The Xtal Set Society
The Xtal Set Society
profile | register | search | main site

email us
Forums | Rap N Tap Clubhouse - Basic Crystal Radio Fun! | toilet paper roll crystal radio Post Reply Send Topic To a Friend
Author Topic
homebrewPosted - 10 September 2013 9:52  Show Profile  Email Poster  Edit Message  
Graphs are easier to understand most of the time. I cannot completely agree with the article but it is a place to start.
gzimmerPosted - 10 September 2013 10:24  Show Profile  Email Poster  Edit Message

I agree that a diode does not have a "threshold" voltage. It actually has an exponential curve, change the bias current and you change the resistance in a smooth curve.
No elbow, no threshold.

However the trick of using a FET (or similar) stems from a different idea. The intent was to create a "more efficient" diode with a steeper curve.
Perhaps if the signal was applied to the Gate, the resistance change between the Source and Drain would change more rapidly that a normal diode.

Most FET's are either hard-on or hard-off without bias, so you need to apply adjustable bias so it can operate in the "sweet-spot" of the curve.
The logic behind using a "Zero Bias" FET as a detector was is was doped to give conduction with no bias, so hopefully it could be used without a bias battery.

My objection was that there is no guarantee that the Zero Bias point was actually optimum for this mode. Additionally the ZB FET was expensive, hard to obtain, was in a tiny package, and had an unwanted protection diode shunting the Gate.

Notwithstanding, the ZB FET detector did seem to work quite well.
However in my experiments, I had found that a humble RF FET was actually more sensitive, and having variable bias was actually a good thing rather than a hindrance.

But I guess that once something is published in QST, it goes down as Gospel.

.............. Zim

Edited by - gzimmer on 9/10/2013 11:05:21 AM

_J_Posted - 11 September 2013 1:14  Show Profile  Email Poster  Edit Message
I agree. The explanation I saw claimed it worked better because the diode had a higher "offset voltage"

John Davidson

_J_Posted - 11 September 2013 1:34  Show Profile  Email Poster  Edit Message
I did that once, I think that paper is still hanging around this site somewhere?

Let me try to break it down. For small signals, diodes don't conduct much more in the forward direction than in the reverse direction. They act too much like just a plain resistor until they get a lot of signal voltage across them.

We have this vision of a 1N34 that is a 10K resistor in the fwd direction and a 100K reverse. That is not reality for small signals, it is more like 49K fwd and 51K reverse, and that just doesn't detect substantially. That acts too much like a plain 50K resistor, not a diode at all.

Diodes that conduct easily in the fwd direction also leak a lot in the reverse direction. Diodes that don't reverse leak much also don't conduct well in the fwd direction. This is because all diodes follow the same basic physics.

So picking a good fwd conducting diode (a low Z diode, (like 1N34) fails because it reverse leaks too much. The only strategy that works is to get more voltage across the diode to drive it harder. With power limited to what the antenna gives, the strategy is to increase impedance to get more voltage. The diode that works best for this strategy is the one that does not fwd conduct well and does not reverse leak a lot. For the same antenna power, we transform the RF up in Z and maybe get a higher voltage across the diode and get 250 K fwd and 500 K reverse, now that can do a little detecting. That is known as a high Z diode. Make any sense?

John Davidson

Edited by - _J_ on 9/11/2013 1:48:12 AM

Garry NicholsPosted - 11 September 2013 10:2  Show Profile  Email Poster  Edit Message

Nice explanation! Thanks for the refresher.


_J_Posted - 12 September 2013 0:23  Show Profile  Email Poster  Edit Message
Another thing, because of their physics, all diodes follow the mathematical e**x equation (often called the diode equation). At first glance, the diode equation and the e**x curve looks like it has a knee, a point of max curvature. That is an optical illusion in the truest sense of the word. You can prove mathematically there is no max curvature. You can also see it by changing the scale on the axis on a scope and watch the "knee" move to a new point. There is simply no knee in any sense of the word. That is another total myth about diodes that misleads crystal radio designers a lot. It has absolutely no basis in fact whatever. Even more puzzling, there are so many engineers who even get emotional when you debunk their precious diode knee.

John Davidson

gzimmerPosted - 12 September 2013 3:42  Show Profile  Email Poster  Edit Message
Or draw it on a Log scale.

It becomes a straight line, the knee vanishes.

......... Zim

Garry NicholsPosted - 12 September 2013 19:22  Show Profile  Email Poster  Edit Message
A mathematical illusion?

