USB-PD™ Power Reserve and You
Part 1: Why “derating” matters and how it affects charge-through
One of the most useful aspects about USB-C™ — that many companies get wrong — is a feature called “charging pass-through”.
You might have seen this on hubs you use everyday: a second USB-C port through which you can charge your host, while simultaneously using other ports for device data, charging, and/or video.
What many people don’t know is how complex this actually is and how tricky to implement properly. Improper Power Reserve logic is the #1 complaint I have about most “bad” charge-through dongles.
This first article in the series will summarize why derating matters. In following articles, I’ll describe methods to properly implement it.
Correction notice:
This article has been updated on (11/23/19) based on feedback from relevant product experts. Thank you. Numbers now reflect typical production values, and various methods have been clarified. Of note, Variable PDO derating is emphasized, and Vconn guidance provided.
Background (… and some Math)
Related listening: Foo Fighters — The Pretender
Believe it or not, the entire concept of the the “3-in-1 charge-through dongle” was spearheaded by one company. Since then others have attempted copying their implementation… to varying degrees of success.
The root of the problem arises because of the Three Laws of Thermodynamics, a simple Thévenin Equivalent Resistance formula, and the problems arising from companies prioritizing “Compatibility” over “Compliance”.
1.) Energy cannot be created or destroyed in an isolated system.
2.) The entropy of any isolated system always increases.
3.) The entropy of a system approaches a constant value as the temperature approaches absolute zero.
— The Three Laws of Thermodynamics
Any combination of voltage sources, current sources and resistors with two terminals is electrically equivalent to a single voltage source V and a single series resistor R.
— Helmholtz–Thévenin theorem
In other words: Dongles cause cable resistance losses, consume some energy themselves, and you can’t get more energy out than you put in. Simple, right?
In the following sections I’ll try to break down why that’s not so simple.
(1) Managing the USB-PD Power Reserve
Related listening: Dire Straits — Money For Nothing
The easiest and first issue to watch out for are the Laws of Thermodynamics. Simply make sure the following statement is always true:
Energy in — (Energy used + Energy out ) ≥ 0
The Power Delivery spec has a section that cautions about this: USB-PD 3.0 v2, “Section 6.4.1.2.1 Management of the Power Reserve”. It is intended to describe USB-PD sources with multiple ports but is equally applicable to hubs.
Dongles must pay attention to an additional term: “Energy Used”.
… and here lies the root of the problem. Many, if not most, dongles don’t “do the math”. They do not properly subtract — or in other words “derate” — for their own internal power use. Nor account for the specific requirements of their hardware design and components.
Almost all dongles I’ve seen simply copy and paste from a reference design with no power budgeting. Sometimes they do not derate at all. There are a few manufacturers that are on my Banned Vendor List (BVL) as a result.
On the flip side, some manufacturers have worked with chipset providers to *freely* provide well-engineered reference designs.
“Derating” means budgeting for maximum possible device loads, cable and connector losses, hub chip power draw, etc.
Hypothetical Dongle A
Related listening: Queen — Another Watt Bites the Dust
(*numbers updated with production values)
Let’s use the hypothetical dongle example below with some ballpark* numbers and see how this plays out:
… what the heck just happened? Where did my 30 watts go?
Resistive losses and maximum device loads ate up your power. Even the device may need to chip in 1.5w “Vconn” backwards into the dongle.*
This is where the “Reserve” in “Power Reserve” becomes quite important.
Device loads:
The dongle has no way to know how much power the Type-A device connected to it will request. Apple adapters use a proprietary Type-A 5.5w signal*. Other USB-IF compliant hubs may use CDP 7.5w signal.
For example, the adapter pictured above must Reserve 5.5w of power from the source adapter at all times to power the Type-A port.
The dongle may need to take a number of actions if the input is less than 8.15w — one example being a 2.5w (5v*500mA) A-to-C cable. It may need to disable the 5.5w Type-A port, use power from the host, or shut off the Type-A port entirely.*
*(The actual Apple hub derates differently than this. This is merely an example.)
Cable and connector losses:
Cables don’t have 0 resistance. The USB-C specification accounts for this by permitting cables up to a 750mV loss from VBUS to GND. The only thing that “doesn’t count” for derating is the original cable resistance.*
*(This is illustrated in diagram above. “Allowed cable IR drop” is blue .)
Losses are normally accounted for by the device getting ‘a little less voltage’ than expected. Device-side voltage at ‘5v’ may droop as low as 4.00v when loaded to 3a. In this naïve example the dongle has to account for it.*
*(Note I use a 300mV drop for the captive cable. Captive designs can design-in a specific cable loss, rather than assume 750mV worst-case drop from a “bring-your-own-cable” situation that occurs by having a receptacle.)
This is likely why most dongles you see use a short captive plug. Receptacle designs are entirely possible… but would require more math, derating, and result in slower apparent charging. Tricky, no?
