Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Selecting RF FET for low frequency LNA

Status
Not open for further replies.

mtwieg

Advanced Member level 5
Joined
Jan 20, 2011
Messages
3,616
Helped
1,272
Reputation
2,550
Reaction score
1,273
Trophy points
1,393
Activity points
27,263
A while back a made a thread about my attempts to get the specified NF out of a AD8331 LNA. That hasn't worked out (despite consulting with a AD application engineer), so I guess I'm going to need to roll my own discrete LNA.

My application is narrowband (maybe 200KHz) with a center frequency between 7-15MHz (not sure yet). I'm hoping for a NF less than 1dB from a 50 source (though I will need some impedance transformation at the input).

Right now I just want to get an idea of what technologies would give good NF at such frequencies. I'm looking at GaAs Fets, dual gate silicon FETs, HEMTs, etc, but none of them are specced below 100MHz. Also before anyone suggests it, I don't have access to good simulation software like ADS or Ansoft.

Also, I've been preemptively investigating methods for noise characterization, but am still unsure how to approach this. I have access to a good signal analyzer with a NF measurement setup, and a decent network analyzer. Do you think it should be necessary to measure the input referred noise currents and voltages, then calculate my noise impedance and match to that? Or should I just try and find the optimal reflection coefficient by trial and error?

Thanks in advance.
 

vfone

Advanced Member level 5
Joined
Oct 10, 2001
Messages
5,347
Helped
1,553
Reputation
3,109
Reaction score
1,166
Trophy points
1,393
Activity points
33,882
Some microwave GaAs FETs have greater NF at lower frequencies (below 100MHz) than at higher frequencies, due to increase of the noise resistance Rn.
Perhaps there are plenty of BJT's with NF below at 1dB at 15MHz, but you can try using the dual-gate MOSFET BF998, which from my knowledge has about 0.7dB NF at HF frequencies (up to 30MHz).

https://www.nxp.com/documents/data_sheet/BF998.pdf
 
  • Like
Reactions: mtwieg

    mtwieg

    Points: 2
    Helpful Answer Positive Rating

BigBoss

Advanced Member level 5
Joined
Nov 17, 2001
Messages
5,112
Helped
1,529
Reputation
3,056
Reaction score
1,376
Trophy points
1,393
Location
Turkey
Activity points
30,746
If you can find SiGe, use it no doubt... ( Infineon..)
Otherwise use BJT but never use GaAs or compound semicon. discretes..They are very noisy below 100-150MHz
 
  • Like
Reactions: mtwieg

    mtwieg

    Points: 2
    Helpful Answer Positive Rating

mtwieg

Advanced Member level 5
Joined
Jan 20, 2011
Messages
3,616
Helped
1,272
Reputation
2,550
Reaction score
1,273
Trophy points
1,393
Activity points
27,263
Perhaps there are plenty of BJT's with NF below at 1dB at 15MHz, but you can try using the dual-gate MOSFET BF998, which from my knowledge has about 0.7dB NF at HF frequencies (up to 30MHz).

https://www.nxp.com/documents/data_sheet/BF998.pdf
Okay the BF998 looks like a possibility. I'm not really interested in using one of the gates for AGC, so should I just use both gates in parallel or tie one to a fixed bias voltage?

If you can find SiGe, use it no doubt... ( Infineon..)
Otherwise use BJT but never use GaAs or compound semicon. discretes..They are very noisy below 100-150MHz
Huh, never thought of SiGe. The only thing I can see available on digikey is the BFP 640, is that a good candidate?
 

BigBoss

Advanced Member level 5
Joined
Nov 17, 2001
Messages
5,112
Helped
1,529
Reputation
3,056
Reaction score
1,376
Trophy points
1,393
Location
Turkey
Activity points
30,746
Huh, never thought of SiGe. The only thing I can see available on digikey is the BFP 640, is that a good candidate?

Yes, it may be used..You'll obtain probably 0.15dB NF or less at those frquencies.
It's good enough..
 

RCinFLA

Advanced Member level 2
Joined
Aug 31, 2010
Messages
606
Helped
187
Reputation
378
Reaction score
189
Trophy points
1,323
Activity points
5,788
Not sure what your application is but if for communication applications at thoughs frequencies the background noise will swamp out a 1 db noise figure receiver so the noise figure of the receiver does not need to be that low.

1 db NF is noise temp of 78 degs K, typical atmospherics and interference in the 7-15 MHz range is 500-1500 degs K.

Intermod performance and dynamic range are more important then NF in this RF comm range.
 

mtwieg

Advanced Member level 5
Joined
Jan 20, 2011
Messages
3,616
Helped
1,272
Reputation
2,550
Reaction score
1,273
Trophy points
1,393
Activity points
27,263
Yes, it may be used..You'll obtain probably 0.15dB NF or less at those frquencies.
It's good enough..
Wow, that would be great. What kind of noise impedances should I expect to need for SiGe? I'm having trouble finding relevant literature on the topic (for low frequency anyways).

