LM2679: Doesn't reach full output at full load

Hi Sam,

Thank you for the quick reply! Please see attached. As of right now, there are a few value changes from the schematic:

1. R186, was 7,5k now 4.64k (attempting to eliminate current limit)
2. C126 was 1000uF, now 100uF
3. L5, C124 removed.

LM2679Schemo.pdf

Thank you again!

Frank De Stasi said:

I think that a 22uF is rather large for a SS cap. The leakage may prevent the SS voltage from coming to its normal value.
I would try to remove the SS cap as a test and see if you get full voltage.
If you need such a large SS time, you may need to use a ceramic cap.

Thank you Frank! Unfortunately, with the SS cap removed, the core issue remains, although, this did appear to eliminate the issue where the voltage remains at 0V for some time. The only reasoning for using the large SS time is because tube heaters have a much lower resistance when cold, long is good in this case.

Frank De Stasi said:

It looks like you have a total of about 2000uF on the output. This is a little large as well.
You may also try removing one of the 1000uF output caps and see what happens.

Yes, I've tried this. At first I lowered them both to 100uF, then I removed the entire second filter and left a single 100uF which is where it stands now. Same results although the issue appears to be slowly improving with each tweak.

I still noticed the 1-1.5V drop across the inductor today. With a max DCR of ~40mohm, that implies 25A of current flowing, no? 

Is there anything else this could be? 

Frank De Stasi said:

I am not sure how you are measuring the drop across the inductor, but you would need to look at it with a scope. It would be good to see some waveforms of the switch node and vout under various conditions.

Thank you again Frank! I'm measuring with a DMM, which I wouldn't think would be ideal but it always shows exactly my calculated figures of (8.1V in standby, and 12.2V in normal mode) with light loads, and with intermediate loads it will eventually get and stop there after a few minutes, which says to me that its accurate enough to show us something. With the heavier loads, I simply go back and forth between the diode and output nodes and the output node always lags by around that one volt. When the diode node maxes out, the output node still slowly rises as there's plenty of time to probe back and forth, although oddly, it sometimes seems that the output node will suddenly make a fast jump and suddenly it will top out at the 8.1 or 12.2V.

Frank De Stasi said:

I would check the current rating of the inductor to be sure it is good to at least 5A.

The inductor is should be fine, listing an Isat 10% of 7A. It was much beefier than the ones Webench kept spitting out.

Frank De Stasi said:

It is also a good idea to double check the values of the other components and that the connections are good.

Ya, I went ahead and just resoldered everything to be sure but I didn't quite finish in time to test tonight.

Just before I did that, I was wondering if maybe the small 100nF ceramics were an issue, maybe trying to draw very high current spikes through the inductor. Oddly, I found that removing those caps actually made things worse. The past day or so, it seemed like I could, if I waited long enough, get to the nominal voltages at full load, but after removing them I found I was back to square one where we were topping out in 9-10V range. Could that imply that we might have too high an impedance at the switching frequency and maybe it would be best to try some big ceramics?

I'll get you some scope shots tomorrow. Anything in particular you want me to try to grab?

Thanks again for your help!

Thank you again Frank!

Unfortunately, we'll have to wait a few days. Today after resoldering everything I was only able to get 2V max at half load (1.8A...300mA was fine), and my desoldering gun's motor pooped out. I'm only damaging the board with the soldapult, so I'm just going to wait until the new tool arrives. I'll get some new parts too in the interim to see if that's not it.

Ok, I've got some new info. First, attached are some scope shots.

These are all 5V/d and .5us/d with a 1x probe with a diy ground spring that I made today with a load current of 1.8A. The 1st pic is in 8V mode, the second in 12.2V mode. What is interesting is that before I made the ground spring, we were getting the same old slow rising output voltage... but with the spring, the voltages reach their nominal values almost instantly. Funny enough, if I remove the scope probe after the voltage has risen to 8V in standby mode and set the unit to a 12.2V output, the voltage will actually fall from 8.1V to about 7.5 and will never rise to 12.2V. The second I put the scope across the diode, the voltage instantly rises to 12.2V.

Now, the third picture is in 12.2V mode but I'm not making great contact with scope probe. I believe this shows us more of what is going on when there's no scope. I notice that when I can make the ringing appear like this the output voltage drops to 11.6V. If the scope shows a clean square wave like the first two pics, then the output voltage immediately jumps to 12.2V again. 

My guess is the spring/probe are adding just enough capacitance to snub any instability in the IC output. Is there anything else that it could be? And if so, what do I need to do in the redesign to ensure I tame this ringing/instability in the first place?

Another "clue" might be that before I took these shots I was playing with the bare-bones setup with only one of the 3300uF caps in place and no Cinx. I was only able to get around 2V output, but I found that by inserting Cinx and an OSCON 47uF Cin I could make the output voltage jump to 4-5V. Any clue what this might be? This suggests that my input layout might be playing a role too, such as how the input/output grounds are only joined at the GND pin trace? 

Ok, I think I've captured it finally by using the 10x probe setting and a different spring: a cycle is missing. See below.

I took this pic while the voltage was doing it's slow ramp up thing (~5V). Last night I was able to get the circuit working as expected (and at full current) if I touched a small film cap across the diode...100pF to 330pF all worked. If I touched the film cap across the diode while observing the scope, we would see another positive cycle where the spike appears in the image above, and instantly the voltage jumps to the correct value. When the cap is removed, we again see the above image and the output voltage drops.

Interestingly the cap must be directly across the diode for the circuit to work properly. From the diode node to ground pin and input ground didn't work (as shown in the EVAL kits). 

What's going on here? Based on the scope, it appears to happen when the diode stops conducting, as if the controller is getting messed up and misses the next switch turn on. Ground bounce?

Or, I'm beginning to suspect the diode (V8PA10) itself might be playing a role as it has a very high capacitance (850pF) and high reverse current, and they do not list the recovery time (which I assume means it is rather poor). Can the diode be playing a role here or is what we are seeing something else? 

Sounds like you may have a bad connection some where.

I would try to get an EVM from TI and see if you can start with that and

get your application working.

Katelyn Wiggenhorn said:

Did you have a chance to look at the TI EVM and see if you could your component selections working?

Katelyn Wiggenhorn said:

Could you also confirm the inductance value and the output capacitor ESR used for this design? Based on the last waveform that you shared, I want to verify that your component selections are enabling a stable applications circuit.

Inductor is either 22uH or 33uH, I've tried both. For the output cap, I have a 47uF OSCON in there now with an ESR of 27mohm.

Another odd thing is that if I bend the input cap C65 towards the IC we get a higher max voltage, so I suspect the issue isn't a problem with the output but rather the input.

Hello,

Did yo get a chance to try this experiment?

Thanks,

Katelyn

Hi Katelyn,

No not yet. I also have boards on the way with a new layout to test. I'll post with the results!