Fuse for USB-powered product?

[Alex]

So I have designed a product that can also be powered from a USB port. I am using a boost converter to boost the USB port voltage to about 7V.

I am considering whether I should add a fuse (any type) in line with the USB port in the event of a short circuit (boost converters offer a low impedance current path from input to output).

What is your opinion on adding a fuse? Does the USB standard specify such a requirement for the plugged device or the host?

Thanks,
Alex

[johnmx]

Use a polyswitch instead. Some are developed for the USB bus specification. The good USB hubs have one on each output.

[Alex]

Thanks.

So if the host has one on the output do you see a reason why I should use one on my input?

[Sionyn]

most usb chipsets have over current detection

[johnmx]

Well, I know that my hub has, your may have too (take it apart to make sure). But tomorrow you can connect your device to another computer or hub and that one may not have.

Don’t forget that polyswitchs for this current rate may have an internal resistance from 0.25 Ohm to 1 Ohm. Take this into account when designing the boost converter.

Pay attention to the peak current value of your converter. Also there is a limit of input capacitance for any USB device, because of the inrush current. So putting a big capacitor next to polyswitch it’s not a good solution.

Try to use a converter designed specifically for USB powered applications...if you can find any

[johnmx]

most usb chipsets have over current detection

That’s true, but it is also true that most of them don’t use that option. Normally that specific pin is not connected. This is my experience with external USB hubs.

[Alex]

Based on the power requirement of the device the impedance of the USB port, including any patchcords must be less than 3.5 Ohms. This is a rather low impedance if you consider the 'light-duty' usb connectors and cables, this is why I am sceptical about adding this specific part. My setup has approx 2.74 Ohms but a friends has 4.16 Ohms...

Having a look at Littelfuse, a typical resistance in the untripped state for their 0.65A polyfuses is 0.7 Ohm and 1.17 Ohm for the 0.5A version. http://www.littelfuse.com/products/Resettable+PTCs/Radial+Leaded/60R/60R050.html On the other hand a standard 'wire' fuse (Littelfuse 208 series glass fuse) has a cold resistance of 0.265 Ohms for the 0.5A version. That is much better.

I could not find any that met my specs johmx, this is a custom design. I needed the capacitance to meet the ripple current requirement of the converter. Unfortunately I don't have a lot of voltage headroom (due to the impedance of cabling) so it is desirable to avoid having one. Unless of course there is a reason to have a fuse there. By the way, this is going on the motherboard connectors directly, not on an external hub.

How would the overcurrent protection be implemented on an external hub? Series resistor sending a signal to the controller?

[NiHaoMike]

Put the fuse on the output instead and use soft start to reduce inrush.

[Alex]

Good idea, although the actual converter will be excluded in that protection scheme.

I was leaning towards the optional inclusion of a wire fuse. This document highlights some of the differences between the PTCs and wire fuses: http://www.littelfuse.com/design/literature/fuse-vs-ptc/selecting-ptc-or-fuse.html

The resettable feature of PTCs set aside, PTCs are rather poor performers.

I decided to have a look at the USB 2.0 spec that can be found here:
http://www.usb.org/developers/docs/usb_20_021411.zip
And look for the file usb_20.pdf.

Three important points from chapter 7.2: 

1) The device is to draw up to 100mA from the host/hub unless it requests 500mA from the host/hub. Yes, you might be able to get 500mA out of a port by loading it passively but you probably do so with very high source impedance (high voltage drop) as well as not being compliant.

2) The host and all self-powered hubs must implement over-current protection for safety reasons, and the hub must
have a way to detect the over-current condition and report it to the USB software. This in effect allows me to not use a fuse and put the blame on the manufacturer of the usb host! It also explains that protection pin johnmx mentioned.

3) If an over-current condition occurs on any port, subsequent operation of the USB is not guaranteed, and once the condition is removed, it may be necessary to reinitialize the bus as would be done upon power-up. This means I can't use a wire fuse, but PTCs.

This info is from the official USB 2.0 spec document, I hope it proves useful.

Thanks for your help.

Alex

[Leo Bodnar]

So I have designed a product that can also be powered from a USB port. I am using a boost converter to boost the USB port voltage to about 7V.

I am considering whether I should add a fuse (any type) in line with the USB port in the event of a short circuit (boost converters offer a low impedance current path from input to output).

What is your opinion on adding a fuse? Does the USB standard specify such a requirement for the plugged device or the host?

Thanks,
Alex

If you are trying to comply with USB requirements - unless your product enumerates on USB bus and states the current demand it would be violating USB standard even if it draws below 100mA.
You can draw 500mA only after the host enumerates you (device reaches configuration state.)

All being well devices are not required to have current limitation and it is the job of the host/hub to cut them off when they draw more current that they are supposed to.

However if you are trying to ride on the margin and after testing on a few hosts find that you can draw more than 500mA be prepared to deal with a lot of customers coming back with complaints.

By the way most of USB cables sold on their own are shit.  They are made in China from who knows what and often have 2-5 Ohms resistance on power/ground pair.  Try yourself and prepare to be shocked.
The USB specification demands the maximum voltage drop for detachable cables on VBUS to be 125 mV which basically means 0.25 Ohms resistance.
If you are selling high power device, find a source of good cables and include one with your product.

