When implementing a Hartley oscillator on a PCB board, using a BF199 TO-92 transistor and the rest 0805 inductors, resistors and capacitors the frequency of oscillation is only 1/5th of the expected value.
The same circuit simulated in Spectre using a model for the BF199 yields accurate results. Why is the frequency of oscillation so far off? Any ideas?

The picture shows the oscillator. The frequency is supposed to be around 200MHz with 6.8nH and 2.7nH for the inductors and 68pF for the capacitor. But as a PCB circuit it just under 50MHz.
BTW, the capacitor Cf, what should its value be? So far I always choose it such that its reactance is small compared to the reactance of L1.
Could it be the 0805 inductors? The data sheet gives a Q factor of 40 and a resistance of 0.2 Ohms.

Cheers

Thanks for the feedback ACWWong.

>>How about changing some values to see the delta freq.?
I did, but could only change them by a little since I have only 2.7nH till 10nH 0805 components laying around. When increasing the inductance I would have anticipated a further decrease of the already too low frequency, but it remained at around 50MHz. Leads me to believe the inductors might not work right. Hmmm.....

>> The  values of L are quite small for such big inductors... i think you'll need wirewound ones.
I will change the board to include wirewound inductors. This should then also tell me whether the 0805 ones were not working properly.
Their self resonace frequency is 4500MHz, well beyond to what I need. The Q-value is stated as 43 at 450MHz. From the Q-F curve it apprears the Q drops to about 25 at 200MHz.

>> Are you loading with a scope probe or 50 ohm ? are you using a digital scope (if so make sure you are sampling enough) ?
The scope is not a digital one. Agilent 54621A 60MHz, 200MSa/s. I don't have a loading resistor at the moment, just the probe. You think it'll make a difference if I put a high impedance load? At the moment, the output is taken via a capacitor from the collector to get rid of the dc component. I chose this c-value as big as the resonance capacitor.

>> Cf... Cres are you using a low loss RF capacitor ?
The capacitors are RF/Microwave caps. Again, the self resonance frequency is well beyond to what I need. (1GHz)
They're low ESR as well.


How big would you chose the Cf capacitor? Literature never tells you that as it's always being ignored. If I play around with it Cf will increase f when it is very small, the bigger Cf gets the less it influences f.

Yes, I know it's only 60MHz. However, I also tested the circuit on a better oscilloscope with the same results. I will fiddle around with the circuit, change components and try a few more strategies and then post any results.

Thanks

trond wrote on Oct 21^st, 2006, 12:57am:

So how about just shorting across the 0805 indcutor placements ? As changing L makes no effect and the L value is already next to nothing at 50 MHz..., I would guess that shorting would also make no impact (expect a modest increase if f)... Then we are sure our ocsillation isn't hartley obeying 1/(2*pi*sqrt(LC)).

In any case if you are measuring the frequency ok (best method would be Active high impedance probe into a spectrum analyser). An analog scope a couple of octaves outside its bandwidth doesn;t represent something you can trust... so if you got it working correctly, you might not see it!!

trond wrote on Oct 21^st, 2006, 12:57am:

Cf should be large (but not add much parasitic), i'd go for >100pF.
Are you using a large value and good quality for your supply decoupling C as well ? remeber that forms the return path for your resonant circuit... especially if your power supply is crummy.
Are you using a good pcb, suitable for RF ? have you looked at extracting what parasitics the PCB gives you ?

anyway best of luck with it...

yes, the L to C balance was too skewed... sometimes its nice to think about V=L*di/dt (and I=C*dv/dt...) using small L means large di/dt is needed for the oscillation, which might not be acheiveable by you circuit at the frequency/slew you need...

but since you need the 68pF.... how about putting is a series C to reduce the actual Cres ? then you can make L a decent value....

One last thing that I noticed. So far I had tested the oscillator without a bypass capacitor. So to reduce power supply noise I wanted to put a 100nF ceramic capacitor between the supply rails and ground (see picture in first post).  However, when doing so the frequency of oscillation drops from the right 125MHz value to something around 40MH. In fact, no matter what capacitor value I use, as long as its reactance it quite small the frequency drops to 40MHz. When simulating the circuit in Spectre Cby does not change the frequency of oscillation. Why is that? What am I missing here?

Cheers.

.........a little later...........

Maybe I was underestimating the effect on higher frequencies on capacitors. The ceramic 100nF capacitor I wanted to use as a by-pass capacitor is just too big I guess. When looking at the S-parameter plot I noticed a Fr frequency of < 30MHz. So at 125MHz the capacitor would have been an inductor. For a 680pF 0805 ceramic capacitor the Fr frequency is at around 350MHz (see picture). So I will try this one.
How would one choose a good by-pass capacitor? Any more tips than just looking at the S-parameter plot? What is the preferred material?

Thanks