Sallen & Key 2-Order unity gain filter distortion and noise

[Tompson17]

Hi everyone,

I designed an active crossover. The function may be - band pass, 4 order, from 2,5kHz to 15kHz. The realization is cascade of 2xsecond order butterworth low pass and 2x second order high pass filter. Before these filters is signal follower for avoiding impedance issues. The op amps I used is MC33079 - so the filter is designed for one chip, the signal follower is some previous chip. Power supply +/- 15V, the filters are unity gain, resistor values are high enough for normal operation, capacitors are polypropylene, the opamps are bipolar input, everything seems to be in good condition, no overload, chip supply pins are bypassed with 100nF ceramics...

But i get distortion and relatively high noise (or intermodulation... don't know) from output of the system. The distortion and noise comes not from whole output, but it is measurable behind the first filter. The distortion ratio is about 70 dB or 80 dB, depended on input frequency and voltage. The distortion significantly rises when the input is more than -10 dBu, and the ratio is "the same" up to 15dBu. Output noise (maybe intermodulation) is filtered by the main function of the system, but rises, when the signal gets higher than -10dBu, like the hardmonic distortion.

In simulations, burrbrown OPA134 gets "some" better results beside of the MC33079, but Douglas Self measured "the same filters" with NE5532 with no issues... Only difference is input bias current, but I dont know if it is the problem...

The schematic and the layout is on attached pictures.

Many thanks for all your replies.

 

[Tompson17]

Schematics

 

[Khron]

"... it is measurable behind the first filter" = pin 14 of IC11(D)? Or pin 7 of IC15(B)? And i guess it's only on that "midrange" (2.5k-15k) channel?

 

[Tompson17]

"... it is measurable behind the first filter" = pin 14 of IC11(D)? Or pin 7 of IC15(B)? And i guess it's only on that "midrange" (2.5k-15k) channel?

pin 7 of IC15B. Other channels get some small distortion too, but it is well under 90dB ratio with 0dBu nominal signal level. Only this filter is problematic... note that this filter is the farthest, the signal follower (IC11D) is about 250mm from this stage and the signal is above power supply planes (on top layer) and on both sides is GND. Signal is still good on C109, the problem is on IC15 output...

 

[Khron]

You could try removing C111 and see what the signal's like on IC15's pin 7 then. That way you'll isolate it, and then you can know whether the problem is inside that particular opamp, or whether something's loading down its output.

 

[Tompson17]

I tried to remove the C111, but nothing has improved. I tested the distortion and (that is my mistake) the distortion comes from the previous section. I have some differencial input, then there is passive volume control with 6 pos. switch. The voltage divider is driven by op amp as signal follower, coupled with polypropylen cap, after the switch is another signal follower which drives next 4 followers (IC11) for filters. The distortion is significant after IC22B (see attached full pdf schematic) - on its output. On its input is distortion wery low. I thought that the problem can be 4 times input capacitance of op amps, which may be 4x12=48pF, but after inserting the 47R resistor, nothing was better.

At the beginning, when I firstly turned the whole system on, the system had only one polarity of power supply active (mistake due to bad transistor library pinout). Can the MC33079 be half cracked so the high levels can be distorted because of internal collision but in other conditions can everything be all right?

 

[Khron]

Mmm... In that context, i'd be tempted to replace IC22 altogether.

So the signal's ok at pin 1 of IC22, and also at pin 5, but "broken" at pin 7?

That being said, that input stage looks a bit 'overcomplicated' to me (to say the least)...

 

[Tompson17]

Yeah, it is bit complicated, I tried to get the best SNR... The input unit may be simpler, but... this time it is so.

Yep, as You said, IC22 pin 12, 14, 5 all right but pin 7 broken... I try to replace the op amp and let You know.

 

[Khron]

I'm sure things would've been just fine with something like Rod Elliott's balanced line receiver design.

http://sound.whsites.net/project87.htm

Further details: http://sound.whsites.net/balance.htm

 

[Tompson17]

So, I have some peculiarities ...

Problem for the MC33079 can be capacitance maybe. I did some measuring before the 47R (op amp output) rezistor and after (another op amp input). It looks like the input capacitance of my measurement system (Audio Precision AP 582) is too high for direct driving from op amp output.

So, can be the problem the capacitance of 25cm long trace from op amp output? Maybe some small value resistor right after IC11D output...?

 

[Khron]

That may well be a possibility. What looks like distortion could (also) be instability / some form of oscillation.

 

[Tompson17]

I replaced IC14 from MC33079 to OPA1654 and the problem is only after the last stage (IC14C). For measuring I use 100R resistor in series with AP analyer. IC15 is still bad at ic15B output.

I have 100nF caps under every IC between supply rails and on top side between supply and GND. But, can be the instability caused by "long" trace between "GND" pins from cap1 to cap2? Both are connected to the ground plane and the real trace can be about 5cm due to another components around op amp. May be the decoupling caps connected to each other as close as possible (can be the ground plane problematic)?

 

[Khron]

Well, opamps don't really have a connection to ground internally, so bypass caps from each rail to ground aren't exactly "necessary" - one between the rails (V+ to V-) should suffice.

http://sound.whsites.net/dwopa.htm#s32

 

[Tompson17]

I fu**ed it... Can be polyester caps in filters the main problem?... I looked one more time into mouser.com shopping card list and... the caps I bought are polyester. Mistake made by my inconsistency while making BOM...

 

[Khron]

And why / how exactly would the cap dielectric be the problem?

 

[Tompson17]

In Douglas Self's books is something about filters and there are measured filters with polypropylene and polyester caps, resp their THD. In that measurements is significantly rising od fistortion when the signal is high. True is, the test conditions he aplied is 10V rms signal and I am talking about 3V...

This is specialy commented for Sallen-Key filters. In low pass filter may be the cap on the + pin of an op amp polypropylene, the other (from output) can be polyester without suffering of distortion...

 

[Khron]

I don't suppose you've considered taking some photos / screenshots of the distorted waveform, have you?

 

[Tompson17]

So, 1) is the 2,5 kHz low pass - output 5 and 6 (the same) when i test THD + N with 1kHz sine wave +14dBu on generator (+10dBu inside system) - spectre

2) THD + N result for 1)

3) 2,5 kHz high pass - output 7 and 8 when i test THD + N with the same 1 kHz sine (the same 10dBu inside) - spectre

4) low res 3)

5) result THD + N for 3)

6) spectre of output 7 with no signal, all other conditions the same, only without signal...

 

[Tompson17]

1

 

[Tompson17]

2