Internal hub draw:
The final concern is how much the dongle itself consumes. If the dongle uses expensive buck-boost regulator ICs, this could be in the milliwatts. If it uses cheap LDO linear regulator ICs, this could be in the watts. And manufacturers love cutting “Cost Down” corners.
I’ve seen poorly architected dongles try advertising 5v/2.4a through a linear regulator, with 20v negotiated pass-through. This means (20–5v*2.4a=) 36w of heat dissapated through a chip! Needless to say, it didn’t last too long before letting out the magic smoke.
Note to manufacturers: please pay note to Vconn. *
Powering the video circuitry using Vconn (in example, 1.5w) is only possible with captive-cable dongles. Receptacle dongles don’t get Vconn, since cables do not pass it though.
Also Vconn availability ranges from 1w (standard) to 6w (very rare). Vconn voltage is allowed to be (3.0-5.5v) at that wattage.
This also means you have to stay within Vconn inrush current, max current, and voltage slew rates! Chromebooks and Macs are strict and will turn off Vconn using specialty chips (PPC) if you violate it.
(2) Managing USB-C Voltage Losses
Related listening: AC/DC — Highway to Hell
The second and more challenging issue is limitations on USB-C voltages. You must meet the following for “Fixed Voltage Power Delivery Objects” (Fixed PDO). Specifically, the mandatory vSafe5v one all supplies must have.
A USB PD-based charger that has negotiated a voltage V at ≤ 3A shall output
a voltage in the range of Vmax (V + 5%) and Vmin (V -5%) when no current
is being drawn and Vmax and Vmin-0.75 V at 3 A.Under all loads, the output voltage shall remain within the cross-hatched area shown in Figure 4–39.
— USB Type-C Specification 2.0, “Section 4.8 Power Sourcing and Charging”
In essence PD sinks are guaranteed a certain level of voltage performance. Anything too far outside this envelope may be interpreted as a disconnect. A sink might not even be able to function if the voltage is too low!
Note this describes the voltage allowed at the sink end of a cable. Clever readers might remember we said cables can have a maximum drop of 750mV. And the formula for resistive losses is E = I * R — making a triangle.
Cable losses are simply “subtracted” from the ±5% PD voltage limits.
What does this have to do with charge-through hubs?
Fact is almost no dongles do internal voltage re-conversion. They merely “pass through” voltage with some FETs. Refer to the simplified Thévenin Equivalent Circuit:
R3 is the the abstracted load of the dongle, internal draw, peripherals, etc. R1 is the cable. R2 is the captive cable portion. As a gross simplification, all of this can be handwaved away as “Req”*.
Pay note this looks strangely like the cable plugged into a charger on the left hand side of the image…
- (In reality this is more complex, and the intercepts vary depending on load.)
We’ve essentially made an out-of-spec cable with the hub, throwing the voltage limits way out of whack!
- Orange is the area where the host device “theoretically should work” since it’s inside the (5%-0.75v) functional limit. The host may still complain.
- Red is the area where the host device “absolutely will not work and might even disconnect” since you are below the minimum limit.
- Note the X-axis is the sum of all loads. It is (the hub’s draw + the host’s draw). If the hub draws >1.2a on its own… you might start in the red!
If you do not “derate” to where the orange line crosses into red (~1.2A in this example), or use an alternate method of re-advertising voltage such as “Variable Voltage Power Deliver Object”, you’re going to have a bad time.*
*(Apple uses Variable PDO method. CT-VPD use Current Limit method, but cannot draw from this voltage themselves as they are “Vconn Powered”.)
Note to manufacturers: please consider the method used.*
Variable PDO based derating theoretically works with any sink supporting “derated” input voltages from “Power Rules” 5–9–15–20v PDO levels.
However the specification-suggested method is requesting the full wattage from the charger, and re-regulating (bucking down) to “Power Rules” compliant levels. This “Re-Regulation” method is discussed briefly in the next article.
(3) Why would ODMs and their Clients do this?
Related listening: AC/DC — Dirty Deeds Done Dirt Cheap
- It’s cheap and easy to copy code. Conversely, custom firmware from an IC vendor is expensive. Many ODMs just don’t have the skill to do it in-house.
- Users will see a “fast charge” indicator and be happy. Most don’t know PD safety relies on explicit contracts and guarantees, not “it seems to work”.
- By passing through charger PD messages as-is, while siphoning off the top, they hope for “compatibility”… mainly with other noncompliant products.
They couldn’t be more wrong about all three points.
In fact, the third is very off-the-mark since Tier 1 companies like Apple and Google take USB-PD limits very seriously in their firmware. At best this will cause random disconnects and errors, and at worst can be outright unsafe.
Perhaps Manufacturers can consider this a heads up of why they need firmware update capability in their products. Or demand ODMs include it as a selling point of the package.
…that is worthy of an article in and of itself.
Linux Vendor Firmware Service
Hopefully this article gave you an insight into the theorycraft behind charge-through hubs from an engineer’s point of view. The next article will go into various possible approaches on how to address these challenges.
A s always: have fun, hack the planet, and don’t trust anyone with a postcount over 30. -Nathan K.