VHF Oscillator design

[Radio Wrangler]

For 100MHz use and low-ish noise something like the 2N5179 is a classic bipolar for the job, the Ft is about 600MHz which means it has a gain of about 4 in the 2-metre band. For low phase noise you actually don't want too much Ft because transistors in that class start to become noisy at audio frequencies, and audio fluctuations in an RF oscillator all too easily turn into AM and FM sidebands. Some of the low noise audio transistors are good choices for HF VCOs.

There is one useful sourcebook of oscillator circuits, and a lot of stuff onPLLs. Ulrich Rohde (DJ2LR) is the son of one of the founders of Rohde and Schwarz. He's inherited his father's 50% stake. He's had two books on the go for a few decades now, one onHF receivers, the other on frequency synthesis. Periodically, he brings out an updated version with a different title. They are both interesting reads. I think the synthesiser one is "Microwave and wireless synthesisers: Theory and design'

I just checked... and saw the price. Yipes! £170 in hardback.

Don't let the title scare you. Go see about a loan from the reference library, or have a chat in the uni.

For anyone interested in communications receivers, his other book in collaboration with T T N Butcher is also well worth tracking down.

David

[PE9ZZ_JO22KI]

Ah, I see now. The 2SC388 is a TV IF transistor. You're lucky to get it oscillating at this frequency. At the transition frequency gain becomes unity. 2N4416 is a better bet. f(T) for BRF96 is 5GHz or so.

Too bad your SA doesn't have a better resolution. It will be indispensible to troubleshoot these phenomena. Then again, mine lacks a TG...

Tjerk, 9ZZ

[G0HZU_JMR]

Hi Wendy
To look at the free running phase noise of a reasonably well designed 128-148MHz VCO you would need one of the very best lab grade spectrum analysers. For home use this typically means a classic old school flagship analyser from the likes of HP or Advantest costing upwards of £45k when new in the 1980s (or something like an old Marconi 2382). These were designed (at huge expense) to have very low close to carrier phase noise.

Typical contenders are the HP8568B or the Advantest TR4172.

However, there is another way that you can do it today. The cheap/modern kitchen sink alternative is to mix your VCO down to 20kHz and look at the IF signal on a decent 16/24 bit PC soundcard using some spectrum analyser software on a PC.

Obviously, you would need a very clean LO for the mixer. I see you have a Marconi 2022 which would be adequate for this. See the attached image of the phase noise of a Marconi 2022 at 144MHz.

I can show you how to do this and it will also show you the level of your reference sidebands once you include the VCO in a PLL circuit.


NOTE:
You can predict the free running phase noise of your 128-148MHz VCO using a few simple sums based on the loaded Q of the circuit and the VCO tuning gain in MHz/V and the varicap steering resistance. I can show you how to do this if you like but the short answer is that you need to use a series choke + low value resistor for the varicap steering to get the varicap noise contribution below that of the VCO itself.

Allowing for a bit of flicker noise I would expect a simple low cost 128-148MHz VCO to therefore achieve phase noise in the order of:

200Hz offset -73dBc/Hz
500Hz offset -83dBc/Hz
1kHz offset -91dBc/Hz
2kHz offset -98dBc/Hz
5kHz offset -106dBc/Hz
10kHz offset -112dBc/Hz


However, once you use this VCO in a simple single loop PLL with 8.333kHz step sizes there will be a lot of divider noise introduced inside the loop bandwidth. So if you chose a low loop bandwidth of maybe 150Hz you might end up with PLL phase noise of:

200Hz offset -66dBc/Hz
500Hz offset -73dBc/Hz
1kHz offset -82dBc/Hz
2kHz offset -92dBc/Hz
5kHz offset -104dBc/Hz
10kHz offset -110dBc/Hz

According to the measurement plot using the SSA, the Marconi 2022 typically achieves:

200Hz offset -75dBc/Hz
500Hz offset -83dBc/Hz
1kHz offset -99dBc/Hz
2kHz offset -110dBc/Hz
5kHz offset -120dBc/Hz
10kHz offset -130dBc/Hz

This is better than the predicted noise performance of your PLL.

So maybe one of the first tools you need to get working is your cheapo mixer + Mi 2022 LO driving a soundcard to let you look at close in phase noise

I can design you a very simple JFET VCO for this range that has a very low parts count. It probably won't quite meet the noise prediction above but it will be fairly close. Within a few dB at worst case?

[G0HZU_JMR]

I quickly marked up the Marconi 2022 phase noise plot with the predicted noise performance of a PLL that uses a simple, low cost VCO with a VCO gain of maybe 8MHz/V.

