Connect with us

Norton noiseless feedback amp calculation

Discussion in 'Electronic Design' started by Jake, May 31, 2006.

Scroll to continue with content
  1. Jake

    Jake Guest

    +10v
    |
    T1 |
    |
    o 1 o N |
    UUU UUUUUUUUUU
    | | | |
    ---| | |----------------- Out
    | \R1 | | |
    | / \ e / c |
    --- \ \ / |
    C1--- | ----- |
    | | | b |
    | |-R2--|-----R3------
    | | |
    ====== In
    Gnd

    The configuration is much the same as the Norton amplifier with the
    exception of the input being the base of the transistor.

    Assume Ic = 10 ma or in the case of a dual gate FET Id = 8 ma

    Since the {1} feedback winding is connected to the emitter the transformer
    {T1} has defined impedance's and gain. This is compounded by the summation
    of Ie, Ib and Ic input and feedback summation with Re.

    How does one calculate this operating point reflected collector impedance
    for the output and the stage gain?

    Obviously the output may be taken from a tap on the collector winding ie
    1:M:N as well.

    Can R1, not decoupled, be used to define the output impedance better? I
    would think so but how is Zout calculated and the gain change if any.

    A net search only finds the Norton configuration but I think Ulrich Rhode
    covered this configuration in an old Ham Radio publication under noiseless
    feedback.

    Any help will be much appreciated as this is beyond my capability for both
    transistor and dual gate FET Gm = 18mS

    I wanted to use this to replace the first RF amplifier for a Yaesu FRG7
    receiver. A dual gate FET is needed because of direct connection to a
    preselector tuned circuit. There is a need to match to the 560 ohm filter
    input impedance. It has many other applications as well. (Wideband flat
    gain and low noise.) Transistor will be a BFR91A and fet BF966 or similar

    Thanks
    Peter
     
  2. Jake

    Jake Guest

    I was not looking for a discussion to the n'th degree and defined down to
    pin point accuracy. That thumbsuck is fine and a whole lot better than
    nothing.

    Some pointers please, a few words to put me on the right track and save
    hours of messing about by being able to ball park the behaviour.

    Thanks
    Peter
     
  3. Fred Bloggs

    Fred Bloggs Guest

    From what I recall of the original Norton article, the main thing is
    noiseless input impedance termination in addition to gain setting. The
    circuit calculation is straightforward. Break it down into the hybrid-pi
    corresponding to your operating point- that would be the operating point
    computed from the equivalent DC circuit. What is your question anyway?
     
  4. Jake (Peter) wrote...
    Write out the equations and solve them. Routine feedback.
    OK. Stop complaining.

    Look up Dallas Lankford's articles (Microwave Journal, May 1975)
    and patents (3,426,298; 3,624,536; and 3,891,934) on common-base
    transformer-feedback (CBTF) amplifiers. Good for low-Z inputs.
    http://www.kongsfjord.no/ http://www.kongsfjord.no/dl/dl.htm
    http://www.kongsfjord.no/dl/Amplifiers/Ultralinear 2N5109 And 2N3053 Amplifiers.pdf

    There's also been some good discussion here on s.e.d. And check
    Lepaisant in RSI 63-3 (1992). "Low-noise preamplifier with input
    and feedback transformers for low source resistance sensors" by
    J. Lepaisant, M. Lam Chok Sing, and D. Bloyet
     
  5. Jake

    Jake Guest

    If I could I would have and not wasted your valuable time and sarcastic
    advice.

    I have the patents and these deal with a common base amplifier and feedback
    via the transformer. Sorry I just don't have the knowledge to write out
    new equations where the transformer is terminated in the emitter impedance
    and combine the feedback.

    While Re maybe 25/Ima this must be combined with the feedback.

    How?
    or Zi in = Zout but my use required a high input impedance. The feed point
    of those amplifiers ia also the junction of the emitter winding and Rs
    which I have decoupled.

