TV preamp FET choice

[John S. Dyson]

Actually, I've just turned up some 3SK88 DG-MOSFETs. Would they
suffice?

The problem with DG fets is trying to maintain wideband response
and maintaining good noise match... (They usually like to see very
high impedance on the input.) This isn't always true, but the
HP PHEMTs will match much better.


John

 

[Ken Smith]

I assume you're referring to the T092's internal bonding wire
inductance?

Not just the internal wire. There is a min. amount of external wire.

Just 2 and 4 pF for the J301? That seems manageable, surely?

1/(2 * PI * F * C) = about 50 Ohms.

50 = 2 * PI * F * L solve for L

L = 50/(2 * PI * F) = 10nH (about)

This sound low for a TO-92.

"sterling cycle device on the idle load"?? WTF does that mean?

Oh sorry the "s" in Sterling is upper case. You need it because according
to the Mannly-Rowe equations, the thermal noise in the idle circuit will
always be the determining factor when it comes to noise. Peltier /
Seeback devices don't work well enough under these levels of gradient.


Hope this makes it clear.

 

[John S. Dyson]

The problem with DG fets is trying to maintain wideband response
and maintaining good noise match... (They usually like to see very
high impedance on the input.) This isn't always true, but the
HP PHEMTs will match much better.

I strongly suggest looking at devices with input impedance that is fairly
close to 50 (or 75) ohms. The greater the amount of needed conversion,
then the amount of circuit bandwidth and sensitivity to components gets
worse. Untuned transformers can help a little in that regard, but it is
best to avoid them also (esp on input, where the signal loss can be costly.)

When choosing a device where you want to use it in low noise application,
there is also an ideal source impedance where the NF will be best. Sometimes
this match requirement isn't critical, and sometimes it is fairly critical.

For low noise RF applications, you probably want to try to find components
whose input impedance is within 2:1 or 4:1 of the source impedance. You
can CERTAINLY use components outside of that range, but life is easier
if you are minimizing the matching issue. Also, the component noise match
is also important.

For DG fets, you'll find that the optimium source impedance for the best
noise match will be a fairly high impedance (at normal frequencies), and
the ideal impedance match will likely be even higher. When casually
looking at the S11 parameters for typical DG fets, you'll find that the
S11 will be something like 0.9998 or 1.0002 over at least part of the
range (ignoring the phase, and really that cannot be ignored in general.)
Most often, a device whose S11 is 0.9998 means that its input impedance
isn't generally very close to 50 ohms. An S11 figure of 1.0002 likely
means that the component is not very stable (which very roughly means
that more signal will be reflected to the input than supplied by the input
device.)

DG fets do have very good applications (relatively wide AGC, good
intermod characteristics, good noise behavior, etc.) For applications
where AGC isn't needed, then there are components that might likely
do much much better. The issue here is whether or not the user needs
'much much' better performance.

If you look at the ATF-54143 at 900MHz, you'll se that the S11 is
about 0.7 (ignoring phase), the S21 is about 14... This VERY ROUGHLY means
that only a very small amount of transformation is desirable at the input,
and the gain almost comes for free... This small amount of needed
transformation (implying a low loaded circuit Q) means that the parts
values shouldn't be very critical.

*When dealing with other kinds of components, it is likely
that the S21 would be much lower, and the S11/S22 are much
closer to '1.0', which implies a more challenging and likely
narrower bandwidth matching circuit. S11 values of nearly
1.0 can mean a very low impedance (e.g. BJTs) or very high
impedances (e.g. DG mosfet inputs.) When the S11/S22 values
show such a serious mismatch with 50ohms, the S21 (the gain)
will likely show much lower than one might expect. This is
because the gain of a 50ohm input/50ohm output circuit will
indeed be fairly low. When building circuits with the HP
PHEMTS, you'll notice that their gain and ease of use is
very high. Some of the PHEMTS have gm of approx 1mho, while
the good old J310 might have a gm of 0.01 or 0.02 mhos!!!

On the ATF54143 at 60ma and 900MHz, the optimum gamma for the input
is about 0.32 (well, with an angle of
60degrees), but shows that its noise match isn't too damned far
from 50ohms (or 75ohms.) Without any amtching, the NF is still
well under 0.5dB.

Bottom line, with relatively careful layout, the ATF54143 cannot
help but give gain and low noise. Most likely, with good layout,
and careful handling of LF gain and impedance (helping to minimize
intermod effects also), a very very simple circuit cannot help but
give 10-20dB of gain with almost 50mw of power output and <<1dB NF.

