Author Topic: Need helping figuring out noise and issues with TPS40210 AND QX9920 + LED  (Read 1960 times)

0 Members and 1 Guest are viewing this topic.

Offline MatCatTopic starter

  • Frequent Contributor
  • **
  • Posts: 377
  • Country: us
I amm trying to drive a CREE LED rated for 36v / 50 watts.   I used TI Power Bench app to come up with the circuit using TPS40210, and someone else had pointed me to QX9920 for current control.  The power supply for this project is a 360W 12V power supply.   The TPS40210 is set to put out 42V, and the QX9920 is set to draw 1.36A of current.  However it is only drawing about 170mA, the voltage supply circuit is supplying 42V, however its extremely noisy. 

This is actually the second version of this board I did, I respun in to add a bunch of caps and a common mode choke on the 3.3V and 5V rails.  Attached are schematics for the QX9920 and TPS40210 circuits, as well as various measurements of the 5V rail, 12V rail, QX9920 current sense, and mosfet drive pin for the QX9920.

Any help possible in getting this circuit to work is GREATLY appreciated.  Any additional info needed and I will gladly post.
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
I've come to the conclusion that the TPS40210 is terminally stupid, especially in how it deals with (the inevitable) noise on the current sense signal.

Try increasing the filter capacitor C23 to achieve a time constant of around 200-300ns; reduce the value of R6 if necessary*.

The bursts of ringing are from the lumped stray capacitance at the drain of the switch (including from the Schottky) resonating with the boost inductance. An RC damper across the Schottky or the switch (preferred) can help here.

* - EDIT - this sounds backwards - and it is - but the key here is that you don't want to arbitrarily increase the impedance of the current sense signal line. For example, you may need to triple C23 but cut R6 in half to get a good combination of time constant and node impedance.
« Last Edit: August 30, 2019, 06:38:02 pm by MagicSmoker »
 

Offline MatCatTopic starter

  • Frequent Contributor
  • **
  • Posts: 377
  • Country: us
I put a snubber across Q1 consisting of a 820pF cap and 33 ohm resistor, attached is the images of current sense for QX9920 and mosfet gate of QX9920, it's better, but not good enough, still got plenty of ringing noise that is higher then the 250mv threshhold for the QX9920 current sense.

I also changed the QX9920 filter cap to a 220pF from 150pF.
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
What is the switching frequency supposed to be* and what kind of resistor are you using for current sense? Note that any stray inductance in the resistor adds a leading edge spike to the current sense signal, and it doesn't take much inductance to equal - or exceed - 9mR.

You can also try increasing the gate resistor to slow down the transitions - and you want to change it, anyway, because no one uses values from the E192 series for gate resistors!?! Maybe try 15R to 22R.

More/better decoupling for the TPS40210 couldn't hurt, either.

Last but not least, the layout of switchers is crucial to good performance. You need to minimize the area and trace length of the connections between inductor, switch and diode (aka, the "switching node") and minimize the loop area of the diode - output capacitor - ground return.


* - I'm too lazy to figure it out from the RC values.
 

Offline MatCatTopic starter

  • Frequent Contributor
  • **
  • Posts: 377
  • Country: us
Switching frequency of the TPS40210 is set to be 565KHz, the current sense resistor is a 2512 PE2512FKE7W0r009L

I tried to lay everything out according to the datasheet recommended layouts and to keep the circuitry outside of the primary current loop, if I have to though I can completely re-lay it out.  I attached an image of the layout, it has a few notes on it.  I went big on certain copper areas because I wanted to help minimize heat, though with a heatsink on the giant overkill diode I am using for the TPS40210 there is no heat, so next spin I will drop that down to TO-220 with heatsink.
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Right, no ground plane... In the immortal words of the Mythbusters: well, thar's your problem. At 565kHz - and especially with transition times in the sub-50ns range - all forward and return current pathways want to minimize their area, which typically means they want to be mirror images of each other. You can either literally mirror the forward and return traces, or else use a ground plane and let the return current automatically mirror the forward current as needed.

A ground plane will also provide excellent E-field (dV/dt) shielding, and even a modest amount of H-field (dI/dt) shielding through the induction of eddy currents, both of which will greatly cut down on the amount of noise picked up from any ringing still occurring.

