2016年7月16日土曜日

The principle of “IQ orthogonal demodulator with 2 phase orthogonal Oscillator” (Preliminary)

The principle of “IQ orthogonal demodulator with 2 phase orthogonal Oscillator”
(Preliminary)
                                                 (C) Noboru, Ji1NZL Jul.13, 2016

IQ orthogonal demodulator has simple architecture which is useful to use in SDR (Software Defined Radio) receivers. This document describes about it simply to understand it.

1. Architecture

Fig. 1 is the architecture of “IQ orthogonal demodulator” which has a “2 phase orthogonal oscillator". “IQ orthogonal demodulator” here inputs various kinds of modulated RF signal and output (I, Q) signals on the complex number plane.



(I,Q) signals as output are digitalized by A/D converters and are fed to Microcomputer(MPU) or DSP. Then MPU/DSP can decode the modulated (I,Q) signals to output as audio signal, digital data/program code or image signal by calculations by Math. 

2. How it works

To simplify the explanation, I assume it receives RF sine wave and what signals go out to output (I, Q).

Vin = Vr*sin(ωr*t) …(1)
Vosc = Vc*sin(ωc*t) …(2)

V2 = Vc*sin(ωc*t - π/2) = -Vc*cos(ωc*t) …(3)

Get V3 by multiplying Vin and Vosc.
V3 =  Vin*Vosc =  Vr*sin(ωr*t) * Vc*sin(ωc*t)
    = Vr*Vc*{ sin(ωr*t) *sin(ωc*t) } 
    = (-1/2)*Vr*Vc*{ cos((ωr+ωc)*t) - cos((ωr-ωc)*t)) } …(4)

Assume using of Down-converted signal for the equation (4).
Then LPF can remove cos((ωr+ωc)*t from (4)  
by specifying cut off frequency of LPF to be set to ωcut < ωr+ωc .
Then
V-i =  (1/2)*Vr*Vc*cos((ωr-ωc)*t)) …(5)

Get V4 by multiplying Vin and V2.
V3 =  Vin*V2 =  Vr*sin(ωr*t) * (-Vc*cos(ωc*t))
    =  -Vr*Vc*{ sin(ωr*t) *cos(ωc*t) } 
    = (-1/2)*Vr*Vc*{ sin((ωr+ωc)*t) + sin((ωr-ωc)*t)) } …(6) 

in (6)., sin((ωr+ωc)*t) can be removed by LPF, then
V-q = (-1/2)*Vr*Vc*sin((ωr-ωc)*t))  …(7)

Thus output voltage signal comes out as V-i and V-q in (5) and (7).

Viewing (5) and (7), they are orthogonal function for time variable “t”.
Shift -π/2 [rad] of V-i by the equation (7)
∵ Shift -π/2 of Vq = (-1/2)*Vr*Vc*sin((ωr-ωc)*t)-π/2) 
  = (1/2)*Vr*Vc*cos((ωr-ωc)*t)) = V-i …(8)

The equation (8) means (I, Q) = (V-i, V-q) orthogonal signals each other on voltages of the complex number plane.
These orthogonal signals (I, Q) are extracted from the “IQ orthogonal demodulator” here. 

3. Examples of how the modulated signals is recovered to “baseband signal”

(a) Receiving SSB (USB)
    
    Assume USB signal Vin = Vrf*sin(ωc+ωs)*t) comes in.
    Input sine wave ωc+ωs instead of ωr,  ωr := ωc+ωs
    Then the equation (5) becomes
    V-i =  (1/2)*Vr*Vc*cos((ωc+ωs)-ωc)*t)) = (1/2)*Vr*Vc*cos(ωs*t) …(9)
    and the equation (7) becomes
    V-q = (-1/2)*Vr*Vc*sin(((ωc+ωs)-ωc)*t)) = (-1/2)*Vr*Vc*sin(ωs*t) …(10)
    
    The equation (10) means “baseband signal” Vr*sin(ωs*t) is 
    demodulated/recovered 
    as the output voltage V-i and that V-q (10) is orthogonal signal for V-i.
  
    Vq (10) recovered the baseband signal then Voltage Vq can be input to the  
    audio amplifier directly to sound the speaker. 
    This means USB is demodulated on V-q output without signal processing by MPU.

    V-i (9) is orthogonal signal that shifted -π/2 [rad] of V-q (10).
    Here -90 deg. shifted is connected after V-i then (9) becomes
    (1/2)*Vr*Vc*cos(ωs*t -π/2) = (1/2)*Vr*Vc*sin(ωs*t) …(11)
    
    This (11)  is the same as the inverted signal of (10) so we can add a adder 
    after (11).

    Then (10) + (11) =  {(-1/2)*Vr*Vc*sin(ωs*t)} + {(-1/2)*Vr*Vc*sin(ωs*t)}
      =  -Vr*Vc*sin(ωs*t)  … (12)
      This signal is two times of Vq at the equation (10) and can be input to 
      the audio amplifier directly to sound the speaker. 

      Signals (9),  (10) or (12) may be digitized by A/D convertor and they can 
      be filtered by the filter program to eliminate the extra signal around 
   demodulated baseband signal.

(b)  Receiving SSB (LSB)

    Assume LSB signal Vin = Vrf*sin(ωc-ωs)*t) comes in.
    Input ωc-ωs instead of ωr,  ωr := ωc-ωs
    Then the equation (5) becomes
    V-i =  (1/2)*Vr*Vc*cos((ωc-ωs)-ωc)*t)) = (1/2)*Vr*Vc*cos(-ωs*t) 
    = (1/2)*Vr*Vc*cos(ωs*t)…(13)
    and the equation (7) becomes
    V-q = (-1/2)*Vr*Vc*sin(((ωc-ωs)-ωc)*t)) = (-1/2)*Vr*Vc*sin(-ωs*t) 
    =(1/2)*Vr*Vc*sin(ωs*t)…(14)
    The equation (14) means baseband signal Vr*sin(ωs*t) is demodulated in the 
  output voltage V-q and (14) is orthogonal signal of for V-i.
  
