Hi Guys,

I'm working on an HVAC Fan unit that uses a simple step-down circuit to bring the supply voltage down to 12V which in its turn powers a bunch of 26kHz PWM controlled fans. It will be supplied by a standard 24VDC bus supply. For reducing conducting emissions bleeding into the 24V bus I added a pi-filter at the input.


On light to medium load the circuit works as expected but at full load L1 starts to sing at around 16kHz. The scope picture is probed at Pin5 of U1. During this test I used 20Vdc as the source input so don't let that confuse you. As you can see I have a gigantic ripple that's present at the input of the DC/DC converter. All capacitors used are either MLCC's or low ESR OS-CON's.


Is what I am witnessing due to too much DCR (caused by L1) being introduced into the supply line? Is L1 saturating (although the current is well below its saturation current)

Thanks!

 

is the frequency you hear relayed to the frequency the PWM is running at ?

Why do you say the saturation current of L1 is not a problem ?

Remember the PWM basics,
They take lumps of current, then nothing.

So if the PWM is say supplying an average of 1 amp, and is on 10 % of the time, then for that 10% its taking at last 10 Amps form the supply.

The Pi filter is stopping that spike, so where is the PWM to get its energy from ?
C3 and C4 look very small to supply much of a load , hence the massive ripple you see.

My thought,

a) the mechanics are resonating ?
b) the inductor current is much higher than you think,

 

Switching frequency is 700kHz - if it were ripple it would be at that frequency.
It's at 8.54kHz, so it's control loop instability. (I know it says that the IC is internally compensated)
Is it stable with a resistive load? I'm wondering if the fans present an odd load and it upsets the control loop. If the fans are speed controlled then they will present a negative resistance load to keep the power constant (as the voltage increases, the current decreases).
Try a capacitor across R1. A few nanofarads would be a good start.

 

Your inductor needs some mechanical damping. High current causes the ferrite or metal powder core to change shape as it magnetized.
https://en.m.wikipedia.org/wiki/Magnetostriction
You could try physically larger inductors with the same inductance, an inductor made of a different material (ferrite, mu-metal, iron-powder, are all option).

 

is the frequency you hear relayed to the frequency the PWM is running at ?

The Fan PWM frequency is 26kHz, the DC/DC runs at 700kHz. When I turn off the 26kHz PWM and apply a DC signal to the fans instead to run them at full speed the problem still exists. At this point the load is also constant.

Why do you say the saturation current of L1 is not a problem ?

My reasoning behind this is that the sine wave looks very symmetrical and remains symmetrical when I introduce a switching load at the output of the DC/DC converter. I would expect there to be skewing of the positive half of the sine wave when there is high current demand.

Is it stable with a resistive load? I'm wondering if the fans present an odd load and it upsets the control loop.

I added a rheostat as a load on the output and at about 720mA of current draw the oscillation at the input starts causing L1 to make noise at +-16kHz. Under lighter loads there is almost no ripple on Vin of U1. There is very noticeable immediate crossover point between almost no ripple at Vin to massive ripple at Vin when driving up the load.

Bridging L1 does solve the problem.

Try a capacitor across R1. A few nanofarads would be a good start.

I will try this later on today.

Your inductor needs some mechanical damping. High current causes the ferrite or metal powder core to change shape as it magnetized.

I think I first need to solve the issue which causes the massive ripple at Vin.

 

Could it be C3 singing?

MLCC is known for this additional property.

 

Could it be C3 singing?

Hard to tell but I notice the sound gets damped the most when I put pressure on L1.

 

I'm just waiting to see if others come up with the solution.

But I remember that twisting a PCB can change the sound from MLCCs.

 

I would short out L1 and R3. I think 2.2uF is just not enough on the input.
Maybe move L1 to the input connector so all capacitors are on the input of the PWM.
What is the part number for L1?

 

What is the part number for L1?

SDE0805A-680K

 

I have had issues with both inductors singing (core magnetostriction) and ceramic caps (piezo-mechanical)

The best remedy is to move the operation frequency up out of the audio range.

 

Forget what I said - I was thinking about the wrong inductor.

There is a peak in the response of the input filter. Do you really need L1?

 

The best remedy is to move the operation frequency up out of the audio range.

The frequency is way above the hearing range.

The pie filter on the input is not doing what you think. It is the trouble.
Also many type of capacitor change value when you put voltage across them. What is the voltage rating on the 2.2uF cap? It could easy be 1uF when 24V is applied.

Just for a test stack more caps on top of the 2.2uF.

 

There is a peak in the response of the input filter. Do you really need L1?

I'm not sure yet. I don't have the proper equipment in-house to measure compliance for conducted emissions. I would need a day at the lab first to confirm.

 

The pie filter on the input is not doing what you think. It is the trouble.
Also many type of capacitor change value when you put voltage across them. What is the voltage rating on the 2.2uF cap? It could easy be 1uF when 24V is applied.

it's a 50V MLCC cap. I am aware of the DC bias but I followed the datasheet's recommendation. I have not taken into account the potential effects of putting an L-C filter in front of that 2.2uF. I'll stack some more on top of C3 that one to see what it does. Could adding some ESR on C3 or C4 help in case of negative impedance?

 

One way to solve the problem is to dip the offending inductor in varnish (I like varathane). More than one coat/dry cycle might be needed.

 

The power supply is oscillating at 15khz. (switching at 700khz) That needs fixed. Turning down my hearing aids is not a good answer. (cleaver as it is)

 

Just for a test stack more caps on top of the 2.2uF.

I was able to solve it by doubling the value of C3. The oscillation is entirely gone.

Thanks everyone for your input.