I have an pulse generated by an op-amp and frequency generator. the output of the op-amp drives a 1 amp LED. The intensity of the LED is important to my system design, in which it must remain constant throughout the high side of the pulse.

I have a problem that there is overshoot and ringing in the output pulse, which affect my LED intensity during beginning of the pulse. I'd like to filter out the overshoot and ringing, but still keep the 'square'd corners' of my pulse output. Can someone recommend circuitry (not software) to do this?

Output amp is Texas Instruments' OPA548, pulsing at about 12VDC and running on 24VDC supply; on-chip current limited to 1 amp out.

I can upload a schematic later today if needed. Thanks.

 

overshoot and ringing are usually caused by inductance in the output for the pulse. lead lengths are also a cause.

 

a schottky diode clamp circuit might help with overshoot.

 

What frequency is this square wave?

Best to reduce the rise and fall time if possible. Also, make sure that you have a low-inductance capacitor as close as possible to the supply pins of the op-amp. A 100nF surface-mount ceramic would be good. Two of these in parallel even better.

A series resistor as close as possible to the output pin will also help. I know this will loose power. Just a fraction of an ohm will help. Must be non-inductive. You could try a small ferrite bead on the IC lead but this might make it worse. The wire between the output and your LED load needs to be as short as possible - no more than a few inches really.

 

What does the output of the op amp look like with no load and with a 10 ohm resistive load directly at the output?

 

You need to post your schematic AND your circuit layout. If you have long wire runs, that is the major problem. With current that high, you will also need multiple bypass capacitors across the opamp supply pins.

 

Wow! lotsa replies... I am new to this forum, so 1st, a big thanks for the replies!

Here is the schematic. Basically, with a pulse at the EN input, the output of the last amp is driven to a voltage level dependent on the frequency input at DIM_MC (which goes through frequency-to-voltage conversion at U1).

I cannot reduce the rise time, and am stuck with long wires to the LED. I assume I'll have to come up with some sort of filtering circuit at the output. (I have already tried putting a 10 ohm resistor in series with the output. This has reduced,but not eliminated the intensity fluctuations.)
At the output of the circuit, I am seeing about 18VDC pulse, with 4 volt overshoot. The ringing is about 312kHz. It appears that there is also some capacitive-like decay after the ringing as well, in the amount of 2.2VDC.

I think it is this lack of symmetry (2.2VDC - see o'scope pic at bottom) that I will need to improve upon, in order to make my 'fluctuations' go away.
Fluctuations: There is a digital camera that takes snapshots during this pulse, so I need to make the pulse as flat as possible during the 'high' state. As you can imagine, if the camera snap varies in duration or pulse width location, it will display fluctuations in intensity related to the pulse's voltage variation.

Ideas how to reslove?

 

So, I was able to get rid of the initial ringing using a couple of clamping diodes, but still have a voltage decay of almost 2V. How can I fix that?

Any more thoughts? Suddenly, there were no more replies.....hope I didn't scare you away with this one....

 

My first impression is that the resistors R10, R11 and R13 should be down in the 1K range, not 100K.

Still looking at the data sheets.

 

Let us tackle first things first. What is the model and bandwidth of your Tektronix oscilloscope?
What is the make and model of the oscilloscope probe?
What is the attenuation setting on the scope probe, x1 or x10?
Have you adjusted the frequency compensation control on the probe?

 

Tektronix TDS 2014B 100mHz bandwidth 10x probe.

Why would those resistor values be to large? recovery time? BTW - the last amp began as a Howland Current Pump, but I changed it... that is where the 100k's came into play.
I didn't like the amp on all of the time.. too much heat to dissipate... so I changed it to pulsed operation, with current limit. I can try 1k's and see what happens...

 

R10, 11, and 13 have no business being in the 100K range because they create offset voltages in the range of 0.05 volts and input offset voltage drift with temperature in the range of 0.001 volt, but I don't think that's going to affect the output sag.

MrChips is investigating whether the measuring device is causing the error reading.

Am I reading this right? The on time resembles 365 usec for a frequency of 2.7KHz...1.35 KHz if that is a half cycle?

and the 470 pf (C1) into 10K has a frequency corner of 34KHz?

Never mind. I thought the LM331 was an op-amp until I looked it up.

 

Looks like classic probe compensation (or lack thereof) to me.

 

If it is probe compensation, there should be a similar overshoot indication on the falling edge. What does it look like?

 

Parasitic capacitance associated with R10 (100K) could cause this droop/over shoot also. Changing R10, 11, and 13 to 1K would eliminate the issue and the offset that #12 referenced.


Also, looking at the waveforms without clamping diodes and with diodes, suggests that it is not a probe compensation issue.

 

I assume you are showing us the output waveform at pin6 of U3?

So what is the waveform at pin1 of U1, pin3 of U2 and pin1 of U3

And as said can we see a complete cycle please?

 

This is not probe compensation.

Droop is still evident with the 1k resistors. Clamping diodes take out the ring, but not the droop.

 

Some things coming to my mind -
Compensation at the output - Fig 14 of data sheet.
Is there a heatsink? Does it get hot? Cant say about Pd and temperature rise without duty cycle, maybe this is not, as thermal cycles could be longer.
Small cap of 220-470pF parallel to the feedback resistor.

 

Could your 24V supply be sagging? 10uF for C8 seems rather mean. Or perhaps the droop is due to the on-chip current limiter kicking in?

 

So I resolved MOST of the problem by increasing the bypass capacitance at U1&U2 to 1uf. .....even had an assembly error in which the original .1uf bypass cap at U1 was not connected to gnd. .duhh
Changed R10, R11 and R13 to 10k. (and removed the 10uf tantalum cap)

Nevertheless, I still see about 200mv droop, and not sure where it is coming from. (There is also a VERY small amount of symmetry disturbance due to ripple of U2 pin 6).