sfogg Posted August 31, 2005 Share Posted August 31, 2005 Was playing around tonight with some equipment plugged into my FFT.... The sum.... Quote Link to comment Share on other sites More sharing options...
sfogg Posted August 31, 2005 Author Share Posted August 31, 2005 The high pass.... Quote Link to comment Share on other sites More sharing options...
sfogg Posted August 31, 2005 Author Share Posted August 31, 2005 high pass 50hz below crossover Quote Link to comment Share on other sites More sharing options...
sfogg Posted August 31, 2005 Author Share Posted August 31, 2005 Low pass.... Quote Link to comment Share on other sites More sharing options...
sfogg Posted August 31, 2005 Author Share Posted August 31, 2005 low pass 50hz above crossover... Quote Link to comment Share on other sites More sharing options...
Deang Posted September 1, 2005 Share Posted September 1, 2005 Must be Al's Extreme Slippery Slope Network. Wait till you see how I measure, you're bound to get a good laugh. Quote Link to comment Share on other sites More sharing options...
WMcD Posted September 1, 2005 Share Posted September 1, 2005 Very interesting. It is always good to see measured data. My first thought was that the little hump was some 60 Hz getting into the microphone circuit. But the numbers (hard to see with my old monitor and the new forum software) don't support that. My guess is that Dean G nailed it. Nice call. Al K's extreme slope X-overs have a rapid roll off to begin with. Then he puts a notch filter just in the middle of the slope. So the bump is where the effect of the notch has ceased. The continuance of the slope is under the resolution of the 'scope. As you may know, I have a great deal of respect for Al.s work. Best, Gil Quote Link to comment Share on other sites More sharing options...
Al Klappenberger Posted September 1, 2005 Share Posted September 1, 2005 Shawn, I see you pulled the old SD380 analyzer out of moth balls! Dean, Yep, it the ES600 network. AL K. Quote Link to comment Share on other sites More sharing options...
Al Klappenberger Posted September 1, 2005 Share Posted September 1, 2005 Here's how an older model of the same analyzer (SD375) plots the very same network on paper. Doing that lets you see both the highpass and lowpass at the same time. I also ploted the insertion phase of each channel. Notice that the older SD375 can only display the plot on an 80 dB scale. Shawn's newer SD380 can magnify the vertical scale. Al K. Quote Link to comment Share on other sites More sharing options...
Al Klappenberger Posted September 1, 2005 Share Posted September 1, 2005 Gil, You are right on. The little hump is called the "arc top". In filter jargon, the little notch is called a "finite zero of transmission" and serves to draw down the slope quicker. If there were several notches with all of the arc tops arcing up to the same level you would have the classic "elliptic function" stopband response. In this case there is only a single arc formed by a single finite zero. You could get the same slope without using notch sections by adding "poles" in the passband (it would take about 20 parts to do it). but that causes the passband loss to increase. That would then be called an "all-pole" filter having a "monotonic" stopband response. Al K. Quote Link to comment Share on other sites More sharing options...
sfogg Posted September 1, 2005 Author Share Posted September 1, 2005 Another picture of the low pass with an expanded vertical scale. At the bottom the jaggies are noise artifacts. Quote Link to comment Share on other sites More sharing options...
sfogg Posted September 1, 2005 Author Share Posted September 1, 2005 Dean, " Must be Al's Extreme Slippery Slope Network. :)" Is that your final answer? Shawn Quote Link to comment Share on other sites More sharing options...
Al Klappenberger Posted September 1, 2005 Share Posted September 1, 2005 Shawn, Something is suspicious about your expanded plot. Here's a computer simulation of the ES600 lowpass. Notice the frequency range between 1000 and 2000 Hz. My plot and the computer show a far lower attenuation than yours. Remember that you need to add a 1 mHy inductor between the output of the netwrok and the 6 Ohm load resistor. You need to plot the output only across the resistor. Even this wouldn't account for the difference though. Al K. Quote Link to comment Share on other sites More sharing options...
sfogg Posted September 1, 2005 Author Share Posted September 1, 2005 Al, "Something is suspicious about your expanded plot. " Thought you would pick up on that on the expanded chart.... The reason it looks different is because I was not measuring the ES600T. My posts are from a line level crossover, not speaker level. I realized a week or so ago I could 'simulate' a high slope crossover along the lines of the ES600 at line level. Last night I got a chance to try to 'design/build' it. I think this one is actually a little steeper then the ES600 as the 50hz points are down a hair over 15dB, I think the ES600 is about 10dB there. Also on the first few dB of attenuation I think it is a little sharper too. Below the arc it is rolling off at 48 dB/octave which is the biggest difference between the two. I could make that rolloff a lot shallower but didn't really the point of doing that. Shawn Quote Link to comment Share on other sites More sharing options...
