It is a shame that your Chinese dual voltage board is cheap with poor performance. But usually cheap means it is cheap.

An RC highpass filter for coupling an audio signal is simple to calculate:
C= 1 divided by (2 x pi x R x f). So for a -3dB cutoff at 30Hz into 4 ohms the capacitor value is 1300uF.
But if the charging and discharging times of this output capacitor do not match the times of the input capacitor there will be a loud POP when the power supply is turned on or is turned off.

If you want to couple the bandpass filter into a grounded 20k ohms volume control then the capacitor feeding 30Hz at -3dB will be 0.27uF.

RC highpass filters add. The input, the coupling into a volume control, the coupling from the volume control to a power amplifier then maybe a capacitor feeding a speaker are a total of 4 RC stages, each producing a drop of -3dB for a total drop of -12dB.
Your already have -3dB from your subsonic filter and another -3dB from your bass reflex enclosure, then your subwoofer will produce no bass sounds.

 

I know what you said, but your simulation results don't appear to agree with that.
I'd prefer to see the input signal and the circuit configuration you simulated.

Hi,

I dont see why not.
Here is the circuit with a test wave of 100Hz.

 

Here is the circuit with a test wave of 100Hz.

And here's the LTspice simulation of the circuit.

It shows a 2.3dB peak for the first half-cycle at the output, which I'm quite sure is not audible (expect perhaps to those with "Golden Ears").

Your simulation showed a larger and longer peak.
It would appear that your simulation circuit used a 400Hz input and had a time constant of about 5ms, giving a corner frequency of 32Hz.
At 400Hz the output is so attenuated (about 22dB), that it would likely be essentially inaudible over the 0dB normal music level.

 

And here's the LTspice simulation of the circuit.

It shows a 2.3dB peak for the first half-cycle at the output, which I'm quite sure is not audible (expect perhaps to those with "Golden Ears").

Your simulation showed a larger and longer peak.
It would appear that your simulation circuit used a 400Hz input and had a time constant of about 5ms, giving a corner frequency of 32Hz.
At 400Hz the output is so attenuated (about 22dB), that it would likely be essentially inaudible over the 0dB normal music level.

View attachment 191734

Hi,

[funny comment removed]

But i used a 100Hz test signal, and besides that you have to look at more than just 400 or 100Hz. Just because 400 or 100Hz may not have much effect that doesnt mean we forget the whole thing. One case that 'works' doesnt spoil the whole apple tree. And that is still assuming that one case doesnt 'work' (which may be valid).

Here is the result with 400Hz.



The nice thing is it all ends in less than 20ms. The bad thing is that the max is almost twice as high as the norm.
There are a lot of other things to consider too though. For another example, physical placement of the midrange speaker in relation to the woofer. Depending, the lobe of the optimum listening area changes and can even go off center.

 

Goody goody :--)

Since you're descended to being condescending, I'll have no further comment on this subject.

 

Since you're descended to being condescending, I'll have no further comment on this subject.

Hi,

Oh sorry i was just joking around. I can remove that comment if you would feel more comfortable, in fact i will right now.

[A tiny bit later]
Ok it is gone, but i am not sure there is too much more to discuss anyway. My main point was "check the filter response carefully as there are pitfalls when dealing with audio". You made some good points also.

What we should probably be doing is investigating more thoroughly. That involves a lot of work however because we would have to compare several types of filters. The one i would like to revisit is the type used in graphic equalizers as they are pretty good, and note the design for each section is more complicated than the simple filter.

 

The minor points aside, in this thread we have successfully corrected the horrible frequency response peaks caused by the original circuit that used two series Multiple Feedback Bandpass Filters with their high Q peaks but 6dB per octave skirts instead of using a Sallen-Key highpass and a Sallen-key lowpass for the sub-woofer.

I just finished watching a recording of today's Formula One car race and my ears are still ringing. The announcers all used the same microphones that seem to have a sharp peak at about 3kHz or 4kHz which is probably advertised to be for "enhanced" presence. I don't like sounds like that.

 

Only question left is, how to power that thing?
Thanks for replies.

 

Only question left is, how to power that thing?

I cannot remember if your 250W into 4 ohms amplifier is efficient class-D or is heat-producing class-AB.
Class-D uses 288W so the transformer must be rated at 100VDC/1.414= 71V center-tapped at 288VA which is 288/71= 4A.
Class-AB uses 363W so the 71V center-tapped transformer must be rated for 363vA or 363/71= 5.1A.

The filters can be powered from their own little 22V center-tapped 5VA transformer which when rectified and filter produces about +14V/-14V DC.

 

It is class D. The amplifier Board uses 2 50v power supplies. How to get negative voltage out of a Transformer? U just do the same stuff i do with my class D board. it is VIRTUAL +/- 14v, in reality it is + 28 v referenced to GND of Earth and + 14v referenced to GND of earth, right? If i want to use that method, i have to bias the signal with +14v. Then there again remains the question how to remove that offset? Or do i not have to do that bias / offset thing?

 

A center-tapped transformer can use its center tap as ground and have its ends rectified to make positive and negative DC voltages. I suggested using a 22V center-tapped transformer then each of its ends have peak voltages of 11V x 1.414= 15.5V and when rectified and filtered produce +14.8V and -14.8V to power your opamps. Then there will almost no DC offset voltage and a virtual ground is not needed.

 

And here's the LTspice simulation of the circuit.

It shows a 2.3dB peak for the first half-cycle at the output, which I'm quite sure is not audible (expect perhaps to those with "Golden Ears").

Your simulation showed a larger and longer peak.
It would appear that your simulation circuit used a 400Hz input and had a time constant of about 5ms, giving a corner frequency of 32Hz.
At 400Hz the output is so attenuated (about 22dB), that it would likely be essentially inaudible over the 0dB normal music level.

View attachment 191734

 

Hi crutschow,
Could you please provide an analytic method for calculation 3-rd order Sallen-Key LPF with one passive + one active sections?

 

Hi crutschow,
Could you please provide an analytic method for calculation 3-rd order Sallen-Key LPF with one passive + one active sections?

Here is the website I use for that.
I let the tool calculate the values as it's kind of a hairy calculation.

 

Here is the website I use for that.
I let the tool calculate the values as it's kind of a hairy calculation.

 

Thanks crutschow
I know this site. I would like to find an analytic way.

 

Thanks crutschow
I know this site. I would like to find an analytic way.

Here's the Laplace transform filter transfer function from the site.


Have fun solving for the values.

 

Here's the Laplace transform filter transfer function from the site.
View attachment 191938

Have fun solving for the values.

 

I have fun with it from 2008.
Thanks,

 

I have fun with it from 2008.

(?) That's a long time to be working on the problem.