This will require a respin, of course, so in the meantime try increasing the gate drive resistor to 22R. The switching loss is going to go up, but you really need to slow down the edges with the current board layout.

 

Offline MatCatTopic starter

  • Frequent Contributor
  • **
  • Posts: 377
  • Country: us
Re: Need helping figuring out noise and issues with TPS40210 AND QX9920 + LED
« Reply #6 on: September 01, 2019, 12:07:16 am »
Yeah I am just re-spinning it, I will make sure to use a ground plane this time.  You recommend it on the bottom or top layer?
 

Offline ahbushnell

  • Frequent Contributor
  • **
  • Posts: 751
  • Country: us
Re: Need helping figuring out noise and issues with TPS40210 AND QX9920 + LED
« Reply #7 on: September 01, 2019, 01:52:36 am »
use four layers.

Top signal
ground plane
power plane
bottom signals

Andy
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: Need helping figuring out noise and issues with TPS40210 AND QX9920 + LED
« Reply #8 on: September 01, 2019, 10:40:01 am »
Yeah I am just re-spinning it, I will make sure to use a ground plane this time.  You recommend it on the bottom or top layer?

This is easily done with 2 layers, and I prefer to make the bottom layer ground; you always want the minimize the number of interruptions in the ground plane by traces and components, after all.

Don't run the feedback trace next to - and definitely not parallel to - the switching node (ie - the common junction of inductor, switch and diode). Also note that the input of a boost converter is relatively quiet while the output is noisy, so extra care is required in minimizing the loop area of output diode to capacitor to ground.

Finally, you need to keep the transition times for the switch under about 10-20ns at a switching frequency of 565kHz to minimize losses (that is, around 0.5% - 1% of the period), but this directly conflicts with the desire to not excite ringing in any stray resonances or induce spikes across stray inductances, especially in the current sense resistor. After all, even 1nH of stray inductance in a current sense resistor - a typical value for a 2512 package - will generate a spike of 0.5V at every switch turn on when hit with a 5A current pulse that has a rise time of 10ns, and note that the overcurrent threshold for the TPS40210 is a mere 0.12V...

 

Offline MatCatTopic starter

  • Frequent Contributor
  • **
  • Posts: 377
  • Country: us
Re: Need helping figuring out noise and issues with TPS40210 AND QX9920 + LED
« Reply #9 on: September 03, 2019, 07:14:58 pm »
I started the new layout, I changed the current sense resistors to 1W 1206, the attached image shows what I am thinking about for new layout, with primary current path traced out, as well as FB for the TPS, bottom layer will be ground plane filled, let me know if you think this layout will do OK.  BTW, return ground is far right pin of J3
 

Offline MatCatTopic starter

  • Frequent Contributor
  • **
  • Posts: 377
  • Country: us
Re: Need helping figuring out noise and issues with TPS40210 AND QX9920 + LED
« Reply #10 on: September 03, 2019, 10:54:14 pm »
This is the completed layout, let me know what you think before I send it off to JLC :P
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: Need helping figuring out noise and issues with TPS40210 AND QX9920 + LED
« Reply #11 on: September 04, 2019, 10:31:38 am »
Looks better, but there is a slot in the ground plane underneath C6, C7 and C18. Also, don't place vias too close to SMT pads or solder will get sucked down into them during reflowing and the component might tombstone.

Oh, and are there no mounting holes, or are they just not shown?
 

Offline MatCatTopic starter

  • Frequent Contributor
  • **
  • Posts: 377
  • Country: us
Re: Need helping figuring out noise and issues with TPS40210 AND QX9920 + LED
« Reply #12 on: September 04, 2019, 04:37:04 pm »
I purposely put that slot there because of the feedback and current sense lines running for a short period on the bottom layer, and I didn't want the current from the voltage regulator current sense resistor flowing between those lines, and yeah mounting holes have been added since that pic :D
 

Offline Apollyon25_

  • Regular Contributor
  • *
  • Posts: 66
  • Country: nz
Re: Need helping figuring out noise and issues with TPS40210 AND QX9920 + LED
« Reply #13 on: December 05, 2019, 12:23:04 am »
Bit late to this party, but is there a reason you didn't simply use the schematic shown in Fig29 of the TPS40210 DS?
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf