😕 Low Pass Attenuation slope levels out at high frequency and then begins to rise :
Hi all,
I have simulated a standard 4th low pass Linkwitz Riley filter as part of an active crossover at 500Hz, but at about 80kHz it levels out and then begins to rise again.
Is this normal for The LR Low pass filter ?
Many thanks for any guidance on this.
Hi all,
I have simulated a standard 4th low pass Linkwitz Riley filter as part of an active crossover at 500Hz, but at about 80kHz it levels out and then begins to rise again.
Is this normal for The LR Low pass filter ?
Many thanks for any guidance on this.
Not normal for the theoretical transfer function, can you post the schematic and response curve?
Last edited:
Maybe it's measurement error, since by 80kHz the signal is very far down, in a 500Hz LP filter.
Can you link to a web site showing the schematic that you used, is it this one?
Active Filters
Can you link to a web site showing the schematic that you used, is it this one?
Active Filters
Attachments
Last edited:
Hi again, I used ESP - Linkwitz-Riley Electronic Crossover - this is the site and diagram.
I use R= 1K2 & C= 188nF [4 parallel 47nF caps]
Hope that helps.
I have also used the original filter designer set up by The LR website. the diagrams and calcs are identical.
I use R= 1K2 & C= 188nF [4 parallel 47nF caps]
Hope that helps.
I have also used the original filter designer set up by The LR website. the diagrams and calcs are identical.
If there's no error, maybe the op amp model is starting to run out of open loop gain at that point.
In any event it must be around -120dB at 80kHz, so not very important.
In any event it must be around -120dB at 80kHz, so not very important.
Thanks rayma, yes that's the circuit. the op-amp might be the culprit - I used an OP27 in the schematic - just to give the simulator something to work with, but will probably use an LM4562 or similar. The simulator is LTspice.
This often happens with Sallen Key topology. It's an effect of non-negligible output impedance of the op-amp. Other topologies, like MFB, do not have this problem. See page 6 of the attachment and esp. Figure 8.
Hi CharlieLaub,
very many thanks for your response, it's an eye opener as I had been led to believe that the cut off slpoe just keeps on going down and never comes back up again. We live & learn.
I will now try to use an MFB filter instead - should be an interesting exercise.
very many thanks for your response, it's an eye opener as I had been led to believe that the cut off slpoe just keeps on going down and never comes back up again. We live & learn.
I will now try to use an MFB filter instead - should be an interesting exercise.
A speaker for which you use a low pass filter will not be capable of significant output at 80kHz to begin with. Even if it was, you wouldn't hear it. You have to ask whether this is significant.
Hi AllenB,
You're right, absolutely, but I was surprised by the plot, and thought that perhaps I had made a mistake (but couldn't find it) or even if the simulation prgrm was at fault. It's still quite a steep learning curve for me.
You're right, absolutely, but I was surprised by the plot, and thought that perhaps I had made a mistake (but couldn't find it) or even if the simulation prgrm was at fault. It's still quite a steep learning curve for me.
The math assumes the opamp is perfect, with zero impedance to infinity.
Opamps we can buy have some output impedance, reduced by NFB, but open loop gan falls, so output impedance must rise. Meanwhile the capacitor from input to output is decreasing impedance. At some point there is more signal short-cutting around the opamp than through.
For audio, we have had "good enough" opamps for many decades. It is always a problem when textbook filters are transposed to radio frequencies. There was a company made a point of explaining all this, but they went bust (bought out by a bigger fish). Comlinear??
Opamps we can buy have some output impedance, reduced by NFB, but open loop gan falls, so output impedance must rise. Meanwhile the capacitor from input to output is decreasing impedance. At some point there is more signal short-cutting around the opamp than through.
For audio, we have had "good enough" opamps for many decades. It is always a problem when textbook filters are transposed to radio frequencies. There was a company made a point of explaining all this, but they went bust (bought out by a bigger fish). Comlinear??
The general result is correct, either from common sense, old books, or (I was bored) old simulators.
Top-left sketch shows opamp with plausible internal impedance and gain. Counting on thumbs, path "2C" against opamp output will tend to flatten-out near 80kHz. What happens beyond depends (Zout is likely to rise).
With an Ideal opamp U4 the filter slopes-off forever.
I ate all my OP27 long ago. This sim had a LF411 which is a pre-TL07x BiFet of modest performance. It absolutely shows a bend, in this case a bit above 100kHz.
And then what?? I tole my sim to run to 100MHz. Peak near 5MHz then flat beyond. None of this can be trusted. The opamp model was only intended to go a few MHz and some parasitics will come into play past there. And we know ALL amplifiers go to zero at infinity. But we are warned the good response to 100kHz is followed by increasing errors as we go higher.
Top-left sketch shows opamp with plausible internal impedance and gain. Counting on thumbs, path "2C" against opamp output will tend to flatten-out near 80kHz. What happens beyond depends (Zout is likely to rise).
With an Ideal opamp U4 the filter slopes-off forever.
I ate all my OP27 long ago. This sim had a LF411 which is a pre-TL07x BiFet of modest performance. It absolutely shows a bend, in this case a bit above 100kHz.
And then what?? I tole my sim to run to 100MHz. Peak near 5MHz then flat beyond. None of this can be trusted. The opamp model was only intended to go a few MHz and some parasitics will come into play past there. And we know ALL amplifiers go to zero at infinity. But we are warned the good response to 100kHz is followed by increasing errors as we go higher.
Attachments
This is very interesting - I used the OP27 because it is one of the few that I recognized in the LT library.
Later, I tried two others: LT1028 which is reputedly 'high end' [very expensive here in RSA], it showed the same leveling out then rising characteristic, but surprisingly not at a very much higher frequency.
Then I tried the very old TL082 op-amp - which did show a markedly sub-optimal performance.
Later, I tried two others: LT1028 which is reputedly 'high end' [very expensive here in RSA], it showed the same leveling out then rising characteristic, but surprisingly not at a very much higher frequency.
Then I tried the very old TL082 op-amp - which did show a markedly sub-optimal performance.
I use OPA1641 devices in my LR-4 electronic crossovers, they're inexpensive, pretty close to audio state of the art, and are quiet enough. (I use high efficiency horns and there is nothing between the XO and the power amps) There are duals and quads in this family as well - all are SMD. (I do most of my designs as SMD these days so that does not faze me.)
If you low pass filter the input to the crossover I don't think you have anything to be concerned about due to the rising response.
I have avoided MFB filters because I find them difficult to design and tune, they also tend to be somewhat more sensitive to component tolerances real world than is the case with low Q sallen-key filters in my admittedly limited experience with them.
If you low pass filter the input to the crossover I don't think you have anything to be concerned about due to the rising response.
I have avoided MFB filters because I find them difficult to design and tune, they also tend to be somewhat more sensitive to component tolerances real world than is the case with low Q sallen-key filters in my admittedly limited experience with them.