    (14) is recovered the baseband signal. Then Voltage Vq can be input to the  
    audio amplifier directly to sound the speaker. 
    This means LSB is demodulated on V-q output.

    (14) is orthogonal signal that shifted -π/2 [rad] of V-i.
    Here -90 deg. shifted is connected after V-i then (13) becomes
    (1/2)*Vr*Vc*cos(ωs*t -π/2) = (1/2)*Vr*Vc*sin(ωs*t) …(15)
    
    This is the same signal as (14) so we can add a adder after (14) and (15)
    Then (14) + (15) =  {(1/2)*Vr*Vc*sin(ωs*t)} + {(1/2)*Vr*Vc*sin(ωs*t)}
      =  Vr*Vc*sin(ωs*t)  … (16)
      This signal is two times of Vq at the equation (14) and can be input to           the audio amplifier directly to sound the speaker. 

      Signals (14) or (15) or (16) may be digitized by A/D convertor and they may       be filtered by the filter program to eliminate the extra signal around    
      baseband.

(c) Receiving AM

    AM RF signal is equivalent to the added signals of Carrier ωc, USB ωc+ωs.
    i.e. 100% modulated AM signal is expressed as sum of these 3 voltages.
    Vin = Vrf*sin(ωc+ωs)*t) + 2*Vrf*sin(ωc)*t) + Vrf*sin(ωc-ωs)*t)
    
    The Carrier signals cancelled as 0 Hz in the demodulator by the equation (5) 
    and (7).  
    The Carrier signals is changed to 0Hz. It means there is no tone for the 
    Carrier signal. 

    Assume ωr := ωc then (5) becomes 
    V-i =  (1/2)*Vr*Vc*cos((ωc-ωc)*t)) ) = (1/2)*Vr*Vc …(17)
    
    Beside from (7) 
    V-q = (-1/2)*Vr*Vc*sin((ωc-ωc)*t)) = 0 …(18)

At the output of V-i side,

    from (5), re-write it here, V-i =  (1/2)*Vr*Vc*cos((ωr-ωc)*t))  {…(5)} 
    for USB : ωr := ωc+ωs then
      V-i for USB becomes  Vi= (1/2)*Vr*Vc*cos((ωr-ωc)*t)) 
                                          = (1/2)*Vr*Vc*cos(( (ωc+ωs) -ωc)*t))
                                          =(1/2)*Vr*Vc*cos(ωs*t) …(19)
    for LSB :  ωr := ωc-ωs then  
      Vi= (1/2)*Vr*Vc*cos((ωr-ωc)*t)) 
        = (1/2)*Vr*Vc*cos(( (ωc-ωs)-ωc)*t))
        =(1/2)*Vr*Vc*cos(-ωs*t)) = (1/2)*Vr*Vc*cos(ωs*t) …(20)

At the output V-i side,  both demodulated LSB and USB are appears, then
V-i = (19) + (20) 
      = (1/2)*Vr*Vc*cos(ωs*t) + (1/2)*Vr*Vc*cos(ωs*t) = Vr*Vc*cos(ωs*t) …(21)

At the output of V-q side, 
    from (7) re-write (7) here, V-q = (-1/2)*Vr*Vc*sin((ωr-ωc)*t))  {…(7)}

For USB : ωr := ωc+ωs then 
V-q = (-1/2)*Vr*Vc*sin((ωr-ωc)*t))
      = (-1/2)*Vr*Vc*sin(( (ωc+ωs)-ωc)*t))
      = (-1/2)*Vr*Vc*sin(ωs*t)) …(22)
For LSB :  ωr := ωc-ωs then 
V-q = (-1/2)*Vr*Vc*sin((ωr-ωc)*t))
      = (-1/2)*Vr*Vc*sin(( (ωc-ωs)-ωc)*t))
      = (-1/2)*Vr*Vc*sin(-ωs*t)) 
      = (1/2)*Vr*Vc*sin(ωs*t) …(23)

At the output V-q side, both demodulated USB and LSB signals appears, then
V-q = (22)+(23) 
        = (-1/2)*Vr*Vc*sin(ωs*t) + (1/2)*Vr*Vc*sin(ωs*t) = 0 …(24)
… USB and LSB are cancelled in V-q side and V-q become 0 V.

V-i = (1/2)*Vr*Vc + Vr*Vc*cos(ωs*t) …(25)
Vq = 0 …(26)

Remove DC voltage value from (25) by some condenser such as 1uF or subtracting calculation by MPU.
Then AM signal is decoded as the signal Vr*Vc*cos(ωs*t) …(25)’ 
Mathematically,  Vq = 0 from (24). Demodulated USB and LSB may not be cancelled completely. So we may not use the signal V-q when we receive AM signal.  

AM signal may change the phase in some distant signal path and the equation (25) may have some phase angle on the complex plane. This means we have signal strength change according to the phase change. It can be cancelled by the phase angle rotation to become 0. It is the similar principle of so called “Synchronized AM detector”.

(d) Receiving CW (Continuous wave)

CW is equivalent signal of a singular case of LSB where ωs=800Hz single tone or a singular case of USB where ωs=800Hz.

Then CW RF wave can be demodulated as ωs=800Hz single tone on USB side or LSB side.

(e) BPSK, QPSK, and QAM for digital communication
{BPSK: Binary Phase Shift Keying, QAM: Quadrature Amplitude and phase Modulation}

To simplify the explanation, I assume receiving RF sine wave with phase shifted φ(t) and what signals go out to output (I, Q).

Vin = Vr*sin(ωc*t+φ(t)) …(e1)
here φ(t) = 45 deg., 135 deg., 225deg., or 315deg.
              = π/4,  π/3, π/5, or π/7 are used if 4 values of QAM (=QPSK) are assumed.

