Colpitts C values?

[samu]

Hi everybody,

I been surfing the web and my electronics books for a long time searching for an answer. I already posted this question 2 years before, but nobody wrote a "full" answer.

http://www.geotech1.com/forums/showthread.php?t=15245


The question is how to determinate the caps values in a common-base colpitts oscillator. I mean, the coil value is fixed because is difficult to vary, and so I had to choose the caps. I know that total C is calculated with

w osc =1/sqr (LC)

C=C1C2/(C1+C2)

Ok, perfect but what is the criterion to choose C1 and C2??

I have an Colpitts oscillator running and if I vary one of the caps (10 nF to 100nF) the signal becomes unstable. Also, with another coils the signal stability got worse.

I guess that is something about the quality factor (the higher, the better) if you know how to improve it ¡¡iplease tell me!!

Thanks in advance

[mikebg]

Samu, Your equipment operates at worst principle for conveyor belt application - "absorption metal detector".

The target decreases amplitude of oscillation because absorbes energy from oscillating tank circuit L1-C1-C2. At this principle, the main problem is how to maintain stabile large amplitude of oscillation without target. This is very difficult to achieve without automatic control.
A P-I controller with large integration constant will solve the problem.

Here is the correct drawing of your oscillator circuit. Note that C3 is not part of tank circuit, that's why C1 is connected direct to GND. For amplitude adjustment, you should decrease resistance of R4 until amplitude stops to increase (for example 22V), and then slight increase resistance (for amplitude 21V).

[samu]

Hi Mikebg,

I read your post about what is the best oscillator circuit and I´m sure that you know what your are talking about.

As you said,

Your equipment operates at worst principle for conveyor belt application - "absorption metal detector".

I use it because it´s sensitive enough for the application and is quite simple. Also, the sensitivity don´t become very small if the detector is mounted in a metal box and surrounded by the conveyor metals.

I tried a PI circuit (HH) and the Modz differential PI (¡¡¡three cheers for Modz!!!), but both of them decreases sensitivity in a metal box (I posted if there is a way to isolate metals from the electromagnetic field and tried to connect the box to ground). Besides, the electromagnetic noise from the motors is amplified and the signal at preamp output is "like a party". In the other hand, the "normal" operation was terrific.

According to my experiences, the best system for noise immunity is a BFO. I connected the oscillator output to a home-made frequency counter and it works very well but the big problem was sensitivity. I reached a very poor one, less than an hertz for a big metal at 15 or 20 cm. Much less deep than the absorption metal method. Maybe my frequency counter is not right, do you know how much frequency shift is expected?. I know that it depends on many factors but please tell me what is the "normal".

Every suggestion about what technology is the best is welcome.

The target decreases amplitude of oscillation because absorbes energy from oscillating tank circuit L1-C1-C2. At this principle, the main problem is how to maintain stabile large amplitude of oscillation without target. This is very difficult to achieve without automatic control.
A P-I controller with large integration constant will solve the problem.

There is no problem, I have the PI controller and works fine

Here is the correct drawing of your oscillator circuit. Note that C3 is not part of tank circuit, that's why C1 is connected direct to GND. For amplitude adjustment, you should decrease resistance of R4 until amplitude stops to increase (for example 22V), and then slight increase resistance (for amplitude 21V).

I know that C3 is not part of the tank I guess that is for filtering the base polarization voltage.

For the same oscillation frequency I can use a big coil with small caps, big caps and small coil and medium sizes of both. That´s only an example, but represents the problem. I tried to change C2 to 100 nF and the result was a change in the oscillation frequency (as expected) and a "funny party" of noise in amplitude and frequency (above all in frequency). I saw another metal detector like this one and there was a 5.6 nF cap in parallel with the 10 nF one, so there is some kind of cap adjustment.

Do you think that simulating the tank circuit and seeing what cap size makes the frequency peak narrower is a solution??

Do you think that I should abandon this technology and try with another one?

Thanks a lot to everybody.

[mikebg]

Your equipment operates at worst principle for conveyor belt application - "absorption metal detector".

The target decreases amplitude of oscillation because absorbes energy from oscillating tank circuit L1-C1-C2. At this principle, the main problem is how to maintain stabile large amplitude of oscillation without target. This is very difficult to achieve without automatic control.
A P-I controller with large integration constant will solve the problem.

Here is an ancient oscillator circuit with P-I controller designed by Robert Podhrasky (Garrett Electronics). You should increase capacitance of integrating capacitor in order to prolong settling time. I will seek for more new variant of this design.

