Can I get a 90 degree phase shift from 2 low pass filters?

[brett]

Hi
I have an oscillator sine wave signal (fixed at 20kHz) that I want to blend with a small amount of signal lagged by 90 degrees (quadrature signal).  The blending will be done with a single JFET mixer.

As I understand things, a low-pass filter gives a phase shift of 45 degrees at the cutoff frequency (and a "gain" of 0.707). 
To get a signal with approx a 90 degree shift, can I simply put two LP filters (with cutoff of 20kHz) in series?  I presume that the shift will be 0.5.

Is there some reason why I would I be better off with an active all-pass filter?

thanks 

PS Nearly every day I think I must learn to use SPICE.

[R.G.]

Yes, you will get a phase shift of 90 degrees with two low pass filters.

It's probably better to use two allpasses to avoid having to muck with changing amplitude as you tune in phase. It will almost certainly require tuning each unit built, because the component tolerances will change the exact cutoff.

You might want to try a single-frequency dome filter. This is two parallel series of allpasses arranged so that the difference in phase between the two series is close to 90 degrees at all frequencies. The component tolerances will still require tuning, but it won't be quite as sensitive as the single-filters setup.

... I think. 

[brett]

Thanks heaps RG! (again)
I'll use a variable resistor in the first stage and "tune in" a total of approx 90 degrees with my scope.
cheers
PS My application is far from exacting (80 to 100 degrees of lag would be OK).  I'm cancelling some or most of a slightly lagged signal (a few degrees of lag?) by mixing it with an inverted mixture of the unlagged and lagged signals.  (I suspect that the best mixture will be about 95% unlagged and 5% lagged) The project already works without this cancelling, but it will be cream on the cake if I can do it.

[Cliff Schecht]

If you're only at one specific frequency, allpass filters are the way to go.

[alanlan]

I take it you want a quadrature shifted signal at all frequencies - now that is hard to do.  If it were digital, you're looking at Hilbert transforms - I don't know how to do it in analogue.

[earthtonesaudio]

Not much to add to what everyone else has said, but if your frequency is fixed like you said, and you are fine with the lagged signal being reduced in amplitude, a 2 pole lowpass like you first mentioned seems like it will work. 

[brett]

Thanks everyone.
It's just one frequency.  About 20kHz.  I get phase-lagged "bleed" from a transmitter coil to a receiver coil.  I think the bleed would be 0 degrees shifted if the coupling was purely resistive and 90 degrees lagged if it were inductive.  Whatever the reason, it seems to be slightly lagged (a few degrees).  Anyway, I'm going to mix X of the transmit (0 degree lag) signal and Y of the 90 degree lagged signal (where X is much greater than Y) and use that to cancel some of the bleed-through.

This type of cancelling gives some ideas for bleedthrough cancelling in ring modulators and aliasers (maybe feed a variable amount of the switching signal into the opposite side of the output op-amp?).

thanks again

[R.G.]

I take it you want a quadrature shifted signal at all frequencies - now that is hard to do.  If it were digital, you're looking at Hilbert transforms - I don't know how to do it in analogue.

Look up "dome filter". Two chains of allpass which are arranged so that each chain causes a linearly-increasing phase shift with frequency. The starting frequencies are staggered so that (if you did your math right  ) the outputs of the two chains are both 90 degrees apart at all frequencies significantly above the lowest start frequencies.

[earthtonesaudio]

This type of cancelling gives some ideas for bleedthrough cancelling in ring modulators and aliasers (maybe feed a variable amount of the switching signal into the opposite side of the output op-amp?).

I used exactly this on a recent (unpublished) version of my proximity wah, with great success.  Total clock null, even with clock and audio on the same chip (quad op-amp).  A very handy technique indeed.

[alanlan]

I take it you want a quadrature shifted signal at all frequencies - now that is hard to do.  If it were digital, you're looking at Hilbert transforms - I don't know how to do it in analogue.

Look up "dome filter". Two chains of allpass which are arranged so that each chain causes a linearly-increasing phase shift with frequency. The starting frequencies are staggered so that (if you did your math right  ) the outputs of the two chains are both 90 degrees apart at all frequencies significantly above the lowest start frequencies.

Thanks for that snippet, it's very interesting that is....

I've seen a "dome filter" circuit using 2 chains of 6 op-amps (but I'm convinced it can be done in half that number) to produce two outputs 90 degrees apart but the absolute phase of each chain moves all the time,.  By chaining these two chains together, you could with a bit of manipulation get a constant 90 degree phase shift over any band you like up to reasonable limits.  So I guess it would be literally a Phase 90.