I have finally got my active crossover together, although still just a very basic highpass/lowpass. Now I intend to add a variable phase correcting allpass filter, but I have some questions regarding this: Looking at the drawings from Linkwitz Labs below, it explains the calculations very well, still, the Ro resistors (see attached picture) from input to in-, and from in- to opamp output, does not come with an description.. Are they for gain? How should I calculate the values of these resistors?
Also a request for opinions: As the filter will only process high frequencies, i figure Vishays will perform nice in this circuit. Any thoughts on this/better suggestions?
Also a request for opinions: As the filter will only process high frequencies, i figure Vishays will perform nice in this circuit. Any thoughts on this/better suggestions?
Attachments
re; all - pass circuit in crossovers
I've tried, measured and built these circuits for use in crossovers and I still have questions!
The resistors Ro are equal, I used 10k, but any would do. If Ro is twice the value of R, then offset voltage due to input bias current will be minimised.
As an experiment, I wired a pot between the input and output, ie, in parallel with the two series Ro's. As the pot was rotated, the phase shift as measured on the pot wiper, went from zero when the wiper was at the input end, through 90 degrees with the wiper in the centre, to 180 degrees when the wiper was at the output end. This adjustment is independant of the change of frequency, or the values of R and C. Interesting, but what does it all mean?
I saw a circuit for a crossover that had a few of these connected in series, to compensate for the tweeter's source being a few cm in front of the woofer's, so providing a delay.There was no clear guide on calculating the R & C values for different delays. Having seen the phase shift of one, I could imagine that several in series would mangle the phase response of the speaker as frequency is increased, how can this be a good thing?
I too would be interested in hearing from someone who has used it effectively.🙂
I've tried, measured and built these circuits for use in crossovers and I still have questions!
The resistors Ro are equal, I used 10k, but any would do. If Ro is twice the value of R, then offset voltage due to input bias current will be minimised.
As an experiment, I wired a pot between the input and output, ie, in parallel with the two series Ro's. As the pot was rotated, the phase shift as measured on the pot wiper, went from zero when the wiper was at the input end, through 90 degrees with the wiper in the centre, to 180 degrees when the wiper was at the output end. This adjustment is independant of the change of frequency, or the values of R and C. Interesting, but what does it all mean?
I saw a circuit for a crossover that had a few of these connected in series, to compensate for the tweeter's source being a few cm in front of the woofer's, so providing a delay.There was no clear guide on calculating the R & C values for different delays. Having seen the phase shift of one, I could imagine that several in series would mangle the phase response of the speaker as frequency is increased, how can this be a good thing?
I too would be interested in hearing from someone who has used it effectively.🙂
Re: re; all - pass circuit in crossovers
Interresting indeed. Pot value from experiment?
BTW I would like to avoid many of these allpass networks in series... It wouldnt be fair to my speakers..
johnnyx said:As an experiment, I wired a pot between the input and output, ie, in parallel with the two series Ro's. As the pot was rotated, the phase shift as measured on the pot wiper, went from zero when the wiper was at the input end, through 90 degrees with the wiper in the centre, to 180 degrees when the wiper was at the output end. This adjustment is independant of the change of frequency, or the values of R and C. Interesting, but what does it all mean?
Interresting indeed. Pot value from experiment?
BTW I would like to avoid many of these allpass networks in series... It wouldnt be fair to my speakers..
The value of Ro is noncritical, but should be chosen within what's reasonable for the opamp. 10-20k would be a reasonable range.
Re: Re: re; all - pass circuit in crossovers
The pot value was not critical. I may have used 10k or 22k, It's driven by op amps, so if it doesn't load them too much it will be ok, like SY says.
In the design I was referring to, the author had to use a few to give the required delay. He did give an explanation, but I was unable to correlate the "time" with the "phase", because it is different for different frequencies. It didn't seem to correlate at the crossover frequency either, where it would have made sense.
I was using different speakers to him, so I needed a different delay, but I couldn't work it out from the information given.
It's operation remains a mystery.
Rocky said:
The pot value was not critical. I may have used 10k or 22k, It's driven by op amps, so if it doesn't load them too much it will be ok, like SY says.