Al Klappenberger Posted September 1, 2005 Share Posted September 1, 2005 Shaw, Ah HA! Trying to pull a fast on on the old guy, eh! You had me spining becasue it will take a filter with more poles at infinity to make that response! I don't think the extra rejection beyone the arc top will improve things any. Being 25 dB down the energy from a driver is virtually gone and can't interfere with the sound from the other driver. Making the skirt sharper may help a bit though. The ES networks I built are made to get the most out of the parts count. An active filter can give you more performace if you are willing to dittle with multiple amps. You can't just tuck it inside the speaker and forget it like a big passive network. Everything is a trade-ff. Al k. Quote Link to comment Share on other sites More sharing options...
Al Klappenberger Posted September 1, 2005 Share Posted September 1, 2005 Shawn, Since the objective here is have some fun with crossovers, here's and exercise in some very silly extremes. Here's an actually buildable lowpass filter. It goes down to 56 dB at 50 Hz away and hits 47 dB at 30 Hz away. The hang-up is that the realizable Q of the inductors (20) dumps 17 dB of loss on you at 600 Hz! This will make a dip at the crossover when it's hooked to the proper highpass half of the crossover. The inductor at the termination is the voice coil inductance of the K33 driver. This filter has 5 Zeros at infinity and 3 finite zeros. Al k. Quote Link to comment Share on other sites More sharing options...
sfogg Posted September 1, 2005 Author Share Posted September 1, 2005 Al, "Ah HA! Trying to pull a fast on on the old guy, eh!" Wanted to see how close its response came to your network. My goal was to try to get it in the ballpark if I could. Based on the original posts (before I extended the measurement range) I'd say it is very close in that range. "I don't think the extra rejection beyone the arc top will improve things any." I agree. In my case though the extra rejection was trivial to implement so I did. I could make it roll off shallower in literally like 5 seconds but have little reason to do that. On a side note DSP can be a wonderful thing! Getting real time feedback from the SD380 as I tweaked the DSP crossover real time was fantastic. It made it *much* easier to setup. I am even more impressed with how good the ES600 (and/or your simulation software) really is after doing this. "An active filter can give you more performace if you are willing to dittle with multiple amps. You can't just tuck it inside the speaker and forget it like a big passive network. Everything is a trade-ff." Absolutely agree with that. The other problem with the active approach is a person taking it at some point needs to just give in, stop fiddling, and enjoy the setup. For a neurotic tweaker they could *easily* drive themselves insane with the active approach trying to test out all the different slope/crossover points available. In a nutshell it is a 8th order LR crossover (spaced away from the actual crossover point... that is what gives the roll off beyond the notch), two parametric EQs performing -15dB notches (Q10) just to the other side of the crossover point (both set to the same frequency) and one additional parametric EQ even closer to the crossover point adding some boost at a wider Q to bring up the response right to the crossover point and to make the slope sharper still. Without that last EQ boost the response right up there looked very very close to your network in that is dropped a couple of dB slower then dropped down very quickly. I can turn each piece off individually and take shots of each crossover components response if anyone is curious. When they all fit together you get the response as shown above. "Making the skirt sharper may help a bit though." Do you mean getting down deeper before the arc or a higher slope? I could add a third parametric EQ (they are limited to -15dB cut each) to the other two and drop the arch down to -45dB and it would be sharper too. I could add even more then that if desired but I tend to be skeptical if it is needed. If I would do that it would roll off the start of the filter... which I could correct too but I tend to doubt it is worth it and it probably wouldn't be quite as smooth there. I also would rather save DSP resources for the bandpass/high pass and such. I might see what it looks like adding a third parametric notch but move it a little further away from the crossover point. That would likely notch down the arc further. Shawn Quote Link to comment Share on other sites More sharing options...
Al Klappenberger Posted September 1, 2005 Share Posted September 1, 2005 Shawn, Ok.. I get how you did it now. I think it would not be possible for the average joe who dones' thave a SD380 lying around! I think the improvement would be in the area of the crossover. Try toget the 10 dB "window" as narrow as possible with the first EQ notch by moving it in as close as you can, then move the second notch away to force the arc top down. You should then see two notches and two arcs. Here's an example Al K. Quote Link to comment Share on other sites More sharing options...
sfogg Posted September 1, 2005 Author Share Posted September 1, 2005 Al, Yes, without some type of FFT it would be much harder to do it the way I did it. Here is adding two more notches on top of the first two to really get it down deep. I'll try moving the two notch filters further apart and see how it works out. The limit for the slope is really in the Q setting of the EQs. They are limited to Q of 10. Shawn Quote Link to comment Share on other sites More sharing options...
Al Klappenberger Posted September 1, 2005 Share Posted September 1, 2005 Shawn, What you are seeing when you move thos notches around is classic to filter people! Q is always the limiting factor. Q as I think of it is "Quality factor" of individual components. The inductors are limited to about 20 at woofer frequency. That is what devermines how sharp the slope will be. If you are "Q limited", moving the notch in closer to the cutoff just makes it disapear. The practical limit is usualy around 1.05 times the cutoff. Below that things start going down hill fast! At microwave frequency you can get inductor Q on the order of 100 to 150. Using dielectric resonators can up the Q to 400 or more. The best of all is a quartz crystal. They can have Q factors beyond 100,000! Al K. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.