Vosc = Vc*sin(ωc*t) …(e2)
V2 = Vc*sin(ωc*t - π/2) = -Vc*cos(ωc*t) …(e3)

Get V3 by multiplying Vin and Vosc.
V3 =  Vin*Vosc =  Vr*sin(ωc*t+φ(t)) * Vc*sin(ωc*t)
    = Vr*Vc*{ sin(ωc*t+φ(t)) *sin(ωc*t) } 
    = (-1/2)*Vr*Vc*{ cos((2*ωc)*t+φ(t)) - cosφ(t)  } 
    = ( 1/2)*Vr*Vc*{ -cos((2*ωc)*t+φ(t)) + cosφ(t) } …(e4)

Assume using of Down-converted signal for the equation (e4).
Then LPF can remove cos((2*ωc)*t +φ(t)) from (e4)  if cut off frequency of LPF is set to ωcut < 2*ωc .
Then
V-i =  (1/2)*Vr*Vc*cosφ(t) …(e5)

Get V4 by multiplying Vin and V2.
V3 =  Vin*V2 =  Vr*sin(ωc*t+φ(t)) * (-Vc*cos(ωc*t))
    = -Vr*Vc*{ sin(ωc*t+φ(t)) *cos(ωc*t) } 
    = (-1/2)*Vr*Vc*{ sin(2*ωc*t+φ(t)) + sinφ(t) } …(e6) 

in (6)., sin(2*ωc*t+φ(t)) can be removed by LPF, then
V-q = (-1/2)*Vr*Vc*sinφ  …(e7)

Thus output voltage signal comes out as V-i and V-q in (e5) and (e7).

Viewing (e5) and (e7), they are orthogonal function for time t.
Shift -π/2 [rad] of Vi by the equation (e7).

∵ Shift -π/2 of Vq = (-1/2)*Vr*Vc*sin(φ(t)-π/2) 
  = (1/2)*Vr*Vc*cosφ(t) = V-i …(e8)

The equation (e8) means (I, Q) = (V-i, V-q) orthogonal signals on voltages of the complex number plane, and that they are extracted from the “IQ orthogonal demodulator”. 

Now we get φ(t) from the output signal V-i (e5) and V-q (e7) by calculations.
V-q/V-i = (-1/2)*Vr*Vc*sinφ / (1/2)*Vr*Vc*cosφ(t) = -tanφ(t) …(e9)
∴ φ(t) = arctan( - (V-q/V-i) ) …(e10)
Phase value φ(t) of baseband is extracted by the calculation here.

Besides, we get a half length of radius r(t) from V-i (e5) and V-q (e7).
  r(t)  = √(V-i^2+V-q^2) = √{ {(1/2)*Vr*Vc*cosφ(t) }^2 + {(-1/2)*Vr*Vc*sinφ }^2 }
          = √{ {(1/2)*Vr*Vc}^2*cosφ(t)^2 } + {(-1/2)*Vr*Vc}^2*sinφ^2  }
          = √{  {(1/2)*Vr*Vc}^2* (cosφ(t))^2 + (sinφ)^2 } }
          = (1/2)*Vr*Vc …(e11)
a half of radius value of baseband is extracted by the calculation here.

Demodulating QAM signals require to calculate (e9), (e10), and (e11). 
They can be calculated by MPU program for digitalized signals of V-i and V-q.
The principle itself doesn’t give limitation to usages only to QPSK (4 digits values) and it can be expanded to 2^n values transmutation/reception. 
( The integer “n” is greater than equals 1 and it has no limitation as Math.
i.e. Values = 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2018, … )

In the actual work, The phase φ(t) of RF QAM at transmitter side and the phase of OSC Vosc at IQ demodulator side is not synchronized. Because QAM itself has no synchronization protocol and transmitters and receiver are located in the distant place. 

There is no synchronization mechanism for phase of QAM itself.  
So synchronizing signals for the initial phase of QAM and the procedures in the MPU calculation for phase synchronization is required.

(f) Demodulating FM (To be continued)

4. Summary 

(1) SSB(USB/LSB), AM, and CW can be demodulated without calculation by MPU.
MPU can do DSP Filtering extra signal around these signal.

(2) QAM can be demodulated with calculations by MPU or DSP via IQ orthogonal demodulator.

(3) IQ orthogonal demodulator can demodulate input of RF signal (Vr-i(t), Vr-q(t)) to convert to output signals (V-i(t), V-q(t)) that are orthogonal functions for the variable time “t” on the complex number plane for voltage.

(4) The principle and architecture is very simple but IQ orthogonal demodulator can be used for many kinds of signal modulation.  It has good characteristics to connect MPU or DSP processing. There may be some unknown applications that has not been found yet.

5. Subject

(1) Low tone of beat in AM and strange sound of music on SSB
This method requires very accurate and stable frequency for the local oscillator.  If the frequency error exists, the error of the frequency may cause low tone of beat when receiving AM signal. Human’s ear is very sensitive for around 8Hz frequency drifted.

It is not so serious in listening human’s voice but sound of music may be strange or uncomfortable feeling even if turned frequency has very small errors such as 10Hz. 

If some PLL as a local oscillator work in every 100Hz steps, it is out of usage to listen to the music in SSB/AM. SSB/AM can not be suitable to send or receive the tone of music.
Old fashioned LC analog oscillator often drifts the frequency gradually by time passed. 
However, PLL, DDS , TCXO or latest devices are changing the situation be better rapidly. 

(2) Strength of decoded AM signals changes by amount of shifted phase of Oscillator
V-q becomes 0 in the equation (24). This means output of “product detector” can be 0V and radio listener may loose the existence of the radio station’s signal when the Oscillator’s phase shift becomes ±90 deg, ±270deg. (=π/2 [rad],  3π/2). 

In general, Carriers of transmitter’s of radio stations and radio’s oscillators can’t be synchronized and signal strength Vr changes from -1*Vr to +1*Vr.  This subject have not always known very well.

