Buck regulator not able to reach full voltage

Thread Starter

liquidair

Joined Oct 1, 2009
192
Hi all,
I've built a regulator circuit using the LM2679 Simple Switcher....my first time using switching circuitry. It was designed to deliver 2.7A into a resistive load at either 8 or 12.2V (controlled by a switch). The circuit works fine at low loads, but as soon I get the load above 1.5A, the voltage will very slowly increase and eventually will stop rising far short of the desired voltage, usually in the 5-6V range. Oddly, throwing the switch from 8 (startup) to 12V will not affect how high the voltage will rise, which seems to rule out problems in the feedback network.

Using a scope across the diode, I can see what looks to be a missing cycle every 2nd time the wave tries to go positive (switch turning on), with only a spike to indicate where it should. I found by accident that setting the scope probe to 1x, the voltages come up immediately (most of the time) and I see a nice square wave. I can also cure the problem just by touching a 100-330pF cap directly across the diode. When I scope this I can alternate seeing a nice square wave and full voltage output, to a missing cycle and slow rising voltage simply by touching or removing the cap.

My question is, what is causing this? Ground bounce? Poor parts choices? I do wonder about the diode that I chose...it has 800pF of capacitance and 4mA of reverse current. Given that the problem occurs on the positive going cycle (diode switching off, switch turning on) this seems to point to the diode but I don't know enough to be sure.

I'd prefer to get the circuit and layout right rather than to have to add a band-aid in the capacitor. Any help/teaching would be appreciated!
 

Thread Starter

liquidair

Joined Oct 1, 2009
192
Is this on a solderless breadboard?
Show us the schematic and board.
No, PCB. Attached are the schematic and board images.

The actual circuit has been tweaked a bit to simply:
  1. C64 is now a 47uF OSCON cap (trying to see if Cin ESR was issue)
  2. Soft-start cap C66 is removed (simplify)
  3. L1 is now 22uH, was 33uH (seeing if saturation was occuring)
  4. C69 is now 47uF OSCON (seeing if Cout too large and if ESR was issue)
  5. L2 and C70 removed (simplify)
    LM2679Schemo.png LM2679Layout.jpg
 

crutschow

Joined Mar 14, 2008
34,285
What is the input voltage when the output voltage isn't correct?
What is the transformer current rating?
 
Last edited:

BobaMosfet

Joined Jul 1, 2009
2,110
Where are your calculations? Usually you have to switch at a specific frequency, and you have to understand power dissipation and have a few other calculations for component choices so that you can establish the potential and the max possible current available coming out of the switcher.

If done properly, you'll get higher voltage out of the switcher than expected with no load (not by much), but with a load it will settle down and you will get the voltage and current where you need it. And it isn't like it 'ramps up' over a long period of time, it should get there within a few ms.
 

Thread Starter

liquidair

Joined Oct 1, 2009
192
Where are your calculations? Usually you have to switch at a specific frequency, and you have to understand power dissipation and have a few other calculations for component choices so that you can establish the potential and the max possible current available coming out of the switcher.
I followed the design procedure in the datasheet pretty closely. TI's Webench software seemed to confirm these values, so I went with it. My deviations were to use a larger output cap (but lowering this down to target values had no effect), used a larger diode than specified, and the inductor value could have been either 22 or 33uH so I first tried the 33uH. AFAIK, the frequency is fixed at 260kHz in these units and max current is 5A. Current limit is set to 4.95A. I have the notebook with my calc's here if necessary.

If done properly, you'll get higher voltage out of the switcher than expected with no load (not by much), but with a load it will settle down and you will get the voltage and current where you need it. And it isn't like it 'ramps up' over a long period of time, it should get there within a few ms.
Interesting. At no load I observe exactly what was calculated, and I see those same figures when I have the cap across the diode at full load. And with the cap, it seems instant for the voltages to come up. It can be minutes however without the cap to get to whatever it want's to peak at.

Whatever low voltage it decides to reach at full load (without the snub cap) seems to depend on both the input and output caps. I've gotten it to reach the target voltages at medium loads, although after some minutes, and I've gotten it to where the voltage won't rise above 2-3V at that same medium load. I can't be sure why, but it seems related to the input and output caps and frequency because the worst performance was when I removed the ceramics.

