Radiated susceptibility problem-PWM dimming

[Winsu]

Hi all,

I have a product which is failing in radiated susceptibility. It is a lamp which is dimmable by a PWM input. When the test is done the flickering is noticeable. I have already shorten tracks a bit and it has improved quite significant but it is still a problem.

The PWM circuit is as follows:

-A shielded cable, a terminal block to connect the PWM input cables on the PCB, an optocoupler to isolate and to convert the signal from 24V to 5V, a comparator(op amp), a TVS ( for transient), and RC network to average the voltage, a voltage follower, after this there is another op amp configured as non-inverter amplifier with an specific gain to adjust for the analogue input of the LED Driver. The order I have announced the components is the order as how they are connected.

-An RC filter was put before the optocoupler with no good result. A capacitor was connected in parallel with a resistor that was already on the design in the non-inverter input of the comparator in case it could filter out the noise, and it didn't do it. Those modifications didn't bring any good results.

-A fixed voltage ( DC) was applied on the comparator ( first op amp) to set the lamp at 50% and the test was carried out. No flickering was seen.

-The tracks after the first op amp are super short, so I don't think noise is coupling after that, but it could be. The RC filter that I mentioned before had a cut off frequency of 2Khz ( the PWM is 1Khz) and it didn't work, so I assume the problem comes from a low frequency. The carrier frequency of the wave radiated is 8Hgz and the modulated is 400Hz,this is as well why I think the problem comes from a low frequency signal. I could put a high pass filter but the cap in series on the PWM line would convert the signal in positive and negative as suppose as the PWM signal is ( referenced to cero, as my op amps are).


Any one have any idea of things I could try.....?,

Thanks a lot!,
Winsu.

 

[albbg]

I would try with a ferrite clamp on the coaxial cable. I suppose the rest of the circuit is placed inside a metallic enclosure connected to a good ground.
In any case could you post the set-up of the test bench

 

[Winsu]

I forgot to say it has already got a ferrite clamp, it helped to pass radiated immunity.

I can't post set up of the test bench but I can explain a bit how it is.

The unit is on a side of a table resting on a foam mate. The cable comes from the other side of the table horizontally. The unit is parallel to the antenna and the cable is perpendicular....

 

[KlausST]

Hi,

Please post a schematic.

Klaus

 

[albbg]

The unit is on a side of a table resting on a foam mate. The cable comes from the other side of the table horizontally. The unit is parallel to the antenna and the cable is perpendicular....

What about the connection to ground ? Under the foam there should be a thick copper sheet connected to ground. I suspect there could be something related to the different grounds at the two sides of the opto. I don't know if you can try (just to understand) to short the two grounds and have a look on what happen.

 

[Winsu]

Sorry Klaus I know an schematic helps to understand but I cant for commercial reasons....

The opto was incuded to isolate the PWM from he driver side as it was creating problems in the driver, once the opto was included in the design it started to work fine...

We suspect know that what is picking the noise are the cable inside the metal housing, we would like to find something to shield those cables, what could it be?, thanks

 

[albbg]

What kind of cable are you using at present ?

 

[c_mitra]

we would like to find something to shield those cables, what could it be?, thanks

I wish I knew! But I can guess.

The flicker is visible hence the noise is low frequency. I guess it is coming from the opto coupler. Or from the op-amp. Which one? I do not know.

The flicker is perhaps due to motorboating. Bypass the power supply pins of the op-amp with decent caps. Does it work? Perhaps not.

Twist all the cables tightly. Is that the line freq pickup? Does that help?

Is the input PWM clean looking on the scope? Is that the villain? Perhaps. Hunting for the noise can be a real experience...

 

[Winsu]

We know the PWM signal is clean when it is at normal conditions, unfortunately we can not test it inside the room where the susceptibility test is carried out.

I am thinking to get something like this.

https://kemtron.co.uk/emc/knitted-wire-mesh-2/

 

[KlausST]

Hi,

The flicker is visible hence the noise is low frequency. I guess it is coming from the opto coupler.

I don't see why an optocoupler should create that much low frequecy noise that one even can see it.

I rather like the idea about motorboating.
Is the PWM control feedbacked somehow?

Klaus

 

[c_mitra]

Is the PWM control feedbacked somehow?

I guess that is the likely culprit.

 

[Winsu]

No, the PWM is not control feedbacked....

 

[c_mitra]

No, the PWM is not control feedbacked....

Then what? The stray pickup by the long PWM cable? That can perhaps explain 50/60Hz flicker but...

High impedance at the receiving end? The PWM cable is feeding a high impedance source? Put a 10K in parallel with a 0.01 uF cap to bypass high impedance noise to ground.

If that does not kill the flicker, then the opamp needs more detailed scrutiny. Particularly bypass at the power supply pins and the signal input pins.

I am out of ideas now. My brain is currently under lockdown.

 

[KlausST]

Hi,

I don´t think that induced signals are likely to cause low frequency flicker.
Without schematic and/or more detailed inflormations about the circuit, it´s function, it´s wiring... all we can do is guessing in the dark. Almost useless.

