(moderator, please move this if it is in the wrong forum)
Hello all. Wondering if those in the know can confirm something. Is there any other [ or better ] way to produce an ANALOG time delay stage then to use a Bessel Low Pass filter with a cutoff frequency well above the upper bandwidth point of the driver?
Think trying to time-align a tweeter horn to a mid-horn where the two are about two feet apart [Klipsch La Scala].
As I'm building a custom constant-current gainclone tri-amp for it, and will be making extensive measurements of each horn+driver in the speaker with a sophisticated tool I have, I'll be able to determine the exact time-alignment difference between the tweeter/mid/bass horns. I wish to design in some compensation into my design ... without resorting to digitizing and using a DSP.
Is making use of the uniform group delay provided by a Bessel filter circuit stage, and cascading them as needed until I reach the desired delay, the only/best solution in the "analog" domain?
Thanx much for your thoughts.
Hello all. Wondering if those in the know can confirm something. Is there any other [ or better ] way to produce an ANALOG time delay stage then to use a Bessel Low Pass filter with a cutoff frequency well above the upper bandwidth point of the driver?
Think trying to time-align a tweeter horn to a mid-horn where the two are about two feet apart [Klipsch La Scala].
As I'm building a custom constant-current gainclone tri-amp for it, and will be making extensive measurements of each horn+driver in the speaker with a sophisticated tool I have, I'll be able to determine the exact time-alignment difference between the tweeter/mid/bass horns. I wish to design in some compensation into my design ... without resorting to digitizing and using a DSP.
Is making use of the uniform group delay provided by a Bessel filter circuit stage, and cascading them as needed until I reach the desired delay, the only/best solution in the "analog" domain?
Thanx much for your thoughts.
There is a thing called an allpass filter which changes the phase and can be used for time delay Mentioned on linkwitz's site here:
http://www.linkwitzlab.com/filters.htm#4
To get the delay you want, you would have to cascade many such stages, its probably worth reconsidering your decision not to go digital.
http://www.linkwitzlab.com/filters.htm#4
To get the delay you want, you would have to cascade many such stages, its probably worth reconsidering your decision not to go digital.
Nice Link! Thank You.
If I have to I may. I'll give it some time trying analog but digital may end up the case.
Is it true that time-alignement really only audibly exibits itself in the cross-over region?
P.S. am I making an incorrect assumption that Siegfried Linkwitz in the same Linkwitz as in Linkwitz-Riley?
If I have to I may. I'll give it some time trying analog but digital may end up the case.
Is it true that time-alignement really only audibly exibits itself in the cross-over region?
P.S. am I making an incorrect assumption that Siegfried Linkwitz in the same Linkwitz as in Linkwitz-Riley?
Nice ... take a look @ all those extra sections on the tweeter:
http://www.linkwitzlab.com/images/graphics/38xo_eq1.gif
I read on another site how using a Linkwitz/Riley without delay compensation negates benefits of using Linkwitz/Riley. I wasn't sure how "right" that was though, or if I could rely on that (or quote it ever if need be). I see now under the "Delay Correction" item on the link above that I can! "Active crossover circuits that do not include phase correction circuitry are only marginally useable. "
There may be hope for the analog route yet....
This page is useful for someone that's been out of this game for decades like me, annotating all the sections of the complete x-over circuit:
http://www.linkwitzlab.com/xo_eq.htm
consort_ee_um, this circuit has -everything- I need as an x-over for my constant-current tri-amp project I'm embarking on for my Klipsch La Scalas (the LM4780TAs arrived yesterday
), including delay compensation and selectable tilt control. I could not be more jazzed! Thank you. And thank you Mr Siegfried Linkwitz if you read this board!
http://www.linkwitzlab.com/images/graphics/38xo_eq1.gif
I read on another site how using a Linkwitz/Riley without delay compensation negates benefits of using Linkwitz/Riley. I wasn't sure how "right" that was though, or if I could rely on that (or quote it ever if need be). I see now under the "Delay Correction" item on the link above that I can! "Active crossover circuits that do not include phase correction circuitry are only marginally useable. "
There may be hope for the analog route yet....
