Although extremely good, the 627s are $25.00 each in the US. Make sure the circuit works with the excellent OPA604, and let the people who have upgrade-itis spring for the 627s
It's not the filter it's the net response of the system. Call it whatever you want, but chances are slim that the electrical filter will have a Butterworth characteristic if the acoustic output of each driver follows a Butterworth characteristic.
Wouldn't that assumption be true for any crossover filter topology? Does this justify ommiting the option to implement? Isn't a logical starting place one that yields theoretically correct power response with minimum phase anomolies?
I haven't the foggiest...
Anyway, those links are to my threads of my builds. That might have something to do with it.
I believe that what you said is incorrect - if "other topologies" have the same off axis response, then they will have the same power response, since power response is just the frequency response integrated over all angles (see the link in a previous post in this thread to the John K analysis of power response for the math).
Uh, no, it's not overlooking the power response, the plot is essentially showing it and all the "dirty details" behind it.
You can't seem to get past talking about flat power response. IMHO power response is of secondary concern and the direct response is most important. This is because the level of the direct response is above the room's response, e.g. the "power response". As you can see from the plot, the off axis is all over the place, even if the sum is net zero (e.g. flat power response). I am concerned with the listener being positioned "off axis" where the direct sound frequency response is anything but flat. Personally, I would prefer that the off-axis responses track symmetrically, and this is the case with the LR4, because you can compensate for this in a more consistent way. But there are others that do this as well, or even better than the LR4 or BUT3. For instance, a crossover based on a 3rd order 0.5dB-ripple Chebyshev type I has a much tighter off-axis pattern within +/- 30 degrees, and has better selectivity than the 3rd order Butterworth filter. The transient response will be only slightly worse than BUT3.
Keep in mind that the power response is just an average over all angles. The speaker response does not need to be smooth or flat at any particular angle for the power response to be perfectly flat. The BUT3 is an example of this "problem". So you have to consider the power response in context, with other responses, both in the frequency and time domain, in order to have a balanced assessment of the loudspeaker. Do not elevate power response above all other considerations that make up the "big picture".
-Charlie
Alex when you posted in the other thread with a "right way to do it" I said that you sounded like a shill. It was that more than the link. Big difference from just posting a link to your own projects.
And, yes, I agree that to be most useful a crossover needs to offer odd and even order electrical filters. It should also offer at least some equalization, whether through a separate block or adjustable filter Q and phase alignment.
Charlie has chosen to offer some of the features as add ons to simplify the basics for those who decide that they don't need them. Let's stop discussing our favorite topology. Charlie asked us what features we would like to see included and we have told him that we want to be able to make odd and even order filters among other things. now it's time to see what Charlie designed.
And, yes, I agree that to be most useful a crossover needs to offer odd and even order electrical filters. It should also offer at least some equalization, whether through a separate block or adjustable filter Q and phase alignment.
Charlie has chosen to offer some of the features as add ons to simplify the basics for those who decide that they don't need them. Let's stop discussing our favorite topology. Charlie asked us what features we would like to see included and we have told him that we want to be able to make odd and even order filters among other things. now it's time to see what Charlie designed.
Charlie :
Your analysis is in direct conflict with the reference I cited earlier (the article in Audio, Aug-78). When the transfer functions for filters are solved for on-axis power and total power, only odd order Butterworth have both constant. Total power is of less importance, but it does color perception of overall response. In my opinion, the anomolies of 3rd order Butterworth are overall less objectionable. The reference I cited also verifies that the "dirty details" occur equally for any summed filters - at worst a 3 dB peak or an infinite dip.
Your analysis is in direct conflict with the reference I cited earlier (the article in Audio, Aug-78). When the transfer functions for filters are solved for on-axis power and total power, only odd order Butterworth have both constant. Total power is of less importance, but it does color perception of overall response. In my opinion, the anomolies of 3rd order Butterworth are overall less objectionable. The reference I cited also verifies that the "dirty details" occur equally for any summed filters - at worst a 3 dB peak or an infinite dip.
I agree. If he can find the equations to implement the filters I posted such that they can be built as either 2nd or 3rd order, then all of the basic filter functions will be possible with minimum stages.
Sorry I did not keep better records for my research, but all the information can be found in back issues of JAES esp. years 1975 - 1985.
The authors of the Audio magazine from 1978 likely assumed that the drivers were point sources in their analysis - their conclusions about on-axis frequency response and power response are correct under that restriction, however you cannot stop there because you will be overlooking the problems of the BUT3 filter: the Butterworth 3rd order filter has very non-flat off axis frequency response. I keep trying to point this out to you, and why this is important compared to the purported flat power response. IMHO this filter is bad because of the peak and near complete null in the off-axis response. Not all filters have frequency responses that vary this much within 30 degrees of on-axis.
Unlike the "model" in the Audio article, in the real world, with real drivers, the BUT3 has flat on-axis response but doesn't have flat power response unless your drivers happen to be point sources. Look at the analysis by John K again and you will see this. And all this is only true if you have time aligned drivers, otherwise the on and off axis response as well as the power response are even worse.
I don't see how you can find my "analysis" to be in conflict with your Audio article. I shows the same thing in the plot as they do, clear as day. We use many of the same conditions in our "model" (like point sources) and we get the same results. You probably have just not seen the data plotted this way, showing the off-axis frequency responses, which implicitly give the power response. The plot shows the flat on axis frequency response, the 3dB peak you mention and the null. All the things that you mention from your Audio article-they are all there!
