Lab 6 - The Power Amplfer and LNA - EE133 - Prof. Dutton - Wnter 2004 1 EE133 - Lab 6 RF Amplfcaton: The Power Amplfer and LNA (revsted) 1 Introducton There wll be no formal Prelab 6 handout. Ths does not mean that there s no work to be done pror to comng nto lab, however. Please read through ths handout to determne what tems you must complete before comng nto lab. In ths lab, you wll be buldng and testng a monolthc power amplfer from Mn Crcuts. Ths power amplfer s the last stage of your transmtter before the antenna. Ths s what wll allow you to get much hgher dstance performance out of your FM system. In addton, you wll be revstng your LNA wth some further characterzaton and mpedance matchng. We wll also look at an mportant characterstc of almost any analog crcut, lnearty. 2 Lnearty and The Thrd-Order Intercept Pont Although there are multple ways to measure lnearty, the ones most commonly used are the thrd-order ntercept pont and the 1-dB compresson pont. Secton 16.5 n the text has a good explanaton of these two characterstcs. If we can represent the output of an amplfer as a seres expanson n terms of the nput voltage, then the output voltage mght be wrtten n ths form: V o (t) = AV (t) + BV (t) 2 + CV (t) 3 +... If we assume that 4 th and hgher order terms won t cause much error, then we can truncate the seres after the 3rd term. Now, f we assume an nput of the followng form, V (t) = cos(2πf 1 t) + cos(2πf 2 t) then the output wll have a number of harmonc and what are called nter-modulaton terms (that s, terms that nvolve more than one dfferent frequency). As stated n the book, these terms wll appear as follows: Second harmoncs: 2f 1, 2f 2 (from V 2 Thrd harmoncs: 3f 1, 3f 2 (from V 3 Second-order ntermodulaton products: f 1 ± f 2 (from V 2 Thrd-order ntermodulaton products: 2f 1 ± f 2, 2f 2 ± f 1 (from V 3 The thrd-order ntermodulaton terms are the ones of concern to us because they appear close to our desred sgnal and because they tend to grow more quckly than the 2 nd order terms as the nput power ncreases. We can test to see how lnear an amplfer s by combnng two dfferent tones (that s, two sgnals of dfferent frequences) at the nput, and measurng the levels of the ntermodulaton products as we ncrease the nput power sgnals. In theory, the thrd order terms wll ncrease 3dB for every 1dB change n nput power (On a lnear scale, these terms ncrease as the cube of the nput voltage). In realty, however, both the fundamental frequency terms and thrd order ntermodulaton terms wll slope off at hgh nput powers. An example plot of fundamental and thrd-order power as a functon of the nput power s shown n Fgure 1. The pont at whch the output fundamental power falls off by 1dB from the extrapolated power s known as the 1dB compresson pont. Therefore, when you measure IP3, you wll have to extrapolate the actual value from
Lab 6 - The Power Amplfer and LNA - EE133 - Prof. Dutton - Wnter 2004 2 values at lower frequences. Input vs. Output IP3: There are two ways to specfy IP3, by referrng to the nput power (IIP3) or by referrng to the output power (OIP3) at whch the extrapolated frst and thrd-order power lnes ntersect. Data sheets wll often quote OIP3 but lst t as smply IP3. Ths s a marketng ploy to make a component look lke t has better performance than t does, snce OIP3 wll tend to be a bgger number than IIP3 (f t weren t, you wouldn t have a very good amplfer). Therefore you must be careful when selectng components based on these specfcatons. Output Power (dbm) Actual Output OIP3 P_24.5MHz P_24.5MHz Slope=1 Extrapolated Intercept Pont P_24.3MHz P_24.3MHz Slope=3 (deally) 30 25 IIP3 Fgure 1: Determnng IIP3 Input Power (dbm) 3 The Power Amplfer 3.1 Buldng the Power Amplfer Ask your TA for a power amplfer chp and solder mount board. Ths chp s a surface-mount part, so we re gong to have to use a slghtly dfferent solderng technque to put t on our board. As you can see, the part tself s too small to ft onto our boards drectly, so we ve made up a header board to solder the part to. Once you ve soldered the surface-mount part on, you can then solder the rest of the crcut to the board. 