Almost of radios in the market may not take care of this phase-shift subject in their designs. So called “AM synchronized detector” can solve it. But many radios may have had this potential subject even now.

Traditional diode detector radios don’t have this subject because it works by the carrier that is derived from received original carrier (not shifted) but this method has the subject that 2nd, 3rd, … N-nd harmonics distortions for the baseband signal are generated in it.

(3) Canceling the fading phenomenon by phase change by time

SSB (USB/LSB) signals don’t have carrier on them. So there is no way to canceling fading. But if we add weak carrier on them we can use it to fix this problem.

As AM signal have a carrier, it appears as phase angle on the complex plane.
We can cancel it by DSP calculations. It is similar as so called “AM synchronized detector” .

This “phase-shift” subject wrote on (2) must occur in transfer path between radio stations and this phenomenon have been so called “fading”, but it has not been fixed yet. It also happens in the ionized layer (D, E, and F layers) paths.
If radios have a function to shift phase of oscillator’s frequency, performance of radios can be improved.  

(4) Synchronizing phase for QAM

Phase of QAM signal is changed by the change of signal path and by the difference of initial phase of transmitter side and receiver side.
Synchronization for them is required. 

(5) Avoiding multi-path distortion for QAM

OFDM is designed to avoid multi-path distortion by using multi-sub-channels of carriers that are orthogonal function for time variable “t”. It may be difficult to remove multi-path distortion of QAM that has single channel of carrier by DSP operations at the current technology. Because it takes long time to process to remove duplicated multi-path distorted signals.
It may be difficult to send fast digital signal on narrow band of HF. So degreasing speed of BPS may be some kind of solution for this subject.  Or we may define some small multi-sub-channels of carriers that are orthogonal on the complex plane signal.


Appendix;
A. Actual work of SSB demodulation
 




B. An example of assembling of "IQ demodulator"


------------
Revision:
Preliminary public version : Jul.13, 2016 ; 
This version may still contain some editorial, mathematical, or technical problems.  
I recommend you read my stories carefully. This is under construction :)



(Return to index)

2016年7月12日火曜日

What was wrong in designing AM transmitters ? (preliminary)


Fig.1 A simulation result of "Low powered AM modulation" circuit on LTspice (4).
(LTspice(4) is a spice simulator developed by Linear Technology inc.)

Fig1. is an AM/FM modulator component that can generate AM modulated signal voltage or FM modulated signal voltage. It's embedded in LTspice (4) .

In this simulation, I used it as an AM modulator. 
The directive "SINE(40m 40m 1K)" means the amplitude signal voltage as input audio signal is 40mV and DC offset voltage 40mV.  Input signal of sine wave 1KHz is swinged on the DC offset voltage 40mV, and is modulated with 1MHz carrier frequency.

In this case, the AM modulation rate = 100% and quality of the modulated output signal voltage is very good and it is almost perfect.
On the FFT of Fig1, there is almost no distortion in this work and is almost ideal work. 


Fig.2  Block diagram of "Low powered AM modulator"

Fig.2 is a block diagram of "Low powered AM modulator" that explains how it works.
I thought it out by myself from the equation (1) of AM modulation signal voltage.

Vout=(Vdc+x(t))*(Vc*sin(ωc*t)) ... (1)

Here,
x(t): baseband (audio) signal voltage to be fed to the AM modulator .
        On the Fig.1, x(t) = Vin = V3 = V0*cos(2*π*f0*t) ...(2)

Vdc   : DC offset voltage V2. Vdc is often defined as DC 1V in other explanations.  
Adder: Analog signal adder input Vin and V2 and output the added voltage signal as V4.
V1 : Oscillator of sine wave to generate carrier signal for AM signal.
        V1 = Vc*sin(ωc*t), ωc= 2*π*f1*t ...(3)
Mixer (Multiplier) : Analog multiplier. It inputs V4 and V1, then does multiply V4*V1, 
                                  and output V4*V1 to Vout  
Vout : Output signal voltage

As described above, the principle of generating AM signal is very simple.
Once we know the principle beforehand, it makes designing or building AM transmitter/transceiver be quite easier. 


Fig 3.  Collector AM modulation method at the Final transistor used in RJX-601
(Copyright by "Matsushita Denki Corp." in about 197x(?)-198x(?))

RJX-601 is very popular products, many people of Amateur radio licensed love it. 
It can send 1W or 3W of AM or FM E/M wave on the frequency 50MHz to 54MHz.
The specification was the best one on those days.

However it has  some subjects on electronic work in sending AM modulated signal.
When RF power selection is set to 1W, AM modulation is good. 
But it is set to 3W, it generates low modulated AM.
This phenomena was so called "Minus modulation" .

Why "Minus modulation" occurs ?
Following is my analysis of the "collector modulation method at the final transistor".


Fig.4 Abstract circuit of BJT transistor's collector AM modulator

Fig4. is an abstract / equivalent circuit of BJT transistor's collector AM modulator.  
Please watch Fig.4 carefully comparing with Fig.1.

(1) Optimized DC bias voltage of VB is fed to base terminal of the final transistor.
      This bias voltage must be set to class AB or class A to amplify with good linearity. 
      Class C is impossible to use in this application. [2]

(2)RF sine wave voltage source such as 50.6MHz is fed to base terminal of the             
    TR(Transistor). Assume Z of Vrf here = 0 ohm and base terminal has some resistor
    value such as 500 ohm.

(3)Audio Power Amplifier is required to amplify small signal from a microphone to become enough power such as 1W to 5W and drives the 1st side of "Audio power modulation trans".

(4) Vcc = 12V to 13.8V is fed to the 2nd side of "Audio power modulation trans" and amplified audio signal AC voltage rides on the DC Vcc line. This audio modulation signal is connected to "Zc". 