The yellow trace is the Mi2022 phase noise at 144MHz
The red trace is the predicted phase noise for a cheap/basic free running VCO at 8MHz/V.

The blue trace is the predicted phase noise if you include it in a simple single loop PLL. You can see the loop bandwidth is about 150Hz because there is a cutoff 'knee' in the shape of the curve at 150Hz.

Below the knee you can see the PLL cleans up (removes) the VCO noise so the PLL is cleaner than the VCO below about 100Hz but the noise is worse above 100Hz. By 10kHz offset the noise is about the same as the free running VCO.

Note that the 2022 also appears to use a similar loop bandwidth of about 150Hz?

Note I have guessed the level of the reference sidebands at 8.333kHz intervals. These appear as the vertical spikes in the blue trace below. Your ref sidebands in your final PLL design could be much better than this or they could very easily be a lot worse. It depends on the performance of your PLL chip and your (active/passive) loop filter.

See below.

You should be able to improve the noise level of the PLL >5kHz by at least 5dB if you design a slightly more complex VCO that is optimised for low noise.

Or you could opt to use an all in one PLL/VCO/Divider chip like an ADF4351 that would give fairly good performance here from a very small footprint with no need to do any discrete design. But it would mean soldering a very fine pitch SMD chip on the PCB.

[Radio Wrangler]

The mix down with a really good sig gen technique will outperform the spectrum analysers. I put a how-to-do-it sidebar in the oscillators chapter in the ARRL handbook (all editions after 1995) The HP8640 is the classic sig gen of choice for this with its suprb cavity oscillator.

The cheat's method is to build two and to lock them apart by an Af offset, then mix to get audio for your soundcard analyser. Random noise won't correlate and you blame half of the noise power on each oscillator.

David

[G0HZU_JMR]

Yes, the downmix method works quite well and there's plenty of spectrum analyser software out there for the PC/Soundcard

I knocked up a quick simulation of a JFET based Hartley VCO that would tune 125-148MHz with a tuning voltage range of 2 to 4.5V and it can show a reasonable loaded Q of about 10-15.

With the above 9MHz/V VCO gain the steering resistance needs to be kept very low so I've used 220R in series with a 10uH choke as below.

This VCO is very much a barebones and minimal component design and it uses no AGC but it will deliver similar noise performance to my previous graphs/tables. eg better than -110dBc/Hz at 10kHz offset. Obviously, it isn't designed to get ultra low noise performance but it will provide 'typical' noise performance for a simple low cost design.

I can list up details of the inductor and the components if that helps. L1+L2+L3 is really just one handwound/tapped inductor.

Or you could just make use of the (Q2) active PSU cleanup circuit shown alongside the VCO and try it with your existing VCOs. This is a classic VCO PSU cleanup circuit that I have used many times although this is a minimal parts version of it. The transistor can be a general purpose NPN part like a BC109 or BC849.

The resistance used in the VCO steering of the varactors needs to be low. I've used 220R + a series choke. A simple sum can be used to predict the phase noise impact of fitting the wrong (too high) resistance here for a given VCO tuning gain of 9MHz/V and you can easily degrade the noise performance of the VCO by 20dB if this isn't right. Get it 'very' wrong and you can cause so much noise you can make the free running VCO sound dirty.

I think the BB405B is fairly similar to your BB121A in terms of performance so I think you could use the BB121A instead. The JFET is the commonly available J310.

[G0HZU_JMR]

Here's a couple of 'back of a fagpacket' calculations to predict the phase noise of your VCO.

However, I've kind of cheated and used an Excel spreadsheet (courtesy of Vizmuller).

But the sums behind the spreadsheet are very basic. You simply enter in a few details about the design and then you get the phase noise prediction at your chosen offset. However, these sums only apply in the region above the flicker noise region. JFETs give low flicker noise so the spreadsheet can be used quite close in. eg 10kHz.

The second part of the spreadsheet calculates the noise impact of the VCO varicap steering resistance for a 9MHz/V VCO.

You can see that 220R is a good choice as it contributes less noise than the basic oscillator alone.
(-118dBc/Hz)

I added a second spreadsheet showing what happens with a 120k ohm varicap steering resistor here. The noise degrades to -90dBc/Hz at 10kHz offset! over 25dB worse! The VCO would be very, very noisy with this resistor value fitted here! All the low noise performance of the basic VCO would be thrown away with this simple error

These sums are based on simple physics (Leeson's equation) and are only an approximation of a real VCO but I have demonstrated many times in the past that these simple sums will get you close to what a real VCO will deliver in terms of noise performance.

