MMIC amps: Input impedance

[Joerg]

Hello Folks,

The BGA2001 from NXP looks like a good deal at around $0.30/1k. Hard to
justify rolling your own amp at that price, plus it's got a really nice
noise figure of under 1.5dB.

http://www.nxp.com/acrobat_download/datasheets/BGA2001_6.pdf

I am planning to use it well under 100MHz, actually down to maybe
5-10MHz. Looking at figure 12 the input impedance seems to exceed
500ohms down there (they don't show anything below 100MHz). Except for
500MHz its input is pretty far off 50ohms on most bands. Is this normal?

I wonder why they spec input intercept and other things in dBm then. Not
complaining here since a higher Rin actually helps my photodiode app
because I won't need a RF transformer. Just wondering. I have always
rolled my own, first time I am considering a MMIC.

 

[Joop]

Hello Folks,

The BGA2001 from NXP looks like a good deal at around $0.30/1k. Hard to
justify rolling your own amp at that price, plus it's got a really nice
noise figure of under 1.5dB.

http://www.nxp.com/acrobat_download/datasheets/BGA2001_6.pdf

I am planning to use it well under 100MHz, actually down to maybe
5-10MHz. Looking at figure 12 the input impedance seems to exceed
500ohms down there (they don't show anything below 100MHz). Except for
500MHz its input is pretty far off 50ohms on most bands. Is this normal?

I wonder why they spec input intercept and other things in dBm then. Not
complaining here since a higher Rin actually helps my photodiode app
because I won't need a RF transformer. Just wondering. I have always
rolled my own, first time I am considering a MMIC.

Well, this particular MMIC is not specified as "a generic 50 ohm
building block" as some others are. I guess the dBm simply refers to
power level in general, not usual RF dBm(50) or telco dBm(600).
That would also mean that you might need more input voltage level to
reach the same input power in dBm than you would if the preceeding
stage is designed for 50 ohm load.

But even a BGA616 which IS specified as "Cascadable 50 ohm gain block"
does not always behave as close to 50 ohm in/out in my simulations
(SPICE is available). Almost a factor of two off can take place
depending on biasing and input impedance.

Joop

 

[Joop]

....

Almost a factor of two off can take place
depending on biasing and input impedance.

That came out wrong. I mean the input/output impedance varies with the
impedance of the stage on the other side of the device. Supply biasing
through resistor only or resistor/inductor combi has influence as
well.

 

[Joop]

Hmm, I also looked at the demo board description.
http://www.nxp.com/acrobat_download/applicationnotes/LNA900MHZ.pdf

That seems to suggest both input and output would be about 50 ohm
(with SWR 1:1.5).

 

[Rene Tschaggelar]

Hello Folks,

The BGA2001 from NXP looks like a good deal at around $0.30/1k. Hard to
justify rolling your own amp at that price, plus it's got a really nice
noise figure of under 1.5dB.

http://www.nxp.com/acrobat_download/datasheets/BGA2001_6.pdf

I am planning to use it well under 100MHz, actually down to maybe
5-10MHz. Looking at figure 12 the input impedance seems to exceed
500ohms down there (they don't show anything below 100MHz). Except for
500MHz its input is pretty far off 50ohms on most bands. Is this normal?

I wonder why they spec input intercept and other things in dBm then. Not
complaining here since a higher Rin actually helps my photodiode app
because I won't need a RF transformer. Just wondering. I have always
rolled my own, first time I am considering a MMIC.

Joerg,
they want you to AC couple anyway. So from the reverse
biased photodiode an inductor to GND and a cap to the
input.

Rene

 

[Joerg]

...



That came out wrong. I mean the input/output impedance varies with the
impedance of the stage on the other side of the device. Supply biasing
through resistor only or resistor/inductor combi has influence as
well.

Thanks, Joop. I'll use a resistor/inductor series combo since the needed
frequency range is wide and there is that dreaded inductor series resonance.

The output also looks a bit high-Z-ish so I'll have to spring for a
BFS17 follower there. The BGA616 looks better but not perfect either and
its noise figure isn't so hot.

 

[Joerg]

Joerg,
they want you to AC couple anyway. So from the reverse
biased photodiode an inductor to GND and a cap to the
input.

Rene

Yes, Rene, that is about how I plan to do it. Looking at the output
chart S22 it seems that this MMIC (and lots of others) doesn't really
like to drive a 50ohm load. More like several hundred ohms. Guess I'll
spring for the usual BFS17 follower. If it was a really high volume
product I'd also do the MMIC part with a couple transistors but not in
this case.

