RE: Simple analog signaling for an AC adapter

Thread Links	Date Links
Thread Prev	Thread Next	Thread Index	Date Prev	Date Next	Date Index

On Dec 13, 2010, at 3:30 PM, Atkinson, Lee wrote:

Edgar, in fact the only guy I know using a digital communication method for ID is Dell. To Gary V's question, have we rationalized the messages that we want to send to the sink, with expected latency ? Have we also confirmed what kinds of devices would use active messaging?

My thinking on a few topics;

a. There is no A/D or microprocessor needed for a simple proportional signal. Though A/D and processors are common in notebooks, simpler devices could use as little as a transistor and resistor (if they need to discern the power capability of the AC adapter at all).

Cost of silicon will go down in the long term (as long as UPAMD-compliant interfaces get integrated with the power supply control circuitry). A transistor-resistor solution would have extremely low information density.

** In most systems the info density need only be low!

b. The only negotiation for power is simply to declare a connected load; we are on the path of a single terminal voltage and not a sink programmable voltage (as was contemplated at first and would have required the more complex signaling).

Don't forget expandability of the standard. Although we have mostly converged towards a single voltage, multiple voltages have always remained as part of the long-term vision, especially for higher (250W+) power devices.

**I would propose a simple analog minimum implementation as the lowest level. I do not think this precludes a more sophisticated overlay.

Spark Free does not need the ID pin, certainly our Cupertino friends vary to terminal voltage as a function of VCC load (the spark in this case is particularly interesting, since they are the only notebook company I know that positions its output voltage when connected).

c. If the communications protocol needs to announce the power capability of the AC adapter, it can do so in a number of ways (by either static or dynamic signaling, via either the ID pin or the VCC mains). We don't really even need an ID pin to confirm basic power capabilities.

One area that has not been confirmed is the need to build UPAMD around infinite gradients of power capability; I agree that a static analog ID is going to have problems discriminating a 53watt adapter from a 54watt adapter. But if we believe that a finite set of power capabilities (eg, 20w, 40w, 65w, 90w, 130w) then an analog method will be sufficient (we only need the equivalent of a 3 or 4 bits of data).

HP Smart actually becomes digital (low impedance to VCC) at the power limit, so an arbitrary power capability can be messaged. If we want messaging of a change in power capabilities we could achieve this dynamically thru the same analog signal.

d. There is nothing to prevent digital communication on the same path as an analog descriptor. A pullup is a pretty benign thing on a digital path.

That is only if the digital path can tolerate impedance mismatches. If we are talking very low speed unterminated communications, then yes, different pull-ups can be used without too much problem (within reason). But for higher speed communications (which could be as low as 100kb/s on reasonable cable lengths), impedance mismatches of more than 40% or so could cause a large amount of problems.

To be able to support higher speeds (1Mb/s+) a 20% mismatch (10% SWR directly reducing signal to noise and eye aperture) could be a bit of a stretch, and that would considerably limit how much of a range you would have for analog signaling.

** Most adapters I’ve seen have as short a cable as possible due to IR drops, so for 9X% of the market we are talking ~ 6ft as a communications distance. Communications in most cases would be negotiated on first connection and start-up and the information density is quite low. Here practically ~500ms to 1000ms is consumer acceptable. (Most adapters start in 2 to 4 seconds so adding 20% to this is no big deal). Unless of course you are talking about altering Vout or power limits during operation dynamically at fairly high bandwidth .... but even here realistically you are talking a few kHz in practical bandwidths. The case for high speed signalling I just do not see.... or am I missing something here?

Edgar

I laughed out loud when I saw Garry's mail (that 'a resistor and a wire are a big deal'). I think Garry has been spying on emails with my marketing, who have perpetually questioned any value towards even our 10 cent solution in HP Smart. To a degree, I cannot argue with my marketing's argument against a "scale of dis-economy;" meaning, volume never erases the cost of real material. I am not HP's AC adapter purchasing or development engineer, but I can tell you exactly how many cents we pay for the connector and wire. I can tell you as well the increased failure rate in the field by just 1 added wire on the connector. Pretty much the world except for Dell and HP don't deploy any kind of ID. Volume cannot erase real costs for copper, epoxy and silicon.