Another oddity is that when I was using the soft-start, the regulator would actually not start for what appeared to be the soft-start duration. In other words, I'd see no switching at the diode node nor any output voltage. I don't know if that helps or not.
 

crutschow

Joined Mar 14, 2008
34,285
What current is the inductor rated for?
I couldn't find a reference for the number you have on the schematic.
I also could find no info for that diode number.
Who did the PCB layout?
 
Last edited:

Thread Starter

liquidair

Joined Oct 1, 2009
192
What current is the inductor rated for?
I couldn't find a reference for the number you have on the schematic.
Mouser says 8.6A, but the datasheet gives a figure of 4.8A at a 40 deg C rise and Isat 10% at 7A. The 22uH I have in there now is 5.3A (40degC) and Isat 10% 8.2A so they *should* both be OK.

I also could find no info for that diode number.
Here's a link to the diode's datasheet: https://www.mouser.com/datasheet/2/427/v8pa10-968328.pdf
Admittedly, I selected the diode way oversized because I didn't quite understand how things worked at first but I wanted to ensure it's survival even if the inductor saturated and the current rose drastically.

Who did the PCB layout?
I did. Please go easy, lol. It is heavily based on the TI design in Webench however. The thing I don't like about it is that I have the input ground connect to the output ground through the IC's ground pad. The location of the feedback resistors and the opto that switches the feedback resistors forced me to not be able to bring the input ground on that side because the loop would be huge. The feedback trace forced me not to be able to make a ground plane on the bottom...based on what I've read since then, it would be best to bring the input ground right to the diode ground, as the input cap to diode loop is the nastiest in terms of changing currents and ground bounce.
 

crutschow

Joined Mar 14, 2008
34,285
Okay, I'm at a loss as to what's causing your problem.
All I can think of, is a layout or ground problem.
Look at the complete high current paths for both when the MOSFET is on and when it is off.
If any of those paths don't have a low impedance connection, you might parallel it with a good sized wire, or a piece of braid (better).

Actually the diode current path is to the output capacitor, not the input.
The MOSFET ON path is from the input cap to the output cap.
So there should be a really good, short ground connection from the input cap to the diode to the output cap.
Also need a good connection from the input cap to the MOSFET input, from the MOSFET output to the diode and inductor input, and from the inductor output to the output cap.
 

BobaMosfet

Joined Jul 1, 2009
2,110
I followed the design procedure in the datasheet pretty closely. TI's Webench software seemed to confirm these values, so I went with it. My deviations were to use a larger output cap (but lowering this down to target values had no effect), used a larger diode than specified, and the inductor value could have been either 22 or 33uH so I first tried the 33uH. AFAIK, the frequency is fixed at 260kHz in these units and max current is 5A. Current limit is set to 4.95A. I have the notebook with my calc's here if necessary.



Interesting. At no load I observe exactly what was calculated, and I see those same figures when I have the cap across the diode at full load. And with the cap, it seems instant for the voltages to come up. It can be minutes however without the cap to get to whatever it want's to peak at.

Whatever low voltage it decides to reach at full load (without the snub cap) seems to depend on both the input and output caps. I've gotten it to reach the target voltages at medium loads, although after some minutes, and I've gotten it to where the voltage won't rise above 2-3V at that same medium load. I can't be sure why, but it seems related to the input and output caps and frequency because the worst performance was when I removed the ceramics.

Another oddity is that when I was using the soft-start, the regulator would actually not start for what appeared to be the soft-start duration. In other words, I'd see no switching at the diode node nor any output voltage. I don't know if that helps or not.
You need to measure more things, and then think about the whole as a system and see what altogether the values are telling you. You also need to understand the relationship between voltage and current is reciprocal. You must consider the behavior of both simultaneously, not apart as they are never independent of one another.

Caps in A/C are used to control the flow of current based on frequency. In essence, a Resistor with no thermal issues. What frequencies are your caps in & out dealing with, and based on voltage and current, are these values in spec based on frequency? If not, what's impacting? Too much load, wrong frequency, wrong-sized cap?

Don't forget your circle equation: C = Lamda * Voltage where C = speed of light, Lamda your wavelength, and V = Hertz.
 

Thread Starter

liquidair

Joined Oct 1, 2009
192
Thank you for your help crutschow!