Rethink whether a "LED dimming circuit" is that rocket science that you have to kill us if you show parts of it .... or if it´s more important to quickly find a solution.

Klaus

 

[Winsu]

The lamp has been tested without the dimming circuitry being connected and it was a pass.

I have already put some capacitance at the output of the last opamp ( the one which is connected to the driver) and it didn't make it better.

I put a low pass filter in the dimming circuit to filter the low frequency ( as the modulated signal is 400hz) and it didn't work. Do you mean to put a resistor in parallel to a capacitor? , I don't understand how it could work?, Mitra could you explain please?

Actually there is already a 10K in parallel to 100nf in the +5 power rail ( which is the one which powers up the op amp)

 

[c_mitra]

I have already put some capacitance at the output of the last opamp ( the one which is connected to the driver) and it didn't make it better.

That is understandable because the capacitor is removing the HF but not bothering the LF. You have (my guess) HF noise modulated by line frequency. That is why I was curious to know the flicker frequency.

The lamp has been tested without the dimming circuitry being connected and it was a pass.

At the lamp side, you will need a BP filter tuned to 400Hz. At least something to remove 50Hz line frequency.

The driver at the lamp side sees 400Hz plus 50Hz line frequency (because you have filtered the RF noise). I guess dimming is working but the flicker is constant (most visible at low setting).

You need to bypass the signal at the lamp end; the modulation input at the lamp is high impedance and is susceptible to noise. Putting a cap and resistor can reduce the noise at the lamp end.

 

[c_mitra]

Do you mean to put a resistor in parallel to a capacitor?

Sorry, I forgot to reply that!

Source: 400Hz desired signal coming from the dimmer unit: this is a PWM signal given to the lamp.

Plus: noise: perhaps RF random with a 50Hz powerline modulation.

Removing 50Hz with a RC filter will take lots of C- perhaps inconvenient. Simple C will pull down the 400 Hz signal also.

How does the R help: noise has a high source impedance and a low value of R will look like a voltage divider with the high source impedance.

But for the 400Hz signal, that has a low source impedance, most of the voltage will appear across this R.

If this does not work, you need to look deep in the op-amp side. Power supply impedance is too high OR power supply is having too much voltage ripple.

The question is high much ripple is acceptable at the op-amp stage? I do not know but you have to dig. and dig more.

 

[KlausST]

Hi,

A lot of guessing is already done...

Flicker frequency usually is in the range below 20 or 25Hz. Everything above is considered not to flicker. (But maybe a combination of two higher frequencies may cause a sub harmonic frequency / alias frequency)

My opinion:
* High freuqencies > 400Hz should no cause the flicker
* I don´t think that induced noise into the (power) cable can cause visible flicker. (at least on an non feedbacked system)

Thus I see just two sources of flicker:
* PWM jitter
* DC bus voltage variations
(since the LED just sees the voltage of the DC bus and the timing of the PWM)

Since the PWM is said to be clean .. then there just remains the DC bus.

Maybe I missed something...

Klaus

 

[Winsu]

At the lamp side, you will need a BP filter tuned to 400Hz. At least something to remove 50Hz line frequency.

We already put a BP in the PWM input and it was disturbing the PWM signal as it has a capacitor in series..., We haven't put extra filtering in the main inputs ( the main input is AC and it has got a line filter) as it is passing conducted emission with very good results....

The dimming circuit is just failing on radiated immunity when the antenna is swiping at frequency of 80Mhzand 200Mhz, the rest of the frequency range it works fine.

The driver at the lamp side sees 400Hz plus 50Hz line frequency (because you have filtered the RF noise). I guess dimming is working but the flicker is constant (most visible at low setting).

There has been a misunderstanding from my possible poor explanation. The flicker is just visible through a camera, it si no visible to the human eye. The human eye just see how the level of light fades...Across the two frequency ranges when we see the problem it fades the same amount of light.

You need to bypass the signal at the lamp end; the modulation input at the lamp is high impedance and is susceptible to noise. Putting a cap and resistor can reduce the noise at the lamp end.

Not sure where you suggest to put a cap and resistor, at the input of the lamp or at the input of the PWM. Adding more capacitors to the main would affect to the power factor

- - - Updated - - -

Flicker frequency usually is in the range below 20 or 25Hz. Everything above is considered not to flicker. (But maybe a combination of two higher frequencies may cause a sub harmonic frequency / alias frequency)

The flicker is just visible with a camera, the human eye just see the light fading a bit. Sorry if my explanation lead to misunderstandings...

Thus I see just two sources of flicker:
* PWM jitter
* DC bus voltage variations
(since the LED just sees the voltage of the DC bus and the timing of the PWM)

The PWM is very clean but it was tested on lab conditions. When it is exposed to a electrical filed it may change.... what I am thinking now is to improve the robustness of my PWM tool ( the external one). At the moment it is just and IC which set a frequency of 1kHZ and duty cycle, it excites the gate of a FET which is connected to a 24V. What could a put after the FET to create a barrier?

 

[KlausST]

Hi,

The flicker is just visible through a camera

This is a complete new situation.

Then it´s quite possible that it´s just an interference between camera_sampling and PWM.

Not even sure if it´s an error at all.

Klaus