With an ATF54143, it is important to take advantage of the two
source leads, and keep them at about 1.6nh each (total of 0.8nh).
This means using good microstrip layout. To help flatten the
LF gain (and to help with intermod producing LF components), then
some parallel negative feedback (between drain and gate) will serious
help performance (at the cost of some NF, but there is some NF to
spare.)

John

 

[Ken Smith]

I strongly suggest looking at devices with input impedance that is fairly
close to 50 (or 75) ohms.

If you can't find 50 or 75 ohms, try for twice that and parallel two of
them.

[...]

For low noise RF applications, you probably want to try to find components
whose input impedance is within 2:1 or 4:1 of the source impedance.

Better yet, look for one where the maker elected to put the noise at
the impedance level you need on the first page of the data sheet.
Component companies always (mostly) try to make their part look the best.
As a result, they put the characteristics that are good on page one.

 

[John S. Dyson]

If you can't find 50 or 75 ohms, try for twice that and parallel two of
them.

Hmmm... I didn't think about this before, but I guess that someone
could use 4 DG mosfets in parallel? The gm would effectively
be multiplied, and the high input impedance would help to mitigate
some of the loading effects. DG mosfets still like high impedance for
good noise match...

[...]

For low noise RF applications, you probably want to try to find components
whose input impedance is within 2:1 or 4:1 of the source impedance.

Better yet, look for one where the maker elected to put the noise at
the impedance level you need on the first page of the data sheet.

I look carefully at those. The philips DG mosfet data sheets seem
to be relatively complete for DG mosfet. data sheets.

John

 

[John Larkin]

Oh sorry the "s" in Sterling is upper case. You need it because according
to the Mannly-Rowe equations, the thermal noise in the idle circuit will
always be the determining factor when it comes to noise. Peltier /
Seeback devices don't work well enough under these levels of gradient.


Hope this makes it clear.

Do people still use parametric diode/circulator amps? I remember when
tunnel diode/circulator amps were the rage (sniffles in nostalgia for
tunnel diodes.)

NEC has some dirt-cheap PHEMTS that are, like, 0.6 dB NF at room temp,
if you tune them right. So the input of such an amp looks like a 50
ohm resistor cooled to 40K or something, a nice little heat pump all
by itself.

John

 

[legg]

The J310 would probably give you a very small gain at 850MHz with

That may be all that is needed.

RL

 

[John S. Dyson]

Do people still use parametric diode/circulator amps? I remember when
tunnel diode/circulator amps were the rage (sniffles in nostalgia for
tunnel diodes.)

NEC has some dirt-cheap PHEMTS that are, like, 0.6 dB NF at room temp,
if you tune them right. So the input of such an amp looks like a 50
ohm resistor cooled to 40K or something, a nice little heat pump all
by itself.

I checked the NEC34018 (it looks like the NEC part that seems to be
reasonably stable to work with at 1GHz.) Unfortunately, it has a
gamma-opt of approx 0.6 at about 1GHz with a noise figure of 0.4dB. This
is not a wonderful noise match to 50/75 ohms (but probably not very
bad either.) It appears that the noise voltage on the 34018 is a little
higher than my own personal choice . Looking at the NEC line, it might
be a good idea to look at the SiGe BJTs. I suspect that if someone is
worried about cost, the SiGe BJTs might be less expensive (less than $1),
and the higher current units with very high Beta (>200) would also have
very low 1/F noise... This would help in applications where phase noise
could be a killer (e.g. oscillators.) Potentially lower noise NEC
components (other than SiGe BJTs) would be significantly higher
freq components, and might be more difficult to tame. The high beta
of the SiGe BJTs and the good processing, makes for very low noise
component that doesn't create much phase noise.

On the other hand, my suggestion of a ATF54143, the gamma-opt of about
0.35 (or better, under higher bias) and the associated NF of 0.20dB
would be easier to deal with. It appears that the 54143 has lower
en (noise voltage) than the NEC component, and would work better with
lower input impedances. With the ATF54143, you could bias it at 80ma
and even end up with a gamma-opt of about 0.25, and associated NF of
about 0.22dB. Running the 54143 at 80ma and 3V seems a little extreme,
and is something that I'd might avoid in production. However, with
that incredibly good
noise match, the 54143 would probably be unexcelled by anything except
for 1/F sensitive applications where SiGe BJTs at 10-20ma would be ideal.
(For the lowest noise application with minimal matching, perhaps the
HP ATF531 type component would be better yet.) That component is designed
to work at high currents.