    The above deals with the case where Rout is determined by Rin or vice
    versa. m*m/(m+n+1)Rs = Rl

    The decoupled Rs common emitter is determined by other factors such as Re
    but that is not all.

    I would also be interested to see what then net effect of additional
    degenerative feedback is on the comon emmitter circuit.
    Sorry this is Africa..... If it is not available on the net I am not going
    to get it in a hurry or without traveling some distance and begging to use
    a U library.
     
  6. Jake

    Jake Guest

    snip

    Correct but the original is a common base with signal injection at the
    emitter winding and external resister junction. Rin determines Rout.

    I have that point decoupled and the signal is injected at the base or
    source as the case may be.

    The use I wanted was for a SW receiver RF amplifier replacement. The
    original design dictates how this will be done. The input is from the
    preselector tuned circuit so this can't be loaded or Q will be reduced. A
    dual gate FET was originally used with Rd = 560 ohms. The output must
    match a filter Zin = 560 ohms. Freq 0.5MHz..30MHz.

    It is also such a useful building block that I can think of at least a few
    other uses even as an output buffer because the transistor version looks
    like low output impedances are quite easy to achieve ie 50 ohms with
    reasonable gain.

    I have two old valve signal generators that desperately need a 50 ohm
    output :)

    It's not for production because I can see that Zo will vary with the
    operating point and that would need to be stabilised at the very least.

    My apology not all are blessed with a higher education and I am totally
    self taught. Hybrid Pi is Greek to me ;-0

    Regards
    Peter
     
  7. Leon

    Leon Guest

    Win's been very helpful to you and I can't see any sarcasm in his
    response - look it up in a dictionary. He was probably niggled by your
    apparent inability to search for relevant stuff for yourself.

    Leon
     
  8. Jake

    Jake Guest

    If I had not done any research and I knew how to model the circuit I would
    not have asked. That much is obvious to anyone who reads my original post.
    Win made the incorrect assumption I was being lazy and got the correct
    response for it. He also incorrectly assumed I had the skills to remodel
    the circuit and I have already stated I did not.

    Possibly I was niggled by his inability to read and understand, for that I
    apologise. His references that I can obtain all point to the common base
    version because I have them already.

    The only reference I can remember to the common emitter is that of Ulrich
    Rhode 1980s Ham Radio article and I don't have a copy and have no source of
    obtaining it.

    I would appreciate some help, not be told to do what I have done or have
    stated I can't do. All the references I can find deal with the common base
    version as detailed in my reply. An available reference to the common
    emitter version would be nice or some help in remodelling.

    I appreciate that the patent papers show a specific application and show
    the math but that is to my mind significantly different (Rin = Rout is not
    device and operating point determined) to what I wanted. If I knew how to
    rearrange it or change it I would have done so and not wasted everyone's
    time. Originally clearly stated.

    I have monitored this group for some 20 years and have learnt much from
    that. I don't ask questions without at least making a large attempt and
    effort to answer it for myself.

    All I wanted was a bit of help from those far more knowledgeable than
    myself.

    Regards
    Peter
     
  9. Fred Bloggs

    Fred Bloggs Guest

    http://ece-www.colorado.edu/~bart/book/book/chapter5/ch5_6.htm
     
  10. Jake wrote...
    Well, that what the rest of us have to do. It always amazes
    me how some folks who are asking for our help, and want us to
    spend our time giving it, can be so demanding, with details
    of what they are or are not willing to do on their part.

    Anyway, I've added the article, Lepaisant_preamp-xfmrs_RSI.pdf
    on my wesbsite, at http://www.picovolt.com/win/elec/articles/

    As for the technique, it's simply substituting a transformer,
    with its noiseless turns ratio, for resistors in a feedback
    circuit. You can use the technique for all kinds of circuits,
    but its best value as a low-noise technique comes when you're
    working with very low impedances. That's because if resistors
    are used, they must have absurdly-low values. For example, if
    you're making a 100 pV/root-Hz amplifier, you may need to use
    feedback and bias-setting resistors, etc, with values that are
    under 0.6 ohms, to keep their Johnson noise from ruining your
    amplifier. On the other hand, if you use a transformer, you
    just have to keep the winding resistance well under 0.6 ohms.

