Group,

I am having some difficulty with a crazy resonace problem. It’s a simple enough circuit. The transistor Q1 is always on, being biased from a bench top power supply at 0.6V. The FET M1 is switched on and off in a pattern. This pattern may be on for 125ns, off for 125ns and then a delay followed by the same. Its not too important at the moment. When M1 is on, the LED is on. The LED is supplied 8V along a 50cm cable and then there is a cable of length 50cm to transistor Q1. Q1 and M1 are on a PCB and very close together. The lengths of the cable are represented by inductors L1 and L2. I have guessed at their value.

When M1 is switched on, the collector of Q1 begins to resonate at a frequency of 4MHz. It starts off a small sine wave and gets bigger and bigger till it reaches 30V. There is no sign of the resonance at the emitter. The amplitude of the resonance gets worse as you increase the base voltage of Q1 a bit. The LED works but the signal looks horrible. The resonance ends when M1 is switched off.

I assume I have hit some resonance point between the inductance of the cable and the parasitic capacitance by drawing the current through the transistor at this switching rate.

Can anyone give some advice on what is going on

Thanks in advance

Chris

Attached Images

resonance.jpg (27.3 KB, 1186 views)

 

Some things that might help:
1) slow the edges down of the source switching the mosfet & put a resistor in series with the FET gate. (10 Ohms)
2) Put a resistor (maybe your 3.3 Ohms) in the collector of Q1. This will help dampen any resonant behavior due to parasitic C's in trans.
3) For your 50cm cable, you could use a twisted pair to reduce the effective value of the inductance. There will remain some but shrinking its value will cause the resonant point to be at a higher frequency making it easier to dampen/eliminate.

 

Thanks Optikon,

I will try out some of your suggestions. I tried slowing the FET gate switching edge to >100us but the resonance is still there. Unfortunetly I can't change the cable. I will try putting a series resistor in the FET gate drive.


In the past I have seen ringing but not true resonant behaviour like this from switching on a FET.

I have attached a picture of the resonance. The blue trace is the collector of Q1. The yellow trace is a slowed down gate enable signal to M1. Nasty eh.

C

Attached Images

tek00003.jpg (63.2 KB, 1170 views)

 

You have a common-base amplifier with positive feedback provided by the transistor's C-E capacitance plus PCB parasitics. The Mosfet isn't turned-on very hard so it represents a high impedance current source.

Ground the emitter. Invert the pulse and feed it to the base through a current-limiting resistor that is missing in your existing circuit.
You also don't have current-limiting for the LED.

 

Have to ask, why you have done it this way? If the object is to switch the led on and off, it could be simpler.

 

First, Thanks for replying to me. I appreciate you taking the time to.

Let me explain a bit more and add a schematic to make things clearer. I just cut the circuit down the last time to maybe make things a bit more clear but I think it would be more useful to explain the rest of the circuit.

The LED Bias signal is just an analogue voltage that sets the bias level of the NPN transistor (now called Q6). A feedback circuit controlled by U2 (AD8061) keeps the current through Q6 constant. The circuit needs to produce a constant LED brightness.

When I want to switch the LED on, I enable the gate of Q4. This switches on the LED. The current is set with the 3R3 resistor and the voltage developed across it is fed back to U2 to control the current.

The enable signal that switches on the LED, sends pulses to switch on Q4. These pulses are of varying length, the smallest being 6us. The rise/fall time of the signal is around 20ns. The resonance does not occur when the gate of Q4 is switched with a signal of 100Hz (surprisingly enough, not).

Fried, I have done it like this because I need to enable an LED very quickly and keep the current through it constant.

The LED is on the end of a cable, however, I am rapidly coming to the conclusion that it’s the small pulse size I using while switching the FET that is causing the bother.

Thanks in advance

Chris

Attached Images

led_992.jpg (30.0 KB, 1103 views)

 

So the opamp is given a suitable input voltage for good LED brightness, but since the FET is not enabled yet, the diode and the negative feedback set the transistor's base voltage too high.
Then you enable the FET with an input pulse, and the first part of the LED current will be extremely high. The cable's tuning is charged.
The opamp sees the very high voltage drop across the 3.3 ohm resistor and even though is a very fast opamp and the transistor is also fairly fast, delays reducing the LED's current. The cable's tuning discharges into the LED.

Yeah, you have a nice oscillator, and a high enough Q in the cable's tuning to amplify the 5V to 30V.

 

Thanks audio guru.

Is there anything that can be done? I breadboarded this circuit before I made it with exactly the same cable, just the interconnections between the chips and components were made from wire instead of PCB traces. Have I just hit the sweet spot? I thought there should be more chance of it resonanting with more inductance.

I changed the enable to a 10Hz signal and during the 10Hz on pulse it still resonated like crazy for the whole time. I can just put a DC signal on the gate enable with a 500ohm resistor in series and it still resonates. I thought it would eventually decay.

Thanks

Chris

 

Have you tried to dampen it out with some resistance in the collector?
When you make an LC scenario like this, getting an R in the right spot will lower the Q and if sufficient, suppress the oscillation.

 

A cap from the collector to ground will certainly dampen it, but use a resistor in series with it so the transistor and/or LED don't blow.

 

The entire device seems rather complicated, what's it actually for?.

Is the supply to the circuit not very stable?, as long as it's stable a simple resistor would keep a pretty constant brightness.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

 

Thanks for your responses,

Optikon,

I placed a resistor in between the collector and the LED. About a value of 47ohms gets rid of the oscillation. You can take it down to 18ohms but you still see some ringing at the edge before it is damped. I changed the 3R3 resistor at the bottom to 18 ohms in place of the one between the collector and the LED and it produced the same result (ringing at the edge).

Audio guru,

I did the snubber circuit 100pF and 51ohms and I am sure I could reduce the resonance if I tuned the cap value.

Nigel,

It is a control circuit for illuminating a reflective microdisplay. There are 3 of these circuits which control a red, green and blue LED. The circuit has to have resonable control of the brightness but most importantly fast switching of the LEDs, down to 1uS, at a resonable current 500mA. Most importantly it shouldn't ring like crazy when you switch it on.

Chris

 

Whats really bugging me the most is why if I put a DC voltage on the gate of 3.3V, why it still resonates. Surely it should damp down?

Is it the AD8061 that keeps it going?

Was it a bit of a bad idea having the FET as an enable?

C

 

Those inductors you have drawn are really transmission lines, and they will resonate and cause your transistor to oscillate. Try putting the resistor in series with the other end of the cable (the +8 volt end). This is counter-intuitive, but it should terminate the cable in something near its characteristic impedance, reducing reflections and hence ringing/oscillation. Putting the resistor near the collector will dampen the resonance, but it is a poor termination.

 

Hi Chris,
You don't need the AD8061 nor the transistor. Just let the Mosfet drive the LED through a resistor. It looks like you are using PWM to the Mosfet to control brightness anyway. 500mA through the LED for only 1us certainly won't make the LED very bright unless the pulse width is about 50ms.