FWIW, I calculated what capacitance to gnd would produce a -1 dB voltage divider. For 1 nF (1st order filter only) it is 120 pF, for 0.5 nF (2nd order with the two 1 nF caps in series) it is thusly about 60 pF. (I did the curves for the 1st order HP and lo and behold, the level only falls off 0.5 dB).
Current jfet input op amps have about 2-6 pF input capacitance, too low to count? But anyway, I saw the same effect with the bipolar NE5532 as well.
What *could* produce significant capacitance to ground is the double ground plane of the MOX PCB's. Now why does that not happen to Jens then
Did he do his test with a prototype board which was single layered? Did he solder the caps at a couple mm higher from the board (that can't really be it though, since the actual connection is still at board level).
MBK
Current jfet input op amps have about 2-6 pF input capacitance, too low to count? But anyway, I saw the same effect with the bipolar NE5532 as well.
What *could* produce significant capacitance to ground is the double ground plane of the MOX PCB's. Now why does that not happen to Jens then
Did he do his test with a prototype board which was single layered? Did he solder the caps at a couple mm higher from the board (that can't really be it though, since the actual connection is still at board level).MBK
I made the measurements on the board pictured on my webpage:
I'm not sure whats wrong to be honest.
The formulas dont really hide anything AFAIKT
\Jens
An externally hosted image should be here but it was not working when we last tested it.
I'm not sure whats wrong to be honest.
The formulas dont really hide anything AFAIKT
\Jens
Mystery...I do have the equipment to measure such small capacitances...
BTW Jens, I must say I was looking upon your Active Filter 1 with respect. I think I need to do something similar, if maybe a little smaller. I now have a total of 7 speakers to build based on MOX-designed prototypes, but I hate to have to use multiple 4"x6" boards. (3 ways) A little GEQ wouldn't hurt on some of the cheaper ones
BTW Jens, I must say I was looking upon your Active Filter 1 with respect. I think I need to do something similar, if maybe a little smaller. I now have a total of 7 speakers to build based on MOX-designed prototypes, but I hate to have to use multiple 4"x6" boards. (3 ways) A little GEQ wouldn't hurt on some of the cheaper ones
Jens,
OK, on thing I have differently is, I installed 47 pF C200 and C500 comp caps. Cutoff well into 200 kHz. Besides, this cap is isolated from the 1 nF caps by the resistors.
Tiroth,
that would be very interesting. I only have a multimeter. No chance below a couple nF (even if it does display random data).
MBK
OK, on thing I have differently is, I installed 47 pF C200 and C500 comp caps. Cutoff well into 200 kHz. Besides, this cap is isolated from the 1 nF caps by the resistors.
Tiroth,
that would be very interesting. I only have a multimeter. No chance below a couple nF (even if it does display random data).
MBK
Nope, I found that they are not needed with the NE5532.
I will have to dig out my old protoboard and redo the measurement for the HP filter. I lost some of the data when my harddisk crashed.
I'll post results during the weekend.
(Regarding the filter one, Tyler mail me your adress and I'll send you a set of boards to try out)
\Jens
I will have to dig out my old protoboard and redo the measurement for the HP filter. I lost some of the data when my harddisk crashed.
I'll post results during the weekend.
(Regarding the filter one, Tyler mail me your adress and I'll send you a set of boards to try out)
\Jens
Hello,
I just finished retesting the HP part of the module and found that I have the exact same problem now. (I must have missed the test on the 1nF cap earlier)
I did some simulations on the thing, and the problem seems to come from the frequency selection resistors dangling in free air when they are not used. The resistors will have some capacity to ground on the board, and when the multiplication cap is about the size of this parasitic capacitance the problem gets noticed.
I will have to do some more tests on it, to see if the problem can be resolved easily by means of removing the connection to ground underneath the resistors on the top layer. I’m still working on that though.
Regarding the C500 + C200 You should not go above 10pF as this will influence the performance of the filter at 20 kHz.
\Jens
I just finished retesting the HP part of the module and found that I have the exact same problem now. (I must have missed the test on the 1nF cap earlier)
I did some simulations on the thing, and the problem seems to come from the frequency selection resistors dangling in free air when they are not used. The resistors will have some capacity to ground on the board, and when the multiplication cap is about the size of this parasitic capacitance the problem gets noticed.
I will have to do some more tests on it, to see if the problem can be resolved easily by means of removing the connection to ground underneath the resistors on the top layer. I’m still working on that though.
Regarding the C500 + C200 You should not go above 10pF as this will influence the performance of the filter at 20 kHz.
\Jens
Well,
Regarding changing the smallest capacitor to 10nF is definitely a good idea, but the 0.1 multiplication cap needs to be 10uF. This is not practical (I think) but if you can live without that option it should work ok.