    OTOH, if you're making a high-impedance amplifier, then it's
    likely you won't need the transformer's low noise-resistance.
     
  11. Winfield Hill wrote...
    [ snip ]
    Now I remember, check out the s.e.d. discussions about
    Jeroen Belleman's amplifier designs, and his web page,
    http://jeroen.home.cern.ch/jeroen/tfpu/LNA.shtml
     
  12. Jake

    Jake Guest

    From: Winfield Hill - view profile
    The feeback is Vo/N
    The sum is Vi-Vo/N
    -Vo is Hfe(Vi-Vo/N)
    Vo/Vi = Hfe/(1+Hfe/N)

    So for N = 10:1 and Hfe = 60
    Vo/Vi = 8.57
    Not quite applicable as can be seen by the difference in the circuit shown

    Zo above is determined almost completely by the impedence seen by the
    feedback winding.
    Correct this is essentially the same but no treatment is given for
    determining Zo.

    Without feed back a signal injected at the feedback point would see Re/10mA
    Thus Zo is reduced from N*N*Re/Ic(mA) by the feedback
    Or for the figures I gave Zo < 100*2.5 = 2K5 ohms

    Intuition says its closer to 50..60 ohms.

    I can't figure how to calculate it.

    Regards and thanks

    Peter
     
  13. Jake wrote...
    Re/10mA, what's that? Do you mean V_T/10mA?

    If you were using a BJT, where re = kT/qIc = 2.5 ohms
    at 10mA, there are several ways to show Zo = N^2 higher,
    or 250 ohms. I measured 150 ohms in a spice analysis,
    using a bc547. For a FET, use 1/gm at the operating
    current. Zo will likely be 5 to 20x higher, or as high
    as 2.5k, etc. You may want to add an emitter follower.
    Explain your intuition.
    You're on the right track if you use 1/gm, rather than hfe
    in your thinking. hfe has little effect other than Zin.

    A month has gone by, why not make one and measure it?

    BTW, did you get all the articles, patents, and web-links
    that I referenced, and even posted for you on my website
    last June 3rd in this thread? Any thanks due? You think?
     
  14. Jake

    Jake Guest

    Eek that should be 250 ohms. My apology.
    Sorry I see I made several typo mistakes, should just be Re derived from
    the 10mA IC ie 2.5 ohms.
    For Zo = 50 ohm with a follower (BJT) implies a IC of 1 mA or less. Is
    that a good idea with an amplifier that needs an IP3 of +20..+30dBm?

    However in this application I required a Zo = 250 ohms and gain in the
    region of 20dB from a FET.

    It is possible to tap the output winding at any integer value between 1 and
    N. Zo of 450 and 200 could also just as easily be handled by a simple
    small additional transmission line transformer on a ferrite bead. In most
    cases Ic can be varied by a factor 2 ie 7 to 14mA to acheive a specific Zo.

    My original application dictated a minimun of components which had to be
    retro fitted to an existing dual gate mosfet RF amplifier feeding a 560 Ohm
    Zin filter. Zin had to be high so as not to degrade a resonant tuned
    circuit.

    But it is an interesting circuit that I wanted to know more about, for
    knowledge and who knows.

    If Zo can be adjusted by IC then small changes are possible for a one off.
    For production it would be a nightmare with unselected FETS.

    Low Zin (BJT) can be set with C-B voltage feedback resister. Here Zin
    would be Rfb/N.

