I had not looked at it this way before, Thanks for making this clearer. Very helpful. As always you are
In this case, being a 4 ways and not a 3, it makes it a bit easier and less bandwidth is demanded from each driver.
And my low-mid driver is a bit special, although it's a 15" that could pass as a woofer, it does have something fairly unusual, with a bandwidth extending to 6khz. Its usable response is 50hz-6khz, and an amazing efficiency of 105db@1w@1m. It sounds great and climbs quite high. I don't really want to use its upper bandwidth too much, but it's good it's there. I have a horn above it with a 1" driver that goes down to 800hz and up to 20khz. The recommended crossover is 800hz, but I want to cut it a little higher, and I'm thinking 1.5khz, since I do have that nice low-mid below, and it won't have to climb all the way up because I have a super tweeter above that.
My tweeter has a usable bandwidth that goes from 6.5khz all the way beyond 20khz, with the recommended crossover at 7khz, and I was thinking about putting that at 8khz.
Even the big 18" woofer for the low can reach 2500hz, and I won't be getting anywhere near that by cutting it at 150hz.
A 4th order is better than a 3rd anyway, and the steeper roll offs can only help.
50-6000 for the low-mid isn't wide enough you think?
I haven't done the calculations for that, but my drivers seem to have wide enough usable bandwidth for the job.
Besides, if I couldn't get that done with those drivers, then there is hardly anything available on the market that would anyway.
Well, what about those doing it with 3 ways only? Even worse, 2 ways?
If I can't do it well enough with what I have in 4 ways, then who can?
Well I do have all my drivers and the cabinets have been operational for years. This is far from being a new project. I'm just trying now to get closer to completion.
What corrections would you think are needed? Any suggestions?
That would be a stage after the limiter then, wouldn't it?
I haven't locked in a choice for that, I was only thinking I would make use of the limiter's VCA to double up as the level adjustment.
Do you have any specific circuit to suggest?
Maybe not "audible", but surely measurable, at least in simulations.
Working on the power amps simulations, that input coupling cap always was a good source of degradation.
Definitely, but minimizing the occurrences of coupling caps should be a plus. I'm not saying eliminate them all, but just to keep them to a strict minimum.
This will likely be part of the discussion on this thread. The calculations for the caps and resistors to obtain the xover frequencies won't be too hard, and the only hard part will be tweaking that to stick with values that we can get. It will take some compromises there, and it's obvious the targeted frequencies will get missed a little, but not too much, if we use tricks like caps in parallel or series, and same thing with resistors, and making use of the e192 series or something...
Reading Self already... I'm making my way through "The Design of Active Crossovers" right now. A lot to absorb and I'm not at my best at all, but some of it should bear fruits...
Good reference as well.
I won't be doing it quite the same because this one doesn't make use of the all-pass with a substraction for each band, but the calculations should be just about the same.
The hard part may not be the filter itself but rather all the fluff around it, like the limiter, level adjustment, signal presence detection on the input, etc...
I have been going through a nice bunch of possible options for the opamps, and right now as a price/performance compromise, I am leaning towards the OPA134.
And I will likely use the INA134 for the balanced line receiver, where I will need to figure out how to do the input signal detection without interfering and deteriorating the quality.
That AD797 surely is looking quite impressive, and so is its price!
When I opened up my mix table to look how it was built and what was used inside, I saw that most opamps are of the TL071 type (and their dual and quad versions of course).
That mixer has been sturdy and always gave good sound over the years. It was designed in the 80s (early), and it's still a good one nowadays.
That is what I will continue using, with an equalizer on its output, then balanced lines going to the 2 amp racks.
The long balanced lines might reach in the 20m range, perhaps even a bit more, but this should be fine with the line drivers nowadays.
Like I said, this is no living room hifi, and it does need to go outside...
That mixer has been sturdy and always gave good sound over the years. It was designed in the 80s (early), and it's still a good one nowadays.
