6 Feb
2012
6 Feb
'12
5:50 a.m.
On Feb 6, 2012, at 12:01 AM, Mouse wrote:
Sorry. Still searching for something I like better than Thunderbird,
which I'll use over Outlook any day but not Mail.app.
Well, it does if you're making emitter-followers. They only drive voltage in
the direction of the emitter (so NPN will source strongly while PNP will sink
strongly). ECL is designed with a complimentary pair of emitters sourcing
a low-differential voltage into a termination pulldown (which is necessary,
or you'll only get high voltage, since the emitter-follower won't pull the
other way).
So in short, yes it does assume the polarity, but ECL defines the polarity.
They could have done emitter-followers with PNP if they'd wanted, but I
imagine at the time the difference between the dopants would have been
extreme enough that you'd get severe enough performance degradation that
NPN with a negative Vee was considered worthwhile. It can still be a factor
of 2 difference between P-type and N-type these days.
It's not. It's just that the long-tailed pair (which ECL uses as the
input buffer) isn't as sensitive to noise on Vee (because that just changes
the common mode voltage) as it is to noise on the inputs (which you'd get
through noise on Vcc). Using ground as Vcc makes for a more stable
circuit in this case, assuming ground is a quieter rail than a positive
Vcc would be.
Again, as above, if you could make PNP transistors that performed as well
as NPN, they might have opted for that instead. These days, we use PECL
(with NPN drivers) for all sorts of things because transistors are better
and the noise margins are good enough that we don't have to worry about
that; we worry about having enough current to drive hundreds of MHz over
long traces with fast edges (which is why you'll see lots of PECL clock
buffers even today).
They're not saying that ground is stabler, they're saying that using
ground as Vcc is stabler than using a regulated rail as Vcc (which, in
any normal system that's not using a daisy-chain of lamp wire for its
grounding, is usually true).
Wikipedia actually has a pretty good discussion of this on its ECL page.
- Dave
And, in the
same sense, earthed ground is very important for lots of things.$
Please don't use paragraph-length lines. There are two
good reasons ECL runs at negative voltages:
- The output drivers are emitter-followers, so
[...]
Um, doesn't this assume which polarity of transistor you're using?
Except for asymmetries in the behaviour of the dopants, I can't see any
raeson why any bipolar technology wouldn't work equally well if you
swap NPN and PNP, reverse all diodes, and negate all voltages. Am I
missing something? - The system
"ground" is typically more stable,
This mystifies me. Why would system ground be more stable if you view
the power supply as ground-and-negative rather than
ground-and-positive? This has a lot
more to do with the way we treat what we want to call
"ground" than any reason relating to its absolute potential, of
course.
I'm still mystified. How do people treat ground-and-negative power
differently from ground-and-positive that makes ground stabler?