1audio, a common mode component should be reduced by the nature of differential topology(two LPF, two outputs, and inputs). Also, very important the opamp used in such an LPF, -100db is probably lm353 level. It is Sallen-Key as is, the schematics are in any school book, no any tricks only not typically low impedance.
The MFB version is a bit more complex, as I remember I did add 100pF between pins 6 and 7 to get a stable FB loop for opa2156.
Need to say that the MFB a lot more sensitive for the PCB layout.
The MFB version is a bit more complex, as I remember I did add 100pF between pins 6 and 7 to get a stable FB loop for opa2156.
Need to say that the MFB a lot more sensitive for the PCB layout.
Last edited:
Thanks for sharing.
A minor detail: C18 in the MFB filter has a typo. It should of course be 3.3n like C30.
An interesting detail is the use of the additional op-amp on the outputs of the MFB filters. Have you investigated how much this changes the performance, primarily the distortion?
I have just simulated the frequency response for the two filters. There is a small difference in the response. Apart from this, is there any difference in performance, e.g., the distortion?
A minor detail: C18 in the MFB filter has a typo. It should of course be 3.3n like C30.
An interesting detail is the use of the additional op-amp on the outputs of the MFB filters. Have you investigated how much this changes the performance, primarily the distortion?
I have just simulated the frequency response for the two filters. There is a small difference in the response. Apart from this, is there any difference in performance, e.g., the distortion?
Thanks IVX for sharing，
but 4.8K high input impedance May be Low frequency noise(<1K) is too high, I simulated Sallen-Key results of 2rd -12db, 3rd -19db，
FFT Low frequency noise Level May be a problem
but 4.8K high input impedance May be Low frequency noise(<1K) is too high, I simulated Sallen-Key results of 2rd -12db, 3rd -19db，
FFT Low frequency noise Level May be a problem
My experience is the same, can't get any acceptable results using LM4562. Research on the cause lead me to
"Distortion in Positive- and Negative-Feedback Filters
OSCAR JUAN BONELLO
Solidyne SRL, Buenos Aires, Argentina
Journal of the Audio Engineering Society, April 1984, Volume 32 Number 4, pp.239-245"
&
"Practical Optimisation of Noise and Distortion in Sallen and Key Filter Sections
PETER J. BILLAM
Filtek A.G., Biel, Switzerland
IEEE Fournal of Solid-State Circuits August 1979, Volume SC-14 Number 4, pp.768-771"
Both papers were discussed quite often in this thread.
I just took my dusty first LPF version(Sallen-Key) from the junk and remeasure the distortions for you guys.
Bfore:
After:
Bfore:
After:
I made a sample based on the IVX Sallen-Key LPF circuit, the blue one is the LPF, you can see that the low frequency noise is indeed higher, the high frequency noise is less, the strange thing is that my 3rd harmonic has not changed. The op amp is OPA1612 CAP is 1206 and 0805 C0G Resistor is SMD 0603 film type
Last edited:
😳I only have 2.4K or 1.2K 0603 thin film resistors. In fact, there are hundreds of precision resistors with 0.1% 25ppm in my studio. I spent a lot of money. Unfortunately, most of them are 0603 size. For this filter I will try a better resistor for the VCR, at least use a lower noise
I haven't checked mentioned above about papers 1979
Last edited:
The Viking 0207MELF is the best, in that LPF I used 2 in serial but 1pcs is ok up to 10Vrms(simply zero VCR). Caps are more sensitive regarding the VCC, so better to go with 100-250V rated C0G and 2pcs in serial. OPA2156 a bit better vs 1612 as well.
CE Distribution "stocks" the Sound Semiconductor ICs: https://www.cedist.com/search/node/sound semiconductor
Mouser, Profusion and myself stock the THAT2180 but they've been hard to get. Last time I looked Mouser had some THAT 2180 in stock.
Cabintech is also a resource: https://cabintechglobal.com/semi
At Cabintech check out the Alfa Rpar ICs.
The key to using a VCA in a super-low distortion oscillator is to degenerate the VCA so that its overall contribution is quite small. The ULDO Oscillator has the VCA running at a level of around -70 dB relative to the output level. Cordell degenerated his FET and IIRC in Viktor's new design he degenerated the FET.
This is my morning exercise to reproduce Sallen-Key filter misbehave.
Input from SG:
Breadboard:
Result 1:
Result 2:
As you can see, Filter increases THD if pin 2 (In -) short to pin 1 (out) with OPA1656. Resut 2 taken with pin2 in series with 2.4k.
Same time I didn't notice this issue when testing with OPA1652 & OPA1688 (both CMOS). And LM4562 works well in both cases - with resistor and w/o this time.
My conclusion, that papers from 1979 & 1984 right, that recursive positive feedback nature of Sallen-Key filter heavily complicates design. Slightest unbalancing at the input resistance or capacitance (and OPA1656 is not like TL72 they may experimenting with at that time) -- and seems correct design (at least theoretically) does just exactly opposite what it should, instead of attenuation - amplifying THD level !
Input from SG:
Breadboard:
Result 1:
Result 2:
As you can see, Filter increases THD if pin 2 (In -) short to pin 1 (out) with OPA1656. Resut 2 taken with pin2 in series with 2.4k.
Same time I didn't notice this issue when testing with OPA1652 & OPA1688 (both CMOS). And LM4562 works well in both cases - with resistor and w/o this time.
My conclusion, that papers from 1979 & 1984 right, that recursive positive feedback nature of Sallen-Key filter heavily complicates design. Slightest unbalancing at the input resistance or capacitance (and OPA1656 is not like TL72 they may experimenting with at that time) -- and seems correct design (at least theoretically) does just exactly opposite what it should, instead of attenuation - amplifying THD level !
I didn't read these papers, I just said that with decent passive parts and opamp, you can reduce 3rd harmonic to 15db(LPF 2nd ord) from -135db to -150db.
I believe, that way to wipe up harmonics is more effective than RC oscillators:
1) F adjustment isn't required
2) 10x-30x less noise, excellent for the THD+N Guinness book records
3) no any variable resistors for volume control, normal DAC's output makes possible a level sweep plot
4) cheaper, about $4 for all, the cost of the DAC isn't included because everybody has it
I believe, that way to wipe up harmonics is more effective than RC oscillators:
1) F adjustment isn't required
2) 10x-30x less noise, excellent for the THD+N Guinness book records
3) no any variable resistors for volume control, normal DAC's output makes possible a level sweep plot
4) cheaper, about $4 for all, the cost of the DAC isn't included because everybody has it
Correct, I agree on 1-4, filtering is most effective way to build UL THD sine wave source. I already show circuits based on tween-T high Q previously, that I used to test mcp3561 & max11270
What I 'm trying to say, is some type of filter is kind of "diversion" technology. It would show all sorts of wrong doing w/o any apparent reason. Sallen-Key is a good example.
BTW, using tween-T you even don't need audio level DAC at all, internal 12-bits (available for free in almost any uCPU) with -70 dB THD would be O'K, post filtering using only 1 single OPA could reduce distortion down to below -130.
-130db it is my start level which I want to wipe up, I did play with MFB BPF and found that worse than even SK LPF. I'll take a look at T-T in a simulation later.