Hi,
It is for a a series first order with an appropriate
bafflle width, then it gets more complicated.
Nevertheless the bass section should be say 4dB higher than mid and treble
sections when choosing drivers if you want low mid / treble attenuation.
rgds, sreten.
Last edited:
Hello John,
Very interesting feature. I understand minimum phase is the system and it inverse are causal and stable.
Can we do this in passive, for example with a LR filter ?
Second question : It is very useful to do this ? Are the records preserve enough phase information to hear the benefits ?
Note I have seen similar projects :
Elsinore Speakers DIY , http://www.diyaudio.com/forums/multi-way/97043-elsinore-project-thread.html
May be a confusion for me here, the phase of the system is limited to +-20°
http://www.diyaudio.com/forums/mult...-linear-open-baffle-speaker-measurements.html
Thanks.
This is all very true as I just found out recently with my own little project. A simple series XO is probably the best way to go for a 2-way setup and have it would nearly perfect the first time out.
Can you pad the fullrange with a resistor if it is too sensitive?
The drivers I have to play with are Fostex FX120s and AE TD6Hs.
Both are around 89dB sensitivity...
-herm
The drivers I have to play with are Fostex FX120s and AE TD6Hs.
Both are around 89dB sensitivity...
-herm
You can pad a fullrange with a resistor but be aware of two things: the level drop will vary a little with frequency. It won't drop as much at frequencies where the driver impedance is highest, generally at resonance and at higher frequencies where the inductance climbs. The net effect at resonance will be to increase Q somewhat.
Secondly, you will be consuming power in the resistor. If you just needed 1 or 2 ohms in series it isn't a big deal. Beyond that, look for a different way.
Regards,
David S.
Thanks for the reply, Dave.
I have attached the impedance curve of the FX120.
As you can see, it starts at ~8 ohms for my application (baffle step
at about 500 hertz) and rises to 32 ohms at 20khz.
Let's say I use an 8 ohm pad resistor. At 500 hertz I have reduced
power delivered to the driver by 50%.
By 20 khz, I am at 20%.
So I would have a rising response, yes?
Can we guesstimate that the rise is on the order of 1-2dB over the
range of the driver?
Herm
P.S. I realize I would be much better off finding a full ranger with less
sensitivity, but those drivers are all I have to play with for now.
I have attached the impedance curve of the FX120.
As you can see, it starts at ~8 ohms for my application (baffle step
at about 500 hertz) and rises to 32 ohms at 20khz.
Let's say I use an 8 ohm pad resistor. At 500 hertz I have reduced
power delivered to the driver by 50%.
By 20 khz, I am at 20%.
So I would have a rising response, yes?
Can we guesstimate that the rise is on the order of 1-2dB over the
range of the driver?
Herm
P.S. I realize I would be much better off finding a full ranger with less
sensitivity, but those drivers are all I have to play with for now.
Attachments
Hi,
Use an L pad on the driver, one series resistor, one parellel.
Try to keep the series fairly low, say 4 ohms and parallel
fairly high say 8 ohms - this is a -6dB 8 ohm L-pad.
This will reduce but not eliminate the response rise.
If the treble rise due to inductance is not wanted at all. (it
may be it is in the overall balance), add a Zobel capacitor
in series with the parallel resistor to prevent the rise.
This will reduce the overall attenuation, for the same
as above you increase the series resistor to 8 ohms.
Above are typical values but you can use any values you
like (keeping overall impedance healthy) to tune treble
and adjust the sensitivity.
rgds, sreten.
Qts at Fs (or Fb) will increase, but not an issue here.
Use an L pad on the driver, one series resistor, one parellel.
Try to keep the series fairly low, say 4 ohms and parallel
fairly high say 8 ohms - this is a -6dB 8 ohm L-pad.
This will reduce but not eliminate the response rise.
If the treble rise due to inductance is not wanted at all. (it
may be it is in the overall balance), add a Zobel capacitor
in series with the parallel resistor to prevent the rise.
This will reduce the overall attenuation, for the same
as above you increase the series resistor to 8 ohms.
Above are typical values but you can use any values you
like (keeping overall impedance healthy) to tune treble
and adjust the sensitivity.
rgds, sreten.
Qts at Fs (or Fb) will increase, but not an issue here.
Last edited:
Since the driver is 8 ohms at middle frequencies, the 8 ohm resistor in series will cut those frequencies to half level or -6dB. At resonance and at high frequencies the 8 ohm resistor is relatively less compared to the rising impedance, so the loss is less.
Easy to calculate: dB loss = 20 log [Rspeaker / (Rseries + Rspeaker)]
At 15kHz you have 10dB over 8 ohms = 3.2 x 8 = 26 ohms
26 / (8 +26) = .76
20 log .76 = -2.33 dB
So the mids lose 6dB and the top end loses 2.33. In relative terms the top end gained 3.67
The better way to think of it is that as the driving impedance rises, as you move from constant voltage to constant current, the frequency response becomes equalized by the exact shape of the impedance curve.
Sretin is right that for higher attenuation an L-pad approach will drop level with flatter response.
David S.
The topic title refers to minimum phase, but the discussion has focused on linear phase. Why would anyone want linear phase?
The OP intent is a little ambiguous. If he meant minimum phase crossover, then most crossover sections are minimum phase. A minimum phase system would be linear phase to the extent that it had smooth and wideband response. That is the definition of minimum phase, that any phase deviations are simply those tied to frequency response shortcomings, so equalizing the system perfects the phase response.
As I've mentioned before, proper crossover topology would not be enough. Interunit time delay would have to be physically corrected also.
Why would we want it? Pretty square waves. No one has yet proven that linear phase is a prerequisite for good sound.