1. The GALI-5: Look at the data page for the GALI-5 (there s a lnk from the EE133 webste) and note down mportant specfcatons for the amplfer. Some thngs to look for are power gan, frequency range, nput and output mpedances, power consumpton, etc. 2. Chp Layout and Connecton: The chp has three pns and one tab on top. The mddle pn and tab are to be connected to ground, the left pn s the nput, and the rght pn s the output (ths wll be connected to the power supply through an nductor and bas resstor). 3. Orentng the Board: Notce that the solder mount board has three plate-through holes, one for the nput and two for the output. Ths wll allow you to connect a couplng capactor to the nput, and t wll allow you to connect the RF choke and couplng capactor to the output. The mddle pn and top pad are connected to the small copper ground plane. Ths ground plane should be connected to the ground of your crcut. Fgure 2 shows the general crcut layout. 4. Solderng a Surface Mount Part: Solderng surface-mount parts takes a slghtly dfferent approach than solderng normal parts. Frst, wthout puttng the part on the mountng board, place a small
Lab 6 - The Power Amplfer and LNA - EE133 - Prof. Dutton - Wnter 2004 3 Rbas + Vcc Cbypass Lchoke From Multpler or Colptts Cby GALI 5 Cby Matchng Network To Antenna Fgure 2: MnCrcuts GALI-5 Amplfer Crcut dab of solder on each of the traces where you want the pns to be soldered to the board (ths ncludes the ground tab and mddle pn, whch connect to the ground plane on the board). Then place the MnCrcuts part on top of the cold solder. You may want to use a par of tweezers to hold the part n place. Now heat up the trace wth the solderng ron untl the solder melts and the pn s secured to the board. Be careful not to overheat the part. 5. Buldng the Rest: Note n Fgure 2 that the output of the amplfer s connected through an nductve RF choke (so-called because t presents a hgh-mpedance to RF sgnals and so chokes them off) and a bas resstor. Ths resstor s necessary for the chp to operate correctly. Do not connect the output of the chp to power wthout ths bas resstor. The value of ths resstor should be about 60Ω, and should be made from a few resstors n parallel, because t has to dsspate a large amount of power. For a Power Supply voltage of 8-9V, the effectve value of Rbas should be between 55Ω and 65Ω (ths s because the actual voltage at the output of the amplfer s around 5V). For more nformaton on ths, see Bassng MMIC Amplfers, whch s lnked off of the GALI-5 web page. As for L choke, you can choose any value of nductor that wll present a hgh mpedance at 24MHz (20-100uH s probably fne - just be sure to measure your actual nductor value at the correct frequency). The bypass capactors should be the usual hgh-frequency values. Don t forget to solder a connecton from the ground plane to the ground of your board. For now, don t worry about the matchng network. 3.2 Characterzng the Power Amplfer Now we wll take a look at how well ths amplfer performs compared to ts purported datasheet values. Note that ths amplfer works n a very large frequency range, and we are usng t n the very low end of ts range, so some of the datasheet values mght not apply drectly to our applcaton. They should gve us a good dea of ts general behavor, however. 1. DC Power: What s the DC power consumed by your amplfer? 2. S-parameters: Usng the 8712E Network Analyzer (be sure t s been calbrated correctly), measure the S-parameters of ths amplfer by connectng the nput and output through BNC connectors to the analyzer. What are the nput and output mpedances (use the Smth Chart format)? What s the power gan ( S 21 2 )? How much do these values change over a frequency range of 10-40MHz? Sketch the S-parameter responses over ths range. 3. Power Effcency: Power effcency of power amplfers s often characterzed by ths equaton: η = (P owerdelveredtoload)/(dcp owerconsumed). Measure the power that your amplfer delvers to a 50Ω load. What s the Power effcency of your amplfer? 4. Two-Tone Measurement Setup: Now we wll measure the IIP3 and 1-dB Compresson Pont for ths amplfer, and obtan a plot smlar to Fgure 1. There should be a spectrum analyzer and two functon generators set up to do a two-tone test on your amplfer. Ask your TA for help wth the set-up.
