EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: joeyjoejoe on May 13, 2020, 12:31:24 am
-
Well, powered up a new circuit and let out the magic smoke from the buck converter.
Hoping to get a 2nd pair of eyes for something obvious I've missed. The 12V buck is based on a reference design from TI's Power Designer.
Honestly, I wish I'd built the board once piece at a time - start with just the rectifier and the buck, do a smoke test. I don't know if it's related, since the TPS54202 took out the surrounding area, but measuring resistance between GND and 12V rail gives 20 ohms, which seems a little low. Removing the 3V3 regulator brought that up to a few kilo ohms which seemed more reasonable. But who knows if that's because of melted traces/planes, blown 3v3 reg, etc.
(https://i.imgur.com/yPRrrAx.png)
As we speak, I am checking all the footprints :)
-
the recommended L for 12Vout is 22uH (page 17) yours is "underhenried" 8.2uH, feed forward C6 is recommended 100pF, yours C4 is "overfaraded" 220pF and directly connected to L, is there possibility for oscillation? have you made probing the TP1 (12V) point? if its oscillate below 3V3 there is possibility reverse internal current in AZ1117 and burnt, next time for experimentation, put Vin protection diode from Vo to Vi of A1117. ymmv.
-
Thanks!
I'm certainly scratching my head at how the heck I got 8.2uH - I definitely don't have the smarts to be running these number by hand so I was using TI Power Designer. I just ran the same thing again, and now it is 22uH. |O
Cff is still putting 220pF.
(https://i.imgur.com/h5E8vCB.png)
I've got another IC in the mail from digikey, although now I'm delayed another day for that coil you found as well. I think I have an 18uH coil though kicking around.
Can I run the buck/12V portion of the circuit with no load? At least to ensure it doesn't blow up? I assume it will either start skipping a lot of pulses, or just shutdown altogether?
-
Thanks!
I'm certainly scratching my head at how the heck I got 8.2uH - I definitely don't have the smarts to be running these number by hand so I was using TI Power Designer. I just ran the same thing again, and now it is 22uH. |O
Cff is still putting 220pF.
(https://i.imgur.com/h5E8vCB.png)
I've got another IC in the mail from digikey, although now I'm delayed another day for that coil you found as well. I think I have an 18uH coil though kicking around.
Can I run the buck/12V portion of the circuit with no load? At least to ensure it doesn't blow up? I assume it will either start skipping a lot of pulses, or just shutdown altogether?
Could you have perhaps inadvertently combined the 80mOhm and the 22uH when noting the values needed and mistakenly marked it down as 8.2?
-Pat
<edit to add - definitely double check your footprints - they will bite you in the butt if given the chance. I go through datasheets and mark pin 1 and also note the orientation of the drawing - sometimes they use bottom view, sometimes top.>
-
Anything is possible :) I was also doing another design as well, but I did check and it's not using 8.2uH.
Same as proofreading text, you can look at the same block a thousand times, once you get used to it your brain somehow can't see the mistakes. I even checked powerbench a few times. Next time, screenshot.. either proof it wasn't me screwing up, or proof it was and to do better next time!
I should be able to bodge a slightly bigger footprint 22uH on the footprint. I do have some non SMT power inductors as well that might also work in a pinch. At least until the next board fab with all other mistakes fixed (I'm sure there will be some)
-
Anything is possible :) I was also doing another design as well, but I did check and it's not using 8.2uH.
Same as proofreading text, you can look at the same block a thousand times, once you get used to it your brain somehow can't see the mistakes. I even checked powerbench a few times. Next time, screenshot.. either proof it wasn't me screwing up, or proof it was and to do better next time!
I should be able to bodge a slightly bigger footprint 22uH on the footprint. I do have some non SMT power inductors as well that might also work in a pinch. At least until the next board fab with all other mistakes fixed (I'm sure there will be some)
I find that it seems 50% of the time I can scroll all over the screen for ten minutes looking at a SolidWorks drawing or OrCAD schematic trying to spot my screw up(s) and not see anything. Print it out on an 11x17 sheet and I'll spot the misspelling or whatever other error as I'm walking back from the printer. |O |O |O |O |O |O
-Pat
-
So I have some 10uH, and 47uH, both through hole.
Can I put two 10uH in series to bodge this? The footprint is SMT, lots of pad area to solder on. Obviously it's wildly unusual to do this, but just to validate everything is OK until ordering new inductors, new PCB with new footprints, etc. As I figure there might be other issues with the board, I'd rather use this for a few weeks and do a respin of the PCB with those changes as well.
-
Obviously it's wildly unusual to do this
thats how we do it during early prototype. but without a DSO to probe, it will be more difficult to reach optimum setting. maybe a DMM with frequency counter and Vpp display may help if no DSO available.
-
I do have a scope I could use to check out, but all I care about is things not blowing up really :) I think these coils are 1A as well, but for the bare minimum testing even 300mA is enough. 2A would only be if/when this is hooked up to an external horn.
-
with small oscillation, 1117 should survive. but since its dead, something serious has happened. not just to avoid any serious oscillation, but DSO usage to reduce it and further small tweaking to optimize PSU spec. tinkering without having a DSO around (while you have it) is something "wildly unusual".
-
It let the smoke out again. I used two 10uH inductors bodged in series, and I did not wire up any load or 3v3 regulator so I could focus solely on the buck.
(https://i.imgur.com/GNGxGB9.png)
(https://i.imgur.com/IvKsSbJ.png)
Pretty close to saying screw it, linear regs and some heat sinking here I come :)
-
Is pin 1 even connected to ground? It doesn't look like it.
-
It is - poorly hidden by silkscreen.
(https://i.imgur.com/jsJBGo4.png)
-
And it makes it under the capacitor to the actual ground plane? Also check the input never hits 30V. Have to check with a scope when power is applied. The tps chips are not joking about that limit and your ideal isn't far from it and you have no protection.
-
It does, but ugh.. what a terrible mistake. Not exactly a nice low impedance ground for the IC.
-
What is the input? Like, some kHz? 20uF isn't enough for rectified mains, not if you're drawing more than a few mA.
Yeah, waaaaaay too much trace length (and not enough width) between input bypass and reg.
More stitching vias. Like, more than ~none.
Follow the application layout!
May be able to salvage the boards by scratching away some soldermask on GND, near the chip, and putting a cap right there, between VIN and GND.
Tim
-
That's the first thing I noticed as well, you need at least an order of magnitude more capacitance on the input. Not only is 20uF not nearly enough, it looks like you've used MLCC's so they're gonna be a lot less than 20uF by the time you put 15VDC across them. Your input capacitance is probably around 5uF, when it should be closer to 1,000uF depending on the current you're wanting to draw.
-
Rushed this design and it shows! I got fixated on the TI Power Designer saying 2x20uF for input of the module, ignoring the fact they assume I'm supplying it DC. Although in this test case the current draw was zero, so I'm hesitant to try to just add some bulk capacitance and try it again.
I could bodge a wire to fix up that grounding issue, but that itself would add some inductance.
To go linear from 24VDC to 12VDC at 2A would be quite a bit of heat to dissipate. Looks like I might have to sharpen the pencil and figure out the kinks in this one and do a version 2.
-
And it makes it under the capacitor to the actual ground plane? Also check the input never hits 30V. Have to check with a scope when power is applied. The tps chips are not joking about that limit and your ideal isn't far from it and you have no protection.
Hmm good point as well, protection would be good to have. Could I do this with a simple Zener and add an inline fuse to the rectified side?