Connect with us

20MHz Wideband Amp

Discussion in 'Electronic Design' started by Tim Williams, Sep 16, 2007.

Scroll to continue with content
  1. Tim Williams

    Tim Williams Guest

    Schematic:
    http://webpages.charter.net/dawill/Images/20MHz_Amp.gif
    Pictures:
    http://webpages.charter.net/dawill/Images/20MHz_Amp1.jpg
    http://webpages.charter.net/dawill/Images/20MHz_Amp2.jpg

    With the 100 + 10pF on the +in transistor and 10pF on the tail, I've
    eliminated the UHF/microwave oscillation as best I can tell. It no longer
    changes gain or bias when I wave my hand around and touch parts. My
    laptop's wifi probably stays connected, too (I didn't have it with me when
    I did this).

    Now I can concentrate on the characteristics. The squarewave response is
    good, but it has a strong bounce on it. In fact, hooking it to an RF
    generator (Eico 322, not much output into 50 ohms, but frequency appears to
    be accurate) I see a substantial rise of perhaps 20dB at 26MHz! It is of
    equal and opposite amplitude on both outputs (viewing with 10x probes, one
    with somewhat more capacitance), so I suspect it comes from the LTP.
    Feeling around the circuit produces no change (aside from loading down the
    2N5179 collectors, which predictably reduces HF gain overall, not
    specifically the 26MHz peak), implying it might be an LC sort of thing.

    Ideas?

    Tim
     
  2. john jardine

    john jardine Guest

    I'd lose the 220p on the output stage.
    The rising response it causes can transmogrify into a gyrated inductance
    that resonates with the output Ccbs and strays, forming a nasty hi Q peak at
    the top end. Sounds like you've already spotted it. (probably then no need
    for the 10p 100ohm)
    There's no overall feedback at the moment and the enormous X300? gain should
    run out of steam about 6-8MHz. This value set by the 1k3 loads and scope
    capacitance break.
    No real alternative to reach the 20Mhz other than inductive peaking (yuck)
    or buffering the output loads.
    Due to the 3GHz GBW, fitting (much) higher ft transistors would be the next
    step.
    (BTW neat video)
     
  3. Tim Williams

    Tim Williams Guest

    Eww, no way! Bandwidth drops to like 5MHz and square waves are SLOOOW!
    Plus, everyone else does it. (And yes I'll jump off a cliff when they do
    too! ;-) )

    Besides, it's present even while puttering around with said capacitor. I'm
    pretty sure it's in the LTP, but where, or from what, I know not.
    No, that's for the UHF+ parasite that I can't see on my mere 200MHz scope.
    The 26MHz hump is there in all cases.
    Come to think of it, I never did bother to measure it, but yeah, something
    like that. Gotta go from 0.5V or less up to full screen 100Vp-p, dontcha
    know :)
    Eh? The '5179's (or are they 6304's in circuit, I forget) are 1.2GHz or
    something, and already well in excess of my needs. I don't even need to
    add emitter compensation. (Though I did find some benefit while
    breadboarding, but I didn't get any UHF parasites then, either!)
    Where, you mean on YouTube?

    Tim
     
  4. MooseFET

    MooseFET Guest

    Sort of working left to right on the design and pointing out each
    thing I would do differently:

    The 22R on the emitters of the 2N5179s lead to a node with a fairly
    large capacitance. I would have used a higher resistance, lets say
    100R to the common point and a 44R between the emitter with very short
    legs. You may want to put a series RC in parallel with this to boost
    the high frequency gain a bit when you make a change I suggest below.

    You show only one 0.1u capacitor for the collectors of the 2N3904s. I
    would use two at least, if I'm following the layout correctly. You
    want the 680R to connect directly to the capacitor and not to the
    collector first.

    The 220pF between the emitters of the 2SC1569s looks to have too long
    of leads. You are stuck with this for mechanical reasons but it
    doesn't take much to make the 220pF into a series tuned circuit. This
    would cause it to give more boost than you want as it comes into
    resonance. I suggest making one of its leads into a 22R resistor to
    spoil out its Q.
     
  5. Bob

    Bob Guest


    Analog stuff with discrete transistors beyond basic switch is not my
    area
    of expertise so I may be totally wrong here but I think the 10pF
    capacitor is
    far too small for circuit operation below about 10MHz.