(5) "Zc" is assumed to use ferrite beez such as FB-225. The ferrite beez acts as resistance of hi-Z and inductance value is very low on RF frequency domain. 
It makes the TR to be stable as amplifying device. 

Traditional circuit had used LC tuned tank circuit in the application of vacuum tubes.  But LC tuned tank circuit makes the TR works to be very unstable.  

Because the TR cannot work on such hi-Z stable,  it causes abnormal RF oscillation and TR goes to heaven. It often causes permanent break downs of very expensive TR.

(6) LC Z-matching section must be adjusted to be terminal impedance properly such as 50 ohm.  We can never use LC Hi-Z tank circuit here. 
(Be careful. LC tank circuit is for vacuum tubes but not for TRs.)

(7) Feeding audio amplified signal rides on Vcc=DC 12V to 2 stages of TRs audio amplified is not correct theoretically according to the principle of AM modulation method mentioned above (See the equation(1)).


Fig.5 Visualized work of AM modulation on BJT TR

Fig5. is visualized picture of AM modulation on BJT TR on the final stage of RF power amplifier.
This explanation is based on the traditional design theory of  TR amplifier  with "Load Line method" that was known in 1964.[1]

In this load line design method, TR bias point is set to saturation area of Collector Current "Ic".
Now assume Vcc=12V and there is no audio amplified signal on Vcc on the Load Line.

50.6MHz RF sine voltage signal comes in to TR and amplified along with the Load Line.
There is still no problem.

In this status, input audio signal. Then the Voltage of collector terminal of TR is swinged deeply by the amplified audio signal voltage.
Since Vcc is swinged by the amplified AC voltage, and Load Line is moved by it from left of Vcc and/or to right of Vcc widely.

This means proper bias point of TR moves dynamically and the proper bias to become bad bias point.
As the characteristics of BJT TR, Ic increases with very small amount and it is not linear work with big distortions. The TR can't work as an analog multiplier on this Ic saturated area.

This is the reason why "Minus modulation" occurs.
There is enough reason traditional designed circuit caused "Minus modulation" when 3W output is specified by the power selectable switch. 
There was a reasonable design error.  :-)

How should we correct the traditional design to become well-designed one?
From the equation (1)',  Fig.6 can be introduced as followings. 

Fig. 6  AM modulation by Analog Multiplier

                     Fig. 6A  AM Collector modulation by BJT TR 2N2222 used as an "Analog Multiplier"
(This types of design works with bad quality of AM with distorted AF tone.
The characteristics of BJT TR as an "Analog Multiplier" is not always good.)


Vout=(U0+x(t))*(Vc*sin(ωc*t)) ... (1)'

Here,
x(t): baseband (audio) signal voltage to be fed to the AM modulator .
        On the Fig.6, x(t) = V3 = Va*cos(2*π*f0*t) ...(2)'
V3 is actually audio power amplifier to supply modulation Voltage source with DC offset U0 by the Vcc line (eg. 12V).

U0   : DC offset voltage V2.  (=Vcc=12V)
Adder: Audio power trans here works as an adder.
It inputs V3 and U0, add them and output the voltage signal Vin to Multiplier.

V1 : Oscillator of sine wave to generate carrier signal for AM signal.
        V1 = Vc*sin(ωc*t), ωc= 2*π*f1*t ...(3)'
Mixer (Multiplier) : Analog multiplier. It inputs V1 and Vin, then does multiply V1*Vin, 
                                  and output V1*Vin to Vout  
Vout : Output signal voltage

The equation (1) and (1)' have the same meaning as AM modulators. The deference between them is that they are used in the low power earlier stage of the transmitter or used in the high power last stage.

I've heard that the principle of the AM modulation method by Fig. 6 had been used in the "Plate modulation method" by Final vacuum tube.

I"d like to think, maybe, understandings for multiplier's characteristics of vacuum tube had not always been understood well in those days.

From as my beginners view, TR collector AM modulation method is simply seemed to be exchanged the vacuum tube of plate modulation to a BJT transistor instead.

However, there are big difference of electric characteristics between the vacuum tube and BJT TR.
BJT TR is known as the active device that has the characteristics of CCCS(Current Controlled Current Sources).
Beside, the Vacuum tube is known as the active device that has the characteristics of VCVS(Voltage Controlled Voltage Sources.)

Unfortunately, I've not heard or read of the characteristics as a Multiplier of the old fashioned vacuum tube.
I'm wondering if nobody has been interested in the equation (1) or (1)' in past so long years ?

Once we notice the characteristics of a multiplier is essential for AM modulation from the equation (1) or (1)', we can fix the problem as a subject of Math.


As next step, I thought is about how can I use a BJT transistor as a multiplier.

Generally, BJT transistor's amplifier is designed on the "Load Line" where the collector current is saturated. [1]
In this saturated collector's current working area, BJT TR cannot work as an analog multiplier because the Ic is saturated and increasing VCE cannot increase collector current.

By the way, how does BJT TR work in front of Ic saturated area ?


Fig.7 BJT TR's characteristics before collector current saturated.

Fig.7 shows the BJT TR's characteristics that Ic increases according to VCE increases before Ic is saturated.
Using this Vce area of BJT TR is not recommended by the old fashioned design book.
In the book, Ic was saturated in very low Vce such as 0.3 - 0.5 V.

How it work in the actual BJT TR such as 2SC1815 very popular transistor in Japan ?
  

Fig. 8   Ic vs VCE characteristics of 2SC1815

From the DC sweep of simulation result of Fig.8, BJT TR 2SC1815 has wide range Vce = 0 to 5V with straight linearity of Ic increase.
This implies one (1) 2SC1815 can be used as an analog multiplier.

Fig.9   1 TR 2SC1815 (as an analog multiplier) AM modulation works

Fortunately, it works well. This circuit has benefit that it is not necessary to use a big audio modulation power trans. LM386 can drive the final TR 1815 directly, and AM modulation works well.



