Crazy idea for a cap multiplier application

It's easier to manipulate the frequency with multipliers and mixers than fool around with tuning.

This kind of circuit has srs probs for small capacitances and/or high frequencies. Sometimes it's not even stable for low ones, either:

Muck with high frequency voltage-feedback cap multipliers at the risk of your sanity; unlike inductance gyrators which tend to be fairly well-behaved all that voltage-feedback cap multipliers for applications other than supply filtering/low frequency really wanna do is oscillate, latch up, or generally jump off.

Don't assume a transistor is just a DC-to-daylight current multiplier or an op amp is an ideal op amp to analyze them, at least.

Basically the simple ones suck for signal processing. There are good high-frequency cap multipliers used in ICs but I think they tend to use topologies you can't do easily with discretes for high frequency like high-performance OTAs and current conveyers.

The issue is not frequency response (at least in a power filtering operation, the transistor is as good as the electrolytic filter) but impedance. The emitter resistance of an ideal bipolar transistor, at low currents (1 mA) is about 0.025V/I_emitter, so that's gonna be 25 ohms. It gets better if you're supplying more current (200 mA, closer to 0.13 ohms) so it's good for a power supply, but that 25 ohms series resistance isn't optimal for high-ish frequencies if you want a pure capacitance.

MOSFETs, while touting low ON resistance, are worse for linear operation than bipolars in this respect.

That's a low frequency application.

It is, what I mean is that embodiment doesn't translate too well to other applications (even with an op amp that isn't as old as I am)

What do you mean by cap multiplier? Got a sketch?

Tube-type FM transmitters sometimes used a "reactance tube" oscillator, which was in fact a cap multiplier.

A cap multiplier works better with a faster transistor, because it reduces the C-B feedthrough.

The main difference is that for a power supply you don't care much about the Q of the capacitance, whereas for tuning you care very much. Since a power supply C-mult is an RC lowpass with an emitter follower hung on it, it's completely unsuitable for tuning.

If you take a 1-pole cap multiplier and move the cold end of the cap from ground to the emitter,

0--*---------* *---------*---0 | \ A | | ----- | | | | | | *---RRRR-----*------| |---* | |

you get a simulated (gyrated) inductor. Using one or more varactors for the capacitor, you could make a tuned circuit. You have to watch out for biasing issues, of course.

There's not a lot of point in using varactor tuning for AM radios nowadays, so the bigger ones such as the MVAM series are long gone.

There are still highish-capacitance varactors around, e.g. the BB201, which work fine for signal-level applications above a few megahertz.

Cheers

Phil Hobbs

Thanks to all who commented. I was pretty sure that if this could be useful it would have been done long ago.

I'd be perfectly prepared to try it for the right application.

Cheers

Phil Hobbs

Ya, it helps if the circuit is expected to be oscillating to begin with which cap multiplier circuits tend to want to do.

A two-pole circuit coupling the plate to grid can make the admittance looking into the plate of the tube appear negative, and can exploit the Miller effect to make either the capacitance or inductance from grid to plate appear bigger. then since the Miller-multiplication is a function of gain you can modulate the frequency with control grid bias.

Pentagrid converter tube works well for a one-tube low-level frequency modulator like this, with feedback on one signal grid and AF on the other.

The current-mode c-mulitplier seems more suited to high frequency and sounds simple in theory, sense current through the capacitor, multiply sensed current by a gain > 1, apply it back to the sense node with a current-controlled current source.

Does this become better-behaved if you add emitter resistance? Or does that kill the Q?

I haven't used one at high frequency, but they're commonly used in telco applications to separate signalling from DC power. For that, I added 3 ohms of Re and it seems to work fairly well.

CH

I haven't gone through the math lately, but I expect that adding emitter resistance will reduces the loop gain, and so will reduce the Q at high frequency. Same sort of deal as emitter-degenerating a current mirror--the matching improves at DC but the high-frequency performance tanks.

The main issue, as with most all simulated inductors, is that it's up to the designer to make sure there's enough supply headroom for the simulated inductive kick when the current changes rapidly. With a high-beta transistor and the right choice of R and C, it can work fine for lots of things. (One probably woouldn't do it in the RF amp of a communications receiver, of course.)

Cheers

Phil Hobbs

See I don't understand that. The gyrator transistor Re ought to be reduced by its current gain. As long as you aren't pushing Ft it should be ok, just servo-ing the output current?

No, but the question is whether it would work in a tuned circuit... perhaps with a varactor tuning the inductance as well as another to resonate with.

Clifford Heath

Gyrators were once all the rage for cheap mass produced high Q bandpass filters in the days before digital signal processing was possible.

These days software defined radios are quite common and more easily implemented. Several TV tuner dongles can be subverted this way.

No chance of working for any *quality* *HF* oscillator.

Capacitor multipliers can only work because they have gain at HF to generate the multiplication. This is a major problem even at 10 MHz.

Secondly, it will generate huge amounts of Phase Noise.

There is no such thing as a free lunch...

-- Kevin Aylward

SuperSpice

We have a couple of products that use a digital capacitor (like NCD2100TTR) for coarse oscillator tuning and then a varicap for fine tune.

At powerup, we center the varicap voltage and step the digital cap to find the code that's closest to our target frequency.

These Ixys parts are deliberately non-monotonic, so we sweep through a lot of steps and then pick the best one. Can't binary search.