Notice the eval board DC1568A uses a 4-layer PCB for the charger.
Is this strictly necessary or can a 2-layer work fine?
Our power demo boards use a minimum of 4 layers, but that does not mean this is always necessary.
There is no single answer to how many layers your design will need. It depends on the power level, copper thickness, and the space available for routing. How many amps are you planning to charge with?
The main concern with these parts is the grounding, and with two layers it often becomes difficult to ensure that grounding is solid. This is due to the fact that you will inevitably need to via some of your traces to the bottom layer. Well, that's also where the GND paddle of the IC connects to the rest of the GND of the board, so it can be tough to avoid cutting off the IC GND from the board if you only have two layers.
The other concern here is thermal behavior. LT3652 has its switching FET integrated, so the IC itself can get hot. Heat is mostly dissipated through the GND paddle which vias down to the bottom PCB layer and then hopefully uses a wide GND plane to dissipate the heat. Again, if the GND on the PCB bottom is cut off by traces (which is somewhat unavoidable), you will have trouble here.
For those reasons, I would recommend at least 3 layers for this part, using the bottom layer as a solid GND and heat sink for the IC paddle. If you are charging with less power, say 500mA-1A, you may be able to get away with two layers. It really comes down to the layout.
Feel free to run your layout by us before sending it out to fab.
That's a relief. Was concerned it was maybe related to switching interference on Vsense. This will be a small corner of a bigger board would rather keep to 2-layers if poss..
Aiming for 19.5v in -> charging 12v SLA backup supply. 1A max probably sufficient.
Found a 2-layer hobby board offering full 2A charging up to 20 Vin for LiPo. Ground plane looks good & circuit probably not much different. SparkFun Sunny Buddy - MPPT Solar Charger - PRT-12885 - SparkFun Electronics
Right, with this part, the concern is mostly thermal dissipation.
Thanks for pointing out the SparkFun board, I actually wasn't aware of it. They spread out the GND connection on the top layer, which is good. It would also help to make a small grid of vias on the GND paddle and extend it onto the bottom layer.
I think with your 1A current and a solid layout you should be able to get this working without much heat.
Any other questions, just let me know!
Here's a quick first attempt (attached).
- spread ground on top + added thermal vias under IC;
- tried to respect switching circuit layout suggestions in datasheet;
- placed sense lines under GND plane;
Is it OK from thermal and noise perspectives?
This is a very good start. I think the grounding will be no problem. I notice that the pour is cut around the connection to pin 6; you may as well make that a solid connection. It could also help to add a few more vias below the IC since it looks like you have room, but don't stress over that.
Your switcher layout looks good. I have a couple of suggestions:
- Remove the copper directly below the inductor such that there is only copper where the inductor's pads are. Let your GND pour on layer 1 flow below the inductor. This will help to shield your sense lines on layer 2.
- I don't see how the sense resistor connects to the inductor? Also, I recommend that you move the sense resistor to the right to give a more direct path and also to avoid having routing directly below the inductor.
- Make sure your BAT and SENSE traces do not route directly below the inductor. I can't tell if they do or not.
I also recommend that you use a pour for your BOOST connections and just make it as thick as possible given the space you have available now. The BOOST to BAT path should all use a pour instead of a trace. It doesn't need to be huge, though. Take a look at how the demo board handled that.
That's all I see for now, I can take another look after those edits. All in all, though, I am not too concerned about thermals given what you were able to do here.
Most helpful, Zack. Thanks!
Have applied the suggestions as attached below.
- pin 6 fully grounded + extra thermal vias added;
- more ground under inductor;
- sense resistor moved out to the right;
- sense lines circumnavigate the inductor;
- boost trace now a pour;
The BOOST line is a bit of a mystery to me: in the datasheet it mentioned being about 8.5v and just for resetting a gate. So understood little current flowed...
Now found and looked at the Demo board see they've done something very clever by making the loop from SWITCH spiral clockwise (the inductor is placed at 180 degrees to my placement). Wonder if I should copy that? However not sure how well that approach respects the current loops (1 & 2) as drawn on page 21 of the LT3652 datasheet?
Looks great. My only remaining comment might be that you could thicken the trace between the inductor and the sense resistor since all of your current will pass through it. Maybe use a small pour there as well instead of a trace.
Yes, BOOST averages very little power, but there can be a reasonable amount of current instantaneously as the cap is charged/discharged to boost to the gate drive voltage.