These days you can use expensive tools like Agilent Genesys and AWR Microwave Office to take it all a few levels higher in terms of accuracy but the basic sums are often a good guide

[Radio Wrangler]

I'd better point out that Jeremy's sums about noise contributions from resistors in the tuning line of varactors assume that good quality metal film resistors are used. Modern 'thick film' resistors may be around 20dB noisier and cracked carbon even worse.

Analogue devices have a rather nice PLL design program you can download. Of course it hinges around their parts, but you can bend it a bit by putting in yor sorts of parts.

For design freebies I use:

ADI PLL designer
Linear Technology LTspice
www.rfdude LLSmith
Agilent AppCAD

David

[G0HZU_JMR]

Hope I'm not overdoing it but here's a simulation of what your ADF4116 PLL will perform like with the little JFET VCO I showed earlier.

SIMPLL also has the option to predict the phase noise of the VCO using a simple Leeson model and you can see that SIMPLL predicts a very similar phase noise plot for the free VCO compared to the Vizmuller spreadsheet (see top right corner of image for free running VCO noise)

However, in the centre of the image is the PLL phase noise at 137MHz and it is very similar to my earlier prediction. However, I think you will see bigger reference sidebands than the simulation predicts.

[G0HZU_JMR]

I think David and I posted at similar times in the last two posts. SIMPLL is the same simulator as ADI Sim PLL that David recommended

For a bit of fun I've tried building the low parts count JFET VCO and it seems to work as per the sums in terms of 2 to 5V tuning range and the phase noise levels across 1kHz to 10kHz offsets.

I measured the phase noise with my old HP8568B and it agrees with the simulation. With the 120K steering resistor fitted the phase noise degrades just as expected to about -90dBc/Hz at 10kHz offset and the signal looks very ragged if I try and zoom in closer.

But it works fine with the 220R + 10uH choke fitted.

I have built it using a leaded metal film resistor here.There isn't much current passing through this resistor and I've modelled it simply as a thermal noise source in my previous posts.

I can post up images of it plus some phase noise plots if it helps? I also recorded the sound of it on an old FT290R (set to SSB) in wav file format.

I did also try a 1K, 10K and 120K resistor in the steering line instead of the series 220R + 10uH choke and the VCO got noisier and became very raspy with the 120K fitted.

I've tried to attach them as zip files. The 120K zip file is the raspy recording and the 220R is the clean recording.

I moved my hand near the VCO during both recordings to let you hear the beat note change on the FT290R in USB mode as the VCO changed frequency.

The VCO is quite stable and I didn't allow lots of time for it to cool after soldering each resistor in so there is still some slight drift on the 220R recording at the start. If nothing else this VCO would make a useful benchmark in terms of simplicity vs performance

[Wendymott]

Well.... good grief, you guys have certainly put some time into this........before I had read all these additional postings, I had re made the "dead bug" oscillator as a pcb, with printed inductor, once it had been screened, it has certainly improved the results I had previously. Until I have tried the testing methods posted above I cannot comment further. The next step is to increase the inductance of the tuning coil as it is about 20 Mhz too high and the loop wont lock at 128 Mhz. This pcb will be made sunday. Incidentally all the passive components except the varicaps are now SMD 0805 types, although I don't know the construction, eg Metal Film ....
When I originally posted the query I never expected such a response from such learned people.. many thanks for all your inputs...
Oh yes... a secondary question.... The PLL and filter is on the PIC controller pcb at present, as this project was to be an "all things to all people" project. I think I already know the answer, but.... should I transfer the PLL and filter to the VCO pcb...this will obviously cut down on leakage, and other nasties.

[Radio Wrangler]

Ordinary SMT resistors are Ruthenium Oxide or possibly Tantalum Nitride thick film jobs and are exactly the ones we're warning about. You can get metal film chip resistors but they take a bit more finding. Tubular shaped 'Melfs' are SMT offspring of metal film throughhole parts only without the end wires and without glaze on the end caps.

David

[PE9ZZ_JO22KI]

This varactor resistor thing I didn't know. Time to test my abandoned LMX2306 monstrosity (PIC controlled!) for phase noise. It has to be said that this device wasn't designed with good noise performance in mind... That with the SMDs I didn't know either.

Tjerk, 9ZZ

[Radio Wrangler]

That resistor thing is worth knowing about for any low noise application... audio, RF, etc. It's called 'excess noise'

Quantum mechanics says that an ideal resistor will produce kT Watts per hertz of noise. k is the Maxwell-Boltzman constant, T is the absolute temperature. In a high value resistor, that power corresponds to more volts and less amps. The impedance of a reverse-biassed diode like a varactor is very high at AF and so a high value resistor will better match all of its noise into the diode, putting noisy FM onto the oscillator. A low value resistor is a bad match to the diode and provides less voltage.