    A test of BFR91A @ 11.4mA and N = 38:3 Emitter resister 56 decoupled by
    47nF I measured a Zo of 56 ohms. Vo/Vin = 6. Sorry the frequency was not
    noted but it is obvious there was some series feedback via the decoupling
    capacitor about 7 ohms worth implying a frequency of about 450KHz.
    Gut feel ;-)
    OK but how does the feedback influence that? I dont know if I have a
    mental block on this but no amount of looking at anything has turned the
    light on. All I really want is a better understanding and throwing high
    level papers at me will not give that. Consider giving a caveman a match
    patent is not going to help him start a fire.
    I did.

    Measure it?

    I wanted a ballpark method to cut down on the iterations of winding small
    ferrite beads and out of general interest.

    I tried a variable resistor (cermit trimmer) across the output winding,
    isolated by a capacitor 0,1uF and adjusted until I had half the unloaded
    voltage swing (600mV) out. Then measured the resistance. Was this what
    you had in mind? Vector impedance measuring equiment does not just ly
    around and the last I had access to was 20 plus years ago.

    For the FET I got 560 ohms or close enough. Device was a BF966 @ 7mA and N
    was 35:3 after several iterations and a check for low frequency response.
    The bead used for T1 was an unknown gathered somewhere in a past life.
    Initial tests used a Mullard FX1115 bead. Most wound with twisted bifilar
    or trifilar 0.198mm wire.

    I got caught by the source resister decoupling and not realising the very
    low impedences to be expected. That wasted some time and effort in
    thinking it was the transformer material and lack of inductance. The
    FX1115 would have done fine.

    Installing in the receiver (FRG7) also produced an oscillation due to poor
    supply line decoupling and layout.
    "Low-noise preamplifier with input and feedback transformers for low source
    resistance sensors" by J. Lepaisant, M. Lam Chok Sing, and D. Bloyet was
    available on subscription only, thanks for a copy but once again the input
    is via the feedback winding in common base configuration so Zo depends on
    Zin. This is not the case for the common emitter version with emitter
    feedback as shown above, please look again. It may seem obvious to you how
    to derive Zo from common base for the common emmitter version but it is not
    to me. You can point me in the direction of 1000 papers on the common base
    version and it will not help one little bit. If it's to much trouble to
    state what one has said was trivial just use a hybrid pi model, then just
    say so. I did try and it is beyond a gap I can't seem to cross.

    I had all the patents and what was available on the net which deal with the
    common base version only. Thanks for the pointer to the SED discussion and
    the paper you so kindly made available. Ulrich Rhode gives some detail on
    the three configurations in push pull but not any way of determining Zo or
    Zin. Yes I found my copy of his Ham Radio article just discussing the
    configurations and possible applications after much digging around in 20
    year old notes. Wes Haywood also had some stuff published in ARRL but once
    again this is based on the patents and common base version.

    I wonder how insensitive you would feel knowing my father had passed away.
    Thanks for your help, effort and time it is much appreciated.

    Peter
     
  15. Jake wrote...
    It's not a good idea to use the output impedance of an emitter
    follower as a matching source impedance for a filter, etc.,
    because the output impedance is a small-signal value that can
    be highly nonlinear with signal swing and transistor current.
    Instead establish a nice low Ro value, with 2 to 4mA, etc.,
    and then add a series resistor to drive your filter.
    Yes, that's reasonable. Add the EF, you'll be happy, then the
    exact Zo of the amplifier stage won't matter.
    I'm sorry to hear of your father. It took me several years to
    get over mine dying, and I was surprised to discover later that
    I had as a result made a direct change just then in some of the
    things I had previously enjoyed. I also had unexpected health
    responses. It's important to try to maintain a good attitude.
     
  16. Jake

    Jake Guest

    Not something I would usually do. I was trying to establish a Zo of 560
    ohms in the drain circuit via the feeback winding in the source.

    That is another typo above. Sorry.

    I was in fact trying to avoid the resistive loss <GB> which would have to
    be added to the filter insertion loss. Since it is the RF front end, noise
    figure was also important as well as large signal handling capability over
    0.5 to 30Mhz.
    N turns are the drain winding but Zo must be determined to a large degree
    by the source winding which sees via feedback the transformed source
    impedance.