The truth is that unless I can prove to my own marketing that we've enabled a really compelling feature for the AC adapter, I will be unable to get their support for the UPAMD effort. What can we do to make a convincing story that we've been sensitive to the value requirements of mainstream consumer products ? Lee

-----Original Message-----
From: upamd@xxxxxxxx [mailto:upamd@xxxxxxxx] On Behalf Of Edgar Brown
Sent: Monday, December 13, 2010 12:03 AM
To: upamd-comms@xxxxxxxx
Cc: garrytomlins@xxxxxx; Piotr Karocki; upamd@xxxxxxxx; Leonard Tsai
Subject: Re: Simple analog signaling for an AC adapter

As an analog and mixed-mode designer I learned quite some time ago that it is preferable not to use an analog system when a digital one can do a better job. Tolerances, repertoire of possible responses, long time constants, and safety are all rather easy to handle and expand through digital means, but can become very complicated and costly through analog ones. My fears in allowing for a completely analog 'communications' system to coexist with a digital one would be:

(1) we compromise overall safety

(2) we make it harder to satisfy the stated goals (e.g., life expectancy of standard, adapter/device compatibility over life of standard, and certification support)

(3) we could be hindering standard adoption rather than promoting it.

(4) we might have to create a full digital communications standard from scratch (thus increasing overall cost)

(5) we would be increasing the cost of adapters and devices that wish to implement the digital parts of the standard.

(6) we could be limiting what can be accomplished by the digital communication system (e.g., speed, compatibility)

We have to keep in mind the point that Leonard and Piotr have made, once devices and adapters are being produced in quantity, the differential cost between an analog method and a digital one will mostly disappear. At that point, all adapters and devices will have to keep around analog communications subsystems just to remain compatible, with no real added benefit.

I would expect the cost curve to follow the same adoption path as many other technologies (1) first the high-cost high-margin devices will adopt the standard (for which functionality would be key), (2) added-value multi-port adapters will follow suit, (3) both of these will create a market for ICs and off-the shelf subsystems for UPAMD thus reducing implementation costs, (4) lower cost devices and adapters will start adopting the standard. I would expect such adoption curve to be followed with or without analog communications, and I actually see that requiring analog communications as part of the standard could hinder rather than accelerate adoption.

Although I can see how a purely analog system can work, I cannot see how it can easily satisfy goal 5 (higher power applied only when it is safe to do so); goal 4, as the aging of analog components will make it harder to guarantee safety or compliance for the life of the adapter; or goal 6, as the certification of analog systems across age and component variations would become harder to accomplish. (An example that comes to mind is the UCD9248 IC from TI, as it states in its errata that it cannot reliably identify a considerable section of its address range because it relies on resistors and analog values to set it.)

Gary and/or Arjan, can you come up with a more fleshed-out proposal on alternatives for how this analog communications would work? A reference design would be ideal, but at the very least it should attempt to address safety and digital communications compatibility.

For the sake of argument, if we assume that digital communications would be following a differential CAN standard over a separate pair, what modifications would be required to make analog communications work?

We can discuss on this Tuesday's communications subgroup meeting.

Edgar

On Dec 8, 2010, at 8:00 PM, <Leonard_Tsai@xxxxxxxxxx> <Leonard_Tsai@xxxxxxxxxx> wrote:

> Coming from ODM point of view, the digital communication cost will disappear

> fast than you expect. Most of the power adapter today already include

> digitally control IC. Adding an interface of digital communication channel

> may increase small $ in the beginning but will disappear once healthy

> competition starts.

> Analog way is a bigger problem as it creates testing requirement for

> production line and room for error.

> Leonard

> -----Original Message-----

> From: upamd@xxxxxxxx [mailto:upamd@xxxxxxxx] On Behalf Of Tomlins, Garry

> Sent: Wednesday, December 08, 2010 8:20 PM

> To: Piotr Karocki; upamd@xxxxxxxx

> Subject: RE: Simple analog signaling for an AC adapter

> I agree with Lee and Arjan:

> This is heading down a path of unnecessary and complex functionality that

> leads in turn to a complex and expensive implementation. I beleive it will

> not be attractive to the major equipment ODM's who will need to adopt this if

> it is to be a success.

