SoftRock RXTX Ensemble RX Mixer

Transcription

1 Reference 1 Purpose and Function The Rx mixer is really two RX mixers with the LO inputs 90 degrees apart giving us a In-phase and Quadrature signals at or near the 0 Hz (direct Conversion) The LO clocks come from the local oscillator Dividers on QSD_CLK_0,1 and the two copies of the RF come from the two balanced outputs from T5. The mixer stage ("QSD" - Quadrature Sampling Detector) acts like two traditional direct conversion mixers operating in tandem. Each takes in half of the filtered RF from the bandpass filter stage and one of the quadrature center frequency signals, then "mixes"/down-converts them to with an output being the traditional mixer products, in this case, two (infra) audio frequency signals that represent the difference between the two inputs (RF and Local Oscillator). These two signals are referred to as the detected I (inphase) and Q (Quadrature) signals and are fed into the high gain Op-Amps stage for amplification and delivery to the audio outputs. The mixer chip (actually a commutating switch, clocked by the two QSD Clock signals from the Dividers Stage) outputs the product and difference signals of the incoming RF against the QSD clock. The effect is to down-convert the incoming RF into its quadrature analogues at frequencies ranging from 0 to roughly 100 khz. These quadrature pairs from the, identifcal in all respects except phase, are fed to the RX OpAmp Stage. 1 William R. Robinson Jr. AJ4MC p1of 17

2 Theory and Design This circuit is a Tayloe or Quadratue Sampling Detector. A really good explanation of how this works can be found at 2 In four articles entitled Fists of Fourier, Part I- IV. It would appear that this is an unfinished series but the four published parts describe the mixer very well. We ve seen that a QSD is really two sample and holds, each sampling at the LO frequency, one delayed by 1/4 of an LO cycle. In this part, I ll show how each of the two sample-and-holds is really a mixer, and how a sample-and-hold can become a balanced mixer. 2 R53 and C41 for an RC charging cuircuit o Smaller τ = R*C will allow the output to follower the input faster To fast and the RF to Audio signal ratio increases To small and there is little Audio signal Note the RX op Amps has a gain off 500, so we are not talking much audio signal out of this stage U10, an FST-3253 is the switching mux I claim R55 and R58 are really part of the Rx Opamps, not the mixer. When one of the Local Oscillator clocks is high then the RF is connected to the sample and hold capacitors which track the RF through the RC time constant R53*C41. This time constant is very fast so the capacitor effectively goes to thellows the RF signal level that is present when the LO clock goes low. When the Local Oscillator clock goes low the capacitor holds this value (less the loss into the RX Opamps which has an input resistance of 10 ohms). This has the effect of sampling the RF at the difference of the LO clock and the RF frequency, so it effectively down converts the RF to the Audio range (as would an analog mixer followed by a low pass filter) A good paper on the design of the mixer is provided in Ultra Low Noise, High Performance, Zero IF Quadrature Product Detector and Preamplifier by Dan Tayloe 3 R53 R53 should be equal to the input resistance (R58) of the RX _Opamps for maximum power transfer from the mixer stage to the RX_Opamps stage. For max power transfer (and easier design of the Antenna low pass filet and the RX band-pass filter ) the impedance the antenna sees should be equal to the antenna impedance of 50 Ohms. The impedance the antenna sees is (R53 + RU10 + R55//XC41)*4 As discussed in the Antenna Band pass filter RU10 is equal to 2 ohms and XC41 is about 35 ohms o Putting all this together yields William R. Robinson Jr. AJ4MC p2of 17

3 C41 SoftRock RXTX Ensemble R R53//35 = 50/4 Skipping the quadratics I get R53 = R55 = 5.64 Ohms 4 o Because we want the I and Q sides to be balanced, then high precession (1%) components are in order Digikey has three choices in stock at the time I write this two are wire-wound and are not a good choice for RfF due to the inductance The remaining part is $3.58 Perhaps this is why 10 ohms was chosen although a few 6.8 ohms are available that would appear to work The Mixer should have a band width greater than the 96KHz that the sound cards are capable of using. Because we want a flat response in the area of interest (up to 96 Khz) it needs to be much greater Reference 3 states actual detection bandwidth will be twice that of the calculated low-pass filter as the the low-pass roll off will extend to both sides of the center detection frequency Simulation shows that the value chosen uf gives 1% attenuation at 48 KHz Conversion Loss Conversion loss of a mixer is equal to the ratio of the IF single sideband output to the RF input level 5 Vif Conversion loss = - 20 log( ) Vrf Dan Tayloe suggests that the conversion loss should be about 0.9dB! 3 Isolation The L to I isolation is the amount the LO drive level is attenuated when it is measured at the IF port 5 Vrf _ at _ if Isolation = 20 log( ) Vrf William R. Robinson Jr. AJ4MC p3of 17

