Boost Converter (5V to 12V)

[Phaedrus]

I'm working on a boost converter to generate 12V from a 5V input, intended to allow you to power 12VDC fans up to ~2A from a USB 3.0 port (or a pair of USB ports, using a splitter). I'm using a TI TPS40210, with an AD4466 for switching. Simulation shows ~90-94% efficiency, which is good enough for my needs (I can get higher efficiency from a synchronous topology, but it's hard to get one the ICs in a hand-solderable package, and the cost is higher).

This is just a prototype PCB, single sided board with 805 SMT parts, so I can hand solder and tweak it easily. If it works well and I want to make more to use/give away/sell I'll probably make a double sided version with 603s.

Anyway, wondering if I could get a sanity check before ordering a PCB.

[Phaedrus]

Come on guys, I know PSU stuff is boring, but all I need is either "looks good to me" or "you're an idiot because x" 

[mariush]

I can't be bothered to check the tps40210 datasheet... so i'm just going to make some general comments.

USB 2.0 gives you 5v 0.5A per port.  Usually, ports are grouped into "hubs" so you have 3-4 ports that in total get about 1.5-2A.    So basically you have 5 x 2A = 10 watts.  Now how do you want to get 24 watts out of that... (12 x 2a = 24w )

I made a quick simulation with LT1308 in LTSpice and even with just 12v @ 0.75a, there's still peaks of about 2.2A pulled from 5A.. so I don't see how you plan to get even more than 1A from usb 2.0.

USB 3.0 will give you more, but will it give you 25w without any negociation? Research that.

 

[Phaedrus]

2A is an edge case, in typical use it will probably see 0.3A - 0.7A (4-9W from USB). For 1-1.5A you use two USB hubs to get up to 20W. 2A (27W) requires USB 3.0.

I assumed that USB 3.0's power management worked the same was as USB 2.0, but you raise a good point. I'll look into that.


In any case, 2A is a maximum rating for the converter, just excess margin. I really don't expect it to be used with fans over 1A or so.

[perfect_disturbance]

The max output of USB3 is .9A at 5V. 
As far as my reading goes your lucky that's per port vs USB1 .5A per host.
ok so 2 ports is 1.8A at 5V = 9W
Not too shabby really as Dave might say you can fly to the moon on 9 watts?

But once it's boosted to 12V your looking at 9W / 12V = .75A.
If efficiency is 90% your max output would .67A.
Keep in mind you have to plan for the start up current of the fans (some capacitance to handle the current spike?) if that goes over .67A you may blow a fuse in your usb host.  But you also may not and in fact most USB hosts can probably handle more but I wouldn't plan on it.

That being said if your real max current draw is less than .67A which is true for alot of 12V case fans you may be able to get this to work.

[Phaedrus]

Planning to use it with desktop systems, and most of those can output up to 1A per port easy. Some can even pull the whole 2.5-3.0A from the standby rail out of just a couple ports!

I wouldn't use it on laptops though. I tried a breadboard prototype on my Lenovo with a .37A fan. On the right-hand ports it runs full speed, but no other devices will function if connected. On the left-hand ports it pulses between full and half speed with a frequency around 1 second. This is probably due to limitations of the USB power circuitry in that laptop.

However on a desktop and higher-end Lenovo it ran at full speed just fine, indicating that these units did not have current limiting, or at least the limit is significantly higher.

I tried to test with a 1.2A fan, but it hit the current limit on my prototype (only built it for 1A).

[Psi]

The max output of USB3 is .9A at 5V. 

USB3 can provide 100W  (20V @ 5A) if the port supports the full USB power delivery specification.

I dont know of any that do though 

http://www.usb.org/developers/powerdelivery/PD_1.0_Introduction.pdf

[tom66]

Your boost converter layout could do with some improvement.

Basically, you want to keep the ground of the input capacitor, switch FET and output capacitor on the same large bus, such that the output current "loops" through from the input capacitor, through the inductor, through the diode, through the output capacitor and then return it into the input capacitor.

Then you want to take a single analog/point ground for your feedback network and tie the switcher's ground to that. All your feedback/sense circuitry should return to this position.

[Niklas]

Add an input capacitor and a lot more copper on the high current traces. Also think about how the current flows through the circuit and adjust the traces accordingly. I usually think like "First it goes here and then over to this component and on to that component..." and follow that order when I do the layout. Even if the components are in the same net, there can be a huge difference in system performance depending on where in that net the component is connected.

Basically you have two loops: Input and output.
The input loop: Input cap positive-Choke-Switch FET-Sense resistor-Input cap negative
The output loop: Input cap positive-Choke-Diode-Output Cap positive and load-Input cap negative

Try to place the component so they match with these two loops and also form as physically small loops as possible. Then you can add  the other components around them.

[Phaedrus]

Thanks for your advice!

How about the revised circuit? I added a 220uF input cap and redid the layout. Getting warmer?

[Phaedrus]

Any comments?