> We should have a simple analog low cost option.

> In my experience in servicing the adapter market for high volume electronics

> adding a pin, a wire, a resistor a pin to an IC is a big deal - let alone a

> separate communications system!

> My vote is for a simple analog option as described. I believe this will have

> a good chance of adoption and would be a success for the project.

> Garry

> Texas Instruments (Cork) Limited, Registered in Ireland under Registration

> Number: 294554, Registered Office: Riverside One, Sir John Rogerson's Quay,

> Dublin 2

> -----Original Message-----

> From: upamd@xxxxxxxx [mailto:upamd@xxxxxxxx] On Behalf Of Piotr Karocki

> Sent: 08 December 2010 11:50

> To: upamd@xxxxxxxx

> Subject: Re: Simple analog signaling for an AC adapter

> I don't agree with Arjan.

> We could either "describe current situation" or "shape the future".

> Our standard could allow resistor-based "communication", something,

> something, and full communication (required voltage etc.).

> But as one of goals is to allow to connect every device to every supply (and,

> in near future, connections in form of grid; power hubs, power storage etc.),

> every supply has to have full communication option. Making provision to

> understand simpler communication (as resistor based) makes supply more

> costly.

> Say, we have hundred million devices. Half of them - resistor based

> communication, and tenth of them - full model of communication (4/10 of them

> some 'in-between' form).

> Or, we could force whole 100 000 000 devices to have full model of

> communication.

> But it is the only way to make this full communication cheaper - as it would

> be "more mass" production. It would be ONE standard...

> And the only way to make possible to connect device from 2010 to power supply

> from 2050 or vice versa. This scenario is not impossible - when standard

> becomes "grid version"... How often you change wiring in your house?

> ________________________________

> From: upamd@xxxxxxxx [upamd@xxxxxxxx] On Behalf Of arjan strijker

> [arjan.strijker@xxxxxxx]

> Sent: Wednesday, December 08, 2010 12:26 PM

> To: Atkinson, Lee; Bob Davis; upamd@xxxxxxxx

> Subject: RE: Simple analog signaling for an AC adapter

> I agree with Lee that UPAMD should also support low cost devices.

> A simple resistor to ground inside the device could tell the adapter what

> voltage it requires.

> More sophisticated device can still do power negotiation etc.

> With regards,

> Arjan Strijker

> From: upamd@xxxxxxxx [mailto:upamd@xxxxxxxx] On Behalf Of Atkinson, Lee

> Sent: Wednesday, December 08, 2010 12:08 AM

> To: Bob Davis; upamd@xxxxxxxx

> Subject: Simple analog signaling for an AC adapter

> Bob, here is a quick presentation on the method that HP uses for

> signaling from the adapter to the notebook. We've had this system in place

> since 2005 or 2006, and have shipped maybe close to 150million systems that

> use the common "HP Smart" system (including notebooks, all in one desktops,

> and small form-factor desktops).

> My suggestion is that we at least baseline a system that will allow some

> scalability of the communication method; at least, allow very low cost

> devices to connect to a UPAMD power source and work reliably even if their

> functionality is limited. I'm not sure that a lot of simple devices that

> would use UPAMD have a need for all the messages that the adapter could

> provide, or would be able to negotiate variable power consumption. I think we

> all agree, there is very little precedence for using sophisticated signaling

> in the common DC powered devices now in the market. Though I agree the simple

> methods are limited, if we can deliver a scalable solution there will be

> fewer reasons for the industry to not adopt UPAMD.

> Thanks again--Lee

> ===============================================================================================================

> This message may contain information which is private, privileged or confidential of Compal Electronics, Inc.

> If you are not the intended recipient of this message, please notify the sender and destroy/delete the message.

> Any review, retransmission, dissemination or other use of, or taking of any action in reliance upon this information,

> by persons or entities other than the intended recipient is prohibited.

> ===============================================================================================================