Switched mode constant current source, huge spikes

I'm designing a buck converter using the HV9918 IC to drive some high current LED's.

I'm seeing some huge spikes at high bandwidth in the current waveform. Does anyone know where they're coming from?

These waveforms are across a 1.7 ohm current sense resistor. So the current in amps is approximately 60% of whatever amplitude is shown in the pictures. The average current is about 250mA. I'm driving a 6.8V zener diode.

In all these pictures the bottom flat trace shows the 0V level for the top trace.

**broken link removed**
The switching frequency is about 1MHz. There are huge spikes that are 700mA peak-to-peak!

**broken link removed**
Zoomed in. Those sine waves are about 300MHz. My scope is a 2465B 400MHz and I'm using a 400MHz probe measuring across the sense resistor with the ground lead.

**broken link removed**
With bandwidth limit of 20MHz on the scope, the current ripple looks much more reasonable. Only 30mA ish.

Is it coming from the wiring inductance? I have the zener diode and external current sense resistor (not the internal sense resistor for the buck converter) connected using alligator clips to each other. But my main intent is to have the LED's connected with a cable.

Should I even worry about it? Will the LED's "see" this high frequency current spike if a 20MHz scope doesn't?

Just like the other forum: Where is your schematic?

Sorry about that, here it is:

**broken link removed**

R4 is a short instead of 1k.

That's the current waveform? Could it just be your switching frequency is too low and your inductance too small?

I would think wiring inductance would cause voltage spikes rather than current spikes (being inductance after all). Of course, never hurts to minimize trace lengths and loop area in your wiring, especially for those loops with high frequency switching or high currents.

But...shouldn't you be using a resistor as your test load rather than a zener diode which would inherently try to regulate the voltage on it's own? That might also give you whacky current readings as it fights your regulator.

It would appear that the LED load is not grounded. If you are trying to measuring current though a floating resistor with the scope connected to ground, then you are not just measuring current but also voltage between the load and the ground. That may be giving you faulty readings.

In any event I don't think the current spikes, whatever they are, will cause any problem with your LEDs.

It's a high side driver, so the switching pin either conducts to ground or floats. I think the LED- node isn't exactly floating as the inductor is supplying current once the switch node floats.

Wouldn't a voltage spike directly translate to a current spike anyway? Especially for a diode where a small change in voltage causes a large change in current.

Since the intent of this is to drive an LED load, I used a zener diode which behaves roughly like an LED.

In my board layout, I was careful to keep all important traces very short. The long traces are in the cables that attach the load.

**broken link removed**

I agree that the spikes probably won't hurt the LED, but I'd like to know why this is occurring. Or perhaps it happens in a lot of LED drivers and people just haven't noticed because they weren't using 200+ MHz scopes?

Oh, you have an LED that has a voltage drop of 6.8V? Your layout looks nice and tight. You could just try increasing your output capacitance. Naybe 1uF is just a bit too small, especially if the LED is at the end of a long cable. 1uF does seem a bit smaller than usual. Isn't it usually at least 4.7uF-10uF?

A couple of possibilities:
Measurement error - What do you see if you place the scope probe on the scope probe ground with both attached to the sense resistor.

Stray capacitance in the inductor. You often see this problem when current sensing dc motors. Is the filter cap ceramic?

Oh, you have an LED that has a voltage drop of 6.8V? Your layout looks nice and tight. You could just try increasing your output capacitance. Naybe 1uF is just a bit too small, especially if the LED is at the end of a long cable. 1uF does seem a bit smaller than usual. Isn't it usually at least 4.7uF-10uF?

allanw said:

It's a high side driver, so the switching pin either conducts to ground or floats. I think the LED- node isn't exactly floating as the inductor is supplying current once the switch node floats.

Click to expand...

That's not good enough because your measurements are happening while it is connected to ground AND while it's floating. That's also assuming your current sense resistor is in the right place to even be half-true which would mean directly connected to the IC pin that switches to ground. Anywhere is no good because there are voltage drops across the inductor and other components.

The oscilloscope has ground hardwired as it's reference. THe ground probe of the scope is connected to the ground of the wall. So unless you're using isolated probes or an isolated power supply you can't just stick the probe and ground clip anywhere you want in the circuit to measure the voltage across anything you want. That ground probe is connected to the ground in the wall and if there's a ground connection between the input and output of your power supply it's gonna give funky results.

You could, however get around this, by measuring the voltage at each end of the current sense resistor with respect to ground and then subtract the voltage measured to get the voltage difference across the resistor. That means two channels are needed - one for each end of the resistor and both channels are being measured with respect to ground.

The output filter cap has to be slightly small so that PWM dimming at low duty cycles still works.

I formed a loop with the scope probe ground and tip and could see the spikes from over a meter away over 100mV in amplitude. So it's radiating EMI.

All the caps are ceramic. I'm going to test without the output filter cap and see if the spikes are there.

The oscilloscope has ground hardwired as it's reference. THe ground probe of the scope is connected to the ground of the wall. So unless you're using isolated probes or an isolated power supply you can't just stick the probe and ground clip anywhere you want in the circuit to measure the voltage across anything you want. That ground probe is connected to the ground in the wall and if there's a ground connection between the input and output of your power supply it's gonna give funky results.

Click to expand...

My power supply is definitely isolated, so it should be okay, right? ie. I can touch the ground probe to any part of the circuit without anything occuring.

Your switching freq is very high you are probably getting reverse recovery issues with the diode.

Try putting a cap across the current sense resistors, it will slow the freq down a little and damp the reverse recovery current pulse and should drop your EMI a bit. You may even need a faster diode if you are determined to run at 1MHz, which seems too high to me too. What does the datasheet say about recommended inductor values and freq? 10uH and 1MHz seem silly values, you will get better efficiency with a large rinductor and a slower switching freq.

Nice PCB layout by the way.

The chip is very fast. It would appear you are just seeing the EMI getting into the scope.

I placed a capacitor across the current sense resistor, which also reduces the switching frequency.

With a 0.5uF cap, there's still a spike but it's very small. Now my ripple voltage is 240mV, with spike 336mV. Previously my ripple voltage was 23mV and the spike was 500mV. The switching frequency went from 1.7MHz to 633kHz.

With 1uF cap, the spike is entirely gone but the ripple is 340mV.

I'm going to order some bigger value inductors and see what that'll do.

Sounding better.

A 0.3v ripple on your 6.8v output to a LED is unlikely to be an issue, I don't think your user will see that at >500kHz.

Bigger inductor (33uH?) sounds like a good idea, along with a bigger Vout cap, like 4.7uF. You should be able to source both in the same package size for the same price so you may as well get the benefits of lower switching freq.