-Charlie
Here it is again, for reference:
Last edited:
Charlie :
Seems we are at an impasse because I do not agree with your interpretation. As BobEllis pointed out, it's time to move on. Hope you will consider offering the options I have suggested since there might be other people who are interested in exploring for themselves.
Seems we are at an impasse because I do not agree with your interpretation. As BobEllis pointed out, it's time to move on. Hope you will consider offering the options I have suggested since there might be other people who are interested in exploring for themselves.
You are right, we seem to be at an impasse. That's OK in my book, to agree to disagree! You are welcome to contact me via PM if you want to talk about it further, privately.
I am considering what you mentioned, about implementing the PCB layout so that one could create a 3rd order section with a single op amp. The core second order section would have to be Sallen-Key however. For the reasons that Bob mentioned, this has to be done with some caution, but it certainly seems to be a valid way to make up a 3rd order section. And I think that this wouldn't cause any issues or invalidate any of the other types of circuits that can be formed using the "universal" board. I will check in to it a little more over the next few days and then will post a follow up.
-Charlie
The equations for the MFB circuits I posted for any Q and Gain are out there in filter cookbooks and JAES articles. You might try contacting Auratron Systems to see if they would offer any help. I don't believe they offer their PC program XOVER unless it accompanies purchase of an xover. Again, my research has revealed that Inverting MFB topology is preferred for audio xover filters.
Last edited:
A Sallen-Key filter has the same interactions with an RC ahead of it. It's doable, you don't need to go high or low impedance, the calculations just get a little harder. You'll need to provide a calculator or plan on answering questions, which you will get anyway. A spreadsheet should make fairly simple work of it. It may take several passes to get standard values in all positions, but that's better than the way we did it in the 80s. You'll want to verify response in a simulator anyway.
Lancaster's Active Filter Cookbook? I own it but haven't picked up in years. Among other options, Ti offers MFB and Sallen Key design with its free filter pro tool Active Filter Design Application - FILTERPRO - TI Software Folder It's too late to search, but isn't MFB fussier about component values than Sallen Key? I prefer independent highpass and low pass sections, but Charlie has valid reasons for going state variable.
Lancaster's Active Filter Cookbook? I own it but haven't picked up in years. Among other options, Ti offers MFB and Sallen Key design with its free filter pro tool Active Filter Design Application - FILTERPRO - TI Software Folder It's too late to search, but isn't MFB fussier about component values than Sallen Key? I prefer independent highpass and low pass sections, but Charlie has valid reasons for going state variable.
I can vouch for TI's Filter Desktop Pro (Filter Pro version 3.1 or later) as a good filter design tool. It's a major improvement on the previous version, "Filter Pro 2.0", which had some pretty bad errors in it!
I could just direct people to the FDP if they needed to calculate component values, but it can't design all the circuit types that the universal board can accommodate, so I would need to provide some kind of design assistance. That's no problem really and I had planned on that kind of thing. I am doing that leg work now as part of the design phase.
Regarding Sallen-Key versus MFB versus SVF:
Sallen Key has the fewest components, but is best when the Q is less than about 3 because above that the circuit becomes too sensitive to component tolerance. MFB can do higher Q's and are better in this regard. I seem to recall that SVF has the lowest component sensitivity, and all parameters (Q, corner frequency and gain) can be adjusted independently via potentiometers.
There is a good TI document discussing lots of dirty details about Sallen-Key filter design here:
http://focus.ti.com.cn/cn/lit/an/sloa024b/sloa024b.pdf
I have realized that I can configure the universal filter board to also accommodate MFB circuits (Alex: yes, inverting!), the only question is if I can fit all of the positions for various components on the target PCB footprint. I think I can, if I stick to resistors and capacitors that follow a 5mm - 15mm pin spacing. I did some checking into this recently while designing the layout of the SVF board, and I think that with this pin spacing range, you can get fit the cap values that you would possibly need, as long as you stick to radial packages. But these are available, at 1% or 2% tolerance, in polypropylene, in these pin spacings.
If you read the TI document linked above, they mention a problem for Sallen-Key (but the same issue comes up with MFB designs) in which a LP circuit has a rising response at high frequencies (above 100k Hz for instance). The solution for this is to add a first order LP filter at the output of the amp, just like you mentioned Bob, with a corner frequency above the audio band. See section 3.5 on page 6 and Figure 8.
I should have more info after the weekend or early next week.
-Charlie
Last edited:
Here is the link to past issues of JAES:
AES Journal Table of Contents
The info we are interested in is focused around 1974 - 1988
Everything you ever need to know about audio can be found here. Most of the authors do not have "hidden agendas" such as promoting a new marketable product, but a occasionally this happens. Take this into account when weighing the validity of the article(s).
Last edited:
When I load this link, here is what I see :
An externally hosted image should be here but it was not working when we last tested it.
How can I see the graphs?
I am not sure why John chose to use such a ridiculous color scheme (with the dark background) for those pages. Maybe he though it would bring out the graphs???
Did you click on the "NEXT PAGE" buttons (at top or bottom, at right side) to scroll though all five pages? The second page shows that the BUT3 does not have flat power response when the drivers are not modeled as point sources.
If you want to get creative you could save the page as html and then edit the code to change the background color to white or something easier on the eyes.
-Charlie
Last edited:
- Status
- Not open for further replies.