LNA Design with the BFP842ESD

Discussion in 'Wireless & RF Design' started by Sparky49, Feb 2, 2015.

  1. Sparky49

    Thread Starter Active Member

    Jul 16, 2011
    834
    417
    Hi all,

    I've got a bit of spare time with exams out of the way and I thought I'd look into some LNA design for fun. :)

    I came across the BFP842ESD, which interested me for having such as low noise factor. Datasheet here: http://www.infineon.com/dgdl/Infine...n.pdf?fileId=db3a3043394427e401394df03a3a27c9

    I then looked at their information sheet regarding a test setup here: http://www.infineon.com/dgdl/AN322.pdf?fileId=db3a30433f565836013f57081aaf0275

    I'd like to, at some point, look at recalculating values for operation at 1.4GHz, but I have a few questions before then. Firstly, does anyone have any reasons as to why they chose in the second link, to bias the transistor with a simple base bias? Wouldn't this be potentially unstable and have a shifting operating point? Why not something like a potential divider bias? Secondly why is there no emitter bypass cap?

    I understand that you guys weren't (maybe!) the designers of this circuit, but may have some insight into this microwave circuit design. Was this circuit just meant to be whisked up quickly, or are there real design benefits to the points above?

    Many thanks,

    Sparky
     
  2. skeptic

    Active Member

    Mar 7, 2010
    51
    9
    "Firstly, does anyone have any reasons as to why they chose in the second link, to bias the transistor with a simple base bias?"

    I don't know what you mean by simple base bias. The base is biased off the collector voltage. At RF anything in series with the emitter can make the amplifier oscillate. That's why the emitter is connected directly to ground. So without anything in the emitter circuit, how do you maintain bias over a wide temperature range in which the Vbe changes significantly. Here as the transistor conducts more, the collector voltage drops reducing the base bias introducing negative feedback. In order to not hurt the gain of the stage, the high frequencies are filtered out with C3 and L1. (There is no emitter bypass cap because there is no emitter resistor) With potential divider bias, the operating point will shift too much over a wide temperature range.
     
    Sparky49 likes this.
  3. Sparky49

    Thread Starter Active Member

    Jul 16, 2011
    834
    417
    Thanks for the reply.

    If anything attached to the emitter can cause oscillations, how come there are a variety of class A RF amplifiers? Example:
    [​IMG]

    I am not trying to be facetious, just really curious. :)

    Thanks again,

    Sparky
     
  4. t_n_k

    AAC Fanatic!

    Mar 6, 2009
    5,448
    782
    @Sparky49
    I agree that an emitter resistor is not necessarily de-stablising - rather it may have a stabilizing effect. In any event, if it is included, it is generally RF bypassed. Both forms of biasing are routinely found in practice.
    Re: your post #3 - I'd be rather surprised that the emitter resistor R5 was 5E6 or 5M ohms in a practical circuit. Probably a misprint.
    Not sure why "simple" base bias is OK for the LNA - it may have something to do with the low noise issue but that doesn't gel with optimizing the stable bias for lowest noise figure. However, both documents in your links appear to indicate that a simple base bias is acceptable. As skeptic points out (#2), there will be some negative feedback effect of any series resistor in the collector circuit which lies between the DC supply and the base bias resistor tap point. So it's not truly "simple" or fixed in that respect. With the 33ohm resistor shown in the schematic (R2 in Figure 3 - AN322), a 1mA increase in collector current would decrease the bias resistor (R1) top voltage by 33mV. Is that significant enough to stabilise the bias? One would have to do some detailed analysis with respect to variation in Beta or Vbe. [See the attachment Agilent AN1293.pdf].
    Further on the stability issue, in Figure 3 there are RF decoupling / matching components (including L1 & C3) which presumably reduce any de- stabilizing RF feedback around the collector-base path.
    Re: your other question about emitter bypass capacitor - Again with reference to Figure 3, why would one need a bypass cap if the emitter is already grounded?

    As I indicated above, it might also be worth looking at the attached document on biasing from Agilent ..

    From the data sheet provided in your link (#1), the stated range in Beta for the device is 150-450. With this range of Beta, a Vcc=3V, a Vbe=0.65V, R2=33 ohms and R1=39k ohms I would expect a variation in Ic of 8-19.6mA. This was for a nominal Ic of 12mA [Beta=240?] per the application note AN322. What effect such Ic variation might have on noise figure would require further interpretation. There's probably a more realistic expectation for the range of Beta which would make the variation in Ic less 'dramatic'.

    WRT redesigning for 1.4GHz, presumably you are aware you can download the s-parameter and other device data files from the Infineon web page.
     
    Last edited: Feb 5, 2015
    Sparky49 likes this.
  5. skeptic

    Active Member

    Mar 7, 2010
    51
    9
    The reason an emitter resistor and bypass capacitor may cause instability is that they drastically change the S11 or input characteristics of the transistor. How are you going to match one stage to the next when you do not know the input impedance of the stage? When the stage is not matched, some of the input power gets reflected back to the previous stage and the next stage reflects some of the output power back to this one and it that reflected power that can cause instability. Also when I talk about stability I mean unconditional stability over the full specified temperature range and over all impedances on the Smith Chart.

    Granted, the amplifiers used in receivers are more voltage amplifiers than power amplifiers and matching stages is not as important it is in power amplifiers. Nevertheless, if you are designing a low noise amplifier you will want to match the input of your amplifier to the antenna and avoid adding components that contribute more noise than gain. You begin by discussing an LNA but your second example is clearly not an LNA.

    If you look at the schematic on page 10/21 you will see a series 10 pF cap and a shunt 3.3 nH inductor to ground. At the center of the band the cap has an impedance of about 5.6 ohms and the inductor about 50 ohms. Now if you go down to page 13/21, the input impedance of the transistor for the 2.4 - 2.5 GHz range is shown as a normalized value of 0.8 - j0.3 or 40 - j15 ohms. I haven't done the math but it appears that those two components will very closely match a 50 ohm input to the transistor input impedance. Would anyone care to calculate the input impedance of the second circuit over the full bandwidth?
     
    Sparky49 likes this.
  6. t_n_k

    AAC Fanatic!

    Mar 6, 2009
    5,448
    782
    @skeptic
    Thanks for your informative comments. You make an interesting argument wrt the inclusion of emitter resistor & its impact on S11 in the LNA design.
    Notwithstanding the uncertainty of this approach it seems there are recommended LNA designs in the literature which advocate or allow for such a design choice. The important caveat seems to be that the bypass capacitor must ensure an effective RF ground at the emitter terminal.

    Re the input matching question with the BFP842 in the amplifier shown on AN322 page 10/21.
    The s-parameter data downloaded from the Infineon website have a specific case for Ic=12mA and Vce=2.5V. At 2.4GHz the device s11 value is 0.2367 @ angle -69.1 deg. For the amplifier topology shown with C1=10pF, C3=10pF and L1=3.3nH I get an effective amplifier input impedance of [31.3+j25.8] ohms at 2.4GHz. That's assuming L1 and C3 are effectively in series and 'shunting' the base to ground at RF. Not sure how this all compares with your observations.

    For what it's worth (probably very little practically), I 'improved' the input match by using C1=2.5pF, C3=10pF & L1=4.7nH - based on 2.4GHz s11 alone.

    I believe that in reality that there are conflicting requirements in LNA design which mean one has to make compromises between input impedance matching and noise figure. This then boils down to a gain vs noise figure trade-off.
     
    Last edited: Feb 7, 2015
    Sparky49 likes this.
Loading...