Hey All,
I am new to implementing buck converter in a design. I recently used AP63356 in a design with the following specification:
Vin = 6V - 30V
Vout = 5.4V
Fsw 450 kHz
While testing the design I noticed the output voltage spikes of as high as 10V for a duration of 20-30 microseconds or so. This only seems to happen when there is sudden interruption of power to the buck converter for a short period.
I am concerned that this could potentially damage the devices such as a microcontroller with absolute maximum voltage of 6V. I implemented this design by studying the typical application examples in the datasheet which does not show a need to add additional protection at the output.
Schematic:
AP63356-AP63357.pdf (1659.52 kB - downloaded 48 times.)
I would love to know from those who are more experienced in buck converter design how big of the problem this is and what are the general design guidelines to mitigate suck output voltage spikes. As far asn the PCB layout is concerned I am aware that everything in the buck converter circuit should have as low imedance as possible.
Component placement:
Layout:
https://ibb.co/3kmVN64https://ibb.co/89DZ33fhttps://ibb.co/JKv7YCRhttps://ibb.co/wSdmTrPMeasurement and Wiring: (Oscilloscope TDS3014 probe: PP215 200MHz)
https://ibb.co/m88G9Ln IMG_20230101_225625_hdr.jpg (2049.52 kB. 4608x2592 - viewed 125 times.)
Update:
The spike only appears when measuring in 10 x mode, and does not seem to apprear at all in 1x mode, so I am starting to think this might be measurement issue.
Thanks,
Kind Regards,
Sparsh
Show input waveform as well, and photos of wiring and layout.
FYI, that TVS at the input won't do anything, it breaks down at much higher voltages. If you need overvoltage protection on power lines, use an avalanche type TVS, rated for maximum nominal input voltage (so, a SMAJ30A or so would probably be fine).
Tim
Thanks for your reply, yes the TVS is there just for the ESD events, the complete input protection circuit is as follows:
In my experience with switching regulator design, physical layout has proven itself the most troublesome aspect. Most of the problems I've encountered were corrected by altering the layout. Pay special attention to grounding and any paths carrying high currents and their relative position to the feedback loop. Study the datasheet for the control IC carefully, often they will have layout examples.
probes used, Zin scope, photos of board with probes?
most likely bad measurement technique
j
Four varistors in parallel? And a TVS? And over and reverse voltage? ...But a unidirectional TVS? There are some inconsistencies here!
F1 should probably be out in front, unless there are reasons (limited voltage rating..?). Varistor/TVS are likely exclusive (either one does all the work, or the other), so pick whichever is suitable and go with that. The TVS (if used) should probably be bidirectional, so the polarity protection circuit has something to do (instead of nuisance blowing the fuse?). Otherwise, the TVS would seem to obviate both protection circuits (OV/RV).
And, I take it this is automotive application, so, load dump? Test the overvoltage circuit and make sure it is able to operate quickly enough; you'll have to rely on either very large TVS or varistor to absorb the brunt of it, while avoiding overvoltage of the regulator which is quite close to the stated nominal range (abs max 40V 400ms!).
Ed: I see the layout screenshots added. What layers are they?
Tim
Hey All,
I am new to implementing buck converter in a design. I recently used AP63356 in a design with the following specification:
Vin = 6V - 30V
Vout = 5.4V
Fsw 450 kHz
While testing the design I noticed the output voltage spikes of as high as 10V for a duration of 20-30 microseconds or so. This only seems to happen when there is sudden interruption of power to the buck converter for a short period.
(Attachment Link)
I am concerned that this could potentially damage the devices such as a microcontroller with absolute maximum voltage of 6V. I implemented this design by studying the typical application examples in the datasheet which does not show a need to add additional protection at the output.
I can Confirm this, these spikes have enough energy to kill sensitive MCU and similar devices.
From my experience, these look like control loop instabilities, caused by input voltage steps. Always take care to test your regulator circuis for such behaviour. Then take each useful measure to stop that. It might not just be layout or input voltage noise causing that, it's an intrinsic issue of the general control loop architecture - you may want to ask the manufacturer for support, or carefully read their app notes and datasheets - these should give you some valuable hints on their designs and how to use them in a successful manner.
If you want to learn something, go and study control loop theory at a university course - that's no simple matter that I'd be able to explain in a few words.
Schematic:
[ Attachment Invalid Or Does Not Exist ]
I would love to know from those who are more experienced in buck converter design how big of the problem this is and what are the general design guidelines to mitigate suck output voltage spikes. As far asn the PCB layout is concerned I am aware that everything in the buck converter circuit should have as low imedance as possible.
Component placement:
[ Attachment Invalid Or Does Not Exist ]
Layout:
[ Attachment Invalid Or Does Not Exist ] [ Attachment Invalid Or Does Not Exist ]
https://ibb.co/3kmVN64https://ibb.co/89DZ33fhttps://ibb.co/JKv7YCRhttps://ibb.co/wSdmTrP [ Attachment Invalid Or Does Not Exist ]
Thanks,
Kind Regards,
Sparsh
[/quote]
Four varistors in parallel? And a TVS? And over and reverse voltage? ...But a unidirectional TVS? There are some inconsistencies here!
F1 should probably be out in front, unless there are reasons (limited voltage rating..?). Varistor/TVS are likely exclusive (either one does all the work, or the other), so pick whichever is suitable and go with that. The TVS (if used) should probably be bidirectional, so the polarity protection circuit has something to do (instead of nuisance blowing the fuse?). Otherwise, the TVS would seem to obviate both protection circuits (OV/RV).
