Hi all,

I'm making a design based on a common base Colpitts oscillator. Like the one in this post:
https://forum.allaboutcircuits.com/threads/common-base-colpitts-oscillator.72724/

It works, but I really don't know how to calculate the values of the different components. I've made several simulations, even I have a sheet with values and results, but still not clear how the two Cs of the LC tank should be chosen. If the values are not well matched to this of the coil (even if the relation between Cs is about 10), the oscillation doesn't start or is distorted. I've been looking for a direct answer for the question but, although there is a few post on this matter in the forum, there's nothing conclusive.

I know F= 1/(2pi sqrt(LC)) but I suppose that you can't make a 5 Mhz oscillator with a mH coil and pF coils or a 5Khz with 1uH coil and 1000uF capacitors (aproximate values).

I'm not sure, but I think that everything has to do with the impedances of the components. The impedances should be match to the amplifier in and out resistances in any way. Maybe, it's more a problem of common base amplifiers.

In this thread, in post #6 Adjuster gives a few clues:
https://forum.allaboutcircuits.com/threads/designing-a-colpitts-oscillator.60308/


Any help is really appreciated.

Thanks a lot!

 

If you want that kind of range in frequency.......go digital. You can get that kind of range with analog.....but it involves more than just switching coils and caps. If you want decent performance.

Check out numerically controlled oscillator (NCO).

 

Thanks for the reply, but those frequencies are only examples. The actual frequency is about 100Khz. I know I can use other technologies, but I have to use this oscillator in this project. Also, it's working, the thread is about knowing which are the relations between component values in Colpitts oscillator not looking for a solution. Maybe it's working but is close to a non-working state and will have problems with drifts. Maybe I can achieve a better signal if I tune the caps, or more current, or a more stable signal... I've been working with Colpitts oscillators for a while but never got to a complete understanding of the configuration.

 

The actual frequency is about 100Khz. I know I can use other technologies, but I have to use this oscillator in this project. Also, it's working, the thread is about knowing which are the relations between component values in Colpitts oscillator not looking for a solution.

See http://www.electronics-tutorials.ws/oscillator/colpitts.html
And one of practical circuit:
EDIT: Because of high feedback fraction F_f =C1/C2=100% and current mode bias of Q1, oscillator works fine with voltage of V1 in range 0.8V...24V.
Reactance of L1 should be about 360 Ohm at oscillation frequency.

 

I'm making a design based on a common base Colpitts oscillator.

Common base Colpitts oscillator:
Same feeding voltage range - from 0.8V to 24V.

 

I had been worked long time with very similar Clapp oscillator circuit. Probably You may use my experience.
About LC tank: at small power it is not important do You apply a 100 pF and 10 uH or 1000pF and 1 uH. Yet at bigger power (like kilowatt part) there should be Q-factor optimization made. As the even most worst capacitor has loss factor much smaller as even best coil, then it is more beneficial to have a small L and large C. However, even the best dielectric materials has tg(fi) about 0,0001...0,00015 but more prost materials like glass or epoxy has 0,05. It means that Q(max) is laying under 20 or in case of good materials 2000. The simple copper coil with electropolished shiny surface under good lack yet has 50...200. My experience is that ANY type of silvering diminishes Q instead of predicted altering. The exact meaning of coil data You may acquire by softwares like Coil32 (http://coil32.net/) or "Inductance coil calculation. ver.1.0" or other products. Or measure Yourself by instruments (I mean not only L but the Q as well). Generally, coil ought be wounded on hard material non-magnetic core with as much You are able for - tiny thermal expansion factor, and be wounded by so much pulling the wire, it stays near to liquid flowing point. Otherhow Your inductance will float, but now it will float so much, as that plastic permits.
About feedback - just apply the common theory, and set it near to 1,0 near the wading-out point, there it gives a best form of signal.
1N2222 is very good choice for small power circuits and works up to GHz range. For really large power there the most best candidate is ixfh42n60 working up to 100 MHz, but what is intriguing, the wonderful and nice device IRF510 working up to 144 MHz is so caprice about generation that makes all nerves shiver, for my experience it is not capable to generate at any frequency. If any, to generate or not is changing by variating the Vcc, what most cases must be condemned.
By the way, I see there some places stays sth similar to serie tank. If any serial resonance happens, from 20 Volts You obtain a 6 kV!!! Be careful not beat out the capacitors and gates (or Your case base).
RE:""Maybe I can achieve a better signal if I tune the caps""
If the temperature stability and phase noise is the main concern, You should choose the Vačkar circuit instead of Colpitts.