That is what I will continue using, with an equalizer on its output, then balanced lines going to the 2 amp racks.
The long balanced lines might reach in the 20m range, perhaps even a bit more, but this should be fine with the line drivers nowadays.
Like I said, this is no living room hifi, and it does need to go outside...
Regarding opamps, I quite like Nwavguy's test results and conclusions:
NwAvGuy: Op Amp Measurements
NwAvGuy: Op Amp Measurements
Duly noted!
And I guess the same kind of comparisons can be done in this design.
I am starting simulations. For now with a model of a OPA134 that I found, and on a single cell of filter, a 1st order, the opamps is basically just a follower...
I made a quick small excel sheet to make quick R/C calculations.
Using this formula:
F = ( 1 + sqrt(2) ) / ( 2 * PI * R * C )
I don't know what parasitics you are including in your capacitor model for your simulations.
A coupling capacitor has no AC audio signal across it, when selected correctly.
Thus it cannot contribute any audio distortion to that signal.
Your simulation should prove this assertion to be correct.
If you select the wrong capacitor, either by value or by type, then you may find the coupling capacitor does develop some audio signal voltage across it. Only then will the parasitics included in your model start to show some distortion of that audio signal.
There are measurements of the added distortion due to filter capacitors, but that ONLY applies to capacitors applied as filters. Coupling capacitors don't fall into that category.
I found this formula in a filter description in an article and the topology has 4 1st order filters in cascade to obtain the 4th order linkwitz.
But now I think I will switch to cascading only 2 2nd order filters instead, which saves a lot on opamps and simplifies a bit the circuit.
For that topology, I am looking at an other description in an other article, unrelated, and the top factor is a bit different.
I am attaching a picture of the simulation that I'm putting together which is for one 1st order cell only. I am rusty with ltspice and going through a re-learning curve, so I don't know yet exactly how to do all of the proper simulations.
That formula with that top factor including a 1+ was what I found in that article, where they used those 1st order cascaded 4 times.
The design I will work on next will be 2nd order cells, and having 2 sets of RC, there are 2 formulas, one with the top factor being sqrt(2), while the other is sqrt(2)/2
I am no math wiz, not even close, so if I'm mistaken or am using formulas that are wrong, this is the time to correct it.
There was an article in an old elektor from april 1987, which I don't have, and I would like to find it. If someone has that old elektor, I would like to get that please.
But now I think I will switch to cascading only 2 2nd order filters instead, which saves a lot on opamps and simplifies a bit the circuit.
For that topology, I am looking at an other description in an other article, unrelated, and the top factor is a bit different.
I am attaching a picture of the simulation that I'm putting together which is for one 1st order cell only. I am rusty with ltspice and going through a re-learning curve, so I don't know yet exactly how to do all of the proper simulations.
That formula with that top factor including a 1+ was what I found in that article, where they used those 1st order cascaded 4 times.
The design I will work on next will be 2nd order cells, and having 2 sets of RC, there are 2 formulas, one with the top factor being sqrt(2), while the other is sqrt(2)/2
I am no math wiz, not even close, so if I'm mistaken or am using formulas that are wrong, this is the time to correct it.
There was an article in an old elektor from april 1987, which I don't have, and I would like to find it. If someone has that old elektor, I would like to get that please.
Attachments
I switched to the 2nd order filter in a simulation (see attached)
And this time the formula reads:
F = 1 / ( 2 * PI * R * C )
using 10k and 100nf for the R and C, the simulation result looks like it agrees with this calculation.
I don't know how to add vertical and horizontal lines on a plot to point to the -6db in the response, so I just pointed manually to it. It's not right on the spot, but close enough to see that the frequency at that point is about 160hz.
And that's basically what I got in the excel sheet with this formula. So I suppose this is correct.
Someone can tell me how to add lines on a plot?
