Non-isolated flyback DC/DC circuit does not work (LT8316)

[Wor3d]

Datasheet: https://www.analog.com/media/en/technical-documentation/data-sheets/lt8316.pdf

Hi there,

I have built a circuit based on the datasheet application Figure 9. Nonisolated 24V Buck Converter with Optional Undervoltage Lockout (with modified feedback and components due to availability at our distributors, and desired PCB footprint size (for mosfet)).

Unfortunately, I cannot get it to work properly... The output is oscillating weirdly at around 3 volts at a low, visible frequency. Feedback is therefore reaching only a few milivolts.

I have checked all the components, replaced all diodes, mosfet, and the IC itself, re-measured all values and replaced them, to be sure I have got them correct like in the schematic, and that they are working. Find below screenshots of schematic, pcb layout (with other stuff around) and oscilloscope screenshots. (Edit: picture states 100 200 .. 500 pF tested, I meant 10 20 .. 50 pF)

Now I am out of ideas, I have debugged this board for two days now... Any more experienced engineers with ideas? Thank you.

[T3sl4co1l]

Huh, weird controller.

The lack of ground plane, and the unshielded inductor, are huge red flags.  I don't know if that's enough to explain the symptoms but it's not going to help.

Tim

[Wor3d]

Well, when dealing with 600V, I am not sure how a ground plane is an option. Do you think that a shielded inductor would help? I also have tried standing the inductor up with no results.

[james_s]

Why can't you do a ground plane with 600V? Obviously anything through-hole is going to need to have clearance around the pins. Oddly I don't see any example layouts in that datasheet, can you find any application notes or an evaluation board using that part to get an idea of how it's done? My experience with buck converters (and switching converters in general) is that layout is really critical, especially in parts of the circuit where there are high circulating currents and their proximity to sensitive parts of the control loop. The layout can be the difference between a circuit that works perfectly and one that behaves wildly or just blows up before you can even see what's going on. I see some rather thin traces around the inductor in particular. Be very mindful of what traces you have running around on the bottom too, when you run a trace under another trace you form a capacitor and at high frequencies in particular you will get coupling between them.

[Wor3d]

Well yes I can do a ground plane, but it would have to have large clearances...

I am also surprised by the lack of PCB layout guidelines, even when they are not 100% perfect, at least they are there... When I was designing the PCB I got little bit inspiration from the eval board (https://www.analog.com/en/design-center/evaluation-hardware-and-software/evaluation-boards-kits/dc2781a.html#eb-overview).

But I thought that the EMI etc. would not be a big problem, because it operates at around 100 kHz and I will be drawing static 20-30 mA (the circuit is just an application-specific high voltage indicator, basically). Everything other than the DC/DC inductor loop is basically DC on the board. Only traces at the bottom are the HV+ supply and GND trace.

Before committing to re-designing and re-spinning the board, I would like to test out as many hacks and mods as I can so I can be sure the re-spin will actually solve the problem. I can see some improvements that could be made, and I even hacked some of them on the board (like input capacitance), but to no successful end.

[james_s]

I'd be tempted to try building it Manhattan style on a piece of copperclad for the initial prototyping. It's a pain with little SMD chips like that but there are ways of doing it, and you can sub through-hole parts for a lot of the passives. That will at least tell you whether you've got the inductor right, for me at least the magnetics are always the biggest challenge.

There is an evaluation board for a non-isolated converter using that chip.
https://www.digikey.com/en/products/detail/analog-devices-inc/DC2793A/10244109

It's a bit spendy, but you can get some clues just by looking at pictures. Note those big fat traces in the area around the inductor, diode, input and output. Those are the areas you have to be really careful with as they will have large currents. You might look and see if AD has a PCB layout for that eval board available, or contact them and ask for one if you can't find it.

[T3sl4co1l]

The schematic only says 100V, is this actually supposed to run at 600?

Please post your complete specification.

Tim

[Wor3d]

Yes, it says I am currently testing up to 100 V, currently I cannot get bigger voltage power supply with current protection in case something goes wrong 

When doing the layout and choosing the component values, I was inspired by the non-isolated eval board (https://www.analog.com/en/design-center/evaluation-hardware-and-software/evaluation-boards-kits/dc2781a.html#eb-overview).

