EveryCircuit - Simplest 1.5GHz oscillator

8 years ago

This circuit is practically impossible to realise. Nothing can be done to stabilize such a free running oscillator. This is why we use resonators, DDS, PLL and other stuff.

zorgrian

8 years ago

As usual, I've bookmarked this for the amazing community comments. Great stuff guys, keep it up...

Robert_Kidd

8 years ago

It’s always good to feel appreciated lol.

LeButch

8 years ago

Anytime for you, zorgrian

nikisalli

8 years ago

I'm working right now on a PLL for over 1Ghz lol

hurz

8 years ago

Zorg, thanx for triggering the trolls again 😂 but you know, a PLL is not an oscillator, a oscillator is part of a PLL. And this is a oscillator, exactly a harmonic LC oscillator. We already had some words about how realistic it is with everycircuit BJT model to do it this simple. And I dont get what you mean with "can not be stabilize" all old superhet tube radios are just having LC harmonic oscillator which they use as stable reference! Sure its not stable as a crystel coupled one with pll, but its stable enough to work for this application.

hurz

8 years ago

@thebugger, have you already tried to make it as common base? I think this is our main issue with the EC model, we cant get rid of the 5pF parasitics and Hfe 10 is even quite high for RF transistors which can multiply the miller capacitance to 50pF!

thebugger

8 years ago

I actually did try it, and it didn't work. Got me wondering why exactly, seeing as how common base amps don't suffer from miller capacitance.

hurz

8 years ago

Ok, I also did and it looks promissing, wait.

thebugger

8 years ago

@zorgrian this is a completely unrealistic model of an LC oscillator (especially at these frequencies). I just wanted to see how high a frequency I can get from it, and wanted to share it with you guys. Also on the low end of the RF spectrum, one may use an LC oscillator, and actually get it very stable, without many considerations. For instance (a tube example coming), early tube radios would often use negative resistance LC oscillators for their high frequency stability - often compared to crystal oscillators - and broad tuning range.

hurz

8 years ago

Check this if you are not a troll http://everycircuit.com/circuit/5284678650298368

LeButch

8 years ago

Why are there two diodes in series?

hurz

8 years ago

To make it useable

thebugger

8 years ago

Wow, that's almost identical to the one I tried :D

thebugger

8 years ago

http://everycircuit.com/circuit/6634408227635200 that's the one I started designing, but didn't quite finish. I actually did make it work at 1.5GHz, but I was aiming higher with this one.

hurz

8 years ago

I stopped that higher and higher long time ago. Doesn't make sense.

thebugger

8 years ago

Yeah, just wanted to see how high I can get it.

hurz

8 years ago

with this one, I did for experimence upto 3.4G working simulations, but that really doesn't help

zorgrian

8 years ago

Yeah, what i mean is with these specific values, undisciplined by anything, this circuit, if implemented using discrete components at this frequency will be totally unstable... OK

zorgrian

8 years ago

So, my comments about PLL, are in relation to discipline of free running oscillation.

zorgrian

8 years ago

I have an MFJ antenna analyzer with a free running oscillator. Runs from audio frequency to 200 MHz. Quite stable.

hurz

8 years ago

You see, and with discrete components we can make it easily upto 433MHz

thebugger

8 years ago

Problem with free running single stage oscillators is that they tend to drift in frequency after some time, especially at high frequencies. Inductors and capacitors tend to shift values as they heat up due to losses and surrounding heat, and the frequency drifts away. There are some ways to make an oscillator considerably stable without PLL, but it requires careful layout design, shielding, proper heat distribution and high Q tank circuits.

hurz

8 years ago

Q is a problem of low frequency. At RF, Q is extremly high, fortunately! Capacitors do have high temperatur drift and cause microphonic effects, they cause the biggest instability. Ls are much less temperatur depending, almost zero. Cuz even metal acts as PTC Q is already extreme high and a little degradation does not take take effect. Example, an 2nH inductor at 1.5GHz is Xl 19Ohm while R for a centimeter of extreme thin wire is within still 10mOhm, so Q is above 1000! For low frequency and bigger coils it get worse for Q, but not for RF in GHz range!

zorgrian

8 years ago

OK, its the 1pF cap that's worrying here. I suggest that planar microstrip style printed components might be best for implementing microwave energy level circuitry. Of course there are other schools of thought here. SAW (surface acoustic wave) oscillator use is often preferred due to the simplicity of use and reliability. There are many ways to skin a cat!