Lab 6 - The Power Amplfer and LNA - EE133 - Prof. Dutton - Wnter 2004 4 You wll need the followng: 2 RF functon generators, a power spltter/combner, and the 4395A Spectrum Analyzer. Connect the RF output from the two functon generators to the power combner nputs (f there are extra nputs, you must termnate them wth a 50Ω termnaton.) Connect the output of the combner (marked wth an S ) to the A or B nput of the Spectrum Analyzer. Set one RF generator to produce a 24.5MHz sgnal at -28dBm. Set the other RF generator to produce a 24.7MHz sgnal at -28dBm. On the 4395A Spectrum Analyzer, select the followng optons: Select Spectrum Analyzer: Format Spectrum Analyzer Select Input A or B: Measure A or B Set Center Frequency: Center 24.6MHz Set Frequency Span: Span 1MHz Set Peak Threshold: Search Peak Peak Def Menu Threshold on, Threshold Value -80dBm (or just above the nose floor) Set Multple Markers to Measure Multple Peaks: Search Multple Peaks Peaks All Dsplay Marker values: Marker Utlty MKR Lst ON 5. Measurng IIP3 and 1-dB Compresson Pont: Now, ncrease both functon generators by 1dB(m) ncrements (the power output of both should be the same) and watch the dsplay on the spectrum analyzer. When you start to see ntermodulaton products appear above the nose floor, you should start takng measurements (you should see two peaks at 24.5 and 24.7, wth a smaller peak on ether sde at 24.3MHz and 24.9MHz). You may have to re-select the peak detectng markers so that you get measurements for all four sgnals. Don t forget to record the nput power as well as these output power levels (you should verfy that the power readng on the functon generator s the actual power beng delvered to the nput of your amplfer). You may want to take sparse measurements at lower powers, but you should take more narrowly spaced ponts as the sgnal powers begn to slope off. 6. Plottng and Extrapolatng IIP3: Plot the fundamental and thrd-order output powers vs. nput power for your amplfer. Extrapolate lnes from the lnear porton of the graph to estmate IIP3. At what nput power s your output power 1dB less than the extrapolated value? 4 Low Nose Amplfer In Lab 3, we desgned and mplemented a shunt-shunt feedback Low-Nose Amplfer. Now we wll revst the concept of nose and the Low-Nose Amplfer. In a cascaded system, the nose contrbuton of the frst stage s the most crucal to the overall performance of the system. Chapter 14 dscusses the varous sources of nose, the concept of Nose Fgure, and nose n cascaded systems. In essence, the nose of the frst stage contrbutes drectly to the total nose of the system, whle the nose of each successve stage s dvded by the product of the gans of the precedng stages. There are a few parameters to control when tryng to mnmze nose: bandwdth, mpedance matchng, and amplfer contrbutons. Because we have already pcked an actve element (the 2SC3302) for our nput gan stage, ths parameter s fxed. We do have control over bandwdth and mpedance matchng, however. Lmtng bandwdth greatly reduces nose, snce whte nose power has a lnear dependence on bandwdth (Nose = ktb). Therefore, we wll nclude an nput flter on our LNA to flter out nose outsde of our sgnal bandwdth. In addton, proper mpedance matchng can reduce nose as well. As s explaned n Secton 14.10 n the book, there s an optmum mpedance match that wll mnmze nose at the nput to an amplfer. It turns out that ths s not always equal to the mpedance match that wll maxmze power transfer nto a load. For our stuaton, however, these two optmal values are relatvely close to each other, and so we wll deal wth tryng to get maxmum power transfer rather than attempt a complex nose matchng calculaton.
