Hey guys,
I have a circuit that I want to be powered from USB. As you may know, USB power source can be anywhere from 4.5V to 5.5V (approximately 5V).
My circuit has some analog components and needs an accurate 5.0V source.
So I decided to try and use a TPS63000 Buck-Boost Converter to create a 5.0V stable power source.

Now for my question: I am confused because the schematic shows 2 separate grounds. Can I just connect all of these grounds together on my schematic/PCB? Does this circuit seem like it will do what I am trying to accomplish? Thanks and any help is greatly appreciated!

On the page 10 of this datasheet you can see the multiple ground planes here (but the first page shows them all as the same ground plane?):
http://www.mouser.com/ds/2/405/tps63000-451768.pdf

On page 8 you can see it here:
http://www.ti.com/lit/ug/slvu156/slvu156.pdf

And here is my circuit:

 

Look at page 16 of the datasheet and look carefully how the GND pin connects to PGND at a single point right under the chip. The point here is to keep the switching noise away from the feedback divider and control circuit. The RC filter going to VINA, EN and PS/SYNC should be done in the same way, meaning the bottom end of the cap goes to the GND pin and not to the PGND plane.

 

The grounds do all connect together but it is critical that this is done properly.
Have you read section 10.1 "layout guidelines" and 10.2 "layout example"?

 

If you look carefully the schematic shows you what you want to do: connect the grounds together at one point ONLY.
In layout terms you want the grounds to be different EVERYWHERE on the PCB, except at a single point.

EDIT: kubeek & blocco beat me to it.

You should also realize that drawing more than 100 mA from the USB connector is akin to dirty pool. You cannot expect more without negotiating with the USB hub you are connected to. If you violate this restriction you may damage your USB source. I mention this because boost converters will often draw more input current than the output current they provide.

In short -- BE CAREFUL, BE VERY VERY CAREFUL

 

Thanks a lot for the info! I don't know how I missed the layout section of the datasheet.
I'm going to give this a shot in eagle, I have not wired up multiple planes like that before so we will see!

 

Ok so here was my attempt at implementing this in eagle. I would greatly appreciate any feedback on this, and if I did this correctly. I was also not sure if it was ok for me to put those vias in the middle of the thermal pad like that. I was also not sure if it was ok for me to run that net from C1 to R1 like I did (and similarly R2 to C14)... the layout guideline shows running it under the ground plane, but I thought this would be better?
Thanks again!

 

Please put it into a PDF, the schematic is not readable enough.

 

D2 is there to prevent reverse connection damage, but it significantly reduces the converter efficiency, especially with only a 5V input.
Unless you think there's a chance of reverse polarity connection, I would replace it with a short.

Alternately for minimum drop with reverse protection, you can use a small logic-level (ON-resistance specified at Vgs of 5V or less) P-MOSFET with one resistor as shown here.

 

Ok looking back at the previous schematic, the design looks quite good. You might just want to add a few vias below C1 and C13/14.

 

So after a couple modifications to the layout I was able to get this to work, and it works really good.

The problem is that these DFN packages are too difficult to solder reliably.

Can someone recommend a chip or circuit that could maybe accomplish the same thing... only not these DFN packages? I really wish these were available in like a SOIC.

 

The USB 5V specification as far as I know is 4.75V to 5.25V but in reality it's usually quite close to 5V with little or no excuse not to be!

 

Well if you pass it through and unpowered hub you will get close to 4.4 due to a diode drop.

 

Yeah thats the problem... for analog circuits that are present on the board, they need a precise 5.0 voltage reference. That is the goal.

So USB standards are that its 5 volts with 5% accuracy. 5 volts with less than 1% accuracy would be acceptable... that is the goal of this circuit.

 

You might try looking at some linear technology chips, they are also big in switching power supplies. Or try doing a parametric search in both TI and LT websites, both work quite well IIRC.

 

I gave that a shot (on both LT and TI websites), these are the best ones that I came up with that are not in a DFN/BGA/something similar package.

Linear Technology - LTC3534:
Datasheet: http://cds.linear.com/docs/en/datasheet/3534fb.pdf
On page 16 it even has a typical application being "USB to 5V at 500mA", so this might do the trick. I was unable to determine the kind of accuracy I might get with this.

Texas Instruments - TPIC74101:
Datasheet: http://www.ti.com/lit/ds/symlink/tpic74101-q1.pdf
It mentions the accuracy is 5V +- 2%.... so that could be a bit better... thats 4.9V to 5.1V thats really not that great if I'm trying to increase accuracy right?

What do you guys think?

 

Hubs weren't mentioned?

If you are going to power something from a USB port you will get anything from 4.75V to 5.25V but on the whole quite close to 5V based on my experience with ATX power supplies.

On the other hand I can't vouch for other types of USB sources. If I was designing a circuit powered from a 5V USB supply it doesn't make sense to me to go for 5V as a reference voltage based on the tolerances already discussed. Why not derive say 2.5V reference (easily done) and work from this?

 

Yeah thats the problem... for analog circuits that are present on the board, they need a precise 5.0 voltage reference. That is the goal.

That's why I use a lower reference of 4.096 volts.

I would never think of using the power rail as a reference, excepting those special times the reading is a fraction of the reference so the absolute value drops out of the equation.

 

Hubs were not mentioned, I thought that the one diode drop I am used to was part of the specification, but now I see it is not. Nevertheless I agree with paulktreg that using a 3.3V or 2.5V reference will solve many issues that a typical 5V powered analog design would have.
Edit: 4.096V ErnieM suggested might be limiting the choice of opamps and droupout voltage of the reference, but if you can get away with that then why not.

 

Hubs weren't mentioned?

If you are going to power something from a USB port you will get anything from 4.75V to 5.25V but on the whole quite close to 5V based on my experience with ATX power supplies.

On the other hand I can't vouch for other types of USB sources. If I was designing a circuit powered from a 5V USB supply it doesn't make sense to me to go for 5V as a reference voltage based on the tolerances already discussed. Why not derive say 2.5V reference (easily done) and work from this?

I understand that USB is 4.75 to 5.25 and that it does not make sense to use that as a reference... this is the purpose of my topic... to take that inaccurate voltage and make it accurate so you can use it as a power source as well as a reference.

What are hubs?

That's why I use a lower reference of 4.096 volts.

I would never think of using the power rail as a reference, excepting those special times the reading is a fraction of the reference so the absolute value drops out of the equation.

What is wrong with using a power rail as a reference?

 

I suppose it all boils down to what you want this reference voltage for and the accuracy of any results produced by it?

Please explain more.