John

 

[John S. Dyson]

That may be all that is needed.

When I said >>>3dB NF, I meant perhaps 5dB or so. I'd try simulating
it with LTspice, making sure that the parasitics are included (because
of the fet models typically not including them for 310.)

A common base circuit might work, where you'd use maybe a 3:1 output
transformer (tuned with the parasitics?) I wouldn't really know how
to deal with pushing the J310 that high. The J310 probably comes
close to 75ohms input impedance with a near Idss bias in a common
base configuration.

My guess is that it MIGHT be plausible to get 3-5dB of gain with 5dB
NF... Maybe, for low cost and reasonable performance, using SiGe
transistor might be good? For 50-75 zin, a good noise match will
likely be at 10ma Ic with an SiGe transistor???

John

 

[Henry Kiefer]

Hmmm... I didn't think about this before, but I guess that someone

could use 4 DG mosfets in parallel? The gm would effectively
be multiplied, and the high input impedance would help to mitigate
some of the loading effects. DG mosfets still like high impedance for
good noise match...

Yes, I saw a app from a very experienced ham applying 4 BF998 for optimum
NF. Unfortunately, applying a couple of them gets you more instability
issues. But it works surely up to 4 items. 2 of them look good in pcb
layout.

You can look into the circuit shematics of european tv sets. All of them use
DG mosfets on the input stage.

I look carefully at those. The philips DG mosfet data sheets seem
to be relatively complete for DG mosfet. data sheets.

John, I don't think you can beat the price-performance ratio with other
devices!
If you need ultimately NF performance, then the DG mosfet may be not the
right canditat. A "practical" circuit works very well with them.

BTW: I'm not from Berlin. I live in Stuttgart near Porsche fab.

cu -
Henry

 

[Tony Williams]

A common base circuit might work, where you'd use maybe a 3:1
output transformer (tuned with the parasitics?) I wouldn't
really know how to deal with pushing the J310 that high. The
J310 probably comes close to 75ohms input impedance with a near
Idss bias in a common base configuration.

Yes, that's about right for a J310. Although early
Siliconix app notes didn't push even a common-gate
J310 much above 200MHz.

 

[colin]

Hi all,

I've decided to build a TV preamp just for the hell of it. There's one
channel on 850Mhz I can't quite pick up without one, anyway. I've got
three JFETs in my junk box that *might* make the grade at this
frequency, but I'm not sure. The parts I have are the BF244, the J310
and J211. Is there going to be any worthwhile gain from any of them,
even in common-gate mode, at the required frequency, or should I be
better off looking at GaAsFETs?

p.

Ive tried several circuits for making a uhf tv pre amp, some with DG mosfet
and other with low noise 5ghz bjt, the spec sheet for both gave aprox 1db NF
at i beleive 50ohm src impedance, wich i beleive is adding hardly any noise.
i didnt see much diference in noise between the two in circuit although the
bjt gave better gain. bf998/bfe520 etc...

I managed to get the circuit small enough to put it inside a coaxial plug
and solder a coaxial socket onto the cable entry end of it.
its easy to bias it so that the input impedance is close to 50 ohm and also
operating at optimum colector curent for lowest noise at 50ohm source.

However i get a slight problem with either interference pick up or
instability, although for weak signals it still does improve it
considerably, however i think i cld do a lot better now with regard to uhf
inductor design etc.

i gues the optimum would be to design your own antena with a higher output
impedance and directly atatch it to a DG mosfet circuit. but i'm too wimpy
to climb up on the roof every time to fiddle about with it.

Colin =^.^=

 

[John S. Dyson]

Yes, that's about right for a J310. Although early
Siliconix app notes didn't push even a common-gate
J310 much above 200MHz.

Sorry for the brain-fart in extremis -- I did mean common gate
instead of common base . That is what I get for posting too
too late .

John

 

[Paul Burridge]

Ive tried several circuits for making a uhf tv pre amp, some with DG mosfet
and other with low noise 5ghz bjt, the spec sheet for both gave aprox 1db NF
at i beleive 50ohm src impedance, wich i beleive is adding hardly any noise.
i didnt see much diference in noise between the two in circuit although the
bjt gave better gain. bf998/bfe520 etc...

Well a bit of research on the web has indicated that the 3SK88s I have
are basically two bf998 gates in a single package. If so, that would
be handy indeed. Shame I've been unable to track down a datasheet for
the '88s, though. :-(

[...]

i gues the optimum would be to design your own antena with a higher output
impedance and directly atatch it to a DG mosfet circuit.

