LM3450: Power Factor needs improving

Hello Steve,

Are you using this as a current regulator?  Normally single stage LED drivers regulate current.  If not and you are just creating a single stage voltage output it looks like the bandwidth may be too high.  Try lowering the bandwidth of the secondary side control loop.  If the bandwidth is close or above the line frequency the LM3450 will lose PFC.  Generally you want the bandwidth to be less than 10 Hz and most likely quite a bit lower due to variances in the optoisolator.

Regards,

HI Irwin,

we are using this as a PF corrected voltage supply, the rest of the system has the LED controlled section via a microprocessor, so this is just the supply to the LED circuitry and does not control the current. The output voltage can have plenty of ripple (+/- 2V is fine).

How do I adjust the bandwidth to less than 10hz?

Steve

Hi Steve,

I would first try changing Rsc to 10 to 20 Kohm from the 2 Kohm it is.  There are quite a few components involved with the bandwidth.  Also RFB2 is 105K and RFB1 is not stated.  I would leave the RFB1 value the same as the EVM and adjust RFB2.

Is your source frequency adjustable?  You can see if it improves or gets worse by changing the input frequency as well.

There can be other issues causing this, it just looks like it's responding too quick.

Regards,

RFB1 is 10K, I'll put it back to the EVM value and adjust RFB2, would adding a cap across RFB1 reduce the bandwidth in the desired way?

Hi Steve,

I wouldn't do that since it can create an instability. I would change Rsc, Csc and Rsbs (this one can't go too high in value otherwise it cannot drive the opto) so I'd work with Rsc and Csc for now.

Hi Irwin,

Sorry I have been on away, please see the attached wavforms.  I adjusted the Rsc 20 20K from 2K, I also changed RFB2 so that RFB1 was the same as the EVM, but still no luck with the PFC at 240V.  At 50V the PF is good, but as the voltage increases past this it goes bad.  The input frequency is Australian mains, so this is fixed at 50hz.

RFB2 105K Rsc 20K 240Vac

RFB2 105K Rsc 20k 50vac

RFB2 27K Rsc 20k 240vac

Hello,

It appears the gain is too high by looking at the oscilloscope pictures.  One thing I noticed that can affect this is how you have Rsbs connected to sbias.  Sbias in this configuration is not a fixed stable voltage, as the error amp responds the voltage on sbias changes causing the opto diode current to increase or decrease.  You can try placing a resistor and zener from sbias to create a stable voltage though you will have to be careful with the range of current the opto diode can get to make sure feedback doesn't go away.  What is your Sbias voltage?  How much current are you pulling from Sbias, 220 uF is a large capacitor and will be part of the control loop.

If you look at the datasheet it has Vop1 and Vop2 created for this reason (Vop1 is the voltage source to the opto diode).

Are you sure Csc is 22 uF?  What part number is the opto-isolator you are using?  Or do you have the CTR range of this opto and the operating specification at very low current?

Regards,

Hi Irwin,

I've actually modified the EVM board to be close to the circuit generated by the XL sheet.  So Sbias is still connected to Rsbs through O2bias.

Hi Irwin,

Csc is 20uf, that is 2 x 10uf caps.

I have just changed Rsbs to 4k7 and this made very little difference (PF now is 0.6).   The regulated voltage at the top of Rsbs is 9.4V and this is very smooth and stable.

The opto is CNY17F-3S as per the EVM board bill of materials, you probably have more information on this than I do.

Regards,

Steve

Hi Irwin,

It seems that most of the problem was coming from the EMC caps on the EVM board, reducing these still passed conducted emissions and now the PF is at 0.88, I should be able to tweak this to improve it with my new transformer with the correct winding's.