    I wanted a insight to the calculation of the drain Zo with source feedback
    transformer so I could see how well established it was and what would have
    the greatest influence on it.

    I still have no idea how well established Zo = 560 ohms was for small and
    large signals. The original circuit used a 560 ohm drain resister and no
    feedback. Unfortunately I don't have equipment to check the filter
    response with both amplifier circuits or I would have done this. The
    possible influence of the filter reflected impedance changes on the
    amplifier drain are also an unknown and my thinking was feedback may
    improve this.
    What does EF refer to? Sorry I am not with you on this.
    Pops was 91 and has a good innings. He died in his sleep at my home where
    he has been staying for the last 5 years since my mother died. Gave up
    driving at 86 <BG> when he came to stay with me.

    61 years is a long time to know somebody and he is very much missed. I
    still look around to see why he is not sitting in his favourite places.

    Thanks for your kind words and help.

    Peter
     
  17. Jake

    Jake Guest

    Not something I would usually do. I was trying to establish a Zo of 560
    ohms in the drain circuit via the feeback winding in the source.

    That is another typo above. Sorry.

    I was in fact trying to avoid the resistive loss <GB> which would have to
    be added to the filter insertion loss. Since it is the RF front end, noise
    figure was also important as well as large signal handling capability over
    0.5 to 30Mhz.
    N turns are the drain winding but Zo must be determined to a large degree
    by the source winding which sees via feedback the transformed source
    impedance.

    I wanted a insight to the calculation of the drain Zo with source feedback
    transformer so I could see how well established it was and what would have
    the greatest influence on it.

    I still have no idea how well established Zo = 560 ohms was for small and
    large signals. The original circuit used a 560 ohm drain resister and no
    feedback. Unfortunately I don't have equipment to check the filter
    response with both amplifier circuits or I would have done this. The
    possible influence of the filter reflected impedance changes on the
    amplifier drain are also an unknown and my thinking was feedback may
    improve this.
    What does EF refer to? Sorry I am not with you on this.
    Pops was 91 and has a good innings. He died in his sleep at my home where
    he has been staying for the last 5 years since my mother died. Gave up
    driving at 86 <BG> when he came to stay with me.

    61 years is a long time to know somebody and he is very much missed. I
    still look around to see why he is not sitting in his favourite places.

    Thanks for your kind words and help.

    Peter
     
  18. Jake wrote...
    Understood. But not the best idea, IMHO.
    It's certainly laudable to be careful with every dB of gain while
    the signal is still weak, or the signal impedance is high, etc.,
    but once you've gone through a healthy gain stage, e.g. 20dB,
    you no longer need to save every dB (or even every 6dB) and other
    considerations take precedence, such as low distortion, accurate
    filter properties, etc. By adding an emitter follower (EF), you
    lower the impedance, so you can then precisely control it with a
    series resistor. The loss of signal is of no consequence. Make
    the stage gain 26dB if it worries you that much. :)
    Yes. But not a very good approach, unless you're a high volume
    manufacturer desperate to save every part. Here, more is better.

    Lucky break.
    Exactly, thereby precisely setting Zo = 560 ohms to insure the
    filter would have its designer-intended frequency response.

    Use the EF with say 2mA for Zo = 12 ohms (and with over 500mV
    filter-drive capability), and follow it with a 560-ohm resistor.
    That's 572 ohms, close enough! Or follow it with 549 ohms 1%
    to be right on the money.
    Improve, yes, but get you where you want to be, no.

    Loading the amplifier's high Zo directly with the filter is a
    dangerous game, creating distortion as you unwisely rob it of
    the excess loop gain it should be using to keep distortion down.
    Emitter follower. Cathode follower (CF) without a filament. :)
     
  19. Jake

    Jake Guest

    Like most things in electronic design there are trade offs. I wanted to be
    able to more fully understand the realtionship between Zo, the feedback
    ratio and operating current.

