From: apogosso@tpg.com.au   
      
   Modulation handling of an AM detector is determined by:   
   1. Carrier level and low level sensitivity;   
   2 AC impedance / DC load resistance ratio;   
   3. Slew rate.   
      
   If say we are aiming at 95% undistorted demodulation, then the carrier level   
   shall be about 20 times greater than the sensitivity threshold. For example,   
   for an unbiased detector with a silicon diode, which has a 0.6V knee, we   
   need at least 12V carrier. (Unless otherwise stated, amplitude is implied.)   
   For a properly designed vacuum diode (Vsens = 100mV) you need at least 2V   
   carrier.   
      
   For an optimum biased semiconductor diode (Vsens = 25mV) you need only 0.5V   
   of carrier for distortion free 95% demodulation. Thus a radio using a   
   PROPERLY designed p-n diode detector can run at low IF levels, have less IF   
   gain, less IF stages, would require less shielding and have better   
   sensitivity.   
      
   It is well known that for large signals maximum modulation index   
   approximately equals to AC / DC load impedance. The ways to make it close to   
   1 at lower frequencies are:   
   - use a DC coupled hi-Z follower (cathode, source, op-amp buffer, high-beta   
   BJT with low base current, etc.);   
   - never AC couple a volume control to the detector. Instead, make the volume   
   control pot *the* load (or a part thereof);   
   - have a high (5...10M) input impedance of the audio amp, so that it does   
   not load the detector even at full volume. (This stuff has been mentioned in   
   a different thread);   
   - never connect AGC voltage RC filter to the detector load. Use a separate   
   AGC detector.   
      
   In an unbiased detector slew rate issues are also related to the AC/DC   
   impedance ratio. For example, a booooring detector with 500K || 100pF load   
   can handle 70% at about 3.5kHz only. To get things worse, a 47K+100pF ripple   
   filter is added straight after the detector. This cap virtually adds to the   
   effect of reducing the modulation handling at the highs. With 100pF load +   
   100pF filter the 70% cutoff comes down to the appauling 1.8kHz. No wonder   
   the sound of a booooring radio is so crappy.   
      
   There is another less known effect. Not only the heavy C distorts HF   
   modulation due to slew rate limitations, it also reduces original modulation   
   index at HF, working like sort of high-cut tone control. To understand the   
   phenomenon without a deep maths, consider that on a steeply rising RF   
   envelope the detector has to charge the load capacitance. This sucks extra   
   energy from the hi-Z IFT on top what is to be dissipated in the resistive   
   component of the load. On the steeply falling RF envelope slopes, R is being   
   fed from a discharging C instead of the diode. Thus the IFT gets unoaded on   
   the falling slopes. It is easy to see that the peaks are thus "cut and   
   rounded" and the troughs are "filled". Modulation virtually reduces. This   
   reduces the slew rate distortion, replacing it with a HF cut. In the end the   
   sound is still crappy. (Those who are familiar with the operation of a ratio   
   FM detector, as opposed to a Sheeley discriminator, will see many   
   parallels.)   
      
   The above HF unmodulation takes place only if the detector is directly fed   
   from a hi-Z IFT. If a buffer (cathode) follower is used, there is no HF   
   unmodulation phenomenon.   
      
   In unbiased detectors there is no high limit to carrier level (until the   
   diode breaks down) -- discharge current is proportional to the carrier   
   level. Slew rate perfomance does not depend on the RF signal magnitude.   
      
   This is not the case with biased detectors, where the discharge current is   
   (almost) constant. Here it is time to analyse the famous Partick's biased   
   detector.   
      
   Schematic values may vary, but here let us assume it is biased to 50V, uses   
   a cathode follower on 12AU7, a semiconductor diode, has 220pF of C,   
   pull-down resistor of 500K and a ripple filter of 100K+100pF. Thus the diode   
   bias current is about 100uA at no signal.   
      
   A DC biased diode, as I hope people know, has differential resistance of   
   (25 ohms / current, mA). In this case, at no signal and at very low signal   
   the diode acts as a 250R resistor. Capacitor of 220pF has about 1.6K   
   reactance at 455kHz. Thus the whole D+C circuit presents itself as about   
   1.7K impedance to the cathode of the cathode follower.   
      
   Now, what happens when a RF signal is applied to the D+C detector? As   
   someone wisely remarked, the detector will begin to detect if the diode in   
   NOT conducting continuously. This will happen when the AC component of the   
   current exceeds DC component (100uA). With 1.7K D+C impedance this will   
   happen when the RF signal reaches 170mV.   
      
   A 12AU7 tube has low transconductance and the followr probably has about   
   500R of output impedance. Thus it will probably require 200...220mV of the   
   signal on the grid for this detector to start working. Not really impressive   
   sensitivity. As has been shown above, to handle a 95% modulation, this   
   translates into at least 4V of carrier.   
      
   Because of the discharge current (100uA) is constant, HF performance depends   
   on the carrier level. For example,   
   - at 4V carrier it can handle 70% modulation to 15kHz and 95% to 10kHz;   
   - at 10V carrier it can handle 70% modulation to 6kHz and 98% to 4kHz;   
   - at 20V carrier it can handle 70% modulation to 3kHz and 99% to 2kHz, etc.   
      
    (Here 220pF+100pF was considered as an audio frequency load).   
      
   The performance is remarkably better than of a booooring unbiased detector!   
   The only disappointing thing is that the stronger the signal (from a local   
   station), the worse is the modulation HF margin. You would expect the   
   opposite from a detector for hi-fi application. The only remedy is to use a   
   strong amplified delayed AGC to maintain 4...5V carrier for a station of any   
   strength.   
      
   (I use an active integrator based AGC amplifier for that purpose, so that   
   all the stations are levelled up to the optimum level, but Partick's AGC   
   into the mixer only is very primitive and inefficient. Besides it introduces   
   more distortion by nonlinearly loading the IFT. However it may be a separate   
   thread to discuss.)   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)