDoc.htm (Loudspeaker, equivalent circuit diagram)
Does anyone know how to implement such a gyrator in OrCAD?
I've also been sneaking around on the tech docs of Basta!, and noticed the speaker expressed in both electrical, mechanical, and acoustic impedances at http://www.tolvan.com/basta/Basta!TechDoc.htm (Loudspeaker, equivalent circuit diagram)
Does anyone know how to implement such a gyrator in OrCAD?
ChocoHolic,
My immediate thought: I think your freestanding issue concerning phase in this case and conclusions is a gross oversimplification.
Maybe my understanding of English is limited but I do not understand this as I don't exclude the possibility to use many mic's capturing the soundfield and multichannel loudspeakers for the corresponding sound reproduction:
Quote: ‘We have exactly the same behaviour already when we listen to the original violine in a live concert. The different frequencies will reach your ears at another phase angle than they were generated by the violine’
I have attached a picture to bring more sense? or better basis for further ‘brainstorming’.
B
My immediate thought: I think your freestanding issue concerning phase in this case and conclusions is a gross oversimplification.
Maybe my understanding of English is limited but I do not understand this as I don't exclude the possibility to use many mic's capturing the soundfield and multichannel loudspeakers for the corresponding sound reproduction:
Quote: ‘We have exactly the same behaviour already when we listen to the original violine in a live concert. The different frequencies will reach your ears at another phase angle than they were generated by the violine’
I have attached a picture to bring more sense? or better basis for further ‘brainstorming’.
B
Attachments
Rocky,
Here is another PSpice document containing a gyrator model:
page 67 at:
http://www.mip.sdu.dk/~pfavr/apswitch.pdf
B
Here is another PSpice document containing a gyrator model:
page 67 at:
http://www.mip.sdu.dk/~pfavr/apswitch.pdf
B
ChocoHolic said:
Here is a simulation of directivity from two point sources of equal magnitude 130 mm apart. The individual sources has a level of + 5 dB (estimated from that frequency response graph) but they are almost out of phase (145 degrees) so that the summed response in the 0 degree direction is 0 dB. Look at the response in the other directions! In dome directions the level is + 12 dB.
Now 2000 Hz as crossover frequency, and 130mm driver distance are common for eg normal 2-way speakers.
I would say that this filter configuration is very problematic since it boosts the response as much as 12 dB only 30 degrees off axis. and also due to the big dip that occurs just a few degrees above the main axis. Moving the head upwards just a little bit will cause great changes in the response.
An externally hosted image should be here but it was not working when we last tested it.
Rocky said:
First I'm sorry, I don't know OrCad so I won't be able to answer your question.
If you don't find a way to do that, you might be helped by knowing that mechanical and acoustical systems also can be simulated by admittance analogies. In this case you will not need the gyrator, you can use a transformer instead. In this case the mass is represented by a capacitor and the compliance by an inductor. Force corresponds to current and velocity to voltage.
...but still, the gyrator is kind of cool, don't you think? 😀
Svante said:
Indeed, it is way elegant. I'm messing around now trying to build one using some current crontrolled voltage sources and similar, I can tell I'm getting closer, but the output indicates I'm still way off a functional setup.. Hard to find good info on this online too, the best clue I've come across is at http://www.geocities.com/f4ier/vented.htm
Svante said:
This is it?
An externally hosted image should be here but it was not working when we last tested it.
What inductance do I model the xformers with?
edit: or should I just mimic the xformers with controlled sources?
Svante & Eric
The simulation was aresult of playing around with transient-improved filter topologies. You can't avoid humps with these. The higher the order and the steeper the slopes the higher those humps get.
So it is advisible to use a filter order as high as necessary and as low as possible.
I am also aware that the polar response wouldn't be very good. Though your model seems to take both drivers as point sources which they aren't in practice (woofer beaming) so it would look a little different.
OTOH a speaker that is not tranisent perfect can't reproduce the input waveform.
If you use drivers that can be placed close to each other compared to the wavelength in the crossover area things look better. So most would be gained when these topologies are used to combine small fullrangers with additional woofers.
Regards
Charles
The simulation was aresult of playing around with transient-improved filter topologies. You can't avoid humps with these. The higher the order and the steeper the slopes the higher those humps get.
So it is advisible to use a filter order as high as necessary and as low as possible.
I am also aware that the polar response wouldn't be very good. Though your model seems to take both drivers as point sources which they aren't in practice (woofer beaming) so it would look a little different.
OTOH a speaker that is not tranisent perfect can't reproduce the input waveform.