I’m not sure about the current draw when you use smaller resistors; I have to do some math on it first to be sure.
I can post a set of resistors, or you can just divide the know values with 10 and use these values with the bigger caps.
\Jens
Regarding changing the smallest capacitor to 10nF is definitely a good idea, but the 0.1 multiplication cap needs to be 10uF. This is not practical (I think) but if you can live without that option it should work ok.
I’m not sure about the current draw when you use smaller resistors; I have to do some math on it first to be sure.
I can post a set of resistors, or you can just divide the know values with 10 and use these values with the bigger caps.
\Jens
Jens,
I just use plain old paper and pen, never got around to learn simulators. So I go by formulas.
I considered only the R200 // C200 combo in conjunction with the R201-220 that set the Q. This in itself should be a shelving LP topology, with F1 (lower corner frequency) of F1=1/2*pi*R200*C200. But haven't considered the influence of the frequency setting network on the other side of the op amp output. That one's a bit heavy w/o a simulator
Ten again the problem we now have would not have shown up in a simulator...
So if you say 10 pF I believe it ...
I'll keep the 47 pF in though , I likely stick to the lower Q's.
MBK
I just use plain old paper and pen, never got around to learn simulators. So I go by formulas.
I considered only the R200 // C200 combo in conjunction with the R201-220 that set the Q. This in itself should be a shelving LP topology, with F1 (lower corner frequency) of F1=1/2*pi*R200*C200. But haven't considered the influence of the frequency setting network on the other side of the op amp output. That one's a bit heavy w/o a simulator
Ten again the problem we now have would not have shown up in a simulator...
So if you say 10 pF I believe it
... I'll keep the 47 pF in though , I likely stick to the lower Q's.
MBK
Hi,
I have checked today if the reported problem "1dB dip on HF" is also present with the crossover I designed and published here on the forum. When I measured the dip on the HF filter, it was approx. 0.2dB. Then I went on to check if the problem would show up when I used a protoboard filter - dip was about 0.3dB, regardless of the opamap used (TL071, NE5534AN,LF35x,...). The dip is apparently caused by the capacitive divider, which is formed of a capacitor 1nF and a parasitic capacitor between the opamp's non-inverting input and the ground plane. Parasitic capacitance must be considerable in Jens's design, mainly due to (unnecessary) implementation of the ground plane on the PCB's top side. If the ground plane were removed, the value of the capacitor in question would probably drop so the capacitor's influence would be diminished. However, I do not believe the existing dip can adversely affect the filter's operation (in the case of LPF, parasitic capacitance may cause a small shift in the crossover frequency). My recommendation is to keep all this in mind when designing the next version Ground plane is not always useful when it comes to low frequency.
The negative effect can be avoided, as someone here has already mentioned, by using larger value capacitors and smaller resistors. Extreme caution is advised, though because that may cause the filter's input impedance to fall significantly in the passband region. Since the application of the crossover we are discussing here is universal, it can be expected to perform inferiorly in some respects to any crossover that has been custom-made for an individual case.
Simulators are not "almighty", as they mostly simulate ideal case scenarios. A more sophisticated board design software (pCAD, etc.) should handle the problem, given enough input data.
Regards,
Milan
I have checked today if the reported problem "1dB dip on HF" is also present with the crossover I designed and published here on the forum. When I measured the dip on the HF filter, it was approx. 0.2dB. Then I went on to check if the problem would show up when I used a protoboard filter - dip was about 0.3dB, regardless of the opamap used (TL071, NE5534AN,LF35x,...). The dip is apparently caused by the capacitive divider, which is formed of a capacitor 1nF and a parasitic capacitor between the opamp's non-inverting input and the ground plane. Parasitic capacitance must be considerable in Jens's design, mainly due to (unnecessary) implementation of the ground plane on the PCB's top side. If the ground plane were removed, the value of the capacitor in question would probably drop so the capacitor's influence would be diminished. However, I do not believe the existing dip can adversely affect the filter's operation (in the case of LPF, parasitic capacitance may cause a small shift in the crossover frequency). My recommendation is to keep all this in mind when designing the next version
Ground plane is not always useful when it comes to low frequency. The negative effect can be avoided, as someone here has already mentioned, by using larger value capacitors and smaller resistors. Extreme caution is advised, though because that may cause the filter's input impedance to fall significantly in the passband region. Since the application of the crossover we are discussing here is universal, it can be expected to perform inferiorly in some respects to any crossover that has been custom-made for an individual case.
Simulators are not "almighty", as they mostly simulate ideal case scenarios. A more sophisticated board design software (pCAD, etc.) should handle the problem, given enough input data.
Regards,
Milan
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.