Not sure what your application is but if for communication applications at thoughs frequencies the background noise will swamp out a 1 db noise figure receiver so the noise figure of the receiver does not need to be that low.

1 db NF is noise temp of 78 degs K, typical atmospherics and interference in the 7-15 MHz range is 500-1500 degs K.

Intermod performance and dynamic range are more important then NF in this RF comm range.
My application isn't communications; but rather NMR coil reception. So throwing the whole thing in a shield is no problem for me, fortunately.
 

mtwieg

Advanced Member level 5
Joined
Jan 20, 2011
Messages
3,616
Helped
1,272
Reputation
2,550
Reaction score
1,273
Trophy points
1,393
Activity points
27,263
I got the transistors in today and threw together a simple circuit with the BF998. I took a shot in the dark and went for a match impedance of 2Kohm, and I'm already measuring NF of between 1.0-1.5dB, with a gain of around 17dB, at frequencies between 17-25MHz. Great so far. The trouble is going to be when I want to drop the frequency down to around to 10MHz, since it will be hard to get inductors with good Q. Also my noise source is only rated down to about 10MHz, so it might lose accuracy around there. I'll keep updating my progress though, since this may be of interest to other people in the future.
 

mtwieg

Advanced Member level 5
Joined
Jan 20, 2011
Messages
3,616
Helped
1,272
Reputation
2,550
Reaction score
1,273
Trophy points
1,393
Activity points
27,263
So today I'm playing with the SiGe BFP 640, and I'm getting poor results so far. The circuit is built in much the same way as my circuit with the BF998 (except for the biasing of the input). The most strange thing is that if I bias the collector current much below 5ma, it starts a strange relaxation on the emitter and collector (though not on the base), with a frequency ranging from 800KHz to 2MHz. I think it might be the CE junction breaking down then recovering repeatedly, but I don't see why this should be the case since the Vce never reaches 4v (it's usually under 3V).

When I increase the bias current to around 10ma, it becomes stable, but the overall gain seems pretty poor (15dB max). And the measured noise figure is also bad (4-5dB minimum). I would take the bias current up higher, but the datasheet implies that above 10ma the NF degrades sharply.

Any advice would be appreciated.
 

vfone

Advanced Member level 5
Joined
Oct 10, 2001
Messages
5,347
Helped
1,553
Reputation
3,109
Reaction score
1,166
Trophy points
1,393
Activity points
33,882
Can you post the schematic you use? At these low frequencies (an bias current) the gain of SiGe BFP640 should be very high (>30dB) and your LNA is possible to oscillate.
 

tsvetelin_velkov

Member level 4
Joined
Jun 1, 2010
Messages
77
Helped
27
Reputation
54
Reaction score
27
Trophy points
1,298
Location
Bulgaria
Activity points
2,009
Hi,

What surprise me is that you are using high frequency transistors, for a low frequency application, i.e. for about 20MHz. Maybe you can try using just low noise transistors. The other thing is how much amplification do you really need, i.e. if you have a 15 to 17 dB amplification, why don't you try to put an attenuator on the output, this will lower the signal, but will also lower the noise.

I have also a small comment about the simulation software. I think you can look at this ( LTspice is currently free ) :
 

Attachments

  • RF_English_01_final.pdf
    3.7 MB · Views: 38

biff44

Advanced Member level 5
Joined
Dec 24, 2004
Messages
4,942
Helped
1,366
Reputation
2,728
Reaction score
1,044
Trophy points
1,393
Location
New England, USA
Activity points
37,131
I would either use a silicon bipolar, or use an instrumentation op amp with a low input noise voltage.
 

mtwieg

Advanced Member level 5
Joined
Jan 20, 2011
Messages
3,616
Helped
1,272
Reputation
2,550
Reaction score
1,273
Trophy points
1,393
Activity points
27,263
Hm? Seems my thread has been brought back...
Hi,

What surprise me is that you are using high frequency transistors, for a low frequency application, i.e. for about 20MHz. Maybe you can try using just low noise transistors.
Only high frequency transistors seem to be specified for low noise. I've looked for lower frequency parts, but they don't tend to specify any noise characteristics.
The other thing is how much amplification do you really need, i.e. if you have a 15 to 17 dB amplification, why don't you try to put an attenuator on the output, this will lower the signal, but will also lower the noise.
15dB is adequate, though 20dB would be nice. As for putting an attenuator on the output, that certainly wouldn't help. I'm concerned about my SNR, not the actual noise on the output.
I have also a small comment about the simulation software. I think you can look at this ( LTspice is currently free ) :
Yeah, I found this back when I was designing the thing and it was somewhat helpful for simulating noise figures, but ultimately I couldn't replicate its results. It always gave absurdly optimistic NF (like 0.3dB). But it was useful for biasing and stability analysis.

