filter Q Question

[mondogenerator]

.......

 

[mondogenerator]

So anyway, i have started my schematic and im seeking a little advice.

Using the RC on the 'front end' of the filter should i buffer its input, as well as the filter output?

Ill post a schematic once ive finished it, but i was unsure whether the RC wad better placed at the front or back of the filter. I settled on the input,although for no particular reason. I also plan an Lpad at the output to trim the spl.

 

[AndrewT]

the RC filter should see zero impedance on it's input.
This is usually done with a buffer or low output impedance amplifier.
But, a low pass filter with R in series, then C shunting to ground can have the input R adjusted to take account of a finite Source impedance. That finite Source impedance should have a fairly constant value over the range of signal frequencies that the filter is passing and cutting. eg an opamp having an output impedance of 10r with 20pF and a little inductance, could be swamped by adding a 100r to give nearly 110r over a wide range of frequencies. The R of the filter is now R= 110r + r'. i.e. r'= R -110r

 

[mondogenerator]

Thanks AndrewT

I think I follow you.

I am somewhat a novice but I have written a BOM for this circuit, and it is staggeringly cheap really, excluding a PSU.

I am not afraid to have to add bits here and there to improve it, but I have an initial schematic done for my low-pass. e.g. R1 and C1 were the design values calculated and not compensated for the RC ahead in the circuit. I'm not sure that's perfect but.....

I'm sure there are ways it could be improved. The picture is annoyingly large and the 'crop' function doesn't appear on my XP paint.....weird.

 

[AndrewT]

the last unity gain is not required as you have shown it.
But you can insert a volume control between 3 & 4 and still retain the low output impedance to drive cables and filters and input of next stage.

 

[mondogenerator]

Ok. Yes that makes sense, and id forgotten to add the level pot! I wasn sure the last stage was needed, and i was going to whack the pot after the last stage, but what you suggest makes sense ��.

Alternatively i guess i could use an extra buffer on the front end to drive both LP and HP sections via a pot, on order to balance their levels.

I shall work some more and see what results. Thanks again AndrewT

 

[AndrewT]

B.Duncan showed the first stage as inverting with optional plus or minus gain adjustment.
I think D.Self also showed this.

 

[mondogenerator]

Ah. I think i get your gist. I really need to buy some books by those authors. Working with an online calc and The active filter cookbook answered a few of my queries, but deal with text book curves only. Now i have my desired curve, i should revisit them. Thanks.

 

[mondogenerator]

So finally I have something Im 99% happy with

SO here it is, my filter (both sections)

Anything to criticise?

Im just wondering if the input Rs for some of the filter stages are a little too high, and that they could present a possible noise issue.

Specifically the R1s for the two LP stages, which are 22k and 24k respectively. (hoping the schematic is legible)

Any opinions regarding this?

Thanks again for reading and any pointers you may be able to give.

 

[AndrewT]

I can't read the values.
The cascaded filters should have appropriate Q for each position. These Q are not the same.

It looks like you have a 5pole filter and a 4pole filter.
How did you arrive at the overall Q for each half?

 

[mondogenerator]

It's ok Andrew. I didn't think it would be clear enough to read.

I haven't calculated a total filter Q. On both filter sections the 1st and 2nd Sallen Key stages are calc'd for Q's of 0.618 and 1
618 respectively.

The lowpass has a unity buffered RC to form the 5th pole. The highpass differs in that the 5th pole is formed by a DC blocking cap at the loudspeaker itself.

 

[mondogenerator]

The only issue i see is that i should try to add an all pass filter to time align the drivers...even if i can only get 360* rotation, and not the full amount.

Hmmm.

 

[AndrewT]

Have a read of the optimal crossover/filter Thread.
They are looking at a quite different way of aligning the drivers' outputs

 

[mondogenerator]

Ta Andrew, ill have a read. Not sure i want a complete redesign though ��

Phase alignment at xo is good as it stands, except the cycle of delay that doesn't show on the 'normal' phase axis scale.

 

[AndrewT]

They are varying the slopes and frequencies and Qs to arrive at a summed output. The drivers end up with an offset that corrects for the physical difference in VC location.

I suspect you can adopt all of that in active, even though they are discussing passive.

 

[Andrew Eckhardt]

If you think there is a possibility of driving long signal cables you may want to add 1 or 2 hundred ohms at the outputs to keep the opamps from ringing into the capacitance. If you're never going to use anything but 1 meter coax cables then you could probably leave it out.

 

[mondogenerator]

AndrewT:

I am certain you are correct, and the thread is interesting. Until I get accustomed to PCD im limited to 5th order as the highest order slope. I don't fully understand how the 6th with 2nd negates the full cycle of phase offset, but i need to sit and digest what ive read and it will probably dawn on me. It usually does, eventually.

Andrew E:

Yes i have seen current limiting Rs in couple of designs. I didn't include them here but i most definitely include a provision for them on the pcb

Its unlikely ill be using long cable runs, but it couldn't do any harm.

I may just have to design a mk2 based on reading the optimal crossover thread, but will probably plough on with mk1, save adding an all pass or two to remove the cycle of delay. A comparison would be interesting.

 

[AndrewT]

A 47r to 220r on the output of every opamp rarely does any harm at the currents normally being handled.
But that defined resistance is very good at attenuating unexpected resonances.

 

[mondogenerator]

Thanks again.

After messing on filter pro for a few minutes i get the point regarding all passes. Its a shame, but 360* looks like needing a 6th order all pass, and might actually end up creating more phase issues than it solves...perhaps a bandpass might be better, though id guess it would still be at least 4th order.

I could just invert one channel and then aim for half a cycle instead. Its a tough one, and i really have to recheck my acoustic phase too....i cant recall if my tweeter polarity is inverted or not.

 

[mondogenerator]

A few experiments later....

Well a big Thank You for your prods of encouragement!

I just decided to throw the LP together on breadboard. Im building up a BOM of the components I will go and buy to make these, but I threw it together with at best 2.5% -10%, and 1-5% Resistors and some 5532s, 470uF e'lytics in each rail, and using a =/-12V supply.

I ran a curve of the electrical response, which I am planning to compare to the simulated electrical response produced in Boxsim, which I have used to simulate the end result Acoustic SPL and phase.

Simulated stuff:

Firstly, a schematic that I hope is legible:

And the simulated Acoustic SPL:

Simulated Acosutic Phase of the individual drivers:

And the Electrical frequency response:

Once Ive gotten it all in excel I will probably post my actual/ measured electrical response, but Ive yet to build the HP prototype and test that.

OdB is referenced to 5 Vrms in my test. Purely to get enough dB headroom for measurement, without going overboard.
I get perhaps 25 dB of useful Readings, so I may repeat with a higher i/p level (nice scope but bad 10:1 probes....) or just find straight probes.......or an old BNC cable.