[Leo Bodnar]

1) The device is to draw up to 100mA from the host/hub unless it requests 500mA from the host/hub. Yes, you might be able to get 500mA out of a port by loading it passively but you probably do so with very high source impedance (high voltage drop) as well as not being compliant.

That's right.  Another gross misconception is that you may draw 100mA "passively" without enumeration at all.  This is wrong.  You have 0.1 second from connection to identifying yourself as a USb device:

Table 7-14. Device Event Timings

Time from internal power good to device pulling D+/D- beyond VIHZ (min) (signaling attach), TSIGATT  <= 100 ms

Then if you have cheated with just slapping a resistor on D+ and obviously failed to enumerate the host will give up and put a bus into idle state.
From that moment you have 0.01 seconds to stop drawing 100mA and fall back to 0.5mA

Maximum time a device can draw power >suspend power when bus is continuously in idle state, T2SUSP <= 10 ms

Pay attention to the peak current value of your converter. Also there is a limit of input capacitance for any USB device, because of the inrush current.

Yes, the limit is 10uF.  Any more than that has to be behind an inductance/resistance or soft-start / inrush current limiter - there are quite a few ICs around exactly for this purpose.

[Alex]

Thanks for clarifying these points Leo.

I am consuming about 1.5 W and my cutoff is 3.5V. Input capacitance is 200uF low ESR. If the voltage is held at reasonable levels say >4V with a good quality cable then I will have plenty of headroom in terms of current limit. It almost sounds silly to have such a low impedance from the host and then ruin it with a 5 Ohm cable, especially when chances are your cable will be of low quality. Thanks for the suggestion of including one.

I am not terribly worried about being fully compliant especially with so many non compliant device around (considering the current draw limitations we just spoke of). However, say I do want to be fully compliant. Unfortunately I cannot afford the footprint of a USB stack in the main MCU. Is there another way of drawing the full current without the standard software enumeration? Is there an off-the-shelf controller that will enumerate as a high power device, or would I have to include a small MCU that does just that plus inrush current limit possibly with an external transistor?


Alex

[johnmx]

Maxim has one battery charger that enumerate autonomously. This is the only IC that I know of.

http://www.maxim-ic.com/company/press_room/product.cfm/id/1768

[Leo Bodnar]

I am not terribly worried about being fully compliant especially with so many non compliant device around (considering the current draw limitations we just spoke of). However, say I do want to be fully compliant. Unfortunately I cannot afford the footprint of a USB stack in the main MCU. Is there another way of drawing the full current without the standard software enumeration? Is there an off-the-shelf controller that will enumerate as a high power device, or would I have to include a small MCU that does just that plus inrush current limit possibly with an external transistor?

I agree that compliance has been blown out of the water by USB cup heaters and mini-hoovers.  
USB forum should have put their foot down but missed the moment.

If you want to go the proper way you'd need some sort of indication from the USB controller that OS has allowed you to draw 500mA.  E.g. if you connect to unpowered USB hub and request 500mA you will be shot down by the OS and have to give up i.e. not to proceed with 1.5W budget.  Just shut down and go into low power mode.  The user should get notified by OS that they have stuck high power device into low power port.

You can take one of Microchip's PIC18F13K50 or PIC18F14K50s, take any of their free USB device examples (e.g. UART or any HID device), modify USB descriptor to show your actual current requirements (e.g. 300mA) and then stick a line there:
if (state == CONFIGURED)
  portX.pinY = high;
else
  portX.pinY = low;

E.g. if you take a mouse-in-a-circle example and just replace desktop page for custom page you will be done.

PIC18F13K50 is less then $1 and is available in SSOP package if you are space-tight.  The only other components are 12 or 6MHz crystal and a few decoupling caps.  I have successfully used ceramic resonators as well.  If you already use 12 or 6MHz crystal for your other uC then you can just piggy-back on it and save a crystal.  6MHz crystal requires low-speed USB mode.

[johnmx]

How would the overcurrent protection be implemented on an external hub? Series resistor sending a signal to the controller?

Usually the overcurrent condition is detected by an external IC.
e.g. Power distribution switches:
http://focus.ti.com/lit/ds/slvs172a/slvs172a.pdf

[seattle]

For all the rules around USB, it has pretty much become a free-for-all standard. The beginning of the end was phones that used USB to charge. Many will consume 500 mA without even negotiating, and most every hub or laptop is happy to oblige. There are a ton of accessories too that use around 500 mA and don't even have a micro. And then there are the devices that have 2-3 USB plugs so that they can pull 1-1.5A.

In short, negotiating anymore seems to be very passe.

The biggest problem with USB is inrush. Stick a few hundred uF of capacitance across the power supply, plug it in, and you'll drop every single device on the hub below 4.5V for a few mS and they'll all reset.

If you are using a boost to get 7V, then if the boost fails you'll be at 5V, which may or may not result in excess curretn being pulled. I'd guess that most of the time it'll result in less current being pulled. As you draw more current, the USB bus voltage will sag and devices will start to drop off. Most powered hubs that I have seen use a 5V 1.5 or 2A supply and just pass that voltage right on through to ever port. So a single device can pull an amp no problem.

As noted, polyfuse can help here.