Next topics is also serous subject on so called "Base AM modulation method".

Fig.10  Base AM modulation transmitter not work at all

"Base modulated AM method" is very famous even now and the circuit of Fig.10 was found on the internet that seemed to be designed by a professional engineer.
Unfortunately, I could not get any good result on the spice simulation for it.

I had some experiences that the base modulation by a BJT TR didn't work well on the actual circuit. It could never get good result also on it.
I knew that these unwanted situations could be re-confirmed on the LTspice.

Fig.11 After improvement of the circuit of Fig.10

I tried to fix the design errors that I found on the Fig.10 base modulated AM transmitter.
And I managed to work it as Fig.11.
But the quality of modulated RF signal was so dirty.

I know Base AM modulation method has been very famous design but I have no solution to get good quality of AM modulated RF wave even now.
I have lost interest on this method until now.


New design by CQ Pub.inc used Gilbert-cell configured BJT TR AM modulation.

Fig.12 First ever Gilbert-cell type AM modulator
(Copyright by "Transistor Gijyutu" by CQ Pub.inc. 2015)

Analog multiplier configured by Gilbert-cell has invented in US, 1968 as far as I know.
It has very good quality as an analog multiplier and implemented in many DBM (Double Balanced mixer) ICs.
Fig.12 circuit  is the best result of AM modulation method configured by BJT TRs I've ever seen.


Fig.13  Works of C class amplifier
(Copyright by USAF 1963, Mr. Jeff Quitney, Mr. Youtube)

I cannot find application examples of C class amplifier recently.  
I know C class amplifier works with big distortion then it is used to generate harmonic waves as distorted RF output signal such as x3, x4, x5 frequency multiplied. 

I think it is impossible to generate good quality of AM modulation signal from the principle of AM modulation, but very famous Radio magazines for radio fans had been told us that C class amplifier worked very well for AM modulation even until now.

I have no experience that C class amplifier without bios voltage got any good quality of AM. I have no idea to realize to use C class amplifier for AM modulation. 

 From the view of the principle and the characteristics of C class BJT TR amplifier, it's impossible. 
But they say "C class amplifier is very easy to make AM transmitters." 
I cannot find any valid evidence for them.

There are lots of circuit examples of C class ( no bias) amplifiers on the internet even until now but I cannot find any reasonable result from my actual experiments and my calculation by Math. And every simulation by LTspice result tells me that it doesn't work at all. 


Fig.14  Works of AB class amplifier
(Copyright by USAF 1963, Mr. Jeff Quitney, Mr. Youtube)

Class AB is set in RJX-601 on Fig. 2.  Maybe, they knew that Bias voltage was essential for AM modulation by their experiences.

Fig.15  Works of A class amplifier
(Copyright by USAF 1963, Mr. Jeff Quitney, Mr. Youtube)

Class A amplifier requires lots of power consumption even when it is idling. So it is not good for AM transmitter.
This video tells me vaccum tube was used in the linear amplifier, and vacuum tube has a characteristics where there are linear VCVS(Voltage Controlled Voltage Source) working area. 

It also implies vacuum tube has a good characteristics as an analog multiplier.
This characteristics gives reasonable reasons why "plate AM modulation method" works very well.  Beside, BJT TR doesn't have a good characteristics as an analog multiplier. This is big difference between BJT TR and vacuum tube.

Vacuum tubes are often low-estimated as old-fashioned devices but they still have very good linearity and strongness for heavy duty usage for high power RF amplifiers. 
Reliability of vacuum tube can be much better than TR's  in the high power applications.

Fig. 16 is product named "PanaSky 6". I have a few experience to use it when I was a school student. It could send 10W high power AM modulation by vacuum tubes.
An elder student was saying "Don't low-estimate it. It has very good performance." 



Fig. 16 PanaSky 6
(Copyright by Mr. hamlife.jp and  the production vendor of "PanaSky6" )


(Copyright by the production vendor of "PanaSky6" )


I have seen the the product named "FDAM-3" once.  I don't know it well.

(Copyright by ICOM inc. "Inoue Denki-seisakusho-san")


(Copyright by ICOM inc. "Inoue Denki-seisakusho-san")

(Copyright by ICOM inc. "Inoue Denki-seisakusho-san")


Reference:

[1] Amplifiers:

Construction of Load Lines 1971 USAF Electronics Training Film



[2] Electronics:

"Basic Amplifiers" pt2-2 1963 US Army Training Film



[3] Edwin Howard Armstrong:

The Yonkers Man Who Made Radio and the Alpine Tower



[4] RCA Radio Model RC Operation



[5] HOW RADIO WORKS - 1943


( [1] to [5 ]Copyright by USAF 1963, Mr. Jeff Quitney, Mr. Youtube)

[6] MIT 6.003 "Signals and Systems", “Modulation part1, part2” by  Prof. Dennis Freeman


Related documents:


(1) Basic experiments for the principles of AM modulation methods


(2) Basic experiments for “AM modulation method at Base-terminal of BJT TR”


(3) The principle of “Low powered AM modulation method” (How it works)


(4) The principle of “AM modulation method at Collector terminal of Final BJT TR” (How it works)


(5) New designed AM transmitter by CQ Pub. inc.


(6) Vizualization of AM modulated valtage signal on gpaphs


(7) Multiplier (Mixer, Frequency converter) related documents


(C) Bakabon Mama and Noboru, Ji1NZL Jul.12, 2016 

Rev 0.1,  Oct.10, 2016
Fig.6 updated: Add an "adder" in Fig.6 to understand easier. 


