[sdiy] Designing 4-pole filters with identical 2-pole stages - why not?

Richie Burnett rburnett at richieburnett.co.uk
Wed Dec 23 00:09:19 CET 2015


Tom, since we're taking about anti-alias lowpass filters, have you looked at the MFB multiple-feedback lowpass filter?

For the added complexity of another resistor it gives you more flexibility in the combinations of achievable gain, Q and cutoff frequency. The performance of the MFB lowpass is also better than Salen-key filter deep into the stop band when used with real op-amps that have rising output impedance at high frequencies. 

-Richie,

Sent from my Xperia SP on O2

---- Neil Johnson wrote ----

>Hi Tom,
>
>> However, using E series values tends to rule out exact multiples-of-2 except in a few notable cases (E24 includes 10K and 20K, but not 40K or 80K, for example)
>
>With resistor arrays you can also do 1/2R, 1/3R, etc.
>
>In the E24 series you're spoilt for choice of 2x and 3x multiples:
>- 100, 200, 300
>- 110, 220, 330
>- 120, 240, 360
>- 130, 390
>- 150, 300, 750 (6x)
>- 180, 360
>
>> All very neat and tidy. I've been experimenting running programs to look at what ratios of "m" and "n" I can generate with common E series values
>
>Yeah, I did that way back around 1994 to generate a big book of
>resistor ratios.  I think I used E12 at the time and had to use a
>small font to fit into the maximum number of pages that the repro
>department could comb bind!  What was most amazing at the time was
>seeing WordPerfect 5.1 crunch on the table data without crashing.
>
>Neil
>
>
>On 22 December 2015 at 21:52, Tom Wiltshire <tom at electricdruid.net> wrote:
>> Using resistor arrays to get 2R or 3R values is a cunning trick too. I'll have to do some experiments and see what Q values you can get like that.
>>
>> I've been looking at this useful paper, "Analysis of the Sallen Key Architecture":
>>
>> http://www.ti.com/lit/an/sloa024b/sloa024b.pdf
>>
>> One of the simplifications it offers is to look at the two R values as a ratio "m" and the two capacitor values similarly as a ratio "n". So the values in the circuit are R, mR, C, nC. All very neat and tidy. I've been experimenting running programs to look at what ratios of "m" and "n" I can generate with common E series values (so E6 for capacitors is a useful simplification, and E12 or E24 for resistors). However, using E series values tends to rule out exact multiples-of-2 except in a few notable cases (E24 includes 10K and 20K, but not 40K or 80K, for example). The resistor array trick is a simple way around that problem, and as such opens up a few other possibilities. Paralleling capacitors would also be a feasible solution if 2C values proved to be useful to have around. E6 certainly doesn't provide them.
>>
>> Thanks,
>> Tom
>>
>>
>> On 22 Dec 2015, at 14:18, Florian Anwander <fanwander at mnet-online.de> wrote:
>>
>>> Hello Tom
>>>
>>> Am 22.12.2015 um 14:35 schrieb Tom Wiltshire:
>>>>  I was looking for ways to simplify that. I'd tried equal-R and equal-C filters, and couldn't get the response I needed, so I decided to try making the two*stages*  identical instead. And it works, or close enough for rock'n'roll
>>> That is, what Roland did in the Jupiter-6 (and later in the MKS-80 too). Beside threee individual resistors they use resistor-arrays for resistors, and use only three R-values for the complete filter circuit. C's are also only two values.
>>>
>>> Florian
>>
>> _______________________________________________
>> Synth-diy mailing list
>> Synth-diy at dropmix.xs4all.nl
>> http://dropmix.xs4all.nl/mailman/listinfo/synth-diy
>
>
>
>-- 
>--
>http://www.njohnson.co.uk
>_______________________________________________
>Synth-diy mailing list
>Synth-diy at dropmix.xs4all.nl
>http://dropmix.xs4all.nl/mailman/listinfo/synth-diy


More information about the Synth-diy mailing list