 

RE:""Check out numerically controlled oscillator (NCO).""
Actually the more correct search phrase will be DDS instead of NCO. Look for it at ebay (https://www.ebay.com/sch/i.html?_fr...l1313.TR0.TRC0.H0.XDDS.TRS0&_nkw=DDS&_sacat=0) and You shall find a plenty of DDS within f range of khz, tens mhz and even hundreds of MHz adjustable by sub-Hz precizity, even the waveform is programmable, and it cost 4 USD...10 USD or in the most demanding case 20 USD. Thats China.....

 

Thanks Danko and Janis for your answers!


Danko, the exact topology I want to understand is the one in the following picture (I wasn't able to post it before). Is very similar to the second you posted although not sure why there isn't a resistor from base to ground, isn't it necessary to set the operation point? Shouldn't the operation point be set at more or less Vcc/2? I've seen in a few designs that the two base biasing resistors are not equal but 80K-40K, 470K-100K or similar.

Also, there is a 200ohm resistance in the feedback path. Is it use to reduce the feedback gain? What for?
In the first picture, why do you say that the coil inductance should stay at about 360Ohm? Is this a value calculated from the oscillation
frequency or is a objective you should achieve with the design?

I already had visited the electronics tutorial page, and a few more, but the only design guide that I've found are the oscillation freq. and the feedback gain.

Janis, thank you for the clarification regarding DDS. I know this and like it very much, but I have to use a Colpitts oscillator. The circuit is a metal detector and I need the signal to be coil dependant. Of course there are another circuits that probably perform better than this, but all I want is to understand how the values should be chosen.
In this way, I thank you for your explanation of the Q values of capacitors and coils. I will take it into account.

Regarding the transistor choice, although it is not in the main topic of the thread, It interested me a lot because just this morning I've been choosing a transistor for the design. I've seen several designs using 2n2222 but don't understand why this is better than any other general purpose BJT. Maybe the high bandwidth is the key, but I assume that, let say 20 MHz would be enough providing the 100KHz frequency (correct me if I'm wrong). I've chosen another BJT because 2n2222 is not available in my components provider (don't know why).

The main topic of the thread is still unclear for me: how to select the values of the components. C1/C3 should be higher than 1 to start oscillation, but for example, if I place a 470nF C3 and 220nFC1 the oscillation is really low (in simulation). I suppose that at this low frequency the impedance of the coil is not enough...
There must be another implication than the feedback gain to make it work properly, a kind of balance...

Thank you for reading and of course, any help is welcome.

 

 

1. For start, oscillator need some initial collector current only. R3 is enough for it. Operational point establishes automatically when generation occurs. You can try to change R3 100K to 1M (R4 removed) with same result. R3 connected between collector and base, therefore Q1 never will saturated in start mode.
2. Your circuit simulated without any load. In metal detector, L1 will radiate energy in the form of electromagnetic waves. R7 and C5 simulate this loss of energy. Ideally, without load, current through L1 is 150mA (peak), but with load, current decreases to 47mA (peak). However we need more strong magnetic field for increasing sensitivity of detector. Resistor R2 limits collector current, so energy distributes between L1 and R2 in direct ratio of their reactance on 100kHz and resistance, so it is 146 / 3300 = 4,4%. Its means that L1 utilizes only 4.4% energy, rest loses in R2. To decrease losing energy, R2 lowered to 100Ohm.
3. With R2 = 100Ohm current through L1 is 150mA (peak) under load, but form of signal is distorted and frequency changed. To eliminate distortion used R5 = 51Ohm connected between C1,C3 and emitter Q1,R2.

