EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: npelov on February 14, 2023, 09:12:33 pm
-
Hi,
I was looking at comparator datasheets and the propagation delay was specified "200 mV Step with 100 mV Overdrive" or "100 mV Step with 5 mV Overdrive". What does this mean.
Example: AD8561 (https://store.comet.bg/download-file.php?id=7723)
-
It is the amount above the reference. Let's say the reference (negative input) is 500 mV, then for "200 mV Step with 100 mV Overdrive" the voltage on the positive input was 400 mV, then jumped to 600 mV making for a total jump of 200 mV with 100 mV overdrive.
For the second one the initial voltage on the positive input was 405 mV and jumped to 505 mV.
-
It means the voltage difference between the inverting and the non-inverting inputs.
Unlike opamps that are usually used in feedback configuration where the input voltages are the same (linear operation), comparators are designed to be used with input voltages that drive them into output saturation.
Your spcifications simply mean that the input voltage (referenced to the other input) is driven from -100 mV to +100 mV or from -95 mV to +5 mV.
-
Thank you! It's clear now. So if I have configure hysteresis it'll switch faster, because once the output starts moving it'll increase the overdrive.
-
Thank you! It's clear now. So if I have configure hysteresis it'll switch faster, because once the output starts moving it'll increase the overdrive.
Hi,
Really the speed depends on the slew rate. Check that spec out next.
-
Thank you! It's clear now. So if I have configure hysteresis it'll switch faster, because once the output starts moving it'll increase the overdrive.
Not necessarily. Hysteresis is used to defeat noise and shifts on slowly varying input signals. If your input signal is already fast, hysteresis won't change that.
-
But doesn't hysteresis provide positive feedback. When you get over the reference voltage set by R3/R4 divider then the positive input is sum of Vref and output voltage (proportions set by R1 and R2). So Let's say the comparator output is zero - the switching voltage will be a little bit lower than Vref because the output pulls positive input down. then when you get over that voltage the output starts to turn on which pulls the positive input up a bit (overdrive?), so the difference between positive and negative becomes even bigger. According to datasheet bigger difference - faster switching.
-
But doesn't hysteresis provide positive feedback. When you get over the reference voltage set by R3/R4 divider then the positive input is sum of Vref and output voltage (proportions set by R1 and R2). So Let's say the comparator output is zero - the switching voltage will be a little bit lower than Vref because the output pulls positive input down. then when you get over that voltage the output starts to turn on which pulls the positive input up a bit (overdrive?), so the difference between positive and negative becomes even bigger. According to datasheet bigger difference - faster switching.
Well the overdrive forces a faster transition but only at the start of the input change. The slew rate is the main issue because you can not go any faster than that allows.
The whole sequence is like this...
There is presumably a constant voltage on one of the inputs and the other input does not yet cause a change.
Then, the other input rises a little bit, which gets a tiny bit beyond the input offset combined with the constant voltage on the other input.
Now that causes the internal part of the device to start rising, but because of limited gain (even though it is high) the internal part can not rise all the way yet because the small difference multiplied by the internal gain does not equate to a high enough amplitude (for example 1uv and an internal gain of 100000). This means the input has to change a little more by itself as the output has not changed yet. Once it changes a little more the internal gain allows the output to start to rise, but only at the limit set by the slew rate, and probably lower than that. As it rises it starts to contribute to the input that is changing due to the feedback resistor, and that forces that input up higher. As it goes higher the gain now allows the internal part of the device to rise higher which in turn makes the output (eventually) to rise limited by the slew rate now. The output slews up and during that time there can not be any speedup because the output rate of rise of voltage is limited by the slew rate. So the only increase in speed comes at the very beginning of the change on the input that changes.
This may be observable in a simulation by allowing the input that changes to rise very very slowly while watching the output voltage. It would look like a very short time when it rises slowly, followed by a longer time as it slews up at the maximum rate.
In any case to understand the speed of these devices as well as op amps the slew rate is a good thing to study. That will always play a part in the speed. For an example, the LM358 slews at about 0.5 volts per microsecond, so to get to 5 volts it would take 10 microseconds. If you use that as a comparator (and that has been done) the significant part of the speed will be due to the slew rate not the input overdrive.
The input overdrive is something that just gets us past the initial response where the internal gain has a profound effect on the speed, but that's not very long. It is also a way to get past any noise so you can specify a given speed without worrying that the input will change state due to the noise. It's pretty much a guarantee that your measurement will be accurate.
-
Well I was looking at propagation delay which does not actually affect the switching speed - which confirms what you said. Rise and fall times are 3.8 and 1.5ns in this datasheet. Only propagation depends on overdrive. I can see the propagation delay is double the rise time. So if the comparator is switching at 14ns period - 71MHz, the output will be shifted at 90deg.
The hysteresis might help for more consistent switching (less jitter). If the overdrive is not much there might be jitter (just guessing) and hysteresis will help to have more consistent transition. Let's say I want to use this comparator for PWM generation from triangle wave.
Did I get it right this time?
3.8 ns is quite good for a comparator, isn't it. This is the fastest I could find locally. I was looking to make a simple function gen using AD9833. It has sine and triangle. But I want to have square at 50% duty and PWM. The square might be better to be generated from sine because it provides faster transition around zero. And if I need pwm I can switch to triangle.
As for opamps I never had to warry about slew rate. I never worked with high frequencies. Most of the time I add low pass to reduce the frequency so I don't have to sample that often. Although I might need it now because I will probably need a buffer the the function gen.