AndrewT said:
There's no such thing as a Linkwitz-Riley all-pass. So you've got me wondering now. The Linkwitz-Riley is a high or low pass filter network.
I guess the clue is the wording of the article "Active All-Pass Crossover Networks"
It is rather confusing to convention wisdom on the subject. All pass says to me; delays and shelves. Whilst crossover networks suggests high, low and bandpass.
Perhaps this is an all pass network that simply is part of a 'crossover' network?
Found a little bit:
From here:
http://www.aes.org/journal/toc/AES-JanFeb2006TOC.cfm
Looks to be a crossover filter from the description.
From here:
http://www.aes.org/journal/toc/AES-JanFeb2006TOC.cfm
Looks to be a crossover filter from the description.
All capacitors and resistors the same hey. That's easy, just use the state variable Linkwitz-Riley design as described by the Rane Corporation here http://www.rane.com/pdf/linriley.pdf
Features:
Fewer resistors and caps if you want to make an adjustable crossover.
Less opamps than say the Rod Elliot Linkwitz-Riley.
High and low outputs from a single "stage" that have identical crossover frequencies. Again nice for an adjustable.
I have only just started to think about active crossovers but I don't see what the problem with phase might be with this. High and low outputs are in phase to withing few milli degrees at all frequencies.
So far I have simulated and built the low and mid section as in the attched schematic.
Features:
Fewer resistors and caps if you want to make an adjustable crossover.
Less opamps than say the Rod Elliot Linkwitz-Riley.
High and low outputs from a single "stage" that have identical crossover frequencies. Again nice for an adjustable.
I have only just started to think about active crossovers but I don't see what the problem with phase might be with this. High and low outputs are in phase to withing few milli degrees at all frequencies.
So far I have simulated and built the low and mid section as in the attched schematic.
Attachments
As some posters already mentioned - an allpass (alone) can't act as a crossover.
But a crossover often acts as an allpass. An LR 4 for instance has a constant summed output voltage but turns the phase by 360 degrees.
Crossovers that behave this way are often called allpass crossovers.
Transient-perfect crossovers (like the ones the poster above deals with on his homepage) are often just called constant voltage crossovers though this naming is a little inexact.
The crossover in question turns the phase by 360 degrees in total for a three-way crossover i.e. half the amount of the 720 degrees caused by a three-way LR-4.
Regards
Charles
But a crossover often acts as an allpass. An LR 4 for instance has a constant summed output voltage but turns the phase by 360 degrees.
Crossovers that behave this way are often called allpass crossovers.
Transient-perfect crossovers (like the ones the poster above deals with on his homepage) are often just called constant voltage crossovers though this naming is a little inexact.
The crossover in question turns the phase by 360 degrees in total for a three-way crossover i.e. half the amount of the 720 degrees caused by a three-way LR-4.
Regards
Charles
All pass crossovers are a general family of crossover, where typically the low pass is derived by subtracting a high pass from an all pass with the same phase response.
The LR is the only one that doesn't create a zero, and maintains full roll off out of bounds. This feature is realized by the fact that LR uses Butterworth (squared) coefficients. The root two in there causes the zeros to cancel.
In 1988 I derived all pass crossovers that used Bessel coefficients instead of Butterworth. The roll of rate of the low pass is reduced as zeroes exist, but the group delay is better since a Bessel all pass is used as the basis.
IME, the best trade off was to normalize Bessel and Butterworth poles to derive the all pass and high pass function. This alignment provides better group delay than LR, sums to all pass, has good lobing behaviour, and the zero created is far enough out that you still get good roll off through the crossover region.
I've posted about this xover for about the last 14 years. I hope they aren't applying for a patent.
Dave
PS The fact that a zero is created when Butterworth isn't used as the basis drives the better system trade off to create low from high pass. A reduced tweeter roll off rate has a larger impact on system distortion.
The LR is the only one that doesn't create a zero, and maintains full roll off out of bounds. This feature is realized by the fact that LR uses Butterworth (squared) coefficients. The root two in there causes the zeros to cancel.
In 1988 I derived all pass crossovers that used Bessel coefficients instead of Butterworth. The roll of rate of the low pass is reduced as zeroes exist, but the group delay is better since a Bessel all pass is used as the basis.
IME, the best trade off was to normalize Bessel and Butterworth poles to derive the all pass and high pass function. This alignment provides better group delay than LR, sums to all pass, has good lobing behaviour, and the zero created is far enough out that you still get good roll off through the crossover region.
I've posted about this xover for about the last 14 years. I hope they aren't applying for a patent.
Dave
PS The fact that a zero is created when Butterworth isn't used as the basis drives the better system trade off to create low from high pass. A reduced tweeter roll off rate has a larger impact on system distortion.
Technical articles on Journal of the Audio Engineering Society are copy righted. They want $25 per copy from non-member. If you find the article in a nearby technical library (university related, etc.), you can copy it for personal use. But do not post it at the forum here.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.