This page is useful for someone that's been out of this game for decades like me, annotating all the sections of the complete x-over circuit:
http://www.linkwitzlab.com/xo_eq.htm
consort_ee_um, this circuit has -everything- I need as an x-over for my constant-current tri-amp project I'm embarking on for my Klipsch La Scalas (the LM4780TAs arrived yesterday
), including delay compensation and selectable tilt control. I could not be more jazzed! Thank you. And thank you Mr Siegfried Linkwitz if you read this board!m8o said:(moderator, please move this if it is in the wrong forum)
Hello all. Wondering if those in the know can confirm something. Is there any other [ or better ] way to produce an ANALOG time delay stage then to use a Bessel Low Pass filter with a cutoff frequency well above the upper bandwidth point of the driver?
Think trying to time-align a tweeter horn to a mid-horn where the two are about two feet apart [Klipsch La Scala].
As I'm building a custom constant-current gainclone tri-amp for it, and will be making extensive measurements of each horn+driver in the speaker with a sophisticated tool I have, I'll be able to determine the exact time-alignment difference between the tweeter/mid/bass horns. I wish to design in some compensation into my design ... without resorting to digitizing and using a DSP.
Is making use of the uniform group delay provided by a Bessel filter circuit stage, and cascading them as needed until I reach the desired delay, the only/best solution in the "analog" domain?
Thanx much for your thoughts.
There is a variant response-type, called "Linear Phase .05-degree Equiripple Error", which can be implemented with allpass, bandpass, lowpass, or highpass characteristics, using many different analog filter circuit topologies.
One trick is to vary the number of poles, so that you can simulataneously get the cutoff frequency you want (or phase-reversal freq, for allpass) AND the group delay that you want. i.e. Using a higher-order filter (more poles) can allow getting a higher cutoff frequency while still increasing group delay.
You can download the free FilterPro software, from TI's website. It lets you specify the response type, frequency, and filter topology type, et al, and then designs an opamp-based filter for you and plots the amplitude, phase, and group delay. Very handy!
I recommend that you also use LT-Spice (free from http://www.linear.com) to simulate the resulting designs, since you can get much better accuracy from the plots, that way.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
A three-way horn speaker cannot be aligned with all-pass filters.
At the risk of repeating myself:
A three-way horn speaker cannot be aligned with all-pass filters.
Why?
The tweeter needs the most delay, then the mid. The time delay is proportional to the frequency. For the high tweeter crossover points you will need so many opamps for the delay that you will have a sea of noise.
Buy a digital crossover.
At the risk of repeating myself:
A three-way horn speaker cannot be aligned with all-pass filters.
Why?
The tweeter needs the most delay, then the mid. The time delay is proportional to the frequency. For the high tweeter crossover points you will need so many opamps for the delay that you will have a sea of noise.
Buy a digital crossover.
m8o You are correct about the phase being important. When you read the works of the John K's (Krutke and Krekovsky) on passive crossovers you get the importance of phase tracking of the different drivers at the crossover frequency. Most active crossover people seem to ignore this see for example:
http://www.diyaudio.com/forums/showthread.php?threadid=6655
where until the fortuitously named "Phase accurate" comes in with the last post it is not mentioned.
http://www.diyaudio.com/forums/showthread.php?threadid=6655
where until the fortuitously named "Phase accurate" comes in with the last post it is not mentioned.
Yes, I'll need 6 - 8 opamps in all-pass ; I have to put it to the LTSpice to see for sure (which is the least # of opamps; other filter topologies mentioned give less delay per section). I appreciate the warning about noise. I was wondering about that. I may just try it anyway to learn using low-noise opamps; hey, it might work after all.djk said:
...the procrastinator might get the best of me though with that thought nagging in my mind.Do you have any suggestions for any monolithic single chips I can build a circuit around, which has a self-contained A/D,X-over,D/A built into one device?
Circuit no 4 here might be the basis for your design:
http://home.comcast.net/~neilrdavis/Plate_amps/
It uses the TAS3004:
The TAS3004 device is a system-on-a-chip that replaces conventional analog equalization to perform digital parametric equalization, dynamic range compression, and loudness contour. Additionally, this device provides high-quality, soft digital volume, bass, and treble control. All control parameters are uploaded through the I2C port
from an outside MCU through the I2C slave port or from an external EPROM through the I2C master port.