(Return to index)

2016年7月2日土曜日

IQ直交変調器を用いたウェーバー法によるPSN方式 SSB変調器/送信機の原理


IQ直交変調器を用いたウェーバー法によるPSN方式 SSB変調器/送信機の原理

 概要

従来方式の課題:

PSN方式によるSSB変調方式の課題として次のものがあった。

(1)ベースバンド低周波信号の-90度位相シフト器が、ベースバンド帯域内で利得が一定にならない。

例:

Poly Phase Filterと呼ばれるRC構成による 300Hz - 3KHz 周波数帯域に設計された位相シフト器は、帯域内で利得が変化し、特に低音領域の利得が高く、高い音域で利得が低下するため、太い声で耳障りとなり、高い音が弱くなり、聞き取りにくい音声になる。

(2)同Poly Phase Filterは、設計帯域を外れた周波数帯域 20Hz - 300Hz の低い音の位相ずれが-90度を維持できず、低音域がおかしな音になる。

こうした課題を解決する方式として、Poly Phase Filterを使用しないウェーバー法(Weaver Method)が知られていたが、その動作原理説明は、実回路開示や、難解なタイミング図と難しい長い文章による解説で、方式を正確に理解することが著しく困難だった。

(3)ウェーバ法による方式上の課題(問題点)が開示されていなかった。

(4)方式が不明または曖昧な理解であるため、SDRの設計ができない。

ここでは、以上の課題を反省し、これまでの課題を克服するため、計算式によりウェーバー法を導出して求め、やさしく、かつ正確に同方式を定義し、SDR設計に直接使用できる設計内容を明らかにする。

2. 構成

図1. による。



図1

3. 方式の計算式導出と説明

説明は次の文書による。

”The principle of SSB modulator / transmitter of “Weaver PSN method 
by using “IQ feeder” and “IQ modulator”  (Weaver’s PSN method)
http://ji1nzl-official.blogspot.jp/2016/07/the-principle-of-ssb-modulator.html


4. 課題

項番号 3.により、ウェーバー方式には設計上、次の課題の存在が明らかになった。

(1) 現在SDRに多用されているIQ直交変調器が、ウェーバー方式にもそのまま使用できる。
ウェーバー方式では、IQ変調器に、90度位相差があるSSBの2信号I,Q信号を与えるだけで、SSB(USB/LSB)の変調・復調器が実現できる。

(2) ウェーバ方式で利用されるLPF(Low Pass Filter)は、LSB,USBの不要成分を除去するため、良好なフィルタのキレ具合と、帯域内一定の群遅延特性を持つ必要がある。

これは、FIR,IIRフィルタの信号処理で実現できる。

(3) 上記LPFは、アナログ回路では、良好な特性を得る回路実現に難易度が伴う可能性がある。

(4) ウェーバー方式は、音声などのベースバンド信号を一旦、40KHz程度の中間周波数帯に引き上げるため、この局部は発振器周波数だけ、第二局発の発振周波数に対し、キャリア周波数がIF周波数分だけシフトする。

受信機側では、SSB復調時に、キャリアが無いため、そのシフト分は気づくことは無いが、送信機の周波数表示処理は、このシフト周波数をずらして表示する加算または減算処理が必要となる。

(5)上記LPFは、HPFで交代するように方式変更も可能である。ただし、キャリアシフト量は変化する。

付録:
A. SSB/CW SDRトランシーバの実現例

       G11 Gunu Radio based SDR Transceiver
                               Designed by Goran in Eu. (Respective copyright reserved) 
               

(C) Noboru, Ji1NZL, Jun.29, 2016

Rev.0.0 July 2, 2016 初版文書作成 Noboru Noboru .AE35.macbook

The principle of SSB modulator / transmitter of “Weaver PSN method by using “IQ feeder” and “IQ modulator” (Weaver’s PSN method) (preliminary)

The principle of SSB modulator / transmitter of “Weaver PSN method 
by using “IQ feeder” and “IQ modulator”  (Weaver’s PSN method) (preliminary)

1. Architecture



“Weaver PSN method modulator / transmitter” has 3 functional blocks.

(A) “Baseband signal generator” … Eg. Microphone and/or with small signal amplifier(s)

(B) “IQ feeder for baseband signal” inputs baseband signal from “Baseband signal generator” , converts it IF frequency stage such as 10KHz to 40KHz and outputs two orthogonal voltage signals “I” and “Q”.

(C)“IQ modulator” inputs two port of orthogonal signals “I” and “Q” and outputs to 1 port of SSB modulated signal.

2. How it works

(A) Baseband signal generator

    Assume V1 is “baseband signal” to be input as audio signal. 
    (eg. We can use microphone to input V1.)

    V1 = Vs*sin(ωs*t) …(1)
    here ωs = 2πfs  …(2) 
          fs [Hz] is 0 to 20KHz Frequency of baseband signal on Audio frequency        ( 2.6KHz to 3.0KHz is preferred in the actual application as the maximum  
          frequency of baseband AF signal. )

(B) IQ feeder for baseband signal

OSC1 generates ωc1[rad*Hz] sin wave voltage. 
ωc1 is fixed frequency and OSC1 is used for first frequency conversion. 

Vosc1 = Vc1*sin(ωc1*t) …(3)
here ωc1 = 2πfc1 …(4) 
 fc [Hz] is set to IF frequency such as 40KHz to covert baseband frequency ω1  
          to ωc-ωs or ωc+ωs.

Shift Vosc1 signal to be -90 deg. by “-90 deg. Shifter” to get V2.

V2 = Vc1*sin(ωc1*t-π/2) = -Vc1*cos(ωc1*t) …(5)

Get V3 by multiplying V1 and Vosc1.

V3= V1*Vosc1 = (Vs*sin(ωs*t)) * (Vc1*sin(ωc1*t)) 
  = Vs*Vc1*sin(ωs*t) * sin(ωc1*t) 
    =Vs*Vc1*(1/2){ cos(ωc1*t+ωs*t) - cos(ωc1*t-ωs*t) }
    =(1/2)*(Vs*Vc1)*{ cos((ωc1+ωs)*t)) - cos((ωc1-ωs)*t)) } …(6)

Set cut frequency ωcut of LPF to be ωcut < ωc1, 
then cos((ωc1+ωs)*t) can be removed. (->*Note1)

Then get V4,
V4 = -(1/2)*(Vs*Vc1)*cos((ωc1-ωs)*t)) …(7)

...Output of “I” came out here.

