Oscillator : Calculation

[dmowziz]

Originally Posted by G0HZU_JMR

Don't be put off by my comments, this circuit has the potential to be highly educational so I think it is worth making a version with good components and a tight layout. The two SMA connectors used for breaking open the loop should ideally be close together.

.

No, not at all.... I know I can learn a whole lot from you

I have smd caps, I'm working on the layout now.
Thanks, I'll do like you said

[dmowziz]

Thanks Jeremy!

Completed it few hours ago...Anyways

Might be hard to see the caps due to big solder but the resonator components are close together

Phase zero is less than 100 MHz. Due to the inductor? I "got" 2 turns for 20nH.

The S21 is about 5.5 dB (calibrated using nanoVna calibration kit). approx. same reading in Spectrum Analyzer mode.
It's about 2.4 dB in simulator but that's with phase zero at 100 MHz

But when I close the loop, no signal showing in spectrum

Thanks.

[G0HZU_JMR]

Ok, things do look better in terms of the layout although the trace length looks to be very long at the emitter where you break the loop open. Maybe this is causing problems. Also, do you know if your 2N3904 is a genuine part? A 2N3904 is a process 23 BJT and this sets it apart from the other jellybean BJTs.

See below for my simulation using Genesys and a 2N3904 model at 7V and 10mA. If I simulate for a point to point connection method and only include the component parasitic inductance and ESR then it appears to want to oscillate at about 84MHz.

In the simulation below I have done a very quick and dirty negative resistance analysis. This shows about -34 ohms resistive across the inductor at resonance at 84MHz. This means the circuit should still oscillate even if I put (say) a 47R damping resistor across the 20nH coil. This is a bit of a surprise but then again the reactance of the inductor is only about 11 ohms at 84MHz.

This assumes a very tight layout. I could try to simulate it to include the metal losses in your revised layout or I could have a go at actually building it on some scrap PCB material here.

[dmowziz]

Originally Posted by G0HZU_JMR

Ok, things do look better in terms of the layout although the trace length looks to be very long at the emitter where you break the loop open. Maybe this is causing problems. Also, do you know if your 2N3904 is a genuine part? A 2N3904 is a process 23 BJT and this sets it apart from the other jellybean BJTs.

See below for my simulation using Genesys and a 2N3904 model at 7V and 10mA. If I simulate for a point to point connection method and only include the component parasitic inductance and ESR then it appears to want to oscillate at about 84MHz.

In the simulation below I have done a very quick and dirty negative resistance analysis. This shows about -34 ohms resistive across the inductor at resonance at 84MHz. This means the circuit should still oscillate even if I put (say) a 47R damping resistor across the 20nH coil. This is a bit of a surprise but then again the reactance of the inductor is only about 11 ohms at 84MHz.

This assumes a very tight layout. I could try to simulate it to include the metal losses in your revised layout or I could have a go at actually building it on some scrap PCB material here.

Thankss... I'll troubleshoot abit on the board, if nothing then will make the emitter trace shorter later today.. Yeah, it's a genuine 2N3904

Can you please share the S2p file?

Why did you put that 47 ohms at the collector?

Also, closing the loop, just connected a jumper wire to both traces. Is that a problem?

Thanksss

[G0HZU_JMR]

The 47R resistor is really just there through force of habit. In this case it doesn't really matter but one way to tap out some energy is to connect a spectrum analyser at the collector. The combined resistance should really be 25R here but I modelled it without the analyser connected. The harmonic distortion will be very significant at the collector because the oscillator will be gulping current in narrow spikes. However, this is a quick way to get some energy from a Colpitts oscillator into a spectrum analyser.

See below for the s2p file for the 2N3904 at 7Vce and 10mA Ic. I've had to shrink it in size so it now covers 10MHz to 300MHz. It has loads more datapoints otherwise and the file size would be huge. You will have to rename it from a txt file to an s2p file as I don't think the forum will accept s2p data files.

This model was taken with my VNA earlier this year. 7Vce and 10mA Ic is the closest I have to your circuit. The transistor is an On Semi / Fairchild 2N3904BU purchased from Farnell with part number 1700648.

It should be OK to just use a jumper wire to close the loop as long as the connections are short and there is a DC blocking cap somewhere inline.

[G0HZU_JMR]

I also managed to find a data file at 5Vce 10mA for the 2N3904 and I've attached it below. However, it's not going to show much of a difference on the simulator in this case. It should be closer to the bias point for the startup condition for your oscillator.

These s2p files were taken with a logarithmic sweep starting at LF and going up into the GHz region and the file size is huge as there are loads of data points in the log sweep. I've shrunk the file size down to cover 10MHz to 300MHz.

[dmowziz]

Thanks Jeremy

My 2N3904 is different (central semi. corp) but ordering the 2N3904BU, unfortunately can't get it till next month.

Thanks for the files!!
I'm studying on bandpass filters right now, later today I will revisit this thread after this oscillator :
https://www.vintage-radio.net/forum/...=187482&page=2

You explained some things on design with s parameters and measuring.
I have the MMBTH10 from ONSEMI so I'll follow along

Thanks thanks

[G0HZU_JMR]

I had a go at building and testing the oscillator during the footie tonight. See the image below. You can see I used a very direct construction method for the resonator section as the caps are soldered together directly. I used PPI SMD caps and these are quite similar to ATC caps. I used a 2N3904 from the same Farnell bag.

You can see I've built it open loop so it can be measured with a VNA for gain and phase response around the loop. I simulated it in open loop and also measured it in open loop as in the image below and the simulation matched the real measurement quite closely as you can see in the plots below. The traces almost overlay perfectly.

