Non-conformal USB devices (voltage drop to ~3.5v)

[snoop33]

Hi all,

USB specification says 4.4-5.25v. When a buck attempts to regulate to this voltage but is connected to a dumb USB device (generally ones with crap li-ion LDO style chargers in them) I have noticed they don't use a UVLO to keep within the USB spec, disregard any current advertising, and drag the voltage down to about 3.5v or thereabouts. Because I'd implemented a UVLO on the buck/supply side this trips. The end result is conformal devices are fine but shitty USB devices tend not to charge.

I just wondered how others have found it and whether they ended up using a UVLO of a much lower value outside of USB specification. I feel I must do so.

[magic]

USB specification says 4.4-5.25v.

Which part of which spec says what about those voltages?

I don't think devices are supposed to watch Vbus, they are supposed to obey current limits and there is about five million ways of specifying/determining those, particularly when you add proprietary charges to the mix.

[tooki]

USB specification says 4.4-5.25v.

Which part of which spec says what about those voltages?

In the USB 2.0 spec (download zip here: https://www.usb.org/document-library/usb-20-specification and then the file usb_20.pdf within), it’s on pages 175 and 178, with a few preceding pages explaining necessary terminology.

[tooki]

Hi all,

USB specification says 4.4-5.25v.

Not really.

In a nutshell, only a device using 100mA low-power mode (keyboards, mice, etc) needs to operate on 4.4V, though all devices need to be able to enumerate (but not operate!) on 4.4V, even if they require >100mA.

So effectively, the actual minimum is the 4.75V required for >100mA operation.

Of course, no device is supposed to draw more than 100mA without negotiation.

[snoop33]

There's the specification and then there's reality. I've got two USB devices in front of me which are dragging the voltage down to about 3.7v. They aren't respecting the current negotiation. I suspect they're quite a simply li-ion charger so as the li-ion charges I think they'll go from 3.0-4.2 + 0.7v, so 3.7-4.9v VBUS (at device).

So enforcing the specification at VBUS (supply) seems out the question to make these stupid devices work.

[magic]

Looks like you don't have too many options.

Figure out which (if any) way of determining or negotiating charging current they use and implement it.
Increase your current limit.
Allow the bus to sag out of spec, which may break other devices.
Shut down the bus to prevent malfunction of other devices.

I'm not aware of any USB spec demanding that devices should UVLO, so your problem is excessive current.

[Berni]

There is no requirement for a USB device to have UVLO like that.

The voltage range in the USB spec just requires the device to work correctly within that voltage range. This means if the voltage sags to 4.4V your device should still be able to communicate normally with the host. Below that voltage the USB device is allowed to malfunction. The actual requirement on the device is to not draw more than 100mA until allowed to do so. At the same time the host is required to keep the USB supply voltage within normal limits while being loaded with any load between 0 to 100mA.

So if you have a device that doesn't draw more than 100mA and the voltage on the port sags below 4.4V then you have a non compliant host that can't provide the amount of current it should. (Keep in mind this is at the port, not after the voltage drop over a long USB cable)

This is more tricky for non powered USB hubs (since too much load could drop the voltage at the hub), but in a lot of every day use cases they are a bad idea anyway. Use powered hubs whenever possible.

[snoop33]

The supplier in this case may not keep up with the current request (so has violated the USB spec) but doing a UVLO then stops these simply devices from charging at all.

However on the other hand smarter devices appear to set a load which drags the voltage down, sometimes 4.4v, sometimes 4.75v. Some may sample the voltage and adjust the load to main the voltage above a point however others don't and then if current availability drops the voltage does so and they register the ~3.5v as a fault. In the case of phones this may lock up charging until restarted so having a UVLO is necessary.

So there's no easy answer as you say.

[tooki]

There is no requirement for a USB device to have UVLO like that.

The voltage range in the USB spec just requires the device to work correctly within that voltage range. This means if the voltage sags to 4.4V your device should still be able to communicate normally with the host. Below that voltage the USB device is allowed to malfunction. The actual requirement on the device is to not draw more than 100mA until allowed to do so. At the same time the host is required to keep the USB supply voltage within normal limits while being loaded with any load between 0 to 100mA.

So if you have a device that doesn't draw more than 100mA and the voltage on the port sags below 4.4V then you have a non compliant host that can't provide the amount of current it should. (Keep in mind this is at the port, not after the voltage drop over a long USB cable)

This is more tricky for non powered USB hubs (since too much load could drop the voltage at the hub), but in a lot of every day use cases they are a bad idea anyway. Use powered hubs whenever possible.

Only ports on unpowered USB hubs are allowed to drop to 4.4V, if I understood the standard correctly. Powered ports are expected to supply the 4.75V minimum.