Is it possible to explain what it is about this equation that makes it seem to have a knee in its graphical rendering?

I always thought that the graph of an equation represented the equation's independent and dependent variables accurately. In this case it would appear to create a misleading representation.

(My math skills are old and rusty.)


golfguruPosted - 12 September 2013 20:39  Show Profile  Email Poster  Edit Message  
One might argue that the theoretical "knee" might be the "transition point" where the "square law" mode kicks in?


Edited by - golfguru on 9/12/2013 8:41:56 PM

gzimmerPosted - 13 September 2013 4:54  Show Profile  Email Poster  Edit Message
Except that the mythical Knee would shift with bias, but the Square Law transition point doesn't do that.

(it would be great if it did, it would drastically increase Sensitivity if you could climb out of the Square Law region)

Of course RF bias DOES do just that, because it is forcing the diode to switch more cleanly. It becomes a Mixer instead of a Detector.

............ Zim

Edited by - gzimmer on 9/13/2013 5:22:05 AM

gzimmerPosted - 13 September 2013 5:3  Show Profile  Email Poster  Edit Message
Another way of looking at the "Knee".

If you apply a very small amount of RF, measure the output, and use that to calculate the diode's Resistance...

And then superimpose a variable DC bias and watch how the diode's Resistance changes as you adjust the bias...

You get a nice smooth change of Diode Resistance V/S Bias. No Knee.

And it is that relationship which decides how your Crystal Set operates.

Adjusting the DC bias, merely sets the Diode's Resistance.
Even at NO bias the diode still conducts a little.

............... Zim

Edited by - gzimmer on 9/13/2013 5:23:04 AM

gzimmerPosted - 13 September 2013 5:19  Show Profile  Email Poster  Edit Message
Yet another perspective:

What property allows the diode to Detect?
It is of course it's ability to Rectify, eg the ratio between the forward Resistance and the reverse Resistance.

With very strong signals, the diode is switching from hard-On to hard-Off, so the forward to reverse ratio is very high and the diode is very efficient as a Detector.

But as the signal gets weaker, the transition becomes less pronounced. At very low signal levels the ratio becomes closer to 1:1 and the detector becomes very inefficient.

If you examined the curve on the graph around zero, with a (hypothetical) microscope, you would see that at low magnification the curve is a sharp right angle.
But as you Zoom in, the elbow gradually changes to a softer curve, and then at very high magnification becomes almost a straight line.

Now go back to the original question, "what property allows the diode to Detect"?

From the above, the magic is in the amount of curvature, and NOT in the steepness of the curve (eg a straight line gives no rectification, it's just a plain resistor).

So here's the punch-line, if you examine the curve with your imaginary microscope, you will see the slope changes along the graph, but the amount of curvature stays constant.

So if you change the bias, you do change the slope (Resistance) but don't change the curvature. Hence there is no Elbow.

Late note: Thanks to interesting correspondence with Richard, he has convinced me that the rate of curvature actually increases as you move up the curve. More on this anon.

......... Zim

Edited by - gzimmer on 9/16/2013 5:57:21 PM

homebrewPosted - 15 September 2013 11:39  Show Profile  Email Poster  Edit Message  
A bit off the subject but this is the only curve of a biased carborundum detector I've ever run across. Plus the sensitivity chart of various 1920's detectors is included.

Edited by - homebrew on 9/15/2013 11:40:16 AM

golfguruPosted - 15 September 2013 15:57  Show Profile  Email Poster  Edit Message  
Nice paper HB.
Thanks for posting.


_J_Posted - 15 September 2013 16:40  Show Profile  Email Poster  Edit Message
Hey GG,
Re: One might argue that the theoretical "knee" might be the "transition point" where the "square law" mode kicks in?

If that were the case, the "knee" would stay put if I change the scale of the graph axis.

You can see where the current accelerates around the axis, but it accelerates way more than that as voltage increases. The increasing "knee" just keeps getting sharper and sharper the higher you go, but it takes off so fast it looks like a straight line to infinity. When a line gets straight up on a graph, you need another scale to see what is happening. Graphing "straight up" doesn't show any features. The features that are hidden are the fact that the knee keeps getting sharper forever.

John Davidson

Edited by - _J_ on 9/15/2013 4:41:40 PM

Topic is 3 Pages Long:    1 2 3

Click Here To Close Thread, Administrators & Moderators Only.

Show All Forums | Post Reply