Roy,

So if I had an adjustment on VG2 for each circuit and adjust for
required Drain current on each product ?

What's the consensus regarding Common Base BJT as LNA ?

Thanks

Regards

David


Roy Lewallen wrote:
The problem with using FETs of all kinds is the wide part-to-part
variation. Look at the specs for the BF998 - many of the critical specs
show only a maximum or minimum, but not both, or just a typical value.
You can be way off if you simply use a "typical" set of curves. If you
want to do an analytical design with a part with non-specifications like
this is to use a curve tracer to generate curves for the individual
part, then use those curves for your design. Pull another part of the
same part number from your drawer, and you'll need a different design.
This exercise is useful for educational purposes, but it isn't a
technique you can use to design something that can be easily duplicated.

That's probably why you don't see a lot of FETs being used in commercial
products, except in applications where there's a lot of feedback to
stabilize the operating point, such as source followers, or when simply
nothing else will do. Even then, the manufacturer has probably paid the
vendor to select parts with a much narrower, and well specified, range
of characteristic values. That's been my experience in designing
commercial electronic test equipment.

Roy Lewallen, W7EL

David wrote:
Tim,

Say we look at VG1s = 0.1V as per your example.

The graph for BF998 shows that if VG2s = 4V and VDs = 8V then ID
approx = 12.5mA

This would mean that unless I applied a negative voltage on the source
I would need to apply 0.1V forward bias to G1 and 4V to G2 ?
As Rs is creating a negative self bias voltage ?

If I set the bias point lower - say 5mA then VG1s is approx. -0.2V
according to the graph.

I can achieve this by using a resistor in the source of 0.2/5mA (40
Ohms) and then set VG1 = 0 (so that VG1s = -0.2V) and then 4.2V on G2
so that VG2s = 4V.

Is this correct ?

The transfer characteristic curve shows that for say 10mA. If VG2s =
4V then gm = around 24mS and if VG2s is reduced to 0V the gm reduces
to about 7mS.

Thanks

regards

David

On Fri, 28 Jul 2006 06:08:01 GMT, David
wrote:

Roy,

So if I had an adjustment on VG2 for each circuit and adjust for
required Drain current on each product ?

Or pick a set of bogy values and accept there will be a range of
operating current.

I've found DGmosFET perfomance is not greatly impacted by
variations in Idss and Gm for practical circuits.

What's the consensus regarding Common Base BJT as LNA ?

Hard matching the input at any decent current, at 4-5ma the input R
is around 5 ohms. Easily overloaded as a result of usually being used
at low current to make the match easier. It's feature is fair
stability and the device is working at at it's alpha cutoff frequency.
The latter was more important 30 years ago when UHF transistors
were harder to get. Noise performance was dependent on device
but even in 1972 I could get TIMX10s down to around 1.5DB at
450mhz.


Allison


Thanks

Regards

David


Roy Lewallen wrote:
The problem with using FETs of all kinds is the wide part-to-part
variation. Look at the specs for the BF998 - many of the critical specs
show only a maximum or minimum, but not both, or just a typical value.
You can be way off if you simply use a "typical" set of curves. If you
want to do an analytical design with a part with non-specifications like
this is to use a curve tracer to generate curves for the individual
part, then use those curves for your design. Pull another part of the
same part number from your drawer, and you'll need a different design.
This exercise is useful for educational purposes, but it isn't a
technique you can use to design something that can be easily duplicated.

That's probably why you don't see a lot of FETs being used in commercial
products, except in applications where there's a lot of feedback to
stabilize the operating point, such as source followers, or when simply
nothing else will do. Even then, the manufacturer has probably paid the
vendor to select parts with a much narrower, and well specified, range
of characteristic values. That's been my experience in designing
commercial electronic test equipment.

Roy Lewallen, W7EL

David wrote:
Tim,

Say we look at VG1s = 0.1V as per your example.

The graph for BF998 shows that if VG2s = 4V and VDs = 8V then ID
approx = 12.5mA

This would mean that unless I applied a negative voltage on the source
I would need to apply 0.1V forward bias to G1 and 4V to G2 ?
As Rs is creating a negative self bias voltage ?

If I set the bias point lower - say 5mA then VG1s is approx. -0.2V
according to the graph.

I can achieve this by using a resistor in the source of 0.2/5mA (40
Ohms) and then set VG1 = 0 (so that VG1s = -0.2V) and then 4.2V on G2
so that VG2s = 4V.

Is this correct ?

