Low profile heat sink, 3.6W, closed space

[Axk]

Looking at dissipating 3.6W in a closed space (ebike controller/battery bag).

I will need to regulate from 54.6V down to 3.3V, with low noise to not affect current measurement (12bit, 5mA-20A), so I think I should use a liner regulator, thus the power dissipation requirement.

Would like the (power meter) board to be as thin as possible to not take much additional space in the bag, ideally 1cm.

Thinking about a metal plate the size of the board (probably 5cmx10cm) as a heatsink to attach a TO220 transistor. Are such heat sinks available?

Any alternative solutions?

Aluminum cases are thick and I have a Bluetooth module that won't work though an aluminum case I suppose.

[Axk]

Aluminum plates of various sizes appear to be cheaply available from ebay.

[jonroger]

I'd use a buck converter,   If you can't get noise low enough, put a linear regulator after it.

[Axk]

It's not horribly difficult I suppose, but still I would need to spend time learning how to use a buck converter chip and it adds potential mistakes to the design. Linear regulators are easy.

I've looked for DC-DC modules with suitable voltage and current and they are quite bulky and expensive.

Maybe I should reconsider and try find a buck converter chip locally, and quickly try it if I can make it work for my input voltage and current...

I'm also worried about radiated noise from the buck converter's inductor, not sure if it will be an issue. Could provision for a can on top of the regulator or the analog circuitry maybe to mitigate this.

I'll be dealing with 1 LSB currents of 200nA and voltages of 600uV.

[Miti]

https://www.ebay.com/itm/DC-DC-Buck-Step-Down-Converter-Regulator-Module-3-3V-5V-9V-12V-24V-36V-48V-60V/173254382932?hash=item2856c41554:g:fdQAAOSwCPdaxZc8&frcectupt=true

Edit: Never mind, it's linear. The title says buck step down. but this looks good:
https://www.ebay.com/itm/Adjustable-6-5-60V-to-1-25-30V-10A-Step-down-Charger-Module-Power-Supply/183130489462?hash=item2aa36d8276:g:124AAOSwiqdaq-au

[mariush]

You could use one or two of these : LR8K4 : https://uk.farnell.com/microchip/lr8k4-g/linear-volt-reg-0-02a-440v-to/dp/2448524

13.2..450v DC to 1.2v or higher, max 20mA +/-5% ...

The DPAK version would dissipate some heat through PCB and you could also use a DPAK heatsink or maybe solder some copper/aluminum on the tab, or maybe use some thermal pads to transfer heat to a regular extruded heatsink.

You could use two to spread the heat over two chips and bigger surface area... something like this:



Datasheet says typically 50C/w on a 25x25mm FR4 board so you'll probably need some heatsink even with two chips spreading the dissipated heat.

[coppercone2]

watch the new EEVBLOG video about it

[jbb]

I’d suggest a buck regulator + filtering (+ linear post regulator if necessary) supply for 3 reasons:
1- reduces temperature of linear regulator
2- reduces parasitic battery current - relevant for long term storage
3- reduces heat in adjacent cells or pack outside

These days there are a number of good buck converter chips with good light load efficiency. If you’re doing a 2 layer (or more) board it should be no trouble.

[Axk]

jbb,

What do you mean by filtering?
Some sort of a low-pass filter to reduce the switching noise?

[jbb]

Yes, exactly. Low pass filtering could be a set of capacitors for the buck output, then a series ferrite bead, then the output caps (ie an LCL filter). Big solid ground planes are a must here or the switching noise will find its way into everything.

However, I just remembered a challenge. A lot of switching regulators these days use a pulse-skipping mode at low power, which leads to a bit of a sawtooth waveform in the output voltage. Therefore if you need lowest noise and ripple a linear post regulator is a good idea.

[jbb]

So, I had a look at some previous designs.  There are some constraints:
- 56V input is quite high
- you need high efficiency at light load (note, many data sheets only talk about shutdown current rather than operating current)

Given those constraints, I would suggest the TI LMR16006 lineup. They aren’t the best and brightest, but they are small and cheerful.

To run Bluetooth, you probably need 100 - 200 mA peak current for the system. Therefore you want a LDO (linear Low DropOut) regulator with 250 - 500 mA rating and low quiescent current. I’ve used the MCP1700T-33 (fixed 3.3V output, 4uA quiescent current) and it seems to work.

The LDO needs some headroom to operate. The MCP1700T datasheet lists a maximum output voltage tolerance of +3% (ie 3.40V) and a dropout voltage of 350mV.  That means we need at least 3.75V from our buck regulator.

The buck regulator has tolerances too. I would allow for 3% reference accuracy, 2x 1% feedback resistors and 5% margin. This sums to 10%, so the target buck output voltage should be 3.75 / (100% - 90%) = 4.16V.  Say 4.2V for safety.

For 60mA @ 3.3V (198mW) output, we need 60mA @ 4.2V (250mW) from the buck regulator. Expected buck regulator efficiency is around 65% (using Webench from TI), which means buck input power is around 250mW/65% = 384mW, or 7mA. Overall losses are around 186mW, and overall efficiency is around 50%.

BT transmit peaks are short and shouldn’t cause much temperature rise.

While I’m pontificating:
- try to have your BT module and microcontroller sleep as much as possible to reduce average current
- I suggest charging cells to 4.1V to improve cycle life (will reduce energy strorage by 10 - 15%) and slightly reduce fire risk
- make sure that your BMS will be safe even if the BT module starts doing weird stuff (or locks up)
- pleas be careful when handling raw battery packs. Safety glasses suggested (I almost got myself in the eye with a spark last year). Also, the heat shrink over those cells is really thin and easy to scrape through.

[elmohandis]

seems like you need a low noise buck converter, like Linear's Silent Switcher.

Bit pricey, but looks decent:

https://www.analog.com/en/products/lt8641.html#product-overview

[jonroger]

> with low noise to not affect current measurement

My guess is that your current value can be heavily filtered - making noise from any fast switcher a non-issue.

[OM222O]

you will be burning through your battery pack using an LDO ...
use a DC-DC buck converter, followed by a chain of caps (tantalums or electrolytics: 100uF , 10uF, ceramics: 1uF, 100nF, 100pF). this will virtually reduce your noise to nothing. you can also follow it with an LDO if you wish. also an inductor can be used as a chocke if the current consumption of the 3.3V rail is steady. this is probably a huge overkill but should be much better than a massive heatsink and constant battery drain.honestly removing the caps and the choke should also not introduce any massive errors. just follow the buck converter with an LDO.