But the actual specs are (in blue) 60-600 V to 12 V (560V nominal, 600V peak, as the datasheet allows. Under-voltage lockout, so under 60V output is guaranteed to be off). It will be in a high voltage battery (electric formula student car), the next generation will be close to 600 V. This generation I am okay with 410 V, but I wanted to have it designed upfront as much as I can.

After the DC/DC there will be an isolated DC/DC 12V to 12V regulator (HV and LV systems must be galvanically isolated) and it will light up an LED at ~20 mA. Currently I am testing it without the isolator, and just with 1k resistor and red LED (10 mA load).

[Wor3d]

Update: I machined a small inductor shielding from Iron on a lathe (about 1mm thick), did help with the low frequency oscillations (visible ones) but not with the fact that I am getting 2-3 V instead of 12V. Feedback resistors are correct, I am not getting desired feedback target on the pin (1.22 V). I also tried to rotate the inductor from horizontal to vertical etc. with no difference.

[cortex_m0]

I had written out a longer reply but lost it...

Anyway, I observe in your lower-left panel that the Gate switching stops for the last 7 ms or so. I think that's a huge clue. The chip is determining somehow that protecting the gate is necessary. There are a number of protections and lockouts in this chip, but I think it would be good to start by verifying all the voltages on the pins of this chip are in the proper rangte and steady when the Gate pin stops switching.

[Wor3d]

I will try to measure all the pins today, because I checked them all, but didn't take screenshots.

I also tried shielded inductor yesterday, no help, unfortunately. Tried several ones (also with different values, and tried to put the unshielded one physically far away from the circuit connected with thick wires), increasing/decreasing output capacitance, increasing DCM cap and decreasing DCM resistance, changing RC values on the OTA output (Vc pin) as well as adding a resistor to Ireg pin.... 

[james_s]

Given the thin traces and lack of a ground plane I'm betting the problem is that ground is bouncing all over the place and wreaking havoc. This behavior is exactly the sort of thing that happens with a bad layout, and the layout is nothing like that used in the eval board. The traces are tiny in the critical areas where they need to be big and fat.

[Wor3d]

Yes the ground plane... I guess I am still a bit scared of pouring a plane on a high voltage PCB, but I will do it. The question is, which ground to pour? The high voltage - ground, or the "Pseudo-ground" of the IC? (See schematic). Because of the transistor switching into the pseudo ground, I guess the HV ground? Going by the reference images (https://www.analog.com/en/design-center/evaluation-hardware-and-software/evaluation-boards-kits/dc2781a.html) they have poured more of a thick trace to the bottom of the board, and thick trace on the top. That is probably the way. Also I thought the bottom ground plane wouldn't matter much because the fact that it is a "thick" 2 layer board, but I guess I was wrong. Also the currents should have been low (20mA output so I thought like 1mA input... so no need to have big traces) -> I was probably wrong too....  I will post a (hopefully) improved PCB design here

[james_s]

The evaluation board is going to be your best guide. The ground plane is not just low resistance, it is low inductance. Any time you have a trace snaking around you have an inductor, and when you are dealing with what is effectively RF, stray inductance can cause you big problems. It may have a low DC resistance but at high frequency it will behave like a much higher resistance. It is vitally important that all nodes that are supposed to be "ground" are at the same potential at all times. Switching regulators are tricky, it took me a while to learn that. Layout is absolutely critical and if it isn't done right, it will not work right at all.

[Wor3d]

Definitely, I can understand that. I just underestimated the current flow in this design. I have done several DC/DC converters and had them reviewed, but in this case I was a bit sloppy.

But I am still a bit unease with this IC, because of its ground connection... You can see, that effectively every time it switches on, the IC's ground potential changes... I hope that it'll be okay, I guess engineers from LT/AD know what they are doing 

[nnikospapadakis]

GOT ANY UPDATES, ALSO FROM A FORMULA STUDENT TEAM I DESIGNED THE EVAL BOARD BUT IM HAVING THE SAME PROBLEM, I GET JUST 2V ON INTVCC