Lab 6 - The Power Amplfer and LNA - EE133 - Prof. Dutton - Wnter 2004 5 4.1 LNA: Redesgn For the fnal project, we wll ask you to make some sgnfcant change or mprovement to your LNA desgn. The way n whch we have gone about buldng our LNA stage s somewhat pedagogcal n nature. We frst bult a broadband 50Ω nput and output mpedance amplfer usng shunt-shunt feedback, and then used mpedance transformatons to change the small output mpedance to the larger nput mpedance of the multpler, thus gettng rd of the broadband match and the work we dd to obtan the low output mpedance. Clearly, ths mght not be the most effcent means of desgnng ths crcut, but t does gve you the tools for thnkng about a better way for buldng the LNA. Your job now s to rethnk ths stage and get the best performance out of t that you can. You should refer back to the LNA lecture notes and dscuss optons wth your lab partners, TAs, and classmates. Here are some optons: Gan: Shunt-shunt feedback reduces gan, so you mght want to desgn your amplfer wthout t. Ths puts your desgn at the mercy of the transstor parameters, however, so there s a tradeoff, and you wll have to desgn more elaborate mpedance matchng. You can also desgn your feedback network to match 50Ω on the nput and 1.5kΩ on the output. If you replace Rc wth a tuned load (LC tank), ths helps get rde of the problem of Rc loadng down your output mpedance. Stablty: Ths s probably the factor most sgnfcantly lmtng the gan of your amplfer. Because of the hgh frequency of our sgnals, and because of the ease of couplng at these hgh frequences, feedback loops can form between reactve components even when they re not desgned nto your crcut. Therefore you need to be careful about wrng, power supply bypass, and ground planes. In addton, attemptng to acheve too much gan can end up creatng nstabltes that wll turn your amplfer nto an oscllator. You may fnd that ths s especally a problem s you decde to go wth a tuned load on the output. Nose: As stated above, from a nose perspectve we have the optons of lmtng bandwdth and reducng the amount of resstance n the crcut. We can do the former by usng a tuned load and placng a flter on the nput. We can do the latter by reducng the amount of resstve bas we use. Wth shunt-shunt feedback, we can reduce the number of bas resstors by allowng the feedback resstor to provde both AC feedback and DC bas current to the base. Another knob to tweak: If you fnd your LNA havng stablty problems, the frst thng to check s probably that your power supples are properly decoupled and stable. If that fals, the addton of a small (AC) emtter resstor can help to stablze your crcut by provdng seres-seres feedback (note that ths has the effect of counterng the mpedance-lowerng characterstcs of the shunt-shunt confguraton). Ths wll also lower the gan of your amplfer (g m becomes g meff = g m /(1 + g m r e )), but t s an opton that can help curb negatve mpedances and oscllatons. The value of ths resstor should be very small (less than 10Ω) or your LNA gan wll suffer more than desred. 4.2 LNA Characterzaton 1. If you have made changes to your LNA, or suspect that your prevous measurements were flawed, redo your measurements from Lab 1 for S-parameters, Q, gan, etc. 2. LNA IIP3: Repeat the IIP3 measurement from above for your LNA. You can use the tap of the output nductors for your output. Because your LNA has a tuned output, ths measurement may have some error due to attenuaton of the ntermodulaton products (just take measurements at a good enough number of ponts to be able to plot the IP3 curves). Therefore you should use a smaller spacng for your fundamental frequency sgnals (try 24.35MHz and 24.40MHz). Note that your LNA output sgnals wll saturate at much lower levels than the power amplfer, so you wll need to use generally lower power levels for ths measurement. 4.3 LNA Input Flter Lastly, we need some way to reject the unwanted electromagnetc sgnals that can enter our LNA and overload t. Imagne a large sgnal (called a blocker) at a nearby frequency, such as 25MHz. Ths sgnal could be
Lab 6 - The Power Amplfer and LNA - EE133 - Prof. Dutton - Wnter 2004 6 so large that the LNA becomes saturated, and our desred sgnal would be lost. Even wth smaller blockng sgnals, non-lneartes can cause ntermodulaton and desenstzaton to occur (see the lecture notes or the textbook for detals). We would lke to flter out these potental blockers as soon as possble n our recever, so we ll be addng a seres LC flter at the nput to our LNA. You can remove the AC couplng capactor from your LNA nput Desgn a seres LC flter for use at the nput of your LNA. You may want to use ether a varable capactor or nductor (we have some, but not a wde varety) to tune your flter. If t turns out you need a very small value of capactance, you can put the varable cap n seres wth a statc cap. 1. Measurng Q: Measure the Q of your nput flter separately (ths s easy to do wth the network analyzer). 2. Connectng to the LNA: Connect your flter to your LNA nput. Characterze your newly fltered LNA. Congratulatons agan on buldng a very mpressve wreless transmtter and recever! For the rest of the quarter, you ll be characterzng, mprovng, and packagng your wreless system.