 

[John Larkin]

Hello Folks,

The BGA2001 from NXP looks like a good deal at around $0.30/1k. Hard to
justify rolling your own amp at that price, plus it's got a really nice
noise figure of under 1.5dB.

http://www.nxp.com/acrobat_download/datasheets/BGA2001_6.pdf

I am planning to use it well under 100MHz, actually down to maybe
5-10MHz. Looking at figure 12 the input impedance seems to exceed
500ohms down there (they don't show anything below 100MHz). Except for
500MHz its input is pretty far off 50ohms on most bands. Is this normal?

I wonder why they spec input intercept and other things in dBm then. Not
complaining here since a higher Rin actually helps my photodiode app
because I won't need a RF transformer. Just wondering. I have always
rolled my own, first time I am considering a MMIC.

Looks nasty. It's not unconditionally stable, and has weird impedances
vs frequency. Whenever you see a mmic with a nf much below 3 dB, you
can be certain that it's for a narrowband, tuned application. It looks
sort of like a transistor with a bias generator.

The datasheet says that the nf is measured at optimum match, which I
interpret as "tuned."

The Sirenza SGA-3586 SiGe part is about the lowest-noise, wideband,
"true 50 ohm mmic" around.

John

 

[maxfoo]

Looks nasty. It's not unconditionally stable, and has weird impedances
vs frequency. Whenever you see a mmic with a nf much below 3 dB, you
can be certain that it's for a narrowband, tuned application. It looks
sort of like a transistor with a bias generator.

The datasheet says that the nf is measured at optimum match, which I
interpret as "tuned."

The Sirenza SGA-3586 SiGe part is about the lowest-noise, wideband,
"true 50 ohm mmic" around.

John

Check this one out cross-ref to sga-3586, specs look a little better too.

http://www.hittite.com/product_info/product_specs/amplifiers/hmc481mp86.pdf

 

[Joerg]

Looks nasty. It's not unconditionally stable, and has weird impedances
vs frequency. Whenever you see a mmic with a nf much below 3 dB, you
can be certain that it's for a narrowband, tuned application. It looks
sort of like a transistor with a bias generator.

I think that's what it is, a transistor with a bias circuit. I hope it's
stable if I provide a nice 200ohm load. The plan was to follow it with a
buffer so whatever else is connected (not under my control) won't upset
the apple cart.

The datasheet says that the nf is measured at optimum match, which I
interpret as "tuned."

But how could they call it MMIC then? Marketeers at work?

The Sirenza SGA-3586 SiGe part is about the lowest-noise, wideband,
"true 50 ohm mmic" around.

It looks nice but the S11 and S22 also take a nose dive around 500MHz.
I'll be staying under 100MHz. What amazes me is how cheap SiGe parts are
now. There goes the art of designing our own amps...

 

[Joerg]

Check this one out cross-ref to sga-3586, specs look a little better too.

http://www.hittite.com/product_info/product_specs/amplifiers/hmc481mp86.pdf

Indeed. Thanks. But that poor little thing has to dissipate over 400mW.
Probably causes a good blister if anyone touches it ;-)

 

[John Larkin]

I think that's what it is, a transistor with a bias circuit. I hope it's
stable if I provide a nice 200ohm load. The plan was to follow it with a
buffer so whatever else is connected (not under my control) won't upset
the apple cart.



But how could they call it MMIC then? Marketeers at work?



It looks nice but the S11 and S22 also take a nose dive around 500MHz.
I'll be staying under 100MHz. What amazes me is how cheap SiGe parts are
now. There goes the art of designing our own amps...

Eventually nasty people (like a certain JT who I won't identify)
integrate anything that's fun, so it's not worth designing ourselves
any more. So we have to move up the abstraction stack.

John

 

[John Larkin]

Check this one out cross-ref to sga-3586, specs look a little better too.

http://www.hittite.com/product_info/product_specs/amplifiers/hmc481mp86.pdf

The Sirenza is a lower power part and the NF is a bit better. But
Hittite is making some kickass parts lately, like their '465 20 GHz
distributed amps and things.

I like the Sirenza because it's really 50 ohms in, and in fact you can
tune the input impedance by varying the bias current, so it's nice for
time-domain signals at the end of a coax. It's a very nice
medium-speed mmic, although is is a bit fragile as regards input zaps.

John

 

[Joerg]

The Sirenza is a lower power part and the NF is a bit better. But
Hittite is making some kickass parts lately, like their '465 20 GHz
distributed amps and things.

A bit less power would be great, I don't like stuff in my designs that
get too hot to the touch. Except, of course, tubes

I like the Sirenza because it's really 50 ohms in, and in fact you can
tune the input impedance by varying the bias current, so it's nice for
time-domain signals at the end of a coax. It's a very nice
medium-speed mmic, although is is a bit fragile as regards input zaps.

