Hi,
TI has appnote for building an active 2 way here https://www.ti.com/lit/ug/tidu035/tidu035.pdf
For time alignment it proposes a delay all-pass filter circuit to delay the tweeter, the delay is a fixed at 155us till 3.8Khz upper cutoff. Please see section 2.4 and 4.2.3.
Can anyone please explain whether this delay is really fixed at 155us from 20Hz to 3.8Khz?
I thought an all-pass filter give a constant delay in terms of phase angles and we know that a constant phase angle (in a band of frequencies) gives varying time delay, not fixed. But the proposed circuit is claimed to have a fixed time delay, how?
Thanks and Regards,
wondeflaudio
TI has appnote for building an active 2 way here https://www.ti.com/lit/ug/tidu035/tidu035.pdf
For time alignment it proposes a delay all-pass filter circuit to delay the tweeter, the delay is a fixed at 155us till 3.8Khz upper cutoff. Please see section 2.4 and 4.2.3.
Can anyone please explain whether this delay is really fixed at 155us from 20Hz to 3.8Khz?
I thought an all-pass filter give a constant delay in terms of phase angles and we know that a constant phase angle (in a band of frequencies) gives varying time delay, not fixed. But the proposed circuit is claimed to have a fixed time delay, how?
Thanks and Regards,
wondeflaudio
Last edited:
https://www.linkwitzlab.com/filters.htm
The delay correction schematic shows that the delay T0, is only dependent on R & C so is constant until Fp.
The delay correction schematic shows that the delay T0, is only dependent on R & C so is constant until Fp.
The delay is only approximately constant, but it can be a pretty good approximation up to a certain frequency when you make an all-pass with the pole positions of a Bessel or a 0.05 degree linear phase low-pass filter.
Since I have no simulation knowledge, I would appreciate if someone simulated it and confirmed. Thanks.
@batteryman, what is the definition of fp? Edit: ok got confused by use of F0 and Fp in linkwitz article. Got it now.
@MarcelvdG i don't get the meaning of 'the pole position of a bessel or a 0.05 degree linear phase low pass filter'. Could you explain it an other way or point to some explanation please?
I could be interested in a tutorial to perform a simulation about this if anyone have time to spend. I've tryed with Ltspice some years ago but never get it to work past implementing schematic...and moved to other things out of frustration.
@MarcelvdG i don't get the meaning of 'the pole position of a bessel or a 0.05 degree linear phase low pass filter'. Could you explain it an other way or point to some explanation please?
I could be interested in a tutorial to perform a simulation about this if anyone have time to spend. I've tryed with Ltspice some years ago but never get it to work past implementing schematic...and moved to other things out of frustration.
Last edited:
Marcel, no.
Something like a drawing i usually understand.
Here is how i understand what you wrote: using a bessel alignement for the allpass you have more or less a constant delay (up to fo or fp?).
The 0,05 degree linear phase lpf is a kind/type of filter profile ?
Something like a drawing i usually understand.
Here is how i understand what you wrote: using a bessel alignement for the allpass you have more or less a constant delay (up to fo or fp?).
The 0,05 degree linear phase lpf is a kind/type of filter profile ?
Ok it's way clearer. Thank you!
Just found an article describing the linear phase filter family. There is ripple like a Chebichev.
Just found an article describing the linear phase filter family. There is ripple like a Chebichev.
Marcelvdg,
I've got another question: is there a rule of thumb to define the range needed (for a given hp order) to go from Rc to 2RC ( in a drawing describing phase change from Fp/F0 to the point where the drawing start to be horizontal at 2rc)?
A bit like the rule of thumb of 2 octave above fc in a 2pole hp should be on a constant phase behavior.
I've got another question: is there a rule of thumb to define the range needed (for a given hp order) to go from Rc to 2RC ( in a drawing describing phase change from Fp/F0 to the point where the drawing start to be horizontal at 2rc)?
A bit like the rule of thumb of 2 octave above fc in a 2pole hp should be on a constant phase behavior.
Analog all-pass delay is typically embodied as either a first order or a second order AP circuit. Neither has a constant delay for all frequencies. Instead, there is a delay at DC, and then the delay changes (generally decreasing in some way, eventually falling to zero) as frequency increases. For first order this is just prescribed by the time constant, which is analogous to Fs in the frequency domain. For second order there is a "quality factor" to the circuit, which changes both the delay at DC and the shape of the delay above DC. When Q~0.577 the delay curve is maximally flat IIRC, starting to roll off and falling back to zero around Fs. Higher Q values cause delay peaking around Fs as well as a lower DC delay. I wrote an Excel program to design analog AP delay stages at one time that I could make available if anyone is interested.
Edit: I found a screenshot I made of the spreadsheet input page. See attached.
Edit: I found a screenshot I made of the spreadsheet input page. See attached.
Attachments
Last edited:
Yes it makes sense that at 0.56 it makes maximally flat. It's the same as Qtc for a closed cabinet ( overdamped).
CharlieLaub i would be interested in your Excel sheets!
CharlieLaub i would be interested in your Excel sheets!
Hmmm, maybe you are thinking of the second order Bessel response, with Q=0.577? I think you are right about that. I never made the connection.
Yes. It's all related, a sealed box is a second order filter too.
Many thanks for the excel sheet. I'm visual with this things. It's a great help.
Last edited:
Apologies if this sounds like abracadabra, but you can derive an all-pass filter from an all-pole low-pass filter by adding right half plane zeros that are at the positions of the poles mirrored in the imaginary axis. The magnitude response then becomes flat, as it should for an all-pass, and the phase shift doubles.
That is, when you have an all-pole low-pass filter (basically any normal low-pass without notches), you can derive an all-pass filter that has exactly twice its delay. Hence, the Q value that makes the group delay of a second-order low-pass maximally flat, also makes the group delay of the corresponding all-pass maximally flat.
That is, when you have an all-pole low-pass filter (basically any normal low-pass without notches), you can derive an all-pass filter that has exactly twice its delay. Hence, the Q value that makes the group delay of a second-order low-pass maximally flat, also makes the group delay of the corresponding all-pass maximally flat.