And, I take it this is automotive application, so, load dump? Test the overvoltage circuit and make sure it is able to operate quickly enough; you'll have to rely on either very large TVS or varistor to absorb the brunt of it, while avoiding overvoltage of the regulator which is quite close to the stated nominal range (abs max 40V 400ms!).
Ed: I see the layout screenshots added. What layers are they?
Tim
Yes D24 shhould be bidirectional, and it is not protected by Reverse voltage protection, only now i realize this, thanks. No it's not a automotive application, but probably an industrial application, four variastor, one for every possoble power input connector and one TVS before the Buck, idea is to keep the varistors as close the power input connectors as possitble. Load is high brightness LEDs. Blue is bottom layer, Orange on top of blue, then green and Red is the top layer
Update:
The spike only appears when measuring in 10 x mode, and does not seem to apprear at all in 1x mode, so I am starting to think this might be measurement issue.
Maybe check probe compensation?
No 10x probe is going to avoid noise due to the large area return path to ground.
Only a direct connection!
NO PROBES!
1. get small dia coax eg RG174/U, add BNC at one end.
2. strip and tin shield and inner SHORT legnth eg 6 mm. Add Téflon tube or heats rink around shield.
3. Connect DIRECT to output or DUT, add 100 nF series mono if AC coupling needed.
4 Connect ditrect o scope set to1 M AC.
5.Use scope at 1..10 mv/ div, BW LIM 20 MHz if desired.
Bon courage
Jon
Are you measuring with a load attached?
Is the ESR of your output caps low enough?
I'm guessing that the 10ppms are what you have around the lab?
The internal ref is drifts a lot more than that- cheaper E12 resistors will probably be fine. 27/4.7 ratio.
Rescale C8.
Beware of mad values shown in app notes!
BTW
If you need a precise 5.4V, a switcher is not the way to go.
I agree with capt bullshot; there is something wrong with the frequency compensation, which includes the input and output capacitors.
+1 for checking compensation. The datasheet has a fairly big calculation section. To me it’s a warning sign they need to do explaining over a potential weakness of the product.
for the probe compensation aficionados:
No 10x probe is going to avoid noise due to the large area return path to ground.
Only a direct connection!
NO PROBES!
1. get small dia coax eg RG174/U, add BNC at one end.
2. strip and tin shield and inner SHORT legnth eg 6 mm. Add Téflon tube or heats rink around shield.
3. Connect DIRECT to output or DUT, add 100 nF series mono if AC coupling needed.
4 Connect ditrect o scope set to1 M AC.
5.Use scope at 1..10 mv/ div, BW LIM 20 MHz if desired.
Bon courage
Jon
I think you've misunderstood the problem a bit... It's not a switching noise thing: look at the time scale, it's 100us long. If that's between grounds, they've got a MUCH bigger problem than just the waveform.
Tim
What happens if you probe the input? Any voltage spikes there?
I have checked the input voltage to the buck converter and it's completely clean, nothing as far as I could see, nothing close to the buck output spike, I think I have to agree here with capt bullshot as well that it seems to be a very typical loop compensation issue, given the over shoot, what I don't understand is why should this be the case since I almost exactly copied the design of AP63356/7/Q-EVM (attached), almost exactly, I am using a different inductor (Bourns SRP1245C-6R8M vs Wurth Elektronik 74439346068), and I don't expect the ESR of the capacitors that I am using to be very different either. I am using 2 x 22uF CL32A226KOJNNNE as output caps
[Usually, MFr. Ref.. Designs are intended for an quick and easy start to evaluate the chip - so they have all the components fitted to make it work at least, but not best optimized. Follow their hints on optimal loop comppensation.
Such kind of ringing often is the result of output load step changes - that's a commonly to be performed test to any SMPS design prior going production. From that test, you can get an OK/nNOK result, alas very few hints what component values to modify. Use the test and external simgle component only modifications as a directional indicator to tune the circuit
Actually I at this moment I am thinking of switching from ap63356 to a simple switcher from TI like LM51420, if that gives me higher chances of success without having to build and test the buck circuit sperately, because I may not have the resources to do that, also the reason why I did not do it with the current design. I chose ap63356 when there were no chips available. Also AP63356 seems to have a rather horrific absolute maximum Vsw rating, which is (Vin + 0.3)
As opposed to Absolute Max of Vsw being untied to Vin as any decent converter would have, plus TI has WEBENCH support
Hey Guys
I did a lot more testing, I am finding the problem very hard to replicate, many time it seems that it's not ever there, I finally decided to probe both the Vin and Vout together today, I am attaching the images, as you can see the Vout look totally normal, and Vin has been purposfully messed with. At this point it really does not look like a buck converter instability, otherwise it wouldn't be absent most of the time, that's my opinion.
https://i.ibb.co/rf3Mmvh/IMG-20230121-032639-hdr.jpghttps://i.ibb.co/4SwYDLP/IMG-20230121-032613.jpgAlso, I am seeing the kind of same problem with LMR51420 buck converter evaluation board that I am currently testing. Could it be that I am just picking up noise because the ground is getting disconnected before the power from the power supply? The thing is I was generating input transient by connecting / disconnecting that DC barell jack on the left, but DC barrel can easily connect without connecting ground first, now that I have a continuous ground going on with the power supply, I as large spikes as before.
https://ibb.co/D1jQ99NRegards,
Sparsh