    The current through both the 2n5179 has to flow to ground through that
    capacitor. At 1MHz it has a reactance of 15000ohms.
    At 10MHz it is 1500ohms.

    If you have a high impedance in the path to ground little current
    will flow through either of those transistors.

    Is that cap supposed to have negligable impedance at intended
    operating frequencys?

    I had to redraw the circuit with the rails at the top and bottom of
    the
    diagram to make sense of it.

    It looks to me that the path through the two transistor at the
    bottom of your diagram is a feedback pathto set the DC bias
    point of the output?

    Bob
    I take it the path through the two tran
     
  6. Tim Williams

    Tim Williams Guest

    - The bulky 1/2W resistors do go over a good spot of ground plane (in the
    picture, below the green 3pF that isn't there anymore). Capacitance would
    lead to increased HF response, though my intuition says that's a
    contributing factor to the UHF parasite.
    Ah, that's a possibility.
    I have one ceramic for the +9V (the red jumper on the underside) and one
    for the -9V. I tried holding one on the underside (of both nodes), with no
    effect.
    Hmm, it's about 1" is about 30nH, so Fo ~ 62MHz. It's in the same order of
    magnitude, but that means estimated L is up by a factor of 4 or more.
    Playing with hundreds of pF around each emitter-to-ground, individually,
    has little effect, aside from a local change in compensation.

    I do recall getting it to oscillate at some point, and IIRC it was from
    playing around there. That could be local or related by a larger feedback
    path.

    Could the 2N3904's be any trouble? Other circuits (e.g., cascode video
    amps) have base resistors; here I have the 3904 emitter followers straight
    into the base. On the breadboard I had found an oscillation that required
    a ferrite bead on one 3904's base or emitter (I forget which), but
    breadboard activities are to be taken with a grain of salt.
    I can try that.

    Tim
     
  7. Tim Williams

    Tim Williams Guest

    Actually, that capacitor isn't even supposed to exist.

    The input current flows between the inputs, because it is a differential
    amplifier. The tail is supposed to be high impedance, but apparently it
    oscillates at UHF, and I need that capacitor to anchor it. Or something.
    Ditto the 100 + 10pF, which in and of itself makes absolutely no sense in a
    balanced circuit at 20MHz. For that matter, with the 100+10pF I might get
    away without the 10pF, I haven't tried that yet.

    As you've just shown, compared to the emitter impedance (<< 700 ohms), the
    10pF has little effect at signal frequencies, so it can be ignored.
    The current sink takes care of DC operation.
    Ah, schematic style. Eh well, it's traditional. A vertically drawn LTP is
    more familiar, but it has a lot more crossings, too. Oscilloscope amps and
    other balanced circuits are often drawn in this way. Just take a close
    look and make sure the upside-down transistor isn't PNP. ;)
    Correct. I got tired of adjusting common mode operating point, so I made
    the damn thing set itself.

    Tim
     
  8. MooseFET

    MooseFET Guest

    I wouldn't expect the capacitance from the resistor to ground to do
    much at the 20MHz. At 100MHz or so maybe.

    Was this "underside" far from the existing 0.1 and did you use really
    short leads?
    I'd expect to see some change if my theory on this part is right. If
    there was no change, my theory is toast.

    That much isn't likely to be the cause a 20dB peak. Still I'd put in
    the deQing resistor just to see what happens.

    [....]
    2N3904s would much rather oscillate at about 100MHz than 20MHz. Their
    capacitances are on the low side for getting oscillations that low.
    The base sees a resistor on the collect of the stage before. This
    will tend to supress the Colpitts common collector type of
    oscillations.
     
  9. john jardine

    john jardine Guest

    [...]

    First stage GBP 520meg, i.e the 1.2GHz trannie is already starting it's
    death rattle. Ouput stage GBP 260meg, which the data sheet suggest is
    already over the top. Kind of a zero-sum power game is running, the
    practical effects are like squeezing dough.
    If you're sniffing at a 70ns risetime then try increasing the generator
    source resistance beyond 50ohms. :).
    For these jobs I gave up on my 100MHz scope and Wavetek and find a spectrum
    analyser with tracking generator vastly superior. Problems are visible at a
    glance.

    Yep. Youtube. Interesting watching people's induction welding approaches.
    Somewhat odd to hear a 'mercan accent, as anyone posting English text here
    (other than the obvious tossers) gets internalised as 'BBC newsreader', no
    matter where they hail from.
     
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

-