picosecond test points

Right! It's 1mA for this example.

John S

follower,

It looks like it's thermal, because it only happens at high collector currents. Down where you'd actually want to use it, it's flat like a ruler.

I'm seriously considering resurrecting that constant-dissipation trick for thermal tracking in noise cancellers, i.e. jack up V_CE as I_C goes down so as to avoid thermal tails and offsets. It was worthless with VAF=12V Si RF devices, but with VAF >= 1 kV, it might be competitive.

Cheers

Phil Hobbs

I would add a high speed connector. The best I've seen (so far) is the Huber & Suhner MMPX line. They are tiny but rated to about 68GHz. However, with coax and end connectors, your bandwidth will be reduced from that.

We're looking at the latest LeCroy scopes to look at 37GHz signals. Their new 10 series has up to 60GHz of bandwidth and 160GSamples/s (real time). It's only about $400K :-|. They use 2.4mm connectors. I'm not sure if you can get a direct MMPX2.4mm without adapters.

Bob

Oh..i like that non-functional circuit on page two..

Yeah; that floating gate is a bit puzzling..

It's even worse if you want a fast PNP that is space-proof. Say, a BFT93 in a metal can. Nothing. If there was one, you can be sure that its qualification has timed out 10 years ago...

Yesterday, I changed a Wilson current mirror from BFT93 to Intersil HFA3096. Used to work nicely with BFT93, but oscillates like hell with the HFA3096 if it gets more than 1 mA :-((

I wonder what will happen with these extra-wide-body hermetic flatpacks in place of the SO-16s. Probably nothing good. But at least there is a rad-hardened version.

Is there a sure way to calm down a Wilson current mirror without too much sandbagging or introducing noise? It MUST work into a capacitive load. A simple mirror seems to be stable.

I think they just relabel semiconductors they buy somewhere else, and they may even switch manufacturers. I think that was the case for ERAs.

And the phase noise of old SRA-1 used to be better than it is now. But it was never specified, so we cannot complain. Still have NOS.

Rrrrrrrrrright!

Gerhard

In december, we had one of these 80 GSPS real time thingies from Agilent for a week to test, made me quite greedy.

Gerhard

Put a 10-100 ohm ferrite bead in series with each of the bases. (including the diode-connected ones).

Cheers

Phil Hobbs

Right. They trashed the ERA5 beyond recognition. Caused us grrrrrief.

I have a Tek 11801 sampler with a 40 GHz head. Costs about $2K on ebay.

That was at my customer's.

I myself have a hp54750a with 20 and 50 GHz plug-ins, 9 ps risetime. But it is not real-time. One cannot trigger on the one bit error in a 10 GB/s fiber optic link that may or may not happen this night.

Wow, that's pretty bad.

Are they suggesting that, since it's beta bracketed, it should be beta biased? Certain parties Would Not Approve. Makes sense for RF amps.

Plus it doesn't run Windows!

Cheers

Phil Hobbs

Well, you can trigger off it--delay lines are easy in fibre--but you only get one data point per trigger. Just choose that point well. ;)

Cheers

Phil Hobbs

A realtime scope would be handy around here for snooping PCI Express signals. Maybe pricing for a, say, 4 GHz scope (with probes!) will get reasonable some day.

You can get a TDS694 for $5k-ish. 3 GHz, 10 Gs/s.

Cheers

Phil Hobbs

No, the simple npn mirror wasn't stable, too :-(

That's what I meant with sandbagging & noise. 50R makes it stable, but it adds to Rbb. I won't get ferrites for fear of noise pickup. As usual, no silver bullet. But many thanks, anyway. And I like your book!

sigh, Gerhard

Ouh, ferrites can add to magnetic pickup. I never thought of that. Thanks, George H. (a mini toroid?)

Probably because it doesn't happen. A chunk of ferrite - aka a ferrite bead - can provide a low permeability short-cut for any magnetic field in the immediate vicinity, but ferrite beads are small and don't have much of an immediate vicinity. The flux concentrations they produce are going to be really hard to detect.

-- Bill Sloman, Nijmegen