And this time the formula reads:
F = 1 / ( 2 * PI * R * C )
using 10k and 100nf for the R and C, the simulation result looks like it agrees with this calculation.
I don't know how to add vertical and horizontal lines on a plot to point to the -6db in the response, so I just pointed manually to it. It's not right on the spot, but close enough to see that the frequency at that point is about 160hz.
And that's basically what I got in the excel sheet with this formula. So I suppose this is correct.
Someone can tell me how to add lines on a plot?
Attachments
The graph you have attached shows lines in both directions.
Can't you see them?
The -6dB looks ~160Hz and when you go down to 90° phase it too looks about 160Hz.
That C in the RC filter is acting as a filter capacitor. it is not a coupling capacitor.
Using a poor capacitor will add distortion.
The resistor and/or capacitor ratios are not correct.
If you want a Butterworth roll-off you need a Q=1/sqrt(2) to create a B2
Two B2 cascaded give an LR4 for a final Q=0.5 and phase at -6dB = 90°
You are using the wrong formula for the 2pole filter. You have shown the single pole RC filter formula, as in post30.
Spice has used the wrong values you inserted in there to create a Q=0.5 filter and that has a ~160Hz -6 dB point but the roll off shape/rate is wrong. It is not Butterworth.
I still don't see where you got that 1+sqrt(2) factor, I think you misread the source doc, or the source was wrong.
Can't you see them?
The -6dB looks ~160Hz and when you go down to 90° phase it too looks about 160Hz.
That C in the RC filter is acting as a filter capacitor. it is not a coupling capacitor.
Using a poor capacitor will add distortion.
The resistor and/or capacitor ratios are not correct.
If you want a Butterworth roll-off you need a Q=1/sqrt(2) to create a B2
Two B2 cascaded give an LR4 for a final Q=0.5 and phase at -6dB = 90°
You are using the wrong formula for the 2pole filter. You have shown the single pole RC filter formula, as in post30.
Spice has used the wrong values you inserted in there to create a Q=0.5 filter and that has a ~160Hz -6 dB point but the roll off shape/rate is wrong. It is not Butterworth.
I still don't see where you got that 1+sqrt(2) factor, I think you misread the source doc, or the source was wrong.
Last edited:
I was referring to adding lines where needed. In this case, I would add 2 lines crossing at the -6db point to project the frequency of the crossover.
Those lines present are just the grid that I turned on.
I tried adding lines, and was able to, but no way to get them precisely where needed and properly vertical and horizontal. The tool to draw lines doesn't give the ability to do so.
This is a filter unit and those caps are of course not meant to be coupling ones. This is where the proper choice of tech and material is needed.
From what I've found, the polystyrene or polypropylene kinds are rather suitable, is this about right or there are better choices?
I didn't use the 1+sqrt(2) for this latest example with the 2nd order. I used it before for the 1st order with only 1 set of RC.
This time with the 2nd order with 2 sets of RC as shown, I used that formula with only the 1 on top.
What Am I missing?
Indeed the goal is to cascade 2 such filters to obtain the LR4.
Then an all-pass will be needed with the same phase characteristics, so the substraction can be made and that result fed on to the next stage, and do it again.
I'm just building this one step at a time and post the results so everyone interested can see this and learn to do it properly. I am learning as I go along.
Those lines present are just the grid that I turned on.
I tried adding lines, and was able to, but no way to get them precisely where needed and properly vertical and horizontal. The tool to draw lines doesn't give the ability to do so.
This is a filter unit and those caps are of course not meant to be coupling ones. This is where the proper choice of tech and material is needed.
From what I've found, the polystyrene or polypropylene kinds are rather suitable, is this about right or there are better choices?
I didn't use the 1+sqrt(2) for this latest example with the 2nd order. I used it before for the 1st order with only 1 set of RC.
This time with the 2nd order with 2 sets of RC as shown, I used that formula with only the 1 on top.
What Am I missing?