That's an interesting suggestion!

 

[Paul Burridge]

I STRONGLY suggest using the HP 54143 type PHEMT. It biases similarly
to a BJT (needs a small, 0.5-0.7V positive voltage on the gate), and
can provide wideband low noise at any frequency range that it makes
sense to use.

[snip]
Thanks, John. Noted.

 

[Ken Smith]

Do people still use parametric diode/circulator amps? I remember when
tunnel diode/circulator amps were the rage (sniffles in nostalgia for
tunnel diodes.)

They still are one of the few noise free methods. Obviously they will
only be used where low noise is an absolute must or on Paul B.'s TV.

 

[John Larkin]

I checked the NEC34018 (it looks like the NEC part that seems to be
reasonably stable to work with at 1GHz.) Unfortunately, it has a
gamma-opt of approx 0.6 at about 1GHz with a noise figure of 0.4dB. This
is not a wonderful noise match to 50/75 ohms (but probably not very
bad either.) It appears that the noise voltage on the 34018 is a little
higher than my own personal choice .

I use NE34018's... as switches!

John

 

[John Dyson]

I use NE34018's... as switches!

That is certainly an application where BJTs would be less advantageous.

BTW, I didn't mean to seem like an HP sales person, and the 34018 is
definitely
an adequate (and in fact, stellar) performer, when compared with the old,
traditional
technology like the J310 or MPSH10 .

Even when I lamented the gamma-opt of 0.6 with an NF of 0.4dB, that will
still
probably give approx 1dB NF at 50ohms. That is really is still a very
useful
component.

My admiration of the HP 54143 component resulted from a 'WOW' that I had
several years ago when I noticed it.

For simple, run-of-the-mill preamp applications, where a few dB NF is okay,
I
believe that the Sirenza components can probably do most of what is needed.

It takes significantly MORE complexity when using the fully discrete
components,
but sometimes people really don't need all of the possible performance.

For my own 'benchmark' application (a UHF TV preamp), I'd tend to bias my
circuit choices more for higher IP3 rather than absolute lowest NF. With
the
preamp sitting directly on the antenna (e.g. indoor bow-tie), then any NF
better
than 1dB is gilding the lilly. The ability to handle LOTS of interfering
sources
is much more important.

John

 

[John Larkin]

That is certainly an application where BJTs would be less advantageous.

Even the superfast (like 45 GHz Ft) bipolar transistors are pig-slow
switches, and like to oscillate on the edges to boot. PHEMTs are
beautiful. Fujitsu makes some astounding parts.

BTW, I didn't mean to seem like an HP sales person, and the 34018 is
definitely
an adequate (and in fact, stellar) performer, when compared with the old,
traditional
technology like the J310 or MPSH10 .

My concern about Agilent is that they tend to discontinue parts a lot.

Their enhancement-mode PHEMTs look cool for time-domain use, except
that the input capacitances are huge.

Even when I lamented the gamma-opt of 0.6 with an NF of 0.4dB, that will
still
probably give approx 1dB NF at 50ohms. That is really is still a very
useful
component.

My admiration of the HP 54143 component resulted from a 'WOW' that I had
several years ago when I noticed it.

For simple, run-of-the-mill preamp applications, where a few dB NF is okay,
I
believe that the Sirenza components can probably do most of what is needed.

It takes significantly MORE complexity when using the fully discrete
components,
but sometimes people really don't need all of the possible performance.

Nature seems to have a rule against low nf's in really wideband
circuits. You have to tune the hell out of a PHEMT front-end to get
noise-free voltage stepup into the gate. Most wideband MMICS run
around 3dB nf. I think it's feasible to do a transformer-coupled
low-noise amp, but the achievable bw is pretty low.

Distributed amps might be better, in theory.

In do mostly time-domain, DC-up stuff, so can't use all those nice RF
matching tricks.

John

 

[Rich Grise]

Hi all,

I've decided to build a TV preamp just for the hell of it. There's one
channel on 850Mhz I can't quite pick up without one, anyway. I've got
three JFETs in my junk box that *might* make the grade at this
frequency, but I'm not sure. The parts I have are the BF244, the J310
and J211. Is there going to be any worthwhile gain from any of them,
even in common-gate mode, at the required frequency, or should I be
better off looking at GaAsFETs?

You'd be better off looking at antennas. No matter how good of an
amplifier you have, if you amplify a crappy signal you get an amplified,
crappy signal. A yagi a those frequencies would fit in the palm of your
hand, and would be a nifty little construction project.

Good Luck!
Rich