4 Calculated Frequency (audio bandwidth) 1 6 Fc 2RC 1 Fc 2R53* C41 1 Fc 2 *10*0.047uF Fc = 338 Khz Conversion Loss Vif Conversion loss = - 20 log( ) Vrf o I do not know of a way to calculate this value Dan Tayloe suggests that the conversion loss could be as low as 0.9dB! 3 Isolation Vlo _ at _ if Isolation = 20 log( ) Vlo Isolation can also be measured from Rf to If but as rf is usually small LO to If is usually more important o I do not know of a way to calculate this value William R. Robinson Jr. AJ4MC p4of 17

5 Simulation V1 = 7.1 MHz at 1 Vp-p V2 = 7.0 MHz at 1 Vp-p Note how the falling edge of Vswitch occurs at lower voltages of Vin as time increases and Vout shows this trend. The output frequency is this the difference of the two frequencies MHz = 100 Khz o 100 Khz is a high for the application but was chosen so that the user could see the effect of the circuit Rload is R55 and is 10 ohms The design is somewhat sensitive to Rload Rload should be selected to match R53 for maximum power transfer. 7,8 o R53 also affects the RC timing constant o Rload also affects the gain of the Rx Opamps The current values provide near the nominal 50 ohm load to the Antenna Low Pass Filter and the Rx Band Pass Filter 9 o The Plot below shows the time domain response with Rload varied from 2 to 20 ohms in 10 Ohm steps William R. Robinson Jr. AJ4MC p5of 17

6 20 Ohms has the most magnitude Little effect on magnitude of Rf frequencies in Vout RX_mixer-Transient-41-Sweep-Graph v(out)[0] v(out)[1] v(out)[2] v(out)[3] v(out)[4] v(out)[5] v(out)[6] v(out)[7] v(out)[8] v(out)[9] m m m m m m m m m m m m m m m m m m u u u u u u u u u u Time u u u u u Component value selection R53 and C41 o R53 and C41 form an RC charging circuit o Smaller time constants give increased audio signal but also increased RF noise on the signal The RF noise grows faster than the Audio signal I do not how to select the optimal value, especially considering that the Rx Opamps will filter out most of the RF The chart below shows the results for three different values of C41 Similar results were found for varying values of R53 Design Value C41 in uf Vp-p at 10 Khz mvolts Vp-p at 7 Mhz mvolts Frequency (audio bandwidth) Fc =335.8 KHz William R. Robinson Jr. AJ4MC p6of 17

7 vdb(out) RX_mixer-Small Signal AC-36-Graph k k k Frequency Conversion Loss Vif Conversion loss = - 20 log( ) Vrf 0.2 Conversion loss = - 20 log( ) 1.0 Conversion loss = dbv Isolation Vlo _ at _ if Isolation = 20 log( ) Vlo Isolation = 20 log( ) 1.0 Isolation = dbv William R. Robinson Jr. AJ4MC p7of 17

8 Real Circuit Results below are with the entire receiver section was completed. Below is the input and output of the Rx Mixer when all of the receiver section was completed. o Local oscillator is at MHz o Antenna is at MHz Antenna Input amplitude was adjusted for near clipping at audio output Time Domain o Channel 1 is the input at U10 - p6 (at rf frequencies) o Channel 2 is the output at U10 p7 Note the switching transition of the LO leaking into the input Note how much the RF frequencies have been attenuated at the output o Same as above (at audio frequencies) Note again the complex waveform at the input showing the LO leaking into the input Note the rf on top of the audio signal in the output and how small the signal is for nearly full output of the audio amps William R. Robinson Jr. AJ4MC p8of 17

9 Frequency Domain o FFT of the output of the Rx Mixer at radio frequencies is almost to small to detect. o FFT of the output of the Rx Mixer at audio frequencies Results below are with the entire receiver section was completed unless otherwise noted. Below is an oscilloscope screen shot (no load) This was taken at RF = 7.5 MHz and LO = 7.0 MHz o So that reader can see how the RX_Mixer works o This is way above the RX_opamp cutoff frequency o Careful examinations from left to right shows the phase between the LO and RF changing one full cycle, and the resulting output William R. Robinson Jr. AJ4MC p9of 17

