Board design advice/review, SEPIC DC-DC converter

[GardG]

Hi all,

I'm laying out a board for a SEPIC DC-DC converter intended to convert the output of a 16-cell NiMH battery to 23V. I've made a few PCBs before, and usually get it right after a couple of tries, but thought I might ask for some second opinions on this one as I don't really have any experience with DC-DC converters, which might be obvious …

The board isn't finished, it will be extended in length to fit in a some MOSFETs for dimming LED strips (that's what the 23V is for), an Arduino Nano 33 and a LumenRadio TimoTwo wireless DMX (stage lighting control protocol) module, plus some terminals. I'm confident I can get that right tough, it's mainly the DC-DC part that's out of my comfort zone.

I used TI WEBENCH Power Designers to get some suggestions for the circuit, and actually ended up downloading an Eagle schematic it made, which is why the attached schematic is full of idiotic cosmetic errors. Haven't gotten around to cleaning it up yet ... Attaching it for reference, but it's mainly the board layout I'm concerned about. Note to self: WEBENCH is handy, but trying to save time on downloading the schematics it generates is false economy …

My approach was to place the components in a way that made it possible to keep all traces on the top layer, to get a completely uninterrupted ground plane on the bottom layer. There are ground fills on the top layer as well, plus a VIN fill. I've tried keeping the traces nice and beefy, may have overdone it slightly …

I guess it's not ideal to have the feedback trace running under the inductor. I can't make the board any wider, as it's for an extremely space constrained application, so I can't really avoid it, but maybe I should put it on the bottom layer?

In case it's useful, some background on the project: this is for some handheld LED tubes built for a contemporary dance performance, to be used as props by the dancers, with the LEDs controlled from the lighting console. The LED strips are controlled in pairs, giving some degree of directional control of the light. The tubes are 40mm (OD)/36mm (ID) satin Plexiglas tubes containing an inner 20mm steel tube with 6 COB LED strips glued on the outside, and NiMH "stick" batteries and control electronics inside. The first revision used an Arduino Nano clone with a (probably knockoff) NRF24, an Aliexpress-style DC-DC regulator and some MOSFETs on a perf board, all bodged and taped together. This hasn't been as stable as I'd hoped, and is a bit of a pain to maintain, so I'd like to get everything built on one long, slim PCB as described above.

Attachments:
- Screenshot of DC-DC regulator layout
- Screenshot showing GND fills only
- Ugly schematic
- Photo of final application


EDIT: I have tried to give the nets meaningful names, mostly dictated by what IC pin they're connected to, but there are a few that I'm not sure what to call, so they're stuck with the non-descriptive names 5, 6, 7 and 8. Any recommended naming practices for those?

[thm_w]

So 16-24Vin and 23V out. What current output?
What is the switching IC part number?

Inductor is shielded which is good.

I would have a via under C8, taking Vout on the bottom of the board, as you said, then R8/R9 beside C3. But realistically here, its for an LED driver, some additional ripple is not going to matter. Should work OK as is.

[mariush]

I'm not gonna comment on it because using 16 nimh just seems silly.

The 20mm inner diameter seems like it's big enough to allow using 18650 batteries, so you could have 4-6 of them in series for ~12-24 volts (4.2v peak per cell)
I'd actually keep it to 4 batteries just to keep the charging easier, and to allow using a simpler boost only led driver IC, which can boost 10-14v to 23v or whatever the leds need, and can pwm the light at high frequencies.
Another option would be to have 8 batteries, but have 4 groups of 2 batteries in parallel , so it's 4 in series, but twice the capacity.

You can use commercially available chargers to charge the batteries without extracting them from the tubes. Simply place a jst-xh or a custom connector with 5 (or more) pins on one end, and you can connect the tube directly to a charger that will also do cell balancing.
For example, here's a 10$ charger that can do 2S/3S/4S charging and balancing (requires separate 9-16v laptop adapter style psu to power it) :  https://hobbyking.com/en_us/hobbykingr-dc-4s-balance-charger-cell-checker-30w-2s-4s.html?queryID=c9be7a10df30a3cbf4ad7d158924aaa7&objectID=39921&indexName=hbk_live_magento_en_us_products


Here's a fancier charger that can do up to 6 in series and has power supply built in : https://hobbyking.com/en_us/imax-b6ac-v2-professional-balance-charger-discharger.html?queryID=c9be7a10df30a3cbf4ad7d158924aaa7&objectID=45235&indexName=hbk_live_magento_en_us_products


Here's examples of boost led driver ics : https://www.digikey.com/short/zb850c

For example the cheapest there Ap3063, super simple to use, pwm signal to control brightness, use one for each color and you're good to go : https://media.digikey.com/pdf/Data%20Sheets/Diodes%20PDFs/AP3036,A.pdf

Or you could use something like LP8867 with up to 4 channels each up to 120mA and 45v : https://www.digikey.com/product-detail/en/texas-instruments/LP8867QPWPRQ1/296-LP8867QPWPRQ1CT-ND/10445184

[GardG]

Thanks a lot for the feedback! The IC is LM5022, max output current 2A. Sorry I forgot to mention those rather significant details!

Regarding the 16 NiMHs etc ... There's a long story behind that solution, where extremely short deadlines was a significant part. (Right now I'm just retrofitting better control electronics) The inner tube is ID 18mm / OD 20mm, so 18650 JUST won't fit, and finding a 1mm larger tube quickly enough proved difficult. I looked for 14650s, but couldn't get them in time, so ended up with "stick" NiMH batteries from Amazon, apparently intended for Airsoft guns ... Two 8 cell packs in series to get the necessary capacity. Charging is done through a connector at the end, using Mascot chargers. Works OK so far, but will look into those lithium chargers. Very little space for balancing wires etc inside though

[GardG]

OK, I've done some adjustments and done the rest of the board, i.e. dimming MOSFETs, LumenRadio and Arduino modules, and connectors – not too worried about those though.

Attaching some screenshots of the current status of the DC-DC portion of the board, as well as a PDF of the entire thing.

The most important to the DC-DC converter is that I've done as thm_w suggested and moved R8 and R9 to the other side of L1. I have also added some thermal vias and some copper fills for heatsinking underneath the MOSFET and the diode. A few passives have been changed for different values with slightly different package sizes.

Regarding the rest of this – both of the large modules have castellated pads and will be soldered directly to this PCB. I have kept the areas below them as clear of top layer traces as possible, to avoid relying on just the solder mask to prevent shorts between the different PCBs. There'a also an area in the corner of the LumenRadio module which is cleared of all copper, as per the module's datasheet. These large unusable areas, combined with the requirement for a very narrow board, leads to some pretty wonky routing, but I suppose it should be fine.

The Arduino's 3V3 regulator will also power the LumenRadio module; as far as I can tell it should manage that with a bit of headroom.

Any glaringly obvious fundamental mistakes here? Or easy improvements, for that matter?

[thm_w]

Looks good.

Does the cutout for RF antenna match what is recommended? Or do they have anything in the datasheet.