Frequency doubling at 1MHz using AD633 multiplier

[Fizdiz]

Hello,

I have a question on the AD633 multiplier, which I'm using as a squarer and with which I'm supposed to see frequency doubling for a sine wave input.
But at a frequency of ≥1MHz, the output is heavily distorted (looks like a sawtooth wave and heavily attenuated) and its frequency is about the same as the input frequency (i.e. no frequency doubling is seen).

I know that one should see attenuation at very high input frequencies (due to the op-amp gain roll-over) but I wouldn't know why frequency doubling is not observed anymore.

Any ideas on why this is would be highly appreciated


Here is the data sheet for the AD633:
https://www.analog.com/media/en/technical-documentation/data-sheets/AD633.pdf

[Paul Rose]

Maybe see the graph at Figure 9 of that data sheet: "AC Feedthrough vs Frequency"
The X and Y feedthrough peak pretty strongly at 1Mhz (and that's on a graph with logarithmic axis).
Maybe your inputs are swamping your intended output.

[Fizdiz]

Thanks, that makes a lot of sense!

[David Hess]

The AD633 is just not suitable for operation at that frequency.  You would do much better with a Gilbert cell even made with discrete transistors.

[Paul Rose]

I kind of miss the CA3054 transistor array.
6 transistors pre-connected just right for making a gilbert cell, but allowing you to bias how you want.

[David Hess]

I kind of miss the CA3054 transistor array.
6 transistors pre-connected just right for making a gilbert cell, but allowing you to bias how you want.

Many of the early Tektronix custom ICs were specialized versions of the CA3054, and this is why their 2213/2215 series of oscilloscopes could use the CA series transistor arrays instead of custom Tektronix ICs.

[Fizdiz]

I've got another question, for the same circuit (i.e. sine wave input on both AD633 inputs, but this time at 1kHz).

When I calculate the scale factor from the input and output signals of the AD633 (which should be 1/(10V)), I get a much lower value when the input is only positive than when the input is only negative. (Scale factor when input is positive: 9.8, scale factor when input is negative: 9.9)

Is there a possible reason that the AD633 behaves differently for positive or negative inputs? (Note: the Ad633 is used as a squarer so both inputs are always on the same voltage).

[David Hess]

I've got another question, for the same circuit (i.e. sine wave input on both AD633 inputs, but this time at 1kHz).

When I calculate the scale factor from the input and output signals of the AD633 (which should be 1/(10V)), I get a much lower value when the input is only positive than when the input is only negative. (Scale factor when input is positive: 9.8, scale factor when input is negative: 9.9)

Is there a possible reason that the AD633 behaves differently for positive or negative inputs? (Note: the Ad633 is used as a squarer so both inputs are always on the same voltage).

That difference is within the 1% error from nonlinearity.  Be sure to adjust the offset first.

[Fizdiz]

Thanks for the answer but I don't fully understand (I'm a total beginner on electronics).

The error on the 1/10V scale factor is given as 0.25% in the data sheet, so I thought that the difference between the measured value would be 1%=4x0.25% which is quite large.

How should I understand the parameter "nonlinearity" in the data sheet, which you were referring to? Is it correct that it is the maximum difference between the expected and measured output?

Also, do you mean that I should adjust the input DC offset? Indeed there was a small input DC offset, but I am not sure how that would affect the scale factor.

[David Hess]

The maximum X non-linearity is specified as 1% and the Y non-linearity 0.4% and since you are using both, they both contribute error.  The similar X and Y feedthrough each also contribute to output error because of how you are using the device.

The scale voltage error and output offset voltage can be calibrated out.