 

RE:""The circuit is a metal detector""
Opsssss!!! The such devices in the physics level are VERY insensitive. Best case the quarter of meter for small objests and meter for large. You should read a bit about PI system what gives a 3 meters for small objects and tenfold for large. And it is even simpler by shemotechnics. Most recommended for first success is to look for `Pirat` circuit http://arximastoras.blogspot.com/2015/04/russian-metal-detector-circuit.html

Once the fishing police boss was calling me to help him save his workers from firing. Those guys in environmental police was received a Japanese government gift of kinda very lightweight and powerful boat motor to better success running after illegal fisherman. The subordinated officers mounted it, patroled all the day long but find no-one to be cached at that day, thus at the evening they decided to end the first day with say `salute` - little `washing around the new motor`. Thus the vodka party begun, and when the last bottles was emptied, they customarily started to turn on the engine. Yet vodka weakened their brain thus they did as custom pushing the gas up to the maximum, other-how it always had been go stop. But new engine gave a ultrasound wheat instead, the boat made a kangaroo jump, so the motor de-hinged and sink. Normal people would mark the position with some boy, but never being in the grand-lord`s mind having ocean shallower of knees. Thus the morning came when boss told - I cant do it another way as You are fired plus going to pay back about 10 000 USD to the State. So my part was to help the guys by construct a 6 meter large coil (as the water there was 6 meter deep) on the sledges, one man pull it over ice at the nearest winter, and all those frequency discrimination thing was rather sensitive to react even on ice temperature fluctuations - no way to get better result. So, they boyed about dozen suspicious places, but when spring came and they got aqualung, the all what they was able to find was just few old boots, metal tin, and hence wire. No any motor. I haven`t understood why - did it contained too much aluminium and no iron at all, or just that was `revenge` of illegal fisherman, caching the motor just next minute after officers left a place? Or just the water having epsilon about 81 instead of 1 destroys all those steel-core effect of coil.

 

And one more thing: For frequency type of metal seeking devices one of best generators I ever seen is not Colpitts but one working very mildly. Its two transistor thing where first transistor base is going to second transistor collector and is too connected to LC parallel tank, what other leads connected to gnd. Second bjt base is connected to gnd. First transistor collector is Vcc (from 1,5 to 24V any as one may like), both emitters are together and via the resistor (1k...10k...100k) goes to gnd. Thats all in 1 cm3 of space - just pure distilled exaltation. It ideally works with kT315, 1n2222 and many more, about T(f) factor only Vackar has better, it is capable to work from Hz up to GHz depending on f(T) and it is practically impossible to kill it.
P.S. OK, I tried to upload the circ, here it is

 

RE:""using 2n2222 but don't understand why this is better than any other general purpose BJT""
Because it beta is variating according collector current, yhus the mild auto-regulation of amplitude is provided. If beta would be stable then generation regime would be much more hard.

 

Hi again!!

Thank you so much Danko and Janis for all your answers and time. I am learning a lot with your help.

Danko:

Your explanation is very nice, thank you very much.
What I use as a load in simulations is a coil in series with a resistor and with different coupling factors. It's very close the same effect that a metal has in a oscillating coil.
Regarding your simulation, I couldn't reproduce it, in LTspice it didn't start oscillating. I'll try on the bench. However, I don't understand well the way the coil current is increased lowering R2. As I see it (probably wrong), most of the current to the coil is provided by the capacitors in the tank circuit thus the transistor only provides the energy lost. Therefore, with a smaller resistor there are more losses and more current through the transistor but not necessarily through the coil. Anyway I run a few simulations and you're right. There is about 15mA more passing the coil and the transistor.