thebugger

8 years ago

Problem with coils is not the Q factor. It's the thermal expansion of the wire. It can be caused by surrounding heat up from the device itself. Same problem with capacitors. Even NP0 caps have some thermal deviation. At GHz range I'd go with alternative methods. Yes, there are PLL's working in the GHz range, but for high power generation in the range - tubes are still the choice of preference.

zorgrian

8 years ago

Then there's the fixed apparent 5pF capacitor in the transistor model. In reality, this would vary with frequency, voltage, temperature etc. As mentioned, you could probably get away with a PCB printed inductor but, without any form of discipline, I'm not sure how stable this would be. I recently built a 10 MHz GPS disciplined clock / oscillator. It uses a register in an attiny 85 which overflows constantly, the remainder in the register at the point of the 1 second pulse represents the phase error. In this way the clock is disciplined to 1ppp in 10^9. Or 1 Hz in 10 MHz. Good enough for my needs.

zorgrian

8 years ago

Q. Factor! Without high Q as @hurz points out, i believe that the resultant effect of expansion of wire would be negligible. This is the case with resistor style chokes. These are low Q inductors. So the effects of temperature on the circuit are lower. Having said this its not advisable to use these as part of say a colpitts LC oscillator due to the low Q. Heat in the oscillator from current effects are to be avoided.

hurz

8 years ago

@thebugger, you said "it requires careful layout design, shielding, proper heat distribution and high Q tank circuits." Now you say its not a Q problem. Anyway, just because we have a high frequency and will have at the end an high absolut drift in frequency over many effects it does not mean its more critical for RF then for low frequencies. In percent its in the same class. @zorgrian, right heat and expansion of wire which change L can be ignored.

zorgrian

8 years ago

I'm unsure that the percentage change would necessarily be the same. If we go back to the example, the variation in the 1pF cap would cause how much deviation in ppp at this frequency?

zorgrian

8 years ago

In Hz drift, change per percent value of the 1pF cap?

hurz

8 years ago

And lucky the condensator to tune a station of tube radios is without special dielectricum, just air, and is very stable! Caps in this size do not need special temperatur depending isolation. Anyway, we should talk about waves and its length and what does it mean if one whats to go higher and higher in frequency. @1.5GHz, the wave lengths is 20cm. Any interconnection wire has to be a tenth of 20cm -> 2cm max, or we get in trouble with terminations and get a tank circuit instead of an interconnection. Or other said, 2cm is the longest way inside a cube with an edge length of 1.26cm. Our oscillator is already quite small. What if our oscillator cause harmonics, which it will/must have to work? First possible harmonic is at 3GHz, so any size consideration shrinks by the factor of 2! Mostlikely we need the third harmonic to run this oscillator. Conclusion this is for my fingers and eyes to small to handel.

thebugger

8 years ago

Exactly, higher frequencies experience more drift, because even 1pF deviation would cause large drift. On the other hand, bigger coils experience higher drift from thermal expansion. Q factor doesn't matter when the tank circuit changes its center frequency. You can have a Q of 1000 and still a drift. Q defines how much energy can you store in a tank circuit and how narrow a bandwidth you can get. It doesn't define the stability of the center frequency.

zorgrian

8 years ago

OK, so looking at your words. "Q defines how much energy can you store in a tank circuit and how narrow a bandwidth you can get"... The Q is part of the filter that does determine the centre frequency!

zorgrian

8 years ago

What have you been smoking? Are you looking at a megawatt transmitter? Which kind of coil are you looking at? An electric heater coil? In general, its best to keep the power going through an oscillator low.

hurz

8 years ago

@thebugger, wrong conclusion. What you do is mathematicaly not allowed. You compare absolut values and make the wrong conclusion. You have to see it relative and calculate in percent and NOT absolut! Now I could use your wrong statement and compare 50mm wire for an RF inductor with hundreds of meter for audio inductor. Which one will expand more? Forget thermal expansion for inductors in general. Only statement which is correct, Q doesn't change the center frequency. Who was first talking about Q in this context? It wasn't zorgrian or me.

zorgrian

8 years ago

Yeah, I should have said Q is the part of the filter that governs the percentage of possibility to STAY on the centre frequency.

zorgrian

8 years ago

And even that statement is not correct without taking into account a load of other factors... Ha! All this from a theoretical unrealisable model.