I would either use a silicon bipolar, or use an instrumentation op amp with a low input noise voltage.
I tried a couple bipolar transistors, but they always had stability issues, even when I tried cascoding. The BF998 on the other hand never had any issues and gave acceptable results.

Here's a schematic of what I've got working right now:


Its gain is about 16.5dB and its noise figure measures at ~0.9dB, tuned for 10MHz. I believe the NF is limited by the Q of the tuning inductor. I could work on lowering it down to 0.5dB, but it's good enough to use right now.
 

tsvetelin_velkov

Member level 4
Joined
Jun 1, 2010
Messages
77
Helped
27
Reputation
54
Reaction score
27
Trophy points
1,298
Location
Bulgaria
Activity points
2,009
mtwieg,

Your results are really good! Congratulations! BF998 is a really good transistor, your results confirm that once again.

May I just ask, what is the current application of this LNA ? Radio astronomy ?

Why not using a different schematic for the drain :

+12V -> R1 -> 2.2uF -> L3 -> 5nF -> Q1 drain

Edited:

I just did a quick check with datasheet of BF998, it is given that NF at 45 MHz is 2.8 dB, how is it possible that you measure 0.9 dB at 10 MHz? I would expect that with lowering the frequency the NF goes high.

Datasheet from Infineon : https://de.sitestat.com/infineon/in...51041&fileId=db3a30431441fb5d011493688fd31042
 
Last edited:

mtwieg

Advanced Member level 5
Joined
Jan 20, 2011
Messages
3,616
Helped
1,272
Reputation
2,550
Reaction score
1,273
Trophy points
1,393
Activity points
27,263
May I just ask, what is the current application of this LNA ? Radio astronomy ?
It's meant for use in small NMR receivers.
Why not using a different schematic for the drain :

+12V -> R1 -> 2.2uF -> L3 -> 5nF -> Q1 drain
Sure, that bypassing scheme would probably work too, though you'd still need some impedance on the drain.
Edited:

I just did a quick check with datasheet of BF998, it is given that NF at 45 MHz is 2.8 dB, how is it possible that you measure 0.9 dB at 10 MHz? I would expect that with lowering the frequency the NF goes high.

Datasheet from Infineon : https://de.sitestat.com/infineon/in...51041&fileId=db3a30431441fb5d011493688fd31042
[/quote]It certainly will start to increase at some point. In general I've seen that most transistors have a bathtub shaped curve of NF vs frequency. But that 2.8dB number at 45MHz seems way too high... I know that when I was first playing with it, I had it tuned to 20MHz at one point and its NF was around 1.5dB.

edit: I've also seen that switching in different BF998s in the same circuit gives significant variation in gain (by 2dB sometimes) and up to 0.5dB in NF. So those datasheet specs are likely worst case.
 

tsvetelin_velkov

Member level 4
Joined
Jun 1, 2010
Messages
77
Helped
27
Reputation
54
Reaction score
27
Trophy points
1,298
Location
Bulgaria
Activity points
2,009
NMR receiver, you mean for MRI application ?

Why not to cool down the transistor? Maybe you can try with cooling spray, just cool down as in the circuit, let's say to -40 degC and measure NF again. As far as I know the nuclear magnetic resonance, cooling is widely used in those machines.

The other possible option is vacuum tube, maybe there is a ready application for LNA with valve for NMR ?
 
Last edited:

rfmw

Advanced Member level 2
Joined
May 10, 2001
Messages
545
Helped
93
Reputation
190
Reaction score
50
Trophy points
1,308
Activity points
3,978
Might be worth trying soldering another BF998 in parallel to the one in your circuit. At VHF the NF is further reduced using parallel transistors. Just another idea to try lowering the NF...
 

mtwieg

Advanced Member level 5
Joined
Jan 20, 2011
Messages
3,616
Helped
1,272
Reputation
2,550
Reaction score
1,273
Trophy points
1,393
Activity points
27,263
NMR receiver, you mean for MRI application ?

Why not to cool down the transistor? Maybe you can try with cooling spray, just cool down as in the circuit, let's say to -40 degC and measure NF again. As far as I know the nuclear magnetic resonance, cooling is widely used in those machines.

The other possible option is vacuum tube, maybe there is a ready application for LNA with valve for NMR ?
This is meant for small scale, low field NMR (not imaging), using permanent magnets, so no cryogens available. And even in clinical superconducting magnet scanners, none of the electronics (except for those monitoring the cryogens) are supercooled (otherwise we'd never be able to tinker with them without quenching the magnet).

---------- Post added at 08:33 ---------- Previous post was at 08:32 ----------

Might be worth trying soldering another BF998 in parallel to the one in your circuit. At VHF the NF is further reduced using parallel transistors. Just another idea to try lowering the NF...
Yeah, I've seen this done and might try it myself if I run into a wall and still need lower NF.
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Top