From: Mike Monett
X-Mailer: Mozilla 2.02 (Win16; I)
Subject: Re: Anybody have a schematic of an ECL NAND gate?
Date: Wed, 15 Jan 2003 09:00:26 -0500
NNTP-Posting-Date: Wed, 15 Jan 2003 08:59:54 EST
Organization: Bell Sympatico
Winfield Hill wrote:
> Hmm, useful as a cheap high-frequency analog multiplier, for
> two ac-coupled signals biased at the vref voltage. :>)
> - Win
The Motorola ECL from MECL I through 10K, 10KH, and 100E make excellent
wideband analog devices. Two, and possibly three, sections of the 100E116
differential line receiver can be cascaded before instability sets in.
But you have to pick the sections and the pins carefully. Not all
combinations will work. The multiplexers have an interesting internal
structure for some applications.
ECL can drive 50 ohms terminated to -2V directly, but you have problems
driving coax when the shield is grounded due to the offset voltage.
Due to the wide bandwidth, input rms noise might be a problem. I would
not recommend using a number of gates as analog amplifiers with shutoff
capability and combining the output signals. They are just too noisy.
The ECLINPS, Plus, and Light versions have too much gain and bandwidth
for the package grounding, and simply oscillate. I have been unsuccessful
in various attempts at bypassing, grounding, and circuit arrangement to
get these to work as analog devices.
The 100EP52 makes an excellent limiter when a small amount of hysteresis
is added. However, the risetime is quite slow (~700nS) due to the
propagation delay of the chip. A second EP52 should be added if risetime
I was able to plot the jitter vs amplitude in this configuration and
discovered the rms input noise of the EP52 seems to be much lower than
the 100E116, despite the wider bandwidth. I did not pursue it since it
oscillates without hysteresis, but I really should spend some time to
analyze it further. Part of it could be due to the improved processing
needed for these devices.
The devices are not cheap. At one time, I was paying US$50 each for some
100E chips. The ECLINPS series are much more reasonable, but still