[samu]

Ok, but do you think that the oscillator design is better than the one I posted?

The PI don´t was a problem it worked all right.

The coil value is important or not?

[Altra]

Samu,

Your circuit is known as a "Loaded Loop" detector. The trick to this type of circuit is to adjust the oscillator loop gain to be slightly above 1. By doing this any change in load current (metal) causes the amplitude to decrease. To acheive critcal bias on your original circuit add a variable resistor to the base of the transistor. Adjustment can be trial and error. The idea is to get the oscillator just stable, on the edge of instability. Try using the original C1 and C2 values (100n, 10n) . If you want to use different values for C1-C2, classical colpitts c1 is 10X c2. See modified circuit

Let us know if it works
Mark

[Qiaozhi]

I tried a PI circuit (HH) and the Modz differential PI (¡¡¡three cheers for Modz!!!), but both of them decreases sensitivity in a metal box (I posted if there is a way to isolate metals from the electromagnetic field and tried to connect the box to ground).

Are you saying that the coil is placed inside a metal box?
Please can you explain the physical setup in more detail?

[samu]

Are you saying that the coil is placed inside a metal box?
Please can you explain the physical setup in more detail?

I know that sounds a little strange. Is a metal box with one side made in plastic. This is the coil side of course.

The box is 40cm wide 80cm long and 12 cm height. The upper side is open and has a plastic cover.

[lonelyWOLF]

I will seek for more new variant of this design.

hi mikebg
pls. see pdf.

[samu]

Simulations

Hi, I´ve simulated the oscillator circuit in spice and I´ve measured the Q factor with several values of C and L.

The simulated circuit is the attached figure.

The results (frequencies in Khz)

L=130uF
C1=15.6nF

Fc=111.69, F1 (Vout 50%) = 110.45 , F2 (Vout 50%) = 113.021
Q=43,44


L=130uF
C1=10nF

Fc=139,58, F1 (Vout 50%) = 138.18 , F2 (Vout 50%) = 140.83
Q=52.67


L=100uF
C1=15.6nF

Fc=127.34, F1 (Vout 50%) = 125.73 , F2 (Vout 50%) = 129.07
Q=38


L=100uF
C1=10nF

Fc=159.03, F1 (Vout 50%) = 157.37 , F2 (Vout 50%) = 160.78
Q=43,44


L=100uF
C1=100nF

Fc=50.35, F1 (Vout 50%) = 48.75 , F2 (Vout 50%) = 52.08
Q=15.12


As you can see, there is a high Q reduction if the C is 100nF and in general, the results are better with the smaller cap.

Also, a Q factor below 50 is too low from my opinion, what do you think??

Also I changed the coil serial resistance. The increment of Q if the resistance is reduced is very high as expected.

Thanks a lot for the new designs!!!

[Qiaozhi]

I know that sounds a little strange. Is a metal box with one side made in plastic. This is the coil side of course.

The box is 40cm wide 80cm long and 12 cm height. The upper side is open and has a plastic cover.

It would be a good idea to remove any large metal parts from around the coil, as these will greatly reduce the sensitivity.
Then a PI design should work fine in this application.

[mikebg]

Mark, you are correct. The emitter of transistor creates problem because represents a load for tank circuit and diminishes its Q-factor. The problem is solved by using a tap in tank circuit to maintain its operating Q-factor enough high. In the last oscillator circuits of Garrett, the tap is formed by three capacitors as shown in the right circuit diagram.

[samu]

I´d like to test both circuits, the Capacitive Tap by Mikebg and the one by Altra.

I suppose that the collector pin of the BJT is connected to the coil and C common point isn´t it? Also, the base pin is polarized like in the typical colpitts circuits.

If I´m wrong please tell me.

Thanks a lot.

[mikebg]

In the last oscillator circuits of Garrett, the tap is formed by three capacitors as shown in the right circuit diagram.

Here is the last oscillator circuit with P-I controller:

[samu]

I tried the circuit from post #12 and without any important improve. I simulated it and the result was worse than the previous ones:

L= 100uH, C=10 nF, 5C=50nF; 2C=20nF

Fc=91.858, F1=90.239, F2= 93.578

so Q= 27.51

I´ll try the last circuit posted by Mikebg but I have some questions:

Is the last circuit capable of operating with metal absorption principle??

In this case, the output of the 358 is the difference signal between the signal and the integral??

Is there any information about the coil??


Thanks a lot.