In the design I was referring to, the author had to use a few to give the required delay. He did give an explanation, but I was unable to correlate the "time" with the "phase", because it is different for different frequencies. It didn't seem to correlate at the crossover frequency either, where it would have made sense.
I was using different speakers to him, so I needed a different delay, but I couldn't work it out from the information given.
It's operation remains a mystery.
Also if they are high quality op amps dont worry too much about the negative effect of the opamps in a long chain. Linkwitz does it I do it and As far as i can tell there is minimal if any degredation to the sound. Remember that the allpass is designed to make the speakers sound better not worse by inflicting opampy badness to the signal path!
One thing to mention is that an allpass may not be strictly necessary although it could help if you dont have measurint and simulation software. I have not had to use any of these in my actives but I have control over where I place the tweeter. But just 15mm back from the front, I dont know if this is time aligned and im not really bothered as the acoustic centre of the driver changes with frequency, but when simulating i have always been able to get good freq responses and good nulls when the polarity is reversed. It takes some experimenting and a few attempts at various topologies and slopes but I get there in the end.
The pic below is of my three way I was decribing before with the movable tweet.
One thing to mention is that an allpass may not be strictly necessary although it could help if you dont have measurint and simulation software. I have not had to use any of these in my actives but I have control over where I place the tweeter. But just 15mm back from the front, I dont know if this is time aligned and im not really bothered as the acoustic centre of the driver changes with frequency, but when simulating i have always been able to get good freq responses and good nulls when the polarity is reversed. It takes some experimenting and a few attempts at various topologies and slopes but I get there in the end.
The pic below is of my three way I was decribing before with the movable tweet.
Attachments
I forgot the point!
IMO unless you cant get good phase with two drivers I dont think an all pass is strictly nescessary.
IMO unless you cant get good phase with two drivers I dont think an all pass is strictly nescessary.
5th
Tip of the day if you don't mind.
Save your screen shots as .gif files. They will be a lot smaller and you won't have to shrink them and lose resolution! This tip works for graphs that have a solid background.
Tip of the day if you don't mind.
Save your screen shots as .gif files. They will be a lot smaller and you won't have to shrink them and lose resolution! This tip works for graphs that have a solid background.
Okedoke i will 🙂 i always did jpeg as whenever i did a pic as a gif it looked really poor, but this was proper pics and not simple graphs.
Another thing thats hard to assess is load times on a 56k i have the luxury of having a 512k line and cant really remember what 56k load times were like with respect to size, how long exactly does it take for a pic like the one i posted to come through?
Another thing thats hard to assess is load times on a 56k i have the luxury of having a 512k line and cant really remember what 56k load times were like with respect to size, how long exactly does it take for a pic like the one i posted to come through?
5th element said:
Certainly, I am aware that it may or may not have a beneficial effect at all, and I do not concider it a necessity. I have no expectations for it to reveal anything my speakers doesn't show already, but I would not be surprised if it did either. For the sake of it, I want to try it now that my crossovers finally are active 😉 See for myself, and make up my own opinion of it's usefullness or uselessness. It's not like it's a compex circuit anywayz, and I have extra OPA2134 chips laying around so... It's going to cost me the Vishays, thats it.
But instead of trying to pinpoint vibrating acoustic centres and such, which IMO cannot be done very well, what I want is a controller pot on my XO to shift phase, probably with a bypass switch, so I can use my ears for tweaking and experimenting..
Calculations of membran offsets, sound velocity, wavelength at XO, etc. are only to be sure my pot-controlled allpass will be able to cover the "right area". I think that's as good as I can do in such a "unprecise science" time aligning really is...
Tell me it's a stupid idea, and I would still go forward with it.. Out of curiosity.😉
If it appears not to be an improvement, it would still be very useful. I could then easily demonstrate the effect to others by a switch and a knob and see if they agree with my findings 😀
5th element said:
Photos are indeed best done with .jpg. That helps keep the file size down especially if they have some compression. Graphics are another matter.
Your picture file took about 45 seconds to load. Some of that was due to a slowish server having stalls.
Take this link and see the large screenshot graphic about halfway down the page. It's a .gif and only 24K.
http://www.kbacoustics.com/visualears/index.html
The reason the image file size is small is because there are large expanses of solid color.