So it looks like the weak link is the input cap to diode connection if I was judging, but it looks like I can get some wire to bridge the planes in there.

The article I was referencing was about ground bounce in buck regulators and made the point that the most dangerous area is the difference of the two loops you mentioned (input to output cap, diode to output cap)...something to do with AC currents causing ground bounce.

What does the fact that a small cap across the diode cures the problem tell us? Could the diode's capacitance be a problem? I do notice that all the parts that Webench gives have capacitance in the 100pF rance (vs 800pF) and leakage currents in the uA and nA range. I would think that the MOSFET would have to overcome both of these parameters to switch on. But what is confusing is that adding more capacitance solves things. Does that point to a certain resonance with the current values?

Also, is there any danger of just soldering the cap in place and calling it a day? I did hold it for a long time to see if there was any excessive dissipation as a result...there didn't seem to be...nothing got hot, just slightly warm, which is impressive for a regulator circuit putting out 30W+!
 

Thread Starter

liquidair

Joined Oct 1, 2009
192
Caps in A/C are used to control the flow of current based on frequency. In essence, a Resistor with no thermal issues. What frequencies are your caps in & out dealing with, and based on voltage and current, are these values in spec based on frequency? If not, what's impacting? Too much load, wrong frequency, wrong-sized cap?
So, we'd be looking at 13mΩ with a 47uF @ 260kHz in series with an ESR of 27mΩ. And these are OS-CON caps that appear to be rated for 100-500kHz, but that is not typical of the electrolytics I was using prior, or the 3300uF set. But even the 3300uF are computer grade rated at 100kHz...but this may not show the same picture at 260kHz where they may appear inductive. Now, the 100nF ceramics I use for bypassing are 6Ω, but with no appreciable ESR. I can see where a 10uF ceramic would be a better bypass as it would have an impedance of 60mΩ at 260kHz. Some of that would surely appear in parallel with the 40mΩ OS-CONs (depending on it's actual impedance vs freq), which would at least guarantee a 40mΩ impedance for the input and output.

Beyond 260kHz I can't tell, as I'm not capturing any ringing or anything that suggests higher frequencies are at play using a probe with a ground spring.

Am I on the right track?
 

BobaMosfet

Joined Jul 1, 2009
2,110
So, we'd be looking at 13mΩ with a 47uF @ 260kHz in series with an ESR of 27mΩ. And these are OS-CON caps that appear to be rated for 100-500kHz, but that is not typical of the electrolytics I was using prior, or the 3300uF set. But even the 3300uF are computer grade rated at 100kHz...but this may not show the same picture at 260kHz where they may appear inductive. Now, the 100nF ceramics I use for bypassing are 6Ω, but with no appreciable ESR. I can see where a 10uF ceramic would be a better bypass as it would have an impedance of 60mΩ at 260kHz. Some of that would surely appear in parallel with the 40mΩ OS-CONs (depending on it's actual impedance vs freq), which would at least guarantee a 40mΩ impedance for the input and output.

Beyond 260kHz I can't tell, as I'm not capturing any ringing or anything that suggests higher frequencies are at play using a probe with a ground spring.

Am I on the right track?
The problem with high frequency is that it requires higher voltages in order to charge/discharge an inductor to the desired level. Your inductor is the heart of your regulator (and any component related to controlling current flow immediately adjacent to that inductor), everything else is a filter.

Your switcher operates at about 260kHz. But what is your input frequency, 60Hz? Your schematic shows 15.5VAC input, but you don't list frequency unless this is a stepdown from MAINs so it'd be 60Hz. How smooth is the DC signal you have going into the switcher after all the filtering on the front end?

I would recommend you first attempt using a car-battery at 12V for your input, so you have a solid DC input that can supply any level of current you desire, and see what happens to your output. Go from knowns. Don't change a lot of things, make an educated decision to change one thing, and then make a correct observation. Restore circuit state and try another educated decision, if the first didn't give you enough information.

Forget things like ESR and all that B/S. That is not in the realm of what you're working with right now. Just concentrate on making it output a steady voltage with whatever load (within your desired output current range) you put on it. Understand that your switcher must also be _able_ to output enough current to be "stiff", or it will sag under a load.
 
Top