10 From top to bottom LO at U10-2 RF at U10-3 IF at U10-7 Rx Opamp out at JP2 William R. Robinson Jr. AJ4MC p10of 17

11 Scoping Issue When a scope probe was connected to the LO it added significant noise to the other three signals o I found this true with 3 different scopes that I tried o This scope had the least interaction probably because it uses active probes o The affect is even worst if the signals are not Band Width limited (20 MHz) inside the scope Below are screen shots from an oscilloscope with and without a probe on the LO o Note the difference in noise o These was taken at RF = 7.01 MHz and LO = 7.00 MHz Left is LO probe connected Right is without LO probe William R. Robinson Jr. AJ4MC p11of 17

12 Below is a screen shot from an oscilloscope without a probe on the LO This was taken at RF = 7.01 MHz and LO = 7.00 MHz Selected to show output at 10 KHz audio frequencies William R. Robinson Jr. AJ4MC p12of 17

13 Below is a screen shot from an oscilloscope without a probe on the LO This was taken at RF = 7.01 MHz and LO = 7.00 MHz Selected to show output at RF frequencies Below is a screen shot from an oscilloscope with the a probe on the LO This was taken at RF = 7.01 MHz and LO = 7.00 MHz From top to bottom LO at U10-2 o Not connected but know to be at About 5 Vp-p RF at U10-3 o About 20 mvp-p o Note noise on RF is greatly decreased if scope if LO is disabled IF at U10-7 o About 3mVp-p (at desired output) To small to measure but working backwards from known RX_Opamp gain o About 70 mvp-p (at Rf frequencies) (with scope probe on LO is removed Rx opamp out at JP2 o About 1.5 Vp-p at 10 KHz (from audio frequency view above) William R. Robinson Jr. AJ4MC p13of 17

14 Frequency (audio bandwidth) RXTX ENSEMBLE RX MIXER FREQUENCY RESPONSE dbv Frequency Difference RF - LO RF - LO Frequency Difference khz The amplitude of the Audio signal at the IF, especially when compared to the amplitude of the RF energy at the IF, is to small to allow accurate measurments o Increaseing the RF amplitude beyond what saturates the RC_Opams did not help o Not saturating and using the output of the RX_Opamps also did not works as the OPamp bandwidth is lower than the IF bandwidth However this yielded a bandwidth of at least 137 khz Attempted to use scope FFT but LO signal is to small compared to the RF background noise Conversion - Vif Conversion loss = 20 log( ) Vrf 3mV Conversion loss = - 20 log( ) 20mV William R. Robinson Jr. AJ4MC p14of 17

15 Conversion loss = dbv Isolation Note time domain clearly shows the isolation of importance is from the LO to the IF Vrf _ at _ if Isolation = 20 log( ) Vrf 70mVp p Isolation = 20 log( ) 5Vp p Isolation = dbv William R. Robinson Jr. AJ4MC p15of 17

16 Comparison SoftRock RXTX Ensemble The table below compares the results o Conversion loss is greater than expected o Isolation is less than expected Both are in the opposite direction of what is desired The small magnitudes make the results acceptable in my mide Real-Measured Simulation Calculated Frequency (Audio Bandwidth) KHz N/A Conversion Loss dbv N/A Isolation (LO to IF) dbv N/A William R. Robinson Jr. AJ4MC p16of 17

17 References 1. Robson, Richard R. Sr., WB5RVZ, Ensemble RXTX Project, online, accessed Goodman, Pete, G1FGQ, Fist of Fourier, PartI-IVt, online, accessed Tayloe, Dan, N7VE, Ultra Low Noise, High Performance, Zero IF Quadrature Product Detector and Preamplifier, online, accessed F%28x%2B35%29+%3D+50%2F4, online, accessed Mini-Circuits, Modern Mixer Terms Defined, online, accessed Horowitz, Paul and Hill, Winfield, The Art Of Electronics Second Edition, (Cambridge University Press 1989) Section 1.19 RC Filters, p Bowick, Chris, RF Circuit Design 2 nd Edition (Elsevier Inc 2008), p Robinson, William AJ4MC, Maximum Power Transfer (another of these short papers). online, accessed Robinson, William AJ4MC, SoftRock RXTX Ensemble Antenna Low Pass Filter (another of these short papers). online, accessed William R. Robinson Jr. AJ4MC p17of 17