Get V5 by multiplying V1 and V2.
V5= V1*V2 
(Vs*sin(ωs*t)) * (-Vc1*cos(ωc1*t)) = -Vs*Vc1*sin(ωs*t) * cos(ωc1*t) 
    = -Vs*Vc1*(1/2){sin((ωc1*t+ωs*t) + sin((ωc1*t-ωs*t))}
    = -Vs*Vc1*(1/2){sin((ωc1+ωs)*t)  + sin((ωc1-ωs)*t))} …(8)

Set cut frequency ωcut of LPF to be ωcut < ωc1, 
then sin((ωc1+ωs)*t)  can be removed. (->*Note1’)

Then get V6
V6 = -(1/2)*Vs*Vc1*sin((ωc1-ωs)*t)) …(9)

... Output of “Q” came out here.

(C) IQ modulator

Input voltage of “I” and “Q” for “IQ modulator (C)” here are given 
as output of “IQ baseband feeder (B)”. 

OSC generates ωc2[rad*Hz] sin wave voltage.
Vosc2 = Vc2*sin(ωc2*t) …(10)
here ωc2 = 2πfc2 …(11) 
          fc2 [Hz] can be set to RF frequency of VFO such as 14MHz.

Shift Vosc2 signal (10) to be -90 deg. by “-90 deg. Shifter” and get V7. 

V7 = Vc2*sin(ωc2*t-π/2) = -Vc2*cos(ωc2*t) …(12)

Get V8 by multiplying V4 and V7.

V8= V4*V7 = (-(1/2)*(Vs*Vc1)*cos( (ωc1-ωs)*t)) ) * (-Vc2*cos(ωc2*t)) 
                  = (1/2)*Vs*Vc1*Vc2*cos( (ωc1-ωs)*t)) ) * cos(ωc2*t) …(13)


Get V9 by multiplying V6 and Vosc2.
V9 = V6*Vosc2 = (1/2)*Vs*Vc1*cos((ωc1-ωs)*t)) * Vc2*sin(ωc2*t)
    =  (1/2)*Vs*Vc1*Vc2*sin((ωc1-ωs)*t)) * sin(ωc2*t) …(14)

When “Adder” used before Vout,
Vout = V8+V9 =  (1/2)*Vs*Vc1*Vc2*cos((ωc1-ωs)*t)))*cos(ωc2*t)      
                        + (1/2)*Vs*Vc1*Vc2*sin((ωc1-ωs)*t)) * sin(ωc2*t)
= (1/2)*Vs*Vc1*Vc2*(cos((ωc1-ωs)*t)))*cos(ωc2*t) 
  + sin((ωc1-ωs)*t))*sin(ωc2*t)
= (1/2)*Vs*Vc1*Vc2*cos( (ωc1-ωs)*t - ωc2)*t) )
= (1/2)*Vs*Vc1*Vc2*cos( (ωc1-ωc2)-ωs)*t )  … here cos() is even function then 
= (1/2)*Vs*Vc1*Vc2*cos( (ωc2-ωc1)+ωs)*t ) …(15)

  … This (15) means USB generated as output signal Vout. 
     (15) is USB signal and Carrier ωc2-ωc1 signal voltage is removed here
      and receivers have to set tuning Carrier frequency to be ωc2-ωc1.
            -> (*Note 2)

When “Subtracter” used before Vout,

Vout = V9-V8 = (1/2)*Vs*Vc1*Vc2*sin((ωc1-ωs)*t)) *sin(ωc2*t)    
             - (1/2)*Vs*Vc1*Vc2*cos((ωc1-ωs)*t)) *cos(ωc2*t)
 = -(1/2)*Vs*Vc1*Vc2*{cos((ωc1-ωs)*t-ωc2)*t)-sin((ωc1-ωs)*t))*sin(ωc2*t) }  
 = -(1/2)*Vs*Vc1*Vc2*({cos( ((ωc1-ωs)+ωc2))*t )  
 = -(1/2)*Vs*Vc1*Vc2*({cos( ((ωc2+ωc1) -ωs))*t ) …(16)

 … This (16) means LSB generated as output signal Vout. 
    (16) is LSB signal and Carrier ωc2+ωc1 signal voltage is removed here
    and the receiver has to set running Carrier frequency to be ωc2+ωc1.
          -> (*Note 3)

“Weaver’s PSN method” by using “IQ baseband feeder” and “IQ modulator” 
is proved to generate SSB modulated signal ( USB or LSB).

-> (*Note1, Note1’) : 
DSB(Double sided band) is generated by the Multiplier 
on the “IQ feeder” and  LPF can remove USB band theoretically. 

However Characteristics of “Shape factor” and “signal group phase delay” 
on the IF signal band is very important to design the LPF.

Otherwise, unwanted signal may remain on the IF signal line V4 or V6. 
This subject is derived from the principle of the “Weaver’s PSN method” 
but DSP filter’s characteristics can solve this subject.
(Analog type of LPF or HPF may be difficult to solve it.)

HPF can be used instead of LPF on this method as the same principle.

-> (*Note 2)(*Note3) ; 
    SSB receivers have to their tuning frequency to become shifted  
    carrier by Vosc1. 
    This subject can solve easily by the tuning frequency adjustment  
    at SSB receivers side.

(C) Noboru, Ji1NZL, Jun.29, 2016

Rev.0.1 : 
Add missing variable "Vc1" in the equation (16) (mistypes) Jul.5, 2016
(No change of the theory itself. I'm sorry for it.)

Rev.0.1A: 
Changed variable ωc2 to ωc1 in the equation (16) (mistypes) Jul.6,2016
(No change of the theory itself. I'm sorry for it.)

Return to INDEX