The TAS3004 device also has an integrated 24-bit stereo codec with two I2C-selectable, single-ended inputs per
channel.
The digital parametric equalization consists of seven cascaded, independent biquad filters per channel. Each biquad filter has five 24-bit coefficients that can be configured into many different filter functions (such as bandpass, high pass, and low pass).
http://home.comcast.net/~neilrdavis/Plate_amps/
It uses the TAS3004:
The TAS3004 device is a system-on-a-chip that replaces conventional analog equalization to perform digital parametric equalization, dynamic range compression, and loudness contour. Additionally, this device provides high-quality, soft digital volume, bass, and treble control. All control parameters are uploaded through the I2C port
from an outside MCU through the I2C slave port or from an external EPROM through the I2C master port.
The TAS3004 device also has an integrated 24-bit stereo codec with two I2C-selectable, single-ended inputs per
channel.
The digital parametric equalization consists of seven cascaded, independent biquad filters per channel. Each biquad filter has five 24-bit coefficients that can be configured into many different filter functions (such as bandpass, high pass, and low pass).
"Yes, I'll need 6 - 8 opamps in all-pass "
Run the numbers. For a 4Khz tweeter crossover you will need about 36" of delay. 1130 X 12 = 13560 inches, divided by 8000hz is 1.695 inches. The best you can get is 180°, so half of 1.695 is 0.8475 inches. 36 divided by 0.8475 is 42 opamps.
With a frequency of 8Khz, the 4Khz crossover will be delayed properly, but will revert back to 36mS delay above there. It would be better to select 16Khz for the delay frequency, but that would require 84 opamps.
You could say the heck with the tweeter, and just delay the mid. That would only require about 12 opamps for 4Khz and the offset of the mid horn.
Considering that you can buy a used digital crossover for about $150, is it worth going your route?
Run the numbers. For a 4Khz tweeter crossover you will need about 36" of delay. 1130 X 12 = 13560 inches, divided by 8000hz is 1.695 inches. The best you can get is 180°, so half of 1.695 is 0.8475 inches. 36 divided by 0.8475 is 42 opamps.
With a frequency of 8Khz, the 4Khz crossover will be delayed properly, but will revert back to 36mS delay above there. It would be better to select 16Khz for the delay frequency, but that would require 84 opamps.
You could say the heck with the tweeter, and just delay the mid. That would only require about 12 opamps for 4Khz and the offset of the mid horn.
Considering that you can buy a used digital crossover for about $150, is it worth going your route?
The means of building a frequency independant delay is a Bessel allpass filter but as some already mentioned it needs a lot of opamps to achive if this is to be done for large delays and high bandwidth.
So the most practical approach for doing time-alignment with allpass filters is either small two-way speakers (cone & dome) or two-way with direct radiating woofer and horn tweeter.
As soon as you wan to align bass-horns your best bet is a digital crossover.
OTOH phase-accuracy was never the primary reason for using a horn speaker so you are maybe better off using a crossover that is amphasizing the horn advantages the most.
Regards
Charles
So the most practical approach for doing time-alignment with allpass filters is either small two-way speakers (cone & dome) or two-way with direct radiating woofer and horn tweeter.
As soon as you wan to align bass-horns your best bet is a digital crossover.
OTOH phase-accuracy was never the primary reason for using a horn speaker so you are maybe better off using a crossover that is amphasizing the horn advantages the most.
Regards
Charles
I'm not sure, just an ideea: using an analog delay line based on spings&mases, etc.? Maybe those used in pedal effect by old quitarists:
http://ccrma.stanford.edu/~jos/pasp/Delay_Lines.html
Or a tape loop?
See also this about the BBD's:
http://www.du.edu/~etuttle/electron/elect39.htm
http://ccrma.stanford.edu/~jos/pasp/Delay_Lines.html
Or a tape loop?
See also this about the BBD's:
http://www.du.edu/~etuttle/electron/elect39.htm
...and the output is usable to 20-30kHz. See here:
http://www.rhombus-ind.com/dlcat/app1_pas.pdf
for design data.
http://www.rhombus-ind.com/dlcat/app1_pas.pdf
for design data.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.