Cascode emitter follower

Using a cascode of fast transistor and HV transistor is a familiar circuit: that's how many video amplifiers drove CRTs. A cascode used as a follower, however, would not gain voltage range because the collector of the input transistor is clamped at some (low) bias voltage. So, all you gain is some relief from Miller effect of using a collector limit resistor, and a simple collector-to-ground bypass cap does the same thing.

Right, but they used a single (asymmetrical) cascode, plus complementary follower. The transistors they made for this application, were probably the absolute best possible in pure Si: typically 100V, 100mA and 1GHz fT.

For the same voltage rating, the best I see today is 400MHz and below.

The impedance was also rather higher than a 50 ohm load (maybe more like 500 ohms?), which can illustrate my point all the better -- if you wanted to build such a circuit today, you'd have no choice but to stack faster, lower voltage parts, to make up the same total voltage capacity.

Or use a MOSFET, like MRF6V2010 looks like it would do a fine job. It's just 100 times more expensive than the original approach...

Right, there's no electrical advantage, or at least there shouldn't be much that I know of. But by distributing the voltage drop over N devices, you can use devices rated for VCC/N, which will have more fT. It's a matter of economics rather than physics!

Tim

I'm not sure what you mean by cascoding a follower. A sketch would help to start.

Tek made some differential plugins that had phenominally wide voltage ranges, like 100 volts or so, with radically bootstrapped front-end tubes, insane common-mode rejection and overload recovery. Worth reviewing.

How about a constant-current sink and bootstrap the collector/drain?

The sink could be a bootstrap too.

We have bootstrapped the supplies of chopper opamps to make followers that could swing way past the opamp capability. It was rather a nuisance.

Sure:

Maybe instead of 10k dividers for base bias, another set of followers would be needed. That's just details.

It's not something you'd ever normally do, because you can get transistors with enough Vceo for, say, audio amplifiers. Or MRI gradient coil drivers.

But you can't do that at 100MHz or more, and into a 50 ohm load, because those transistors don't exist [anymore?].

I'm kind of tempted to make such a beast, too. I doubt it'd be good for much beyond 100MHz bandwidth, though. Semiconductors can do a great job at improving tubes, but there are some things they're just not good at, and high impedances are one of them.

I'd also like to make a galvanically isolated differential probe, like the dongles with the banana cables on 'em. But those things are only 10MHz or so, tops. 100MHz or more bandwidth, with super high CMRR, would be great for checking GaN gate drivers, what with their voltages and power capabilities going up as they are these days.

I'm thinking power, so you wouldn't be using sinks, but complementary followers instead. Though I'd be okay with 'hot' class AB if it means maximizing bandwidth. It's lab stuff, it's supposed to get hot. ;)

The basic idea is to bootstrap the followers' collectors to half the difference (output to +V/-V), thus reducing each transistor's voltage range by half. Which means you can double the supply voltage, for a given transistor type.

Or N transistors and N times the supply, etc...

Tim

• Aren't C1 and C4 a bit on the high-frequency output bypassing side?

Hmm, why call out C1/C4, and not C2/C3, or all four of them?

Note that the C2-C3 "tap" can be driven by the input just as well (but that would be worse), or Q1 and Q5 are cap-coupled from a half-input node (which follows from the realization that they have equal AC voltages, which is half the input AC).

Or replace C1-C4 and R1, R2, R7, R8 with a pair of half-VAS stages, so that you get DC bias and AC drive to all of them. But that would be complicated. It may be worthwhile anyay, to avoid drive problems (the cap values shown are on the small side already, given the 300pF+ capacitance of transistors this size).

Tim

This is a series-connected amplifier, not a cascode. It is a good thing for high voltages. I once made one with 4 series-connected FETs in each leg for a +/- 1000 V controlled voltage source. You may be limiting the upper frequency limit with the caps, they seem quite large.

In a cascode, the grid/gate/base of the second stage is at a fixed voltage, to isolate the input stage from the ouput voltage.

Could you kindly sketch this configuration? I'm having a hard time intuiting it from your 'purpose'. -F

There was a posted schematic... looked pretty useless to me. ...Jim Thompson

Sorry I hadn't seen this link earlier. I'd call this a bootstrapped EF rather than cascoded.

Rather than looking at Tek's differential plugin design, I'd look at their vertical deflection amplifier designs. Some of the modest-frequency units (mostly

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

I thought you weren't looking at my posts anymore. ;-)

That means I have your IP now! I know where you live!!!

...

Nevermind. :^)

Tim

I didn't say that, I said you were an amateur >:-}

Finally figured out where you got the name, "Seven Transistor Labs", 7 transistors is your cognitive limit ;-) ...Jim Thompson

I think it's a quote from an essay in one of Jim Williams' books, "in praise of small circuits" or something like that.

Cheers

Phil Hobbs

That is not cascode.

I just said that.

Not really for that reason so much as to get the extreme voltage gain. The DC doesn't have to really be fixed all that hard, it just has to be well by passed at the operating frequency.

Now taking that output into a common collector stage (emitter follower) now you got something. Serious gain with some balls behind it, and only three transistors.

Actually, the Ampzilla is series connected, a configuration I do not like. I got the Phase Linear 400/2 which has parallel outputs, not series. I pref er that design quite a bit. All this arranging things in series gives me th e creeps. When I was younger I worked out the topologies and architecture o f these things and found out that there are simply too many variables. In o ther words, f*ck all that.

I am not into making things hard on myself, I got enough problems.

Common emitter feeding a common base. If it ain't that, it ain't cascode.

I didn't make a specific reference, but it's in the same spirit!

Tim

It would be fun to teach a course where we design verious things from, say, 7 parts.

Agreed. Let's do it.

But no white nylon breadboards!

Cheers

Phil Hobbs

Teach'em how to solder. Everybody should learn to solder.