This is why Jeremy suggests using low value feed resistors.

Good opmps are available which make less noise than the resistors around them, if you don't exercise caution and keep their values down.

Some RF amplifiers use transformer ratio controlled feedback rather than resistor ratio controlled feedback because it adds less noise.

The tuning lines on PLL VCOs are very sensitive, and the noise it created is audible in radio applications.

You don't want resistors with excess noise except in non-critical places.

The audiophiles are quite right that there are problems with resistors that most people awe unaware of. What mystifies me is that some of them seem to like the ones with lots of excess noise.

David

[Wendymott]

Hi David. I appreciate all your comments and will look at them one by one in the following days. The problem for us "hobbyists" is holding stock of all the different types of components. I have just invested in a E24 range of SMD types, thinking that was the way to go... for this application,, apparently not.
Obviously the way not to hold a large stock of components is to prove the circuit, then substitute the devices for low noise types.
However ordering from such as CPC or R/S, minimum order charges or quantities are sometimes prohibitive.
I will look at your comments in more detail in the next few days and decide what action to take.

[Bazz4CQJ]

Declaring my "amateur-amateur" status straight away, can I raise the issue of temperature sensitivity wrt this project? Wendy started off by saying she was getting a warbly note from the BFO, and the responses have concentrated mainly on the "purity" of the signal. But especially with dead-bug construction on an open-board, isn't the oscillator going to be quite temperature sensitive, and perhaps even respond to radiated body-heat, and change frequency, unless there is good temperature-compensation built in?

B

[G0HZU_JMR]

It will certainly be prone to pickup from nearby electrical items such as a laptop battery charger/PSU or a noisy lamp etc.

The little minimal parts count VCO I built was in dead bug style and it drifted a lot if I went near it. Partly I guess due to capacitance and partly due to heating the air. However, I could get it to sound clean by using the active PSU cleanup circuit and also the 220R + 10uH choke for the varicap steering (and by keeping well away from it and keeping still ).

Note: when I was tuning it across 2 to 5V I used a regular PSU with an extra section of RC filtering for the Vtune. I think it was 10k ohm and 10uF. This worked quite well and it still gave a clean note.

I'm not sure what the realistic noise performance requirements are for an airband receiver are but I'd expect that -112dBc/Hz at 10kHz offset would be quite adequate. I really don't know how busy the channels are. maybe it gets really busy at an airshow?

should I transfer the PLL and filter to the VCO pcb...this will obviously cut down on leakage, and other nasties.

It depends how you arrange the layout and what other signals might be sharing the same current return paths etc.

Many commercial VCOs sit on their own little PCB so one option you could offer is a blank PCB area in your main PLL board that would support a few common VCO footprints.

A similar thing is done with OCXO/TCXO options in high end test gear. The PCB is laid out with blank space to hold big or small oscillator options. The silkscreen often shows the outline for every option, big or small and you plug your option into the correct hole pattern etc..

If this approach suits your project then maybe you could also leave quite a big footprint with pin/via holes to support (filtered) VCO power, Vtune, ground, FM AF drive, 5V (for an external prescaler) and also the VCO RF output.

You could typically get all this in an area of 40mm x 40mm and people could then build their own VCO daughterboard to fit into this footprint.

[Radio Wrangler]

Hi Bazz,

lassic amateur radio designs and professional ones used to concentrate heavily on achieving low temperature drift and compensating it as much as possible. THings changed and phase-locked synthesisers started to dominate everything. This fixed the temperature drift issue, but they needed wide-range voltage tuned oscillators and the old problem of drift was replaced by one of unwanted noise sidebands.

You can show that an oscillator is a noise processing machine and that the output is entirely noise. We cannot really make a sinewave, but we can make tightly filtered noise. If we are lucky we can get the noise so tightly filtered that it looks sufficiently like a clean sinewave that it will do for our purposes.

Once upon a time shortwave radios were limited chiefly by inadequate selectivity until crystal filters fixed that, then it was drift, then it was intermodulation, now it's phase noise.

David

[Wendymott]

Hi Jeremy. Thanks for the return on the PLL/VCO shared board.......... GOOD IDEA !! I will adopt the system of "satellite modules on a common pcb, then that does not restrict the PLL board to just one VCO.
Bazz... The "dead bug" was just to verify the circuit......it is now on a pcb with the tuning inductor printed on........ its a bit fiddly getting the inductor correct, but when it is.. its as stable as the "substrate" material allows, stops floppy coils etc...

[Bazz4CQJ]

Hi Wendy; could you post a block diagram of some kind to show the whole thing with the VCO and the PLL? Perhaps I need to find your orginal thread?

Thanks
B