    My thought was that it had to be be better than a 560 ohm resister in the
    drain circuit and amplifier stage with no feed back at all. The objective
    was to improve the signal handling capability and dynamic range without
    messing up the match by to much.

    For the parameters with IC and N I need an equation that provides al least
    a ball park figure for Zo. Try as I may I can't see how to solve this.
    Gain has its own problems of dynamic range with high Zo with a Vs of 10Volt
    as well as IP3 for the stage. Feedback has the advantage of added
    linearity reducing other undesired products of ajacent and large signals.
    Sure but I'll bet IP3 and overall linearity are considerably worse.
    1% resisters of such values are like hens teeth. ;-)

    I would be more tempted to run at 10mA ;-) and 2.5 ohms with RE of 470E.
    Then Rseries match of 560E. Should Zin/hFE not be added to this?

    Sorry I have never been much of a fan of resistive matching for RF unless
    absolutely no other better way exists. One needs to achieve a good design
    balance for the full spectrum of requirements most times.

    Retro fitting to existing circuits has it own problem of never enough space
    or remote mounting adding even more problems. But if the transformer
    feedback proved inferior I'll give it a try.
    A perfect match is not the only requirement but one of a set involving
    noise, dynamic range, large signal handling, linearity....

    It takes a considerable amount of time and often one has to think real hard
    to find a way of evaluating what you need to know. More frustrating is
    perhaps better because I know how but lack the equipment. For messing about
    there is no chance of recovering a large outlay on seldom used equipment.
    But a sage with no feedback has no such protection or is the IC current
    drive to the resister match drain load sufficient to overcome the reflected
    impedences?

    I am happy to reduce the gain down to 6 dB or whatever the noise figure is
    at the filter input plus 1 if needed because it is far easier to add a good
    low noise 50 ohm in/out stage at the antenna input ahead of the preselector
    or the preselector input which is 50E and 1st RF amp.
    Ok but out of interest and because it has now become a mission I would
    still like to know how to calculate the approx Zo for the emitter feedback
    circuit given.

    I am busy putting the final touches to ALC on my cheap and cheerful Chinese
    signal generator and that will allow me to more easily check the filter
    under both conditions. I gave up trying to control the tubes large output
    with junction fets or cathode degeneration and instead regulated the plate
    voltage. Pair of signal diodes + TL431 + 4N35 optocoupler and IRF830 as
    the series pass element feeding a 150E resister ahead of the smoothing
    capacitor. Rectified and regulated heater voltage was used to supply the
    TL431 and opto diode. The 4N35 CE shunts the gate and 15 volt protection
    zener driven by a 100K resister by the plate voltage connected to the
    drain. Crude but works well once the TL431 is stabilised. Then to modify
    my old AVO signal generator with 6J5 tube and figure out how to mount the
    stuff. ;-) A BF495 and J310 cascode works fine as a 6J5 replacement.

    Regards and thanks for your adviced and insight to the problem.

    Peter
     
  20. Jake wrote...
    Your design criteria seems to me to harken back to the days of
    expensive components and assembly costs, minimal parts count,
    and painful "optimization" (which, in many celebrated cases,
    absolutely did *not* work out). Not being disrespectful, but
    thankfully I left that design approach behind 38 years ago.
    You're comparing apples and oranges: A common-source stage with
    a highly-linear predictable Zo and an uncontrolled-distortion,
    with a precisely-controlled low-distortion gain stage with a
    fragile "don't-tamper-with-me" Zout. That's how I see it. I
    prefer the second choice, because I can easily fix its problems
    and thereby achieve its low-distortion low-noise Nirvana.

    IMHO, the low-distortion gain stage is the winner, hands down.
    You're very welcome. I'm going to retire from the conversation,
    because your interests seem to move quite separately from mine,
    and I have too many pressing projects on the burner now. By all
    means, enjoy yourself. Let us know how your project progresses.
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-