One limitation of breaking the loop open is that it is never quite the same as having it connected up as an oscillator. Also you can see that the s22 measurement shows negative resistance at the emitter even in open loop.

When connected up as an oscillator it oscillates at about 85MHz. I used a 22nH SMD inductor from Coilcraft and this has an unloaded Q of about 100 at 85MHz. This means the ESR is just 0.12 ohm. The reactance is just under 12 ohms at 85MHz so you can see that it won't take much in terms of metal losses in the connections to spoil the Q of the resonator. Even with my tight layout with point to point connections the loaded Q of the real circuit was slightly worse than the simulation.

It does also show how good my s-parameter model is for the 2N3904. I found the circuit drew about 14mA so the simulation below is done with the 5Vce 15mA Ic s2p model of the 2N3904 although this isn't that critical.

[G0HZU_JMR]

Note that it is possible to post process the s2p data in an RF simulator to improve the accuracy of the open loop measurements to achieve a more accurate estimate of open loop gain. Randy Rhea describes this process in one of his oscillator presentations.

However, I'm not sure this is necessary in this case. The circuit does oscillate quite strongly at about 85MHz.

[dmowziz]

I am not impressed with my speed at all right now...
Anyways, just to post my latest on it so maybe can get feedback from you sooner

Originally Posted by G0HZU_JMR

I had a go at building and testing the oscillator during the footie tonight. See the image below. You can see I used a very direct construction method for the resonator section as the caps are soldered together directly. I used PPI SMD caps and these are quite similar to ATC caps. I used a 2N3904 from the same Farnell bag.

You can see I've built it open loop so it can be measured with a VNA for gain and phase response around the loop. I simulated it in open loop and also measured it in open loop as in the image below and the simulation matched the real measurement quite closely as you can see in the plots below. The traces almost overlay perfectly.

One limitation of breaking the loop open is that it is never quite the same as having it connected up as an oscillator. Also you can see that the s22 measurement shows negative resistance at the emitter even in open loop.

When connected up as an oscillator it oscillates at about 85MHz. I used a 22nH SMD inductor from Coilcraft and this has an unloaded Q of about 100 at 85MHz. This means the ESR is just 0.12 ohm. The reactance is just under 12 ohms at 85MHz so you can see that it won't take much in terms of metal losses in the connections to spoil the Q of the resonator. Even with my tight layout with point to point connections the loaded Q of the real circuit was slightly worse than the simulation.

It does also show how good my s-parameter model is for the 2N3904. I found the circuit drew about 14mA so the simulation below is done with the 5Vce 15mA Ic s2p model of the 2N3904 although this isn't that critical.

Thankssss

Originally Posted by G0HZU_JMR

One limitation of breaking the loop open is that it is never quite the same as having it connected up as an oscillator.

Yes professor..


I re did it, and making the emitter trace shorter like you said makes it oscillate (Thanks! for next time) : The open loop s21 is higher
But the power output is so much less than expected...

My aim is to have the measurement precisely (to a degree) like expected (simulation) like yours..


Havent measured the inductor and it's Q... Did you use your VNA to know the Q? (The inductor to ground ?)... I've done this before but in my mind is how to get crystal motional parameters which uses s21


Thankss... I want to get this right with your help before moving to next oscillator

If you look at the gain, phase plot. The phase crosses 0 again. When I initially did the plot in post #22, it was like this but it became right. What can be the problem?

Also, it does not oscillate when DC input is at 7.2. It does from about 8.5V. and I noticed the power supply leads has an effect on this. The SMA cable lead plays a little part as well.... How are you powering the circuit? Guessing should add a regulator next time? (then adjust voltage on simulator accordingly)



Thanks Jeremy

[G0HZU_JMR]

I got the Q for the 22nH inductor from the Coilcraft website. I used a 22nH from their 1812MS midi spring range.

https://www.coilcraft.com/getmedia/c...8fef2/midi.pdf

You can see the Q of the 22nH version is about 100 at 100MHz.

In my case I powered the oscillator from a bench supply and decoupled the supply with a 100nF ceramic cap. Normally I would use an active noise cleanup circuit at the power feed if I also wanted to measure the phase noise. The cleanup circuit stops any additional noise from the PSU from degrading the measurement. However, in this case I just used a very basic connection from a clean and stable linear power supply.

[dmowziz]

Thankss Jeremy.

I used a TL431 with stable output
used a different transistor (fake) but it oscillates even at 3V
The 2N3904 I use doesn't work.

Made progress from my last post but not home to show.

Please can you share a s2p file for MMBTH10 ? Thanksss
want to use it tonight

Sorry for this question : 5 Vce. Your built circuit is not 5 Vce


Thankss

[G0HZU_JMR]

Sadly I don't have s2p data at the operating point of your circuit. 5Vce is about the nearest I have. It shouldn't matter much in this case though. Any difference in the small signal analysis will be negligible with a change in Vce for your circuit.

I don't have an MMBTH10 model at your operating point but there is one bundled with Genesys that is fairly close. This is at 10Vce and 10mA and it is quite an old model made by Philips and I think this is the same as the MMBTH10. This model covers 40MHz to 1000MHz.
See attached.

The classic way to clean up the supply for an oscillator is to use a simple BJT pass transistor with an RC time constant at the base to provide the noise filtering. In this case the BJT would be a PNP device. Something like a BC557B would work fine here as an active noise filter. It does drop about 0.7V so the circuit would have to be powered from about -8V to get -7.2V.

[dmowziz]

I see..Thankss

Yeah, I used TL431 with a transistor. W7zoi uses the circuit you mention alot

Thanks for the file..

Ordered 2N3904BU, the price is "unreal". maybe fake
https://www.futureelectronics.com/p/...onsemi-7213325