L298 current sensor pin, convert pulsed current signal to consumption flag

Hi again,

Thank you very much to all of you for your support. I think I finally get it work using the concept proposed by Peter, and also changing the filter configuration and inverting the first comparator, so the consumption flag now is 'active low', which is OK. I don't need the reverse operation signal because I have the information from the signals inputed to the LM298.

Attached is what I tried this time. As you can see, I also changed the voltage threshold to 5mV to avoid problems working near GND in the 1rs comparator. It limits minimum current detection to 10 mA but I think will be OK. With this configuration, the VFILTER voltage is below the reference voltage VREF in the 2nd comparator down to 0,4% PWM Duty Cycle, which is a 1/256 resolution, enough for me, during all the time while PWM action is 'on'. After PWM goes to low, the VFILTER voltage recovers and changes the consumption flag when it goes above 3,4v (VREF). This takes about 110 ms after the last pulse sent, enough for the intended application.

In the simulation below, the PWM signal starts at 50 ms (the flag goes low) and stops at 250ms. After 110 ms, the flag goes high.

Do you see any problems with this concept?

I have to test it in the protoboard to confirm it is OK.

Thank you very much again!

Hi John, glad it's working. By the way what simulator are you using? I like the look of it.

In the real circuit you may need to add some hysteresis to the LM339s. 1Meg between the + pin and the output pin. Also if the input signals are a bit spiky you can add a 100pF cap between the + pin and -pin. That will kill off any spurious output glitches. You may not see glitches and your filter may kill them anyway, but normally that is what I'd do...Peter

Not to be an arch-pedant but remember that the LM339 has an OC output.  The output of the left-hand comparator will discharge the capacitor much more swiftly than the pull-up will charge it.  Maybe that is the behavior you desire, IDK.

As drawn, you are being kinda noodnik regarding PS fluctuation.  You can shunt a cap around the lower legs of your reference strings for a cheap fix.  Or maybe you don't care or maybe the current behavior is what you seek.

Or use a reference instead of a ladder across the supply.

The addition of the hysteresis resistor as suggested above is a good way to prevent jitter.  The next step up the path in that regard is to use two comparators, two references and a flip-flop.

overall - feedback yields stability

overall + feedback yields oscillation or saturation

here we want saturation

If you have a spare -5V supply you could try something along these lines, this might be faster but a little less accurate.  I think Peter had the idea awhile ago, IIRC.

Here are some chips just made for this:

Linear Technology - Parametric Search for Reference Plus Comparator or Amplifier

Hi,

@Peter,

Software: I use Proteus ISIS. It has enough power for what I need, and also it is interesting the interactive controls (switch, motor, LED,...) during simulation. On the other hand, some components models are missing.

Hysteresis resistors: if I add a 1M resistor between +pin and output in the 1st comparator, then the voltage threshold (connected to +pin) is bouncing due to the effect of the capacitor, which produces abnormal functionallity of this comparator. How can I avoid that? In the 2nd comparator there is no problem to add it, I think.

Spiky inputs: I added the caps as suggested.

@Don,

OC comparators: yes, precisely this effect has solved the problematic of making the consumption flag remaining active (low) even at low Duty Cycles, thanks to the quick discharge of the capacitor.

PS fluctuation: if I understood your suggestion right, I can solve this by adding a parallel capacitor to the lower side of the voltage divider, isn't it?

Hysteresis resistors: yes, we want saturation to provide a 'clean' binary flag. From that I understand that Peter's proposal is the right way, but still I got the problem with the 1st comparator as told above.

Alternative circuitry: in the application there will not be a -5v supply. By the way, I need 'accuracy' (in terms of stability of the flag during PWM action) rather than speed of the signal, for my particular application.

Thank you again!

ah, I see, there are too many 1M resistors in that chain that includes the reference, the hysteresis and the pull up. Perhaps you could leave out the hysteresis.

Alternatively, you might improve the circuit with an amplifier on the current sense resistor and you bring the operating current signal up to ~1volt. As Don suggests you could use a Vref for the comparator reference so that you are not dependant on the power rails, which will vary. The comparator reference could be derived with resistors around the 10k level. The 1M pull up could be down to <10k.

Peter

Rather than trying to deal with the tiny current sense signal by amplifiers you could increase the value of the sensor resistor to 10R. It would then drop far too many volts at high currents so you would need to shunt it with a schottky diode capable of handling the full load current.

The loss in the diode would probably be not much worse than in the 0.5R resistor but you would have  a much bigger sense voltage to work with.

Of course the linear range of current sensing would be reduced but since you seem only to want on/off detection this shouldn't matter.

MK

HI john

i need same ckt for my application .i need to use amplifier + comparator + plus shaping ckt to make sense signal to 5 V TTL compatible to feed it to FPGA.

Can you help .