The transfer characteristic curve shows that for say 10mA. If VG2s =
4V then gm = around 24mS and if VG2s is reduced to 0V the gm reduces
to about 7mS.

Thanks

regards

David

Sorry, I have almost no experience in using dual-gate MOSFETS, for the
reasons I mentioned. Beginning somewhere around the mid-80s, when
manufacturing became highly automated, my employers strongly discouraged
designs which included any adjustments. Besides the labor required to
make the adjustment, the variable component lowered the product's
reliability, so we'd often design in a lot of parts to get around having
any tweaks. But individual adjustment is still a viable option for some
products. What I don't know is whether you'd get satisfactory
performance with widely differing devices all running at the same drain
current. That would depend on your design and application.

If you're considering making a product, I'd certainly do some modeling
with extreme component parameters to see what happens. And you might
consider some sort of device selection and/or incoming inspection or
sorting to make sure you don't get any truly extreme parts. One thing to
be careful about is that when a part is so poorly specified, other
companies might be buying large numbers of selected parts. That leaves
you with the leftovers. I've seen some really strange distributions
resulting from this -- parts with extreme characteristics on both ends,
but nothing anywhere near the "typical" values. It used to be common
with zeners, until they got better at making -- 10% tolerance zeners
would all be between 5 and 10% from nominal, in both directions, with
none closer than 5%. Those had been selected out and sold as 5%
tolerance parts.

Just a few of the things I've picked up in 30 or so years as a design
engineer.

Roy Lewallen, W7EL

David wrote:
Roy,

So if I had an adjustment on VG2 for each circuit and adjust for
required Drain current on each product ?

What's the consensus regarding Common Base BJT as LNA ?

Thanks

Regards

David

On Sat, 29 Jul 2006 00:00:12 -0700, Roy Lewallen
wrote:

Sorry, I have almost no experience in using dual-gate MOSFETS, for the
reasons I mentioned. Beginning somewhere around the mid-80s, when
manufacturing became highly automated, my employers strongly discouraged
designs which included any adjustments. Besides the labor required to
make the adjustment, the variable component lowered the product's
reliability, so we'd often design in a lot of parts to get around having
any tweaks. But individual adjustment is still a viable option for some
products. What I don't know is whether you'd get satisfactory
performance with widely differing devices all running at the same drain
current. That would depend on your design and application.

If you're considering making a product, I'd certainly do some modeling
with extreme component parameters to see what happens. And you might
consider some sort of device selection and/or incoming inspection or
sorting to make sure you don't get any truly extreme parts. One thing to
be careful about is that when a part is so poorly specified, other
companies might be buying large numbers of selected parts. That leaves
you with the leftovers. I've seen some really strange distributions
resulting from this -- parts with extreme characteristics on both ends,
but nothing anywhere near the "typical" values. It used to be common
with zeners, until they got better at making -- 10% tolerance zeners
would all be between 5 and 10% from nominal, in both directions, with
none closer than 5%. Those had been selected out and sold as 5%
tolerance parts.

Just a few of the things I've picked up in 30 or so years as a design
engineer.

Roy Lewallen, W7EL

Roy,

There are a large number of VHF and HF radios both past and present
that use the DGmosfets in place like the RF amp sometimes a balanced
mixer. A recent example is the TenTec 526 (6n2) 6 and 2 meter radio
as the RF amp. Another current example is the TT 1208 transverter
also as RF amp. If your not pushing the device for max gain or
ultimate IP3 possible they perform well and offer low noise for their
power needs. Neither require pots or other tweaking (other than
tuneable circuits). Can other deivces be used to do better, yes.
But, engineering is always about understanding and compromize.

I'd never use DGfets for something like a scope amp or other
instrumentation where DC operating point or balance is a
requirement. I have use them as a high impedence (1mohm)
AGC'd input for high input senstivity frequency counters.
Another place where I've used them is IF amps, they are just far
easier to AGC than CA3020 or MC1350 and quieter.

My favorite line from years back.

Good, Fast, Cheap, Pick any two.


Allison

In article ,
wrote:
I'd never use DGfets for something like a scope amp or other
instrumentation where DC operating point or balance is a
requirement.

Interestingly enough, in Jim Williams's book "The Art and Science of
Analog Circuit Design", there's a chapter by Steve Roach of Tektronix,
which shows the use of a BF996 consumer-grade dual gate FET in a 1GHz
oscilloscope preamp.

http://books.google.com/books?vid=IS...VGMx8v3yDD4at4