From: apogosso@tpg.com.au   
      
   There seems to be some interest in an AM detector, so I decided to write   
   something up on this subject.   
   We can skip more obvious things related to high level carrier operation,   
   such as:   
   - effect of the internal resistance (permeance);   
   - effect of the RF ripple;   
   - effect of the diode capacvitance.   
   By these specs semiconductor diodes outperform tubes.   
      
   Rather we focus on the less obvious and more controversial issues like:   
   1. Sensitivity to low signal;   
   2. Handling deep modulation;   
   3. Handling high frequency modulation (slew rate issues).   
      
   ==== PART 1 =====   
   Which diode is better?   
      
   Common perception is that a silicon diode is "very bad" for small signal   
   because it has a knee as high as 0.6V. Germanium diode, which conducts at   
   lower bias of 0.1...0.3V, is better, people say. However, some complain   
   about leakage of germanium diodes which sometimes kills a detector.   
      
   Opinions about vacuum diodes greatly vary. Some say that a vacuum diode   
   detector is the best for a small signal, since the diode is always slightly   
   conducting even at no signal because of the thermionic emission. Thus, it   
   has no "knee" and a detector they say is extremely sensitive. Yet others   
   blame a tube diode for it reverse biases itself! Hot electrons reaching the   
   plate deposit negative charge on it which reverse biases the diode   
   preventing it from demodulating weak signals. A load resistor, which is   
   rather large (500K) is not sufficient to fully "drain" this emission   
   current.   
      
   Where is the truth?   
      
   Well, small signal sensitivity of a detector diode is determined by   
   sharpness, or curvature of the Volt-Amp characteristic of the diode where   
   transition occurs from a non-conduction (horizontal flat line) to conduction   
   (rising curve). Position of the knee is not so important if we assume that   
   the diode can be biased right to the knee point or to the point of the   
   maximum curvature. After that, the sharpness of the bend is all that   
   matters.   
      
   Conduction of a diode, whether vacuum or semiconductor, depends on the   
   electrons overcoming a certain energy barrier, whether a work function of a   
   cathode or bandgap. Even if the holes are the major carriers, still it is   
   the electrons which actually carry current. If temperature was low (close to   
   0K) and the electrons did not have thermal movement, then once the bias   
   (field) exceeded a certain level, all the electrons would start moving. That   
   would be an ideal diode with the extra sharp transition from non conduction   
   to conduction.   
      
   In reality the electrons are in thermal motion, some faster, others slower.   
   With the bias increasing, higher percentage of the electrons, assisted by   
   their thermal kinetic momentum, are able to overcome the barrier. Thus a   
   real diode gradually goes fron non-conductance to conductance. The higher   
   the temperature, the less distinctive the transition.   
      
   Physisists have shown that the conduction assosiated with overcoming a   
   barrier is exponential, approximately:   
      
   I = Io * EXP ( eV / kT)   
      
   where Io -- some kind of a coefficient;   
   e -- charge of the electron;   
   V -- bias voltage;   
   k -- Boltzmann constant;   
   T -- absolute temperature in degrees Kelvin.   
      
   At room temperature (T = 300K) conduction current in a diode would increase   
   by e times (approx. 2.7 times) per 25mV bias increase. In other words,   
   within 25mV of bias change we see nearly 3 times difference between   
   "forward" and "reverse" conduction. This is enough for a detector to rectify   
   25mV signal reliably. Thus let us assume that low signal sensitivity of a   
   semiconductor diode is 25mV.   
      
   In the tubes the catode runs at 1200...1300K -- about 4 times hotter.   
   Obviously, for a tube diode it will take about 100mV bias increase for the   
   current to increase by 2.7 times. Thus a vacuum diode theoretically 4 times   
   less sensitive than a p-n diode. Low signal sensitivity of a vacuum diode is   
   100mV. Some people try to improve the vacuum diode by running its cathode at   
   lower temperature. It helps, but not that much, because you can not reduce T   
   to 900...1000K, otherwise the rising internal resistance will make the diode   
   useless at strong signals and cathode poisoning will increase. Perhaps you   
   can squeeze 75mV sensitivity of a vacuum tube.   
      
   As I said, it does not matter what the VA curve does at large signal. In a   
   semiconductor diode it remains exponential, and in a vacuum diode it turns   
   into a law of 1+1/2. Important is that the very beginning, the bend itself   
   is always exponential, sharpness of which is temperature dependant.   
      
   Note that the exponential law applies only to the very beginning of the bend   
   of a V-A curve which is essential to rectification. At higher currents the   
   V-A curve will change into the power 1 1/2.   
      
   So the conclusion is that a vacuum diode is worse than a semiconductor. Note   
   that the semiconductor diode needs to be properly biased right to the bend.   
   This bias is critical and must be temperature compensated. I will cover this   
   in the third part of the article. Also the issues of source loading have   
   been ignored. In other words, a perfect cathode follower is assumed.   
      
   A grid detector (grid leak detector) has the same sensitivity as the vacuum   
   diode. Those who are into regenerative radio design -- try to reduce the   
   cathode temperature as much as practical. So called "anode-bend" and   
   "infinite impedance cathode" detectors are in theory as sensitive as vacuum   
   diodes. In practice however, they are much worse. If someone is interested,   
   I can explain. But now I will just give a hint. Can you expect a good bend   
   if you are using a remote cut-off tube? Of course, not! There will be no   
   distinct bend at all. Now consider that practically any tube, even a sharp   
   cut-off one, has a certain component of "remotecutoffness" because of   
   inconsistency of its grid pitch.   
      
   ====================   
   In the next part I am planning to discuss benefits and drawbacks of biasing   
   the detector. As an example, for that analysis I will be using a famous   
   Patrick's detector biased to 30...50V with a weak pull-down.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)