"flipper" wrote in message
...
On Wed, 05 May 2010 16:01:07 -0500, flipper wrote:

On Wed, 5 May 2010 20:58:30 +0100, "ian field"
wrote:


"flipper" wrote in message
news On Wed, 5 May 2010 15:53:19 +0100, "ian field"
wrote:


"flipper" wrote in message
om...
On Tue, 4 May 2010 21:20:41 +0100, "ian field"
wrote:


"flipper" wrote in message
news:8vo0u5lcehppt1if7msgt9gujo0qe87ag7@4ax .com...
On Tue, 04 May 2010 09:11:47 -0700, John Larkin
wrote:

On Tue, 04 May 2010 01:03:43 -0500, flipper
wrote:

On Mon, 3 May 2010 23:02:27 -0500, "Tim Williams"
wrote:

"flipper" wrote in message
news:1k4vt5pctqpqush7g57aappqgt1e4ooubd @4ax.com...
So you really don't think it matters whether one tries to drive
the
gate through a 10 ohm or 100 megohm resistor?

I said cutoff frequency, which means voltage source. Do you have
a
0
ohm
source at ~1GHz? No? Then you won't get to see fT. ;-)

Tim

I know we won't see it so I have no idea what the heck your tap
dance
is for.

Gm/Co was a popular figure of merit for tubes used in video amps
and
oscilloscopes. They took a hockey-stick turn upward when frame-grid
tubes were invented, just before semiconductors took over.

Mosfets have impressive Gm/Co values, at the cost of high input
capacitance.

John

I understand but that doesn't answer the question at hand because
we
do not have unlimited gate drive capability. FOM, Rdson x Qg, isn't
the issue as the device is 'capable' of more bandwidth if we could
drive it but, because the 'driver' already exists, we do not have
that
choice. At least not without buffering it with something and if we
do
that we might as well use the jfet under the source version. But,
in
either case, it's no longer 'plain Jane'.

For a given MOSFET technology FOM is roughly constant and I was
asking
if the same kind of thing applied to our problem of trading off gm,
with source feedback, vs Ciss, because that is what the driving
impedance works into.

Unfortunately, Mouser doesn't let me search by Ciss, or gate charge
either, so it's not a simple task to find a low one but I seem to
remember that when I originally latched onto the STP2NK60Z it was
'pretty low'... but I wasn't taking gm into account.

As I said, it's probably moot for a (AA5) 'table radio', because we
don't need much bandwidth to begin with, but I was pondering it's
potential suitability as a more general replacement. I suspect we
need
the more complex version if one wants 'full capability'.


Just a thought, but if the common source section of the cascode is
boosted
by a bipolar, then a really low Idss FET can be used - like a 40673
or a
BF991, even with both gates tied together its only a few pF.

Sure, you can.

A 'typical' (if there is such a thing) tube in this power range looks
to be about 11 or 12 pF and the BF244 looks to be about 3 pF so I
wouldn't think that's a problem. The BF246 looks to be about 15pF so
it's not as good.

Being depletion MOSFETs, they should be even better.

Better than what in what way?

More similar in character to a grid.


If you mean it seems 'more similar' physically then I suggest the SS
device, as a whole, is sufficiently different as to make focusing on
one bit not very illuminating but if you like them there's nothing to
prevent it's use. At least, I don't see one off hand.

I think it better to focus on device parameters, like the BF244 has
higher breakdown but the BF991 has more gain. And then there's that
one is through hole and the other is surface mount.

I wanted to buy some of the old BF9xx "pill" devices but couldn't find
any,
I got hold of a few BF991 SMDs and the only UHF types had Vds 7V abs-max,
there's some very old TV tuners in a tea chest in the garage, so I do
have a
very small selection of "pill" types.

7V? Boy, that's not a lot to play with. I don't know what values would
be needed to make that one work but, for example, in the BF244 v8
circuit your 7V job wouldn't survive because there's up to 16V across
the jfet (a bit over 9V even just at idle).

Remember, the MOSFET gate is fixed and there has to be enough
headroom for the jfet *plus* current swing through it's source
resistor (plus the NPN).

One might try a swinging gate on the power MOSFET but that creates
stability and frequency response problems for the same reason we went
cascode to begin with.

I found that if you cut a 4-hole square of Veroboard and carefully
V-groove
across the 2 tracks with a scalpel, making 4 copper lands with a hole
through each, the 4-pin SMD sits between the lands nicely, add 22SWG lead
wires through the holes and the device is ready for general
experimenting..

That's clever but I found TO-92 devices just sort of come "ready for
general experimenting." hehe

On the other hand it's irritating to pay shipping for a 10 cent
device.


There's no reason the same can't be done directly on the underside of a
project stripboard, it just requires a bit of carefull planning of the
layout.


P.S. If you want to just try something why not start with the simpler
'plain Jane' version?

As explained at the start, I don't have an old radio to experiment on, so
its a purely theoretical excercise.

Thanks to the suggestions in the replies, I shouldn't have any trouble
getting started should the opportunity for practical arise.