thebugger

8 years ago

Exactly Q doesn't determine the center frequency. A Q of 1 would have the same center frequency as a Q of 1000. It doesn't also determine the stability of the center frequency. Yes it helps in free running oscillators to have a higher Q, but a PLL synthesizer doesn't need high Q to keep the center frequency locked. As I said, the Q factor determines how much energy the tank circuit can store, and the narrowness of the bandwidth it has, nothing more. It doesn't explain the stability of the components. For instance you can have a silver wire and air cap and have a Q factor of 100, and still a nice little drift in your oscillator. The component stability is measured in a different parameter - how many parts per notation (ppn) does the value deviate from the ideal value at 25dgs. It is often used to describe temperature related deviations. Also ppm describes dimensioneless units, so you can use it to compare absolut values without transforming them into percentages.

hurz

8 years ago

@thebugger, I know what ppm is but what is your point? Oscillators with 1GHz do drift 'more' in frequency then oscillators of 1MHz? You must know this is not true. Again, who was first talking about Q in this context?

zorgrian

8 years ago

Thoughts on inductors (heating coils)... The measure of physical deformation, or derived values from such deformation (such as expansion and contraction of a coil of wire) are defined as a change in length divided by the initial length. Since both quantities have the dimension length, their ratio is dimensionless. As such, the whole argument on change due to heat in the oscillator L is dimensionless and therefore = 1! Naturally, I'm partly joking here. However, its interesting to note that ppn is a non SI notation representing ratio x:y. It immediately occurs, that percentage % is also a represented form of ratio.

zorgrian

8 years ago

@hurz, I believe what @thebugger may be saying, is that at 1 GHz the proportion of drift may be the same as at 1 MHz. However, the stability of any oscillator, when referenced to say 1 Hz may be lower at 1 GHz, than when the same design is configured for operation at 1 MHz. This does seem to be true when modeling oscillators using SPICE or better QUCS where it is possible to simulate temperature and other perturbations. Maybe, my assumption on what @thebugger is actually saying is misinterpreted?

hurz

8 years ago

We dont know as long @thebugger clarifies what the root statement is suppose to be he made: "Problem with free running single stage oscillators is that they tend to drift in frequency after some time, especially at high frequencies". The second statement about Q is even more irritating. To bring Q to the point, Q is importante for the phase noise and is measured in dBc which gives for an GHz oscillator for example -130dBc at offset 10kHz so from center frequency amplitude plus 10kHz we drop 130db to the noise floor. This is a value which highly depends on Q and low phase noise is importante, cuz a PLL can not work against a noisy oscillator and make it better.

zorgrian

8 years ago

It seems logical to suppose that a single stage free running oscillator would have less drift than an oscillator with more 'stages'. I cannot think of a two or more stage oscillator. This is in the sense of stages, where each stage has a specific function. There are the multivibrator oscillator types of which EC plagued, but the bilateral topology of these doesn't provide less drift.

hurz

8 years ago

But it sounds so much more interesting 😂

thebugger

8 years ago

But Q doesn't determine the center frequency. It determines the bandwidth limitations in respect to the center frequency. But the center frequency may drift due to heat expansion of capacitors and coils. Zorgrian, that's exactly what I meant. When both oscillators are referenced to 1Hz, a 1MHz oscillator would drift much less than a 1GHz oscillator. A heat related deviation of 1pF in a tank circuit of 1nH/1pF (completely theoretic), would bring a deviation of more than 1.4GHz. The same deviation applied to a tank circuit of 100nH/30pF would bring a deviation of 1.4MHz, which is actually still a lot but in the first example the deviation is around 30%, and in the latter, it's 1.5%.

hurz

8 years ago

Fortunately, heat does cause a change in percent and not absolut.

thebugger

8 years ago

Yeah, but this deviation can be expressed in a dimenaionless units as well. Anyway, I'm way below 1GHz. Neither have I got the equipment to measure and tune such circuits, nor the components to build them. The highest I've got was 100MHz or so in home

hurz

8 years ago

Im at 352.4MHz in real with BRF90A and very very stable at room temperatur. All this with an extreme simple inductor, just 5turns copper wire on a 1mm diameter. I will try to reach 432Mhz and second harmonic will be on 864.6 for a wireless headphone.

thebugger

8 years ago

You can make a power amplifier to boost only the second harmonic of the wave. I have a few very handy formulas written somewhere to calculate a class C power amplifier with transformation networks and all. Anyway, recently I bought a 0.5W FM transmitter, and I'm currently making a 6W power amp with a 2SC1971. Problem is, I can't find anywhere on the datasheet the input reactance of the transistor, so I'm kind of winging it right now :D