Re: re; all - pass circuit in crossovers
So a simple pot replacing R will not do the trick? I should have a stereo (quad for both channels) pot on the Ro as well, to keep offset voltage to a minimum?
But hey.. now I'm thinking about putting new pots in my signal path
that sux. Good pots are soo expensive 🙁 Well, I guess I will not make this a permanent component in my system, so I'll go for cheapo components all over and do it as a test project.
If I find results pleasing, I'll just build a non-variable delay with my favourite setting from the test circuit!😀
johnnyx said:
So a simple pot replacing R will not do the trick? I should have a stereo (quad for both channels) pot on the Ro as well, to keep offset voltage to a minimum?
But hey.. now I'm thinking about putting new pots in my signal path
that sux. Good pots are soo expensive 🙁 Well, I guess I will not make this a permanent component in my system, so I'll go for cheapo components all over and do it as a test project.If I find results pleasing, I'll just build a non-variable delay with my favourite setting from the test circuit!😀
A constant delay is characterized by a phase shift whose value is proportional to the frequency
Each all-pass filter produces a constant time delay [dependent on RC] that is valid from (1/infinity)Hz up to the frequency where the phase shift reaches 90º. Then the phase shift is no longer proportional to the frequency and the delay progressively decreases for higher frequencies until it reaches no delay at infinity Hz [theoretically]
Look at this example with C=1n and R=10K. The graph shows the phase shift versus frequecy using a linear scale [not logarithmic]
F_90º would be 1/(2*pi*1n*10K) = 15.9Khz
And effectively, the graph shows a phase shift proportional to the frequency until 15.9Khz [cursor position] and then linearity is lost
How to calculate the effective delay in the linear range? We know that the filter produces 90º of true delay at 15.9Khz so this is a quarter wave length. The period of 15.9Khz is 62.8uS and the delay would be a quarter of that, 62.8/4=15.7uS
The sound travels through the air at about 345 m/s so the effective distance of that delay would be 345 * 15.7u = 5.42mm
To know the effective delay at any frequency F just calculate the slope of the phase shift at that frequency [in degrees/Hz], and then calculate : delay = slope/360
Sometimes you may need to cascade 4 or more allpass filters to get the required delay constant up to the required frequency
Each all-pass filter produces a constant time delay [dependent on RC] that is valid from (1/infinity)Hz up to the frequency where the phase shift reaches 90º. Then the phase shift is no longer proportional to the frequency and the delay progressively decreases for higher frequencies until it reaches no delay at infinity Hz [theoretically]
Look at this example with C=1n and R=10K. The graph shows the phase shift versus frequecy using a linear scale [not logarithmic]
F_90º would be 1/(2*pi*1n*10K) = 15.9Khz
And effectively, the graph shows a phase shift proportional to the frequency until 15.9Khz [cursor position] and then linearity is lost
How to calculate the effective delay in the linear range? We know that the filter produces 90º of true delay at 15.9Khz so this is a quarter wave length. The period of 15.9Khz is 62.8uS and the delay would be a quarter of that, 62.8/4=15.7uS
The sound travels through the air at about 345 m/s so the effective distance of that delay would be 345 * 15.7u = 5.42mm
To know the effective delay at any frequency F just calculate the slope of the phase shift at that frequency [in degrees/Hz], and then calculate : delay = slope/360
Sometimes you may need to cascade 4 or more allpass filters to get the required delay constant up to the required frequency
Attachments
It does seem to me to be a rather hit and miss thing this. Applying a certain amount of time delay at one frequency could be benaficial but by the nature of how the circuit works the phase is not constant with frequency although the drivers acoustic centre is not fixed with frequency either. You could improve the time alignment at the xover freq but infact make it worse at others.
Im not at all condeming the use of these as they can be very useful if a good phase response cannot be had from a certain set of drive units on a baffle.
Im rather surprised at Mr Linkwitz statement that active xovers are only marginally useful without delay stages etc. I dont understand this whatsoever.
Im not at all condeming the use of these as they can be very useful if a good phase response cannot be had from a certain set of drive units on a baffle.
Im rather surprised at Mr Linkwitz statement that active xovers are only marginally useful without delay stages etc. I dont understand this whatsoever.