If you use drivers that can be placed close to each other compared to the wavelength in the crossover area things look better. So most would be gained when these topologies are used to combine small fullrangers with additional woofers.
Regards
Charles
Rocky said:
Yes. You'll have to think a bit about the resistances, though, for example Rm is rather 1/Rms than Rms. And the series equivalent for the radiation admittance corresponds to a parallel equivalent radiation impedance. Everything is "the other way around".
phase_accurate said:
Yes, true, but the big dip just a few degrees upwards would definitely be there (given my conditions, see below)
...true, but what purpose is that? If you have ever tried a blind test on an allpass filter @ 2 kHz there is no doubt that waveform exact reproduction is not necessary.
Yes, ok, there we agree, for lower crossover frequencies the problem is less. Still, even if it is less, the polar plot would look like this att 200 Hz and 200 mm driver distance. I think this design has problems too with a big null in the 30 degree upwards direction and + 6 dB downwards 60 degrees.
An externally hosted image should be here but it was not working when we last tested it.
PS. I do not intend to be offensive about this, and I realise from your nick that phase accurateness is something you feel strongly about, so I hope I'm not being too blunt here. And as you say, your example is an example of playing around with different designs, rather than an actual final version filter, so maybe I'm overreacting. Anyway, the point that I am trying to make is that having two drivers play loud at nearly opposite phase is not a good idea.
Rocky,
Your e source should be of a constant current type feeding the electrical side causing a voltage drop over primary side of transformer.
At the mechanical side of the transformer you will have a corresponding velocity as drop over the secondary side sourcing the mechanical mobility branch with a force that acts like a current.
Similar transformation will occur from the mechanical side to the acoustical, but the force entering the transformer will translate into a volume velocity sourcing the acoustical side with a pressure that acts like a current.
In this case I think you should use transformers and simply program your Orcad ideal transformers to Bl:1 and 1:S.
Before you proceed with your circuit simulations, my recommendation for you is to study a good book of electrical an electronic engineering.
My own source and 'Bible' in this matter has been Acoustics (1954); Leo.L Beranek, since the late sixties.
B
Your e source should be of a constant current type feeding the electrical side causing a voltage drop over primary side of transformer.
At the mechanical side of the transformer you will have a corresponding velocity as drop over the secondary side sourcing the mechanical mobility branch with a force that acts like a current.
Similar transformation will occur from the mechanical side to the acoustical, but the force entering the transformer will translate into a volume velocity sourcing the acoustical side with a pressure that acts like a current.
In this case I think you should use transformers and simply program your Orcad ideal transformers to Bl:1 and 1:S.
Before you proceed with your circuit simulations, my recommendation for you is to study a good book of electrical an electronic engineering.
My own source and 'Bible' in this matter has been Acoustics (1954); Leo.L Beranek, since the late sixties.
B
No problem with that. I am able to to see disadvantages and advantages of transient- optimised and -non optimised systems. I am NOT a salesman, neither cars, insurances nor audio so I don't have to praise what I sell as "the best there is".
If you use a fullrange driver that is capable of delivering some (though restricted) bass then you can get away with a 1st /2nd (LP/HP) order topology which has much smaller humps. In this case you will get away with deviations of +- 2dB 30 deg off-axis with a driver-distance of 20 cm and an x-over frequency of 200 Hz.
Regards
Charles
bjorno said:
Of course it is a massive oversimplification. Ideal wave model in open air and neglecting damping function of the air. It is becoming a bad attitude of me to make up my mind about the basics, because I am observing since 6 months that most critical errors are based on ignoring the basics...
I still do not see any error in the simplified model.
Even more. Today I simply followed the model to it's result... earlier I stopped thinking on half way...
If you go through this model with some math, you will notice that:
sin(2 x pi x f1 x t1 + 2.588 x 2 x pi) + sin(2 x pi x f2 x t1 + 5.176 x 2 x pi) = sin(2 x pi x f1 x t2) + sin(2 x pi x f2 x t2)
Herein we have:
t1 = 0 (delay at the violine)
t2 = 8.82ms (delay at the ear)
f1 = 440Hz
f2 = 880Hz
In the terms with f1, there pi is reflecting 1.136ms.
In the terms with f2, there pi is reflecting 0.568ms.
...or just run a spice simualtion and add the phase shifted signals and compare with the sum of the originals.
You will see that both wave shapes are identical, but sififted by the 8.82ms delay, which is given by distance/speed.
OK, now I am ready to think about head related transfer functions....
And also I fully agree that stepped front panels can make sense, if
acoustic speed in the speaker materials and in air are both taken into account.
- Status
- Not open for further replies.