Robert_Kidd

8 years ago

I’m sure you’re aware there are several circuits online for 6W power amps using this transistor - inspiration:-)

hurz

8 years ago

573.7MHz now a little to high for second harmonic, but maybe this opens a chance to go directly on 864.6MHz - and thanx buggzy, you said yourself you never tried above 100MHz. The input reactanc is mostlikey given as smith chart and or s-parameter, or in this case on the first page as complex value Zib=1.3 + j3.2Ohm @175MHz ubfortunately only for one frequency.

thebugger

8 years ago

I couldn't find any reference on the input reactance in the datasheet. I figured, I'd just skip the input matching. Just a 50ohm input resistor to ground, and a variable capacitor to tune the needed attack for the given output power.

thebugger

8 years ago

Ah, yeah you're right, I found the datasheet you're speaking of

thebugger

8 years ago

So, I'd have to put 5nH coil in series with the base, to match the reactance. Is it really worth it?

hurz

8 years ago

Input is already inductive the imaginer part 'j' is positive and equal to 2.91nH and the first value is the resistance 1.3Ohm if you want is to congucated you can use an 284pF capacitor 270p is ok.

hurz

8 years ago

For 144 is 470ip and for 100 its 1n, but only with this impedance for measured 175MHz. As I said, normaly there shloud be a smith chart which gives you the impedance over frequency for in and out.

thebugger

8 years ago

I don't know how to read smith charts :D. So what are we speaking about - 284pF in parallel with the base-emitter junction, or in series?. Or maybe I can use a transformation network altogether.

thebugger

8 years ago

http://everycircuit.com/circuit/5967983251030016 like so

thebugger

8 years ago

Actually, I've been meaning to ask you, since we're working on similar projects right now, and the things are still fresh in our heads, is it really necessary to match the reactive part of the input, being only 4nH?

hurz

8 years ago

Maybe not at this frequency, cuz have you noticed 175MHz seems to be f_transist. So at 100MHz I dont expect a lot gain anyway. Maybe at 7MHz but not much higher.

hurz

8 years ago

Why do you need so much power at this frequency 100MHz. You know this will cause trouble.

thebugger

8 years ago

Yeah, I just want to see how far I can push it. I can always jump down on the power. The transmitter I bought works at 0.5W and still covers the neighborhood good and clear. I had a university project to build something RF and I chose this.

thebugger

8 years ago

http://everycircuit.com/circuit/5967983251030016 this is basically what I had myself built.

hurz

8 years ago

For this frequency I would use a diode instead of a balast coil at base, for a more agressive class c design.

hurz

8 years ago

Im at 432.3MHz with a 440Hz test modulation around +-100kHz

thebugger

8 years ago

A diode might be a good idea after all, I'll look into it. Coils are a bit sluggish at power up, whereas diodes are a bit more aggressive as you said. I'll make a combo with a coil and a diode, because diodes are fault prone.

thebugger

8 years ago

Anyway, you're using FM on your transmitter I presume. What are you transmitting - audio? If you're using it as a controller of some sort, may I suggest a crystal oscillator with ASK modulation.

hurz

8 years ago

Crystals are good for upto 50MHz not more

thebugger

8 years ago

There are crystals on 433MHz. They work on overtones of a lower fundamental frequency. 10p / lot 433.92MHZ 433.92M 433MHZ SAW oscillator crystal filter strip legs http://s.aliexpress.com/faUv26rI?fromSns=Copy to Clipboard (from AliExpress Android)

thebugger

8 years ago

Alternatively, you can use a 108MHz crystal, and amplify the fourth harmonic.

hurz

8 years ago

Check this and forget everything above 50MHz which is possible just a resonator. https://en.m.wikipedia.org/wiki/Crystal_oscillator_frequencies

thebugger

8 years ago

The first sentence says it - from a few khz up to several hundred megahertz

thebugger

8 years ago

I just finished my project by the way. Got the final tuning (with resistive load) right, and got very surprising results. The output power is higher than it's supposed to be with the given input power.

hurz

8 years ago

And you saw the table which ends at <200MHz crystals. Everything above 20 or 50MHz is overtone, and they are less pullable in frequency and they are worse in phase noise, fifth overtone crystals are 5 times worse in phase noise then fundamental ones! Thanx, I will stay with fundamental crystals and dont like the overtone rubbish crystals.