David S.
Use DSP such as the Ultimate Equalizer and you can at least eliminate the arguments that the response suffers because it's not linear phase. The focus can then move on to the areas most likely to be issues that are known to affect perceived response. 🙂
Dave R.
Dave,
That might well be true if one were to regard AES-papers and such as the only source of truth. In this specific question we are nowadays in the lucky position that anyone who is really interested whether "phase matters" or not can readily test this himself/herself, assuming you have access to a PC-based system for measurements and manipulation (convolvers inserted in the audio chain, in this case).
Mind that it does not you cost anything, all the needed software and knowledge is in the public domain... somewhere in the HolmImpulse-Threads I described how to basically phase-linearize a given multiway-speaker, only other SW you need is excel (or any spreadsheet SW) and a convolver (eg the one intergrated in the Foobar2000 SW-player). So you have a powerful FIR-machine running and of course you can further manipulate the kernels and/or use analog/IIR-based additional EQ. For example I dialed in a linear-phase bass-rolloff instead of the natural minimum-phase response, to "speed up the bass" even more....
With a bit of ingenuity in the procedural handling you can set up fully blind testing conditions, including fake tests etc and to my knowledge anybody who has tried this came to the conclusion that, while generally subtle, phase issues -- including polarity -- are audible (with nothing else changed in the speaker, of course, another merit of the source-manipulating approach). The changes manifest in many ways, different timbre, different perception of space/reverb, different dynamics. With some recordings it matters more than with others, and it all is very subjective (as to where preferences go) as well as a matter of taste.
Given this powerful means I think a good approach is to let go of linear phase XO in the analog domain, better use one of the many known good ways to make a decent speaker without panicking about the allpass behaviour. To make full use of adjustable phase response one needs DSP-eq'ed listening conditions anyway and then phase linearisation to any target is part of the cake (with the better systems at least).
- Klaus
One of the biggest problems in assessing linear phase loudspeakers in the past has been that such tests have not been based on all other things equal. With the PC software available today, like that contained in SoundEasy V17 or the Ultimate equalizer, a speaker can be designed from the ground up and listened to in linear phase mode of "normal" nonlinear phase mode with absolutely no changes to frequency or polar response. Or the signal feed to an existing speaker can be preprocessed to correct the output to a linear phase response. These are truly all other things equal approaches which allow the listener to reach his own conclusion regarding phase linearity.
Has it not yet been blind-tested? It should be relatively easy to set up. Unfortunately I don't have a configuration here that allows me to test linear phase.
All true. The question is not whether phase matters, but where the threshold of perception is. Certainly larger delay times will be audible, this going back to Hilliard and theater speakers in the 30s. Also certain test signals have been found that can be more revealing than natural sounds. My belief is that, on music, our currently typical well-under-a-millisecond discrepencies are below that threshold. Tests on all-pass networks e.g. Blauart and Laws, tend to confirm that
My problem is that once "linear phase" is made a priority, in my opinion, more important areas of performance are usually compromised. Just reading this thread you'll see that most here think that all they need is to stick with a first order network or subtractive network and they'll be good. Clearly interunit time delay has to be dealt with, which means complex cabinets with physical driver offset built in. This generaly gives a few more folds in the cabinet to create diffraction and reflection issues. The first order network will compromise power handling and its excess overlap will compromise off axis frequency response. If it is pursued correctly then the rolloff of the drivers, always greater than first order, will have to be equalized upwards beyond the crossover points. This will further complicate matters and compromise power handling.
For a DIY project you can set your own priorities and have at it. You will likely enjoy the results and confirm that all your design choices were enlightened. But for commercial systems, pursuing phase linearity will always have a cost and once you pursue it you will inevitably cut costs in other areas. I think this is the primary reason why linear phase was essentially a fad of the 80's: cabinets cost more and nobody ever heard the difference.
I do put credence in AES papers and follow what Floyd Toole and Sean Olive write. Numerous times they have done rigourous tests where they have not found phase linearity to be a factor in listener preference. Most interesting would be Sean's latest work where he has been able to come up with the numerical weightings of measurable performance factors that allow prediction of product ranking. In the end the required measurements are various forms of frequency response: on axis smoothnes, on axis flatness, bass extension. Phase performance is not a required factor. I would maintain that pursuing flatter phase will typically degrade those factors that are important.
I'd agree with this. Once you go with DSP networks then phase linearity is a fairly simple attribute to achieve and won't force compromise on other performance areas. I just don't believe it is a requirement.
David S.
If you are referring to Toole's latest tome, Sound Reproduction, I'd add that it's evident he didn't seem much concerned about HD in any form as being a factor in ratings. It's hardly mentioned anywhere. FR in it's various forms seemed to 'rule the roost'.
BTW Dave, your comments here (@ DIY) are much appreciated. At least by me 🙂
Here are two tests showing how the AR-SXO series xover behaves on axis and off axis. The dash lines are the phase plots.
Note how the impulse response changed betwee 0 deg. and 45 Deg. Green, if you can see it, is the off-axis test.
The speaker under test is a DIY project still in development; utilizing two SS high end drivers. Smoothing is 1/6th oct. Not too bad a response I'd say.
Note how the impulse response changed betwee 0 deg. and 45 Deg. Green, if you can see it, is the off-axis test.
The speaker under test is a DIY project still in development; utilizing two SS high end drivers. Smoothing is 1/6th oct. Not too bad a response I'd say.
Attachments
Last edited:
very good response I'd say! The low frequency disturbances will almost certainly be room related. very nice 45deg as well!
Tony.
Tony.
- Status
- Not open for further replies.