R2=100Ohm

R2= 3K3



Janis:
First of all, I love your story about the vodka-marine-patrol ;-) It's clear that vodka doesn't make smarter people... it only make people around you look much better. Thank you.

About PI metal detectors, I know the technology and have made a few boards to play around. They certainly work nice and are the way to go in metal detection, but right now I want to make a metal detector easy to build, without a balanced coil and capable of working with a metallic case. This only leaves BFO and damping detectors, I'm trying to make a kind of hybrid measuring frequency and amplitude. In the bench it works as expected, not so much sensitivity but enough for me.
I'll try the oscillator you posted, will tell you my impressions.

Regarding the 2n2222 BJT, thank you so much for the explanation. I hadn't even though of it. It seems reasonable, now I will look for a 2222 type beta curve BJT because unfortunately the max Vce is 40 volts and the circuit presents a little more than this.

 

RE:""I will look for a 2222 type beta curve BJT because unfortunately the max Vce is 40 volts""
Let look for the mosfet, they have a similar curve and may work up to 600...1200 V.
Only the thing I wonder is IRF510 what are generating with a difficulty, all other are better than best.

 

Regarding your simulation, I couldn't reproduce it, in LTspice it didn't start oscillating.

Here it is.

 

As I see it (probably wrong), most of the current to the coil is provided by the capacitors in the tank circuit thus the transistor only provides the energy lost.

Look at schematic and use logic. Circuit feeds by direct current from power supply.
How can direct current (minus) go into circuit? Through capacitors? No, definitely not!
Only through resistor to emitter of transistor.
By the way, transistor turns ON for short time, exactly in moment, when capacitor is fully discharged.
Transistor connects lower end of coil to minus through resistor (upper end of coil is always on plus) and provides coil with some portion of energy.

 

Thank you again for your kind replies.

Janis, I will try mosfet as soon as possible, but before, I'll revise how they work (as you can see, I'm a little oxidized).

Thank you Danko, I'm simulating it as soon as possible. With this new aclaration, I understand the issue much better. I failed to think in DC ...

 

And one more thing: For frequency type of metal seeking devices one of best generators I ever seen is not Colpitts but one working very mildly. Its two transistor thing where first transistor base is going to second transistor collector and is too connected to LC parallel tank, what other leads connected to gnd. Second bjt base is connected to gnd. First transistor collector is Vcc (from 1,5 to 24V any as one may like), both emitters are together and via the resistor (1k...10k...100k) goes to gnd. Thats all in 1 cm3 of space - just pure distilled exaltation. It ideally works with kT315, 1n2222 and many more, about T(f) factor only Vackar has better, it is capable to work from Hz up to GHz depending on f(T) and it is practically impossible to kill it.
P.S. OK, I tried to upload the circ, here it isView attachment 143416

It is bad.Error!
It is good:

 

1)Yeppssssk! You`re right. Mea culpa.
2) I had to write ``Do You are serious about 10 Ohms of inductive losses??`` But then I put the data in my software, what shows that indeed 1 mH will takes a coreless multilayer coil of 15mm inner diam and length 10 mm (now see the different choices):
winded by 0,23 mm wire 300 turns, 9 layers, wire length 17,3 meters, resistance 9,3 Ohm, Q-factor 68, frequency bandwidth dB=0,7 kHz.
Alternative - wire 0,38 mm 350 turns, 18 layers, 46 Ohm, Q=136, dB=0,4 kHz
Mini version - wire 0,10mm turns 279, layers 4, R=51 Ohm, Q=12, dB=4,1 kHz
My choice would lay to the 0,38mm wire or at least 0,23mm.
P.S. capacitor 2520 pF, so use the 2400 and wind slightly more or use 2700 and use less turns.