thebugger

8 years ago

Fifth overtone is still better than free running oscillators. If you're using FM you'd need PLL to pull the frequency. AM should work fine though.

hurz

8 years ago

Funny idea to compare an LCR discret resonator with any crystal resonator, what is this for good? AM is fine and FM is not? From were do you get the carrier frequency in both cases? FM from crystal and AM from LCR, what are you meaning and its getting now very confusing.

thebugger

8 years ago

AM is simpler. If you're using the transmitter as a controller or something, it might be simpler to use AM. And AM works directly with crystals. https://goo.gl/images/27d1xV (for instance)

hurz

8 years ago

Buggzy, and the carrier frequency stability is still not solved. Anyway, I told you this is for a FM headphone. Can you still follow this thread, doesn't look like.

hurz

8 years ago

Really to many wrong statements you made and you not clarified or even you do not tries to answers. You just make bad statements, and then you ignore futher explanations. We better stop here, nothing useful coming out for the communuty. Dont hang out to much out of the window, and it might dont look that stupid in future. Ohh boy.

Robert_Kidd

8 years ago

Same pattern as usual. Interesting, useful discussion ending with insults because someone won’t accept all you say as correct. Shameful.

thebugger

8 years ago

:D and it isn't. Anyway i hadn't noticed that it had to be FM. With FM best way to go is buy a walkie talkie at 433MHz, tweak it a little so you match your desired bandwidth and replace the microphone with an aux lead.

hurz

8 years ago

As I said buggzy, you are leaving plenty of open statements and have said nothing. It does't bring anything useful to continue as long old statements are clarified.

Robert_Kidd

8 years ago

Your honour, I rest my case.

thebugger

8 years ago

Which statements? The stability comparison between any crystal and a free running oscillator, or the AM/FM comparison in a purely technical way, when made via a crystal. I haven't made any other statements.

hurz

8 years ago

If im really bored I will go through all 120 comments and collect all open statements, so far if you are interested in answers then you have to make a list of all by yourself. I have really tried hard with repeat again and again ask you about your statements, but nobresponse. Only ONE example, please make for us a calculation how much a 10nH coil does change in inductivity when it gets from 25°C to maybe 50°C. How much in percent does its inductivity will change. Thats about your thermal expansion theorie/statement.

thebugger

8 years ago

This is not a theoretical thing, it's a practical thing. One can't consider all the factors that determine the thermal expansion and displacement of the coil. You can only try to minimize these factors - for instance, avoid narrow diameters. From the tip of my head - more turns - - - > more wire - - -> more resistance - - - > more heat dissipation - - - > more thermal expansion. Anyway in the project I'm working on, this effect causes problems with my tuning capacitors, not inductors. During heat up, my circuit needs readjusting to stay in tune

hurz

8 years ago

Hello? Now its not the inductor, but the cap. Right caps does have this effect noticeable, but for inductors its not relevant. So this statemante about "coils and expansion" fails, right.

thebugger

8 years ago

No, this is the way it's with my circuit. I'm using low quality variable capacitors, and they're drifty.

hurz

8 years ago

Coils are not caps!

Robert_Kidd

8 years ago

@thebugger, I admire your patience.

hurz

8 years ago

I do not admire your patience to poison a discussion. Stop that or you get reported.

Robert_Kidd

8 years ago

Oh no!

thebugger

8 years ago

Both coils and caps are susceptible to thermal expansion.

hurz

8 years ago

Yes, but as I said, caps with low thermal expasion are used to tune oscillators and those caps are simple air isolated ones. So you agree inductors do not have this expansion problem, the same is valid for caps when its dielectricum is air!

thebugger

8 years ago

Inductors expand as well. And in doing so, the diameter and height of the coil also expand, thus changing its parameters. My air capacitors are air isolated with metal spacings, and they still drift

hurz

8 years ago

We turn in circles, right. Please let us know your calculations how much inductance/capacitance we have in percent per degrees, and remember there were times no crystal exist and radios were stable enough for reception of music.

zorgrian

8 years ago

At 1.5 GHz? I don't know about such tube radios from the past that were operating at these frequencies and part of a very secret military operation.

zorgrian

8 years ago

I do like thermally expanded coils. However with enough current it may be possible to create a linear motor, without thermal expansion being such a problem

hurz

8 years ago

Zorg, be a little more creative and think about FM radio at 100MHz for example. I have an old tube radio receiver and its working fine without the need of a crystal!