As for the SW and inductor layout, I'd say it really comes down to your board layout. I think it is just fine how you have it.
Thanks a lot, Zack, it's been super-helpful to have some feedback to get the first iteration "right".
Don't know how I missed that last point about the trace to the sense resistor. ;-)
Now REALLY looking forward to charging a battery.
Anytime! Just let me know if you run into any trouble once you get the boards built up.
Quick circuit question. I took the lead-acid circuit on page 26 of the datasheet as a reference.
Is the diode 1N4148 strictly necessary (a protection?) in the circuit as it is drawn?
Or is it so that you can then put, say, a 5v pullup on the CHRG pin in order to drive a LED, or wire to a microcontroller for monitoring purposes?
The 1N4148 and the series 1M resistor are not necessary. The function of these components is to provide a different recharge threshold than the default 97.5% of the programmed float voltage. This works by changing the equivalent resistance of the Vfb divider based on whether the IC is in a charging state.
As for the diode, correct, it is not really necessary unless you are also pulling up the /CHRG pin to use it as an indication additionally to this function. The /CHRG pin is either high-Z or pulled to GND, so it will not interfere with the divider as shown in the circuit on page 26. I'd say you can go ahead and remove it if you are not using the /CHRG pin for any other function.
That's good to know. More functionality for v2. Thanks!
Incidentally, in this configuration for the Lead-Acid, have the impression that the modified resistor divider is primarily to create a two stage charge process in which the there's a Fast Charge at a higher level (14.4v) - then when fast charge is over (charge current <= C/10) CHRG goes high-Z and the Float Charge voltage as programmed by the divider becomes 13.5v. The 97.5% threshold in a sense remains the same because it applies to the unmodified divider (so 13.2v and never 14.4v). Glad I don't have to learn any more battery types!!
I opted for C/10 charge termination rather than using the timer just because it was in the datasheet as an option and seemed straightforward. Now apparent this doesn't work with lead-acid chemistry since at that point the battery is insufficiently charged to hold the float voltage. The battery voltage drifts down to 97.5% float after maybe 20secs and the charge cycle re-initiates only to cancel immediately because it's already <C/10.
So will now put a capacitor in!
In the datasheet it says that 0.68uF normally used => 3hrs charge cycle. However in the only Lead-Acid example in the document (p26) they've put a 4.7uF which corresponds to 20+hrs of charging. a) that SEEMS quite long given that after a couple of hours fast charging the battery seems mostly charged; b) that circuit is for a solar charger and don't know anywhere on this planet with 20hrs of useable sunlight!!! Should I slavishly copy 4.7uF from the circuit (my app is plugged, not solar), should I go with the "standard" 0.68uF (but is it really good for all chemistries?)? Or is an appropriate value to be determined experimentally?
Having written that I guess 0.68uF is going to work because after 3hrs it will have easily reached C/10 and be in constant voltage mode. If after that it doesn't hold above 97.5% then the cycle will start over, switching immediately to constant voltage which will procede for another 3hrs before terminating. Ad infinitum until voltage self-maintains. Maybe it will total, 20hrs, maybe rather less.
But the question still begs why they put 20hrs on a solar charger! ;-)
An appropriate value can be determined experimentally or by consulting with the battery manufacturer/datasheet. I would say that the 3hr charge cycle is a good rule-of-thumb, so maybe start with that.
Some places do get 20+ hrs of sunlight at times, believe it or not! I would agree that you don't need anything like a 20hr charge cycle in your app, though.
I remember setting off to climb a mountain in the summer in Lappland (north of Norway near Kiruna): we started by taking boots off and wading through a stream in the valley at 3am in the morning. Our strategy was to get up and be coming down by the time time sun was high in the sky at midday at which point the ice on which we were climbing would become prone to spontaneous cracking. Yes there was daylight more or less 24/7 but mostly not enough to drive a solar cell.
Of course that design with the 20hr charge cycle might have been for a space mission? But then again, whilst a reliable analogue part would be an excellent choice, Lead-Acid is probably not the best for a multi year venture in what's likely a very cold environment. Remain perplexed!
Perhaps all that matters here is that 20hrs is greater than the amount of sunlight time (usually) and less than a daylight cycle. You will charge when the sun is up, and if you are not receiving sunlight for part of the charge cycle, worst case, you will simply not charge. As long as it is less than 24 hrs, your charger will be ready to kick back in when the sun comes up again!
Retrieving data ...