ecl comparator hysteresis

Are those measurements in millimeters? They would seem too big to be in cm, but the smallest boxes would be too small to be useful in mm.

Confused...

m,

millimeters

There are some (very) rough measurements in here.

The TW600 stuff looks like about K = 7 w/mK uncompressed, which is still pretty good. The real advantage of compressing is to reduce the gap. Heat flow goes as Dt * K * Area / Gap.

Still air is about K = 0.025.

Thanks

Nonsense. ECL uses an NPN emitter follower output, with the pull-down being the terminator. There is no push-pull.

te:

e pins

A, and

ould I

esis?

. To adjust hysteresis Fig. 10 shows LE pulled hard high by connecting to G ND. Then the resistor divider voltage is applied to /LEB ==LEB for the adjustment. LE-LEB negative means /LEB is pulled so many tens of mV above G ND. Not much hysteresis to speak of pulling LEB below GND.

an AND. Input abouve lower thresh AND below upper thresh gets a logic "1" o utput, etc.

n though you don't need the complementary output, it's still a good idea to load it to maintain current balance on the power supply distribution in th e chip and eliminate inductive ripple you can't decouple.

complementary output, so I just extended the terminology to this case.

Oh wow- thanks for the correction. Now I know ALL there is to know about EC L.

On Monday, December 2, 2019 at 3:59:02 AM UTC+11, snipped-for-privacy@gmail.com w rote:

e:

rote:

ven though you don't need the complementary output, it's still a good idea to load it to maintain current balance on the power supply distribution in the chip and eliminate inductive ripple you can't decouple.

ll complementary output, so I just extended the terminology to this case.

ECL.

Trust krw to completely miss the point.

You're welcome. Obviously you were clueless before.

te:

even though you don't need the complementary output, it's still a good idea to load it to maintain current balance on the power supply distribution in the chip and eliminate inductive ripple you can't decouple.

ull complementary output, so I just extended the terminology to this case.

ECL.

"Obviously" only if looked at from krw's bizarrely narrow point of view.

Krw's cognitive deficit means that he can't recognise that anybody could ha ve a valid point of view that isn't identical with his.

I'm doing a new optical-electrical converter. It's just on a whim, since our ski trip was canceled, it's cold and rainy outside, I finished my book, and I feel like hell. May as well try something. I'm hoping for 1 GHz.

A more serious comparator is the single ADCMP582, but it has reduced swing outputs so won't drive my fets downstream.

What happens if you don't terminate CML? Do you get a clean source-terminated 800 mV swing?

Should work. Your source-termination would have to include the output impe dance of the CML driver, and you'd be driving both the source terminated line and the pull-down resistor. It could look good into a short and tolera bly non-dispersive 50R transmission line.

The FET gate wouldn't give a particularly clean reflection, but that probab ly wouldn't matter.

But why bother? You won't get any more voltage swing.

These are pretty weird too. If "latch enable" means what it usually means, then feeding the outputs into LE should either permanently latch it or make it oscillate. How that gives it hysteresis isn't clear.

It's an old tradition to use a differential ECL latch enable to also program hysteresis. Maybe AMD invented that, back when they did analog.

External comparator hysteresis, with resistors, can have problems in fast circuits. Or most any circuit.

But I've never seen the outputs of a comparator fed back into the LEs of that same comparator.

You haven't got around much.

Sloman, A.W. and Swords, M.D. "A fast and economical gated discriminator", Journal of Physics E: Scientific Instruments, 11, 521-524 (1978).

The Am685 introduced the feature of latching up the input stage of a fast comparator (which could happen remarkably quickly) and at least one other author published a paper exploiting the feature back in the late 1970s.

A much more recent single-photon-avalanche photodiode detection circuit used much the same trick, but was careful not to mess up the comparator input when a photon pulse might be hitting it.

IIRR Sergio Cova's team at the Milan Polytechnic worked up and published the circuit.

I've sometimes AC-coupled the output to the latch (e.g. with the late lamented MAX900). It works a lot like a Schmitt trigger but without pushing crap back out the inputs. You do need enough delay between LE and the output so that it doesn't oscillate.

Cheers

Phil Hobbs