EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: orinocopaul on July 19, 2021, 02:53:10 am
-
I'm powering some LEDs via a rf controlled PWM dimmer. As a result of the 7.8 kHz PWM frequency, there is 7.8 kHz audible whining inside the SM power supply. Below is how the ripple looks like btw the dimmer and the PS. How do I attenuate the ripple beneath audible values?
The current ranges from 1A to 4A at maximum brightness.
-
Whining is usually from vibration of the magnetic core, megnetostriction. Eliminating it depends on how bad it is.
Often clamping of the core or impregnation of the transformer can help. Another approach might be to eliminate the ripple with passive or active elements.
-
Bonjour, the ear is most sensitive at fréq in that region.
The dimmer has a poor design, the PWM and ripple should be at ultrasonic rate > 20 kHz.
The magnetostriction of the transformer or inductor core will always be audible.
A toroid rather than EE/EI or other shape may be less noisy.
Suggest to get a well designed PWM dimmer with >20kHz.
Jon
-
Not sure how to approach clamping the transformer. I would rather try filtering the ripple pictured in post #1. How would I target that particular curve?
-
You can try to use a CLC filter after your power supply. The last (right) C will buffer current spikes, the L softens (or rounds) the edges of the current spikes and the first (left) C acts like an additional buffer to dampen current spikes further. Of course, this is an explanation in laymen terms but it describes pretty much how it works. In my situation, I chose (I didn't do the math) beefy parts ;-)
The caps are 10.000uF and the inductor was the largest I could find in my stock.
This filter handles a maximum current of 6A and supressed the whining almost completely.
-
Additional filtering on the output will not reduce the ripple current in the inductor. This is controlled by the fundamental design of the supply.
-
Hi,
you're right about the ripple itself you cannot eliminate it - that's due to the principle of a SMPS. But you can change the waveform of the current through the PSU's inductor with an additional filter if the PSU is under load. Just hook up the inductor to your scope and observe yourself.
Whining is mechnical movement of the inductor (turns). The movement is caused by electro magnetic force and this force is depending on the amplitude and the dI/dt (rate of change) of the current through the inductor. More current and a steeper curent curve will cause more whining. It get's worse if you load a SMPS with a PWM controlled load as you change the current flow through the inductor with the PWM frequency while the PSU's inductor is fed from the switch.
So, in my understanding you can eliminate (or at least attenuate) the whining by "stretching" the dI/dt or in other words, by changing the current wave form through the inductor of the SMPS. That's at least what I observed and measured by "fixing" the whining of many cheap SMPS that way.
If I'm wrong, I'm curious to get the right physical/electrical explanation. Thanks!
-
It get's worse if you load a SMPS with a PWM controlled load
You are absolutely right. See the the signal before and after the PWM dimmer. The whining is at its loudest at 50% duty cycle and fades away completely when approaching 100%.
Made a CLC filter with 1.000uF (the only capacitors I have) and it definitely reduced the whining amplitude. I'm just a bit lost in the terminology:
Is this a decoupling situation rather than a straight forward CLC filtering application? If so, than what should I go for regarding the decoupling capacitors: electrolytic, film, low ESR, etc.? I have noticed that some decoupling schematics use different values capacitors in parallel and no inductor.
-
The whining is at its loudest at 50% duty cycle and fades away completely when approaching 100%.
That's because the whining is not caused by the intrinsic current ripple of the SMPS but by the current wave form of the PWM driven load.
Made a CLC filter with 1.000uF (the only capacitors I have) and it definitely reduced the whining amplitude.
The more capacity you add after the filter's inductor (direction towards the load) and the larger the inductor, the flatter the current curve and the better your result gets. You can try to add caps after the L to further improve your result.
I'm just a bit lost in the terminology:
Am I suppose to use so called decoupling capacitors? If so, than what should I go for: electrolytic, film, low ESR, etc.? (I have noticed that some decoupling schematics use different values capacitors in parallel and no inductor)
Decoupling or bypassing capacitors serve as nearby charge reservoirs to decouple devices (mostly CMOS ICs or other high speed devices) from the power rail. This is needed as all conductors act as inductor. The effect of the inductance gets more noticeable when dI/dt gets larger. As e.g. CMOS devices switch very fast and need high currents for short periods of time (=large dI/dt), the track's inductance may lead to a voltage drop. To avaoid that, you place the mentioned bypass/decoupling caps near to the power pins of the device so they can act as "current buffer" in the periods of large current flows. The result is a cleaner supply voltage from the perspective of the device.
Depending on the device's decoupling requirements, a combination of different capacitors is used. You often see a combination of tantalum and ceramic capacitors near ICs. The ceramic one is mostly a small one which takes care of the high frequency current fluctuations and a larger tantalum one handles medium frequency current fluctuations. Both types of caps are "low ESR" compared to "conventional" aluminium electrolytic caps.
In general you can lower ESR by simply paralleling (conventional electrolytic) caps, this also takes away stress from the caps as the current flow is shared between them.
In your case, the steep current curve of your PWM load causes some (minor) voltage drop too along the leads, but that's not a problem for LED lighting. It could make the current amplitude "seen" by the SMPS even worse, as the capacitors would need to be loaded in parallel to your load during PWM "high" states.
The CLC filter softens/rounds (or "averages") the rectanglur current wave form seen by the SMPS, thus reducing the whining. It can be seen as some sort of decoupling of the SMPS from the load.
I hope that cleared things up.
By the way, I build the CLC filter for the exact same purpose: PWM dimmed LED lighting supplied by a cheap SMPS. I use 15kHz PWM, peak load is 6A and was able to reduce the whining from insufferable to unhearable. It's "just" a matter of the sizing of the CLC filter's components.
Regards,
Kryp
-
Thanks Kryp. Quite a lot to wrap my head around, but that sums it up pretty well for me. Will report back once I get the appropriate components.
-
By the way, I build the CLC filter for the exact same purpose: PWM dimmed LED lighting supplied by a cheap SMPS. I use 15kHz PWM, peak load is 6A and was able to reduce the whining from insufferable to unhearable. It's "just" a matter of the sizing of the CLC filter's components.
I've put together two types of CLC filters: one with low ESR capacitors (pic#1) and one with less expensive parallel capacitors (pic#2). They both seem to do a great job. However, even though the switched mode PS is not whining anymore, I can still hear some 7,8 KHz coming from the LED strips. The LEDs are mounted inside a metal housing and that seems to vibrate a little as well.
Now that the SMPS is a lot quiter, I'm thinking about filtering the dimmer's PWM output as well.