# Switching window in comparators

1. Sep 14, 2007

### Gokul43201

Staff Emeritus
I'm looking for a complete definition of the switching window in the context of comparators.

In particular, I want to make sure I'm not misunderstanding how hysteresis is implemented in the AD630 (Analog Devices) chip.

See page 2, in the spec table for the comparator, where the switching window is specced. Also see page 6, under 'Circuit Description', where it says:

Does this mean that no switching will occur until the 1.5 mV differential signal is exceeded? In other words, if one input is grounded, and the other input has a stable sine wave, then switching will occur exactly at 1.5 mV above (below) the zero crossing during the increasing (decreasing) half-cycle? Also, this should introduce a phase shift of roughly 10-3 in the output for say, a 1V input signal. Is that right?

And finally, what is the point of this? The only sense I can make out of it is that it avoids instability around the zero-crossings due to noise in the input signal. But this is already taken care of by the recommended hysteresis circuit in (figure 7) page 8. If I understand that part correctly, it makes the output switch at a differential signal of 0.5 mV (above and below zero). Why would they recommend a 0.5 mV hysteresis circuit, when the chip has a built in 1.5 mV "safety window."

I think my understanding of the term 'switching window' is flawed. Does it in fact mean that switching could happen anywhere within the 1.5 mV window beyond zero, with an essentially 100% probability of switching beyond the window, and some non-zero (maybe sigmoidal) probability distribution within it? What other mistakes am I making in my reasoning?

Please be gentle with me - I'm not traditionally trained in electronics.

Last edited: Sep 14, 2007
2. Sep 14, 2007

### dlgoff

It looks like the switching window is the range of voltages that can be applied to the inputs based on the supply voltages ( -Vs to +Vs) with a ±1.5mV variation from chip to chip.

"And finally, what is the point of this? The only sense I can make out of it is that it avoids instability around the zero-crossings due to noise in the input signal. But this is already taken care of by the recommended hysteresis circuit in (figure 7) page 8."

As shown it looks like you are correct in it being about 0.5mV but you might want to use the circuit to get a larger hysteresis by changing the resistors. It's just an example.

Regards

3. Sep 14, 2007

### Gokul43201

Staff Emeritus
I don't think I'm following you here. This doesn't seem consistent with the bit I quoted from page 6 or with the line in the spec table that says the window itself is ±1.5mV. Or maybe, I've misunderstood what you're saying.

Yes, though in my case, I don't think I will need any more than that - my signal is very clean with noise < 10ppm. But this raises a related question. If I had say, 10 muV (pk-to-pk) noise on my input, then that's a maximum time interval between noise spikes of dT = V(noise)/(slope at V=0) = 10 muV/ (2Vo*pi*f) which is only about 200 ns even at low frequencies (~20Hz). Do comparators even switch that fast? Figure TPC 8 (bottom of page 5) doesn't seem to give me a conclusive answer. I recall that gate charging time in a typical FET is of order 100 ns. So the question is this: Do I even need a hysteresis adjustment if my noise is this low?

Last edited: Sep 14, 2007
4. Sep 14, 2007

### Staff: Mentor

I read the post earlier today and looked at the datasheet, but to be honest, I was not understanding the circuit well at all, and gave up trying With some more effort I may be able to help, but I've honestly never seen that topology for a "mixer" before, so there would be a learning curve.

The only small thing that I can try to answer (and I'm not sure that this applies to this particular part, since I don't understand it yet), is that the positive feedback around a comparator is not just for handling input noise. Often with comparators, even with good PCB layout, grounding and decoupling, the transient from the output switching is enough to cause a transient in the input circuit, and you can get multiple output buzzing switches from that feedback path. The positive feedback of the external hysteresis is meant to prevent this internal transient feedback from causing output buzzing. Not sure if that helps your questions, Gokul.

Hopefully one of the other EE folks will have seen this topology of doubly-balanced mixer before, and be able to deal with your questions more intuitively. Mostly I've used Gilbert cells in all of the doubly-balanced mixers that I've worked with.

5. Sep 15, 2007

### Gokul43201

Staff Emeritus
I believe this is quantified somewhat in the datasheet. I'll think about it a little more. Thanks, berke & dlg.