Unkown Low Pass Filter Design

[tridac]

Hi,

Recently been restoring Fluke 5100 and 5200 calibrators, which both use similar design low pas filters to clean up a pwm generated voltage. For a reference, the key thing is that the filter op amp is not in the signal path, so op amp offsets have no effect at all. A handful pf parts and one op amp, a very cool circuit. The 5200 filter is similar just another opamp and caps to produce a 5 pole filter.

Looks like some sort of gyrator design. Have searched the web for design info, including the Fluke patents list, but can find no info anywhere. Anyone have any ideas ?. Rough 5100B filter schematic attached...

Thanks,

Chris

[mikerj]

There is a whole paper on it here: https://ir.library.louisville.edu/cgi/viewcontent.cgi?article=2077&context=etd

[macaba]

A fantastic technique (in fact, I last built one of these filters a few days ago). I’m also interested to see original sources. The oldest one I’ve found so far is in a book “Applications of Operational Amplifiers - Third Generation Techniques” published in 1973. Fig 3.35.

[moffy]

Interesting post, I wonder if the LTC1062: https://www.analog.com/media/en/technical-documentation/data-sheets/1062fd.pdf  even though switch capacitor works on the same principle since it uses a resistor in series and it has zero offset.

[tridac]

Hi,

Thanks, just found a copy of that on ABE. Afaics, Fluke were using that technique as early as  1975, years ahead of the pack at the time. Can learn so much reading some of the old schematics...

[tridac]

Thanks for the link, excellent...

[David Hess]

Tektronix used the same circuit in their multimeters which were part of analog oscilloscopes.

Update: Added example from Tektronix DM44.

[tridac]

Here's the one in the Fluke 5200 AC calibrator, reference board, 1979 vintage. A 5 pole filter, analog magic. Don't have spice here, but could this be modelled and how to calculate the R and C values ?.

[tszaboo]

It is an excellent capacitive multiplier design, I used it myself to clean up PWM dividers in very high accuracy reference voltages. There are a few thing to keep in mind when using it though.
The leakage current of those capacitors can significantly change the output voltage, so low leakage maybe even film capacitors might be required. The opamp selection is relatively easy, you want a low noise amplifier, DC accuracy doesn't matter, even a 5532 is a good choice. The output impedance of the circuit is high, so you will want to buffer it before using, which is not easy. And be mindful of vibrations. I was surprised when the small SMD ceramic caps picked up with microphonics a lot of noise from fans, and drove it directly as noise.
Where did I see it? Datron 4910. They also have some unobtanium matched JFETs on the opamp input, I think to turn it into a JFET input amplifier.

Here's the one in the Fluke 5200 AC calibrator, reference board, 1979 vintage. A 5 pole filter, analog magic. Don't have spice here, but could this be modelled and how to calculate the R and C values ?.

It's easy to simulate it, and it's very insensitive to the values as I recall, you can change it and the only effect was moving the filter cutoff frequency.

[tridac]

The interest here is to clean up low cost and possibly noisy voltage references, and yes, an output buffer may be needed. The common cap values make sense from a production point of view, perhaps polypropylene, but the resistor values are pretty specific, metal film. Input resistors are 35.885 MF + 34.8R. so how did they arrive at those values ?. Too early for cad, unless they had their own package running on a mini.

Edit: A bit more digging turns up an article from EDN in 2019. Includes equations, but fails to tells us what the S parameter is in the equations is :-(.

https://www.edn.com/zero-offset-active-lowpass-filter-part-1/

[Alex Nikitin]

Here is one more example, from Keithley 263 Calibrator/Source (1987).

Cheers

Alex

[Kleinstein]

The interest here is to clean up low cost and possibly noisy voltage references, and yes, an output buffer may be needed. The common cap values make sense from a production point of view, perhaps polypropylene, but the resistor values are pretty specific, metal film. Input resistors are 35.885 MF + 34.8R. so how did they arrive at those values ?. Too early for cad, unless they had their own package running on a mini.

Edit: A bit more digging turns up an article from EDN in 2019. Includes equations, but fails to tells us what the S parameter is in the equations is :-(.

https://www.edn.com/zero-offset-active-lowpass-filter-part-1/

The original spice is quite old - it started running on a mainframe / minit computer. The circuit is also not that complicated - with enough patience paper and pencil can also do the trick.
The resistors values are not that critical. If changed a little one is a bit off the standart ( likely butterworth ) filter function, but still no bad. For just filtering a reference only does not really care about the exact cross over-frequency or settling speed. In this case polyster capacitors may be good enough and not need for faster settling PP capacitors.

[TimFox]

The interest here is to clean up low cost and possibly noisy voltage references, and yes, an output buffer may be needed. The common cap values make sense from a production point of view, perhaps polypropylene, but the resistor values are pretty specific, metal film. Input resistors are 35.885 MF + 34.8R. so how did they arrive at those values ?. Too early for cad, unless they had their own package running on a mini.

Edit: A bit more digging turns up an article from EDN in 2019. Includes equations, but fails to tells us what the S parameter is in the equations is :-(.

https://www.edn.com/zero-offset-active-lowpass-filter-part-1/

The original spice is quite old - it started running on a mainframe / minit computer. The circuit is also not that complicated - with enough patience paper and pencil can also do the trick.
The resistors values are not that critical. If changed a little one is a bit off the standart ( likely butterworth ) filter function, but still no bad. For just filtering a reference only does not really care about the exact cross over-frequency or settling speed. In this case polyster capacitors may be good enough and not need for faster settling PP capacitors.

"s" in the equations in the EDN article is a variable, technically from Laplace transform equations.
For complex algebra and AC, you can substitute s = jw, where w = 2(pi)f (should be lower-case omega).

[tridac]

Tim, thanks, afrer decades in embedded real time, analog design rusty and was never more than average in that area anywy :-)...