Here the only way to create a zap would be a really bright flash. And
when that one comes then I guess the performance of this amp will be of
secondary concern...

 

[Joerg]

Eventually nasty people (like a certain JT who I won't identify)
integrate anything that's fun, so it's not worth designing ourselves
any more. So we have to move up the abstraction stack.

There seem to be some more good analog chip guys out there. But even 30c
a pop is too much at times. Not in this app but when it's a mass product
I'll always be designing around a 5c RF transistor. Especially in cases
when, as you had mentioned, the MMIC isn't much more than a glorified
transistor with a bias. You can get a real noise figure of under 1.5dB
for less. Just like with burgers, if you make your own from ground beef
they cost less and taste better (we don't do them any other way).

 

[maxfoo]

The Sirenza is a lower power part and the NF is a bit better. But
Hittite is making some kickass parts lately, like their '465 20 GHz
distributed amps and things.

I like the Sirenza because it's really 50 ohms in, and in fact you can
tune the input impedance by varying the bias current, so it's nice for
time-domain signals at the end of a coax. It's a very nice
medium-speed mmic, although is is a bit fragile as regards input zaps.

John

I like a cheap Mini-circuit's ERA-33SM when I need to boost an rf signal.

 

[Joerg]

I like a cheap Mini-circuit's ERA-33SM when I need to boost an rf signal.

I'll check that out, Max. Haven't looked at Mini-Circuits much because
they are often single-sourced. But then again they have never
disappointed me in my whole career and always delivered (mixers etc.).
Unlike another company where the name also starts with "M" ;-)

 

[John Larkin]

I'll check that out, Max. Haven't looked at Mini-Circuits much because
they are often single-sourced. But then again they have never
disappointed me in my whole career and always delivered (mixers etc.).
Unlike another company where the name also starts with "M" ;-)

The ERA-series parts are fine. They are absolutely stable and fairly
rugged for an exotic part. I use them as fast pulse amps and as
sub-100 ps photodiode amps, although I do have a fiber-coupled optical
trigger pulse that's measured in watts! Their input impedances tend to
run low, in the 30's maybe, when the output is matched. And they do
run hot, although InGaAs melts at a higher temp than silicon!

All the classic darlington mmics have a Zin that depends strongly on
output loading. They are very handy parts.

Mini-Circuits now has some faster, 8-GHz range, parts, and some nice
higher-power SOT-89's. W-J makes nice SOT-89 mmics, too, probably the
highest power ones around.

John

 

[Joerg]

The ERA-series parts are fine. They are absolutely stable and fairly
rugged for an exotic part. I use them as fast pulse amps and as
sub-100 ps photodiode amps, although I do have a fiber-coupled optical
trigger pulse that's measured in watts! Their input impedances tend to
run low, in the 30's maybe, when the output is matched. And they do
run hot, although InGaAs melts at a higher temp than silicon!

Sub-100psec? Wow, that is performance. I just don't like semiconductor
stuff that gets really hot.

All the classic darlington mmics have a Zin that depends strongly on
output loading. They are very handy parts.

A little higher Z-in would be good in my case. Although I am afraid I am
going to be flooded with DFB noise but that's what I can't know until
the circuit runneth.

Have you seen pathologies or instabilities with the darlingtons? The old
BGA2001 still looks enticing with it's high Z-in. Intercept is a bit
marginal because it's a low current device.

Mini-Circuits now has some faster, 8-GHz range, parts, and some nice
higher-power SOT-89's. W-J makes nice SOT-89 mmics, too, probably the
highest power ones around.

Some of the SOT-89 parts I looked at appeared to be able to unsolder
themselves...

 

[John Larkin]

There seem to be some more good analog chip guys out there. But even 30c
a pop is too much at times. Not in this app but when it's a mass product
I'll always be designing around a 5c RF transistor. Especially in cases
when, as you had mentioned, the MMIC isn't much more than a glorified
transistor with a bias. You can get a real noise figure of under 1.5dB
for less.

My observation is that if you want a NF below 3 dB, from a 50 ohm
signal, you've gotta tune the input. Some of the NEC phemts get below
1 dB, since the Cin is very low and the matching network can give a
bunch of free voltage gain.

Just like with burgers, if you make your own from ground beef
they cost less and taste better (we don't do them any other way).

I just had a burger with Mo at the Zuni Cafe, and I'd be astonished if
any mere mortal can make their own burger that good. They serve it on
fresh focaccia with homemade pickles and pickled onions and homemade
aioli. $12.

John