- First step : Find out the real-world phase difference [on axis] between drivers in their crossover region [there should be a pure delay component plus other phase shift artifacts]
- Second step : Compensate this phase shift moving the drivers if possible, adjusting the crossover and using all-pass filters to get the best summing on axis or the desired directivity pattern
As Mr. Linkwitz states, all-pass filters are required since you will never be able to add the required compensation delays and phase shifts without affecting frequency response by means of only low-pass and high-pass filters [crossover]
'Acoustic center' and 'time alignment' concepts are only valid for ideal point source drivers and are useless in the real world. Real drivers have infinite effective acoustic centers depending on frequency and axis
- Second step : Compensate this phase shift moving the drivers if possible, adjusting the crossover and using all-pass filters to get the best summing on axis or the desired directivity pattern
As Mr. Linkwitz states, all-pass filters are required since you will never be able to add the required compensation delays and phase shifts without affecting frequency response by means of only low-pass and high-pass filters [crossover]
'Acoustic center' and 'time alignment' concepts are only valid for ideal point source drivers and are useless in the real world. Real drivers have infinite effective acoustic centers depending on frequency and axis
Thanks Eva!
I can only say I agree with you 5th. I remember reading the same statement from him. With all due respect, I guess mr. Inventor himself has a somewhat coloured point of view on his own inventions, don't you think?
...And thanks Eva! your comments was really useful!
5th element said:
I can only say I agree with you 5th. I remember reading the same statement from him. With all due respect, I guess mr. Inventor himself has a somewhat coloured point of view on his own inventions, don't you think?
...And thanks Eva! your comments was really useful!
It makes sense that somone who deisngs their own things as in the delay stage, they are not exaclty going to go hey look at this circuit i designed, dont it look pretty, but its entirely useless! I can see him saying these little ditties are useful in certain applications, but not go as far as saying active xovers are as good as bad if you dont use one.
SL is somewhat cryptic about how he uses the allpass filters but it's all there if you study his examples. Things I've noted:
He uses the same value for all the resistors. R and Ro are the same.
The pic Rocky posted shows the group delay calcs. At DC it's 2RC. At Fo, the pole frequency of the allpass, it's RC and, at the crossver frequency Fc, it's 2RC/(1+(Fc/Fo)^2).
You want to keep Fo well above Fc so you stay on the flat left part of the group delay curve in Rocky's pic but not so far above that you have to use too many stages - it's a tradeoff.
With multiple stages, SL alternates the 0 to -180 and -180 to -360 configs.
Here's a snippet from the Phoenix design showing a practical example how it works.
His crossover frequency is 1440Hz and the design goal is to delay the tweeter 190 usec at the 1440Hz Fc.
First stage:
0 to -180 phase shift
Fo 3052 Hz
DC delay 104 usec
Delay at 1440Hz 85 usec
Second stage:
-180 to -360 phase shift
Fo 2035 Hz
DC delay 156 usec
Delay at 1440Hz 104 usec
Total delay at 1440Hz is 85 + 104 = 189 usec. Pretty close to the design goal.
He uses the same value for all the resistors. R and Ro are the same.
The pic Rocky posted shows the group delay calcs. At DC it's 2RC. At Fo, the pole frequency of the allpass, it's RC and, at the crossver frequency Fc, it's 2RC/(1+(Fc/Fo)^2).
You want to keep Fo well above Fc so you stay on the flat left part of the group delay curve in Rocky's pic but not so far above that you have to use too many stages - it's a tradeoff.
With multiple stages, SL alternates the 0 to -180 and -180 to -360 configs.
Here's a snippet from the Phoenix design showing a practical example how it works.
An externally hosted image should be here but it was not working when we last tested it.
His crossover frequency is 1440Hz and the design goal is to delay the tweeter 190 usec at the 1440Hz Fc.
First stage:
0 to -180 phase shift
Fo 3052 Hz
DC delay 104 usec
Delay at 1440Hz 85 usec
Second stage:
-180 to -360 phase shift
Fo 2035 Hz
DC delay 156 usec
Delay at 1440Hz 104 usec
Total delay at 1440Hz is 85 + 104 = 189 usec. Pretty close to the design goal.
- Status
- Not open for further replies.