Indeed the goal is to cascade 2 such filters to obtain the LR4.
Then an all-pass will be needed with the same phase characteristics, so the substraction can be made and that result fed on to the next stage, and do it again.
I'm just building this one step at a time and post the results so everyone interested can see this and learn to do it properly. I am learning as I go along.
By the way, I was wrong about my 18" woofer's response, as I mixed up the info from an other different driver.
My 18" has a Fs at 57hz, and usable response up to 6300hz.
Then my 15" above, for the low-mid does go from 50hz to 6000hz.
The 18" efficiency at 103db@1w@1m and the 15" at 105db@1w@1m.
However the 18" in its cabinet with the rear expo vent horn is likely to see its efficiency going up from the base driver's.
Overall the 4 drivers' spl are rather close to each other, and only the mid-hi 1" horned driver has a higher efficiency of at least 110db@1w@1m. Having a level adjustment on each of the filter's output can help even out all this, despite the power amps having a different max power capability.
My 18" has a Fs at 57hz, and usable response up to 6300hz.
Then my 15" above, for the low-mid does go from 50hz to 6000hz.
The 18" efficiency at 103db@1w@1m and the 15" at 105db@1w@1m.
However the 18" in its cabinet with the rear expo vent horn is likely to see its efficiency going up from the base driver's.
Overall the 4 drivers' spl are rather close to each other, and only the mid-hi 1" horned driver has a higher efficiency of at least 110db@1w@1m. Having a level adjustment on each of the filter's output can help even out all this, despite the power amps having a different max power capability.
Your new formula is wrong.
F=1/{2PiRC} is for a single pole RC filter.
try using Texas pdf as your sources.
http://www.ti.com/lit/an/sbfa001c/sbfa001c.pdf
http://www.ti.com/lit/an/sloa049b/sloa049b.pdf
Filter Designs in Minutes with Filter Designer | WEBENCH® Tools | TI.com
for a simpler overview read ESP.
Active Filters
And for more complex read Linkwitz
Active Filters
F=1/{2PiRC} is for a single pole RC filter.
try using Texas pdf as your sources.
http://www.ti.com/lit/an/sbfa001c/sbfa001c.pdf
http://www.ti.com/lit/an/sloa049b/sloa049b.pdf
Filter Designs in Minutes with Filter Designer | WEBENCH® Tools | TI.com
for a simpler overview read ESP.
Active Filters
And for more complex read Linkwitz
Active Filters
I took that off of an old elektor article from the french elektor of mai 1987.
And that was from the same topology as my last simulation, to which they provide that formula. I wonder what I missed there or did they make a mistake? Perhaps in the translation from the original article in english that was published in april 87, which I don't have.
Good info. I'm going to go through all this now. Hoping to better understand things. And finally, hopefully get it done right.
I can see now one thing that I did miss. In that previous simulation, I was looking at the -6db point, which was apparently in agreement with that calculation and thinking like it was the LR4, but it is not the LR4, it's only the SK2 2nd order that makes only half of the LR4, so I should not be looking at the -6db point but rather the -3db one.
I used that excel sheet from that linkwitz site and used that in a new simulation (attached), aiming for a 150hz frequency, which I do now see on the plot at the -3db point.
The slope should now be fine right?
I'll cascade 2 of those next and then we'll see what an all-pass and the subtraction does.
I used that excel sheet from that linkwitz site and used that in a new simulation (attached), aiming for a 150hz frequency, which I do now see on the plot at the -3db point.
The slope should now be fine right?
I'll cascade 2 of those next and then we'll see what an all-pass and the subtraction does.
Attachments
You now have a set of resistor and capacitor ratios that give the B2 roll-off. The formula for that circuit will be in some or maybe all of those links I posted for you.
There is an infinite variety of C & R ratios that give a B2 but the one in the sch is the easiest to get accurate ratios since identical, or 2:1, (1:1 & 1+1:1) is all you need to ensure.