The Cyber-Spy.Com Usenet Archive Feeds Directly
From The Open And Publicly Available Newsgroup
This Group And Thousands Of Others Are Available
On Most IS NNTP News Servers On Port 119.
Cyber-Spy.Com Is NOT Responsible For Any Topic,
Opinions Or Content Posted To This Or Any Other
Newsgroup. This Web Archive Of The Newsgroup And
Posts Are For Informational Purposes Only.
Subject: Re: Future power supply concepts for automotive electronics
Date: Wed, 25 Sep 2002 00:28:59 +0200
Organization: Ingenieurbuero Erckert
References: <3D878889.98EC63A6@erckert-ibe.de> <3D8E3436.32FD08A6@webaccess.net> <3D8F86FC.CBC83B71@erckert-ibe.de> <3D8FDF51.A028E654@webaccess.net>
NNTP-Posting-Host: dialin-212-144-152-009.arcor-ip.net (126.96.36.199)
X-Mailer: Mozilla 4.72 [de] (X11; U; Linux 2.2.14-5.0 i686)
Dies ist eine mehrteilige Nachricht im MIME-Format.
How do you interface the 12V actuator drivers? Directly from your logic
IC (today 3.3V port, future even lower)?
Serial bus (SPI or SCI) is nice looking at pin count but carries a lot
of logic into a high voltage technology and there it costs money.
Actuator drivers generate high dI/dt which cause inductive supply bounce
and ground bounce. Is there acceptance to interface the low voltage
signals with differential lines (similar to old RS-423-A or CAN twisted
pair drivers) or do you just low pass filter a single wire to get rid of
signal integrity problems. In automotive I have seen ground bounce up to
+-3V on one board and up to +-15V from board to board when a 0.05 Ohm
switch hits a short circuit!
Do you power down low voltage logic in stand by or do you keep it alive
although it leaks even while it is not clocked (I have to come down to
less than 100uA at 85 Celsius with assynchronuous interrupts still being
I think of moving the interrupt handler into the power IC so that I get
rid of leakages. Problem is that I have no clue how much programability
I will need. (simple micro or is a finite state machine and some
configuration bytes sufficient?)
How about bus wake up? Where should the protocol handler sit? I think it
must be on the same chip as the interrupt handler because it must
monitor the bus while the CPU is off.
Blue lasers need about 3.6V. So it looks like 5V will stay as a mass
market technology (cheap, that's good!). But I don't think you will
drive it directly from your logic IC. The lasers I know need to be kept
close to inversion or in inversion (otherwise pumping them up from 0
will steal you some ns until you have a good beam). So they require
quite some analog circuitry anyway.
Analog on 0.18um is expensive. You have to pay submicron mask allignment
without getting the benefit of shrinking. This only makes sense if you
can save more interfacing bond pad area than the area you need putting
the analog stuff on the 0.18um.
My current conclusion is in the future we will see digtal technologies
operating between 1V and 1.8V, Analog signal processing operating at 5V
and power drivers operating at 17V (your 12V+margin), 40V (12V
automotive plus transient margin) and 60V (42V automotive + transient
Interfacing could be either
digital --- 5V analog --- power
(let's call it topology 'chain')
digital --- 5V analog
(let's call it topology 'star')
Chuck Simmons schrieb:
> I've thought about some of the stuff you mention but there are problems
> that require bridges between technologies. I am more familiar with the
> mess I have to deal with in DVDs. I presently have to have +3.3 volts
> (0.35 micron), +5 for some glue and +12 for motors and actuators. By
> using 5 volt tolerant pads, we are OK for most of the system but the 12
> volt stuff gets interesting because the 3.3 volt parts have to control
> the 12 volt parts and we are eliminating glue logic as fast as we can
> (it gets sucked inside of our LSI). Looking at shrinks, we get into
> trouble with mixed signal parts (analog doesn't shrink). Moreover, going
> to 0.18 with 2 volt power is murderous for signal to noise in the mixed
> signal parts. The 12 volt supply is not going to go away in any case
> because, for example, lowering the force constant of the actuators to
> work on 5 volts would make them too anemic to meet bandwidth
> requirements. So a shrink means I have to have all of +2, +3.3, +5 and
> +12. The 5 volt tolerance has to remain however it can be done because
> even after a shrink we still have an ATAPI interface (for several years
> to come anyway). Another unhappy situation is that the lasers have about
> a 2.4 volt drop in operation. If we get blue, I think it is even higher.
> That sticks us with +5 for the lasers. It's physics.
> This is a lot like the car electronics problem where you really want to
> minimize chip count by shrinking and consolidating functions while
> having to deal with 17 volt ugly power and controlling several amps (up
> to about 75 in typical cars). You have to either proliferate chips to
> keep technology optimized for the task at hand or do mixed technology
> which is, to some extent, possible. The latter allows total power down
> of power hog technologies while context and power up logic can be in the
> less power hungry technology.
> You asked about separate or on board flash because I mentioned that
> practical flash life works pretty well with car life. Depending on the
> ASIC vendor, it may make no difference. People do get upset about on
> board memory. It is huge. It is not unusual for memory to take up most
> of the chip area in a chip with MPU(s). One could argue to pull the
> memory out but this increases pin count and interconnect (bad).
> In any event, the goal has to be to minimize chip count, minimize
> interconnect and still get the job done without compromise.
> ... The times have been,
> That, when the brains were out,
> the man would die. ... Macbeth
> Chuck Simmons email@example.com
Content-Description: Visitenkarte für ricardo
tel;cell:+49 179 5308708
tel;fax:+49 8061 392214
tel;work:+49 8061 392208
org:Ingenierbuero Erckert;IC design
adr:;;Harthauserstrasse 36b;Bad Aibling;Bayern;D-83043 Bad Aibling;Deutschland
Go Back To The Cyber-Spy.Com
Usenet Web Archive Index Of
The sci.electronics.design Newsgroup