23cm Swept Oscillator

If you read my earlier post, here, I covered the background to a 23cm VCO I built. While this vco is likely to be used in a transceiver one day, it also became part of a useful piece of testgear - the swept oscillator. While there are other ways to align filters, using a swept oscillator rates highly on the bang for buck scale.

 Principle behind use

Principle behind use

The basis concept behind the use of the swept oscillator is shown above. The frequency output of the swept oscillator is connected to a radio or a stand alone filter. Somewhere from the radio a response signal, perhaps the AGC line, is picked up and fed to the Y axis of a CRO. The CRO doesn't have to be fast, since the signal has been detected and the rise time of the detected signal is slow.

The sweep speed has to be fast enough to create the persistence of vision of the CRO's trace, but not so fast that the detector has not had time to react. A sweep of 25 to 40 times per second works for me.

My implementation

At 2m and 70cm I took the expedient path and hacked a commercially made oscillator. But at 23cm I had no alternative but to build a swept oscillator. A picture of my swept oscillator, working but not quite finished, is below. The maximum sweep is from 1230MHZ to 1406MHz.  You will notice the piece of copper foil I soldered onto the middle stripline to lower the frequency range it was covering.

The three controls are sweep speed, sweep width and centre frequency. The bnc connector connects via a patch cable to the X axis of the CRO. The N connector is where the RF frequency is sourced. I use an external attenuator of at least 15dB to ensure the filter or radio thinks it is connected to a 50 ohm source.

VCO
I covered this previously, here, so I'll just repeat the schematic.

Ramp Generator
I have used this ramp generator many times. It was based on something I saw at Hans Summer's website but with a small correction. It took me a long time to work out why, at the highest point of the waveform, the voltage did not reverse but instead momentarily spiked to the supply rail. It was caused by something called phase reversal and is the result of the op amp driving a capacitive load.

The fix was a small value of resistance, 330 ohms, in series with the output of the op amp to isolate the capacitance.Works a treat now.

Sampling a response

There was a time when sampling the rf coming out of a filter with a diode detector was the norm and it wasn't possible to worry about the response being linear or logarithmic.

Today there are logarithmic detectors that work on low level signals. The AD8307 is useful up to, and including, 70cm. At 23cm I use the AD8314. I will write these projects up soon because my implementation of these logarithmic detectors has enhancements I have not seen anywhere else.

If I am sweeping a stand-alone filter then the output of the filter, with perhaps 10dB of attenuation to improve the match to 50 ohms, is connected to the AD8314 chip. The output drives the Y axis of the CRO.

Where I am trying to sweep a filter already in a radio then I modify my approach. I haven't done this at 23cm yet because I don't have a radio to align. But I have aligned numerous 70cm and 2m radio's with the following approach and I expect it to work at 23cm. I have a dud bnc patch cable with the dud connector cut off. This open end of the coax becomes my sniffer probe. I simply probe around the radio after the filter until I see a response on the CRO. Then I align the filter. I am not making a direct connection. The AD8307 or AD8314 are sensitive enough that capacitive coupling is all that is needed.

PCB
I have a pdf of the artwork if you would like to build one. I don't know how to put this into the blog so a jpeg image is shown below to whet you're appetite. Note that it is not actual size!

Voltage Controlled Oscillators (VCO) for 23cm

One of my long term projects is to build a 23cm  FM transceiver. This post summarises a few experiments I conducted with a view to making a wideband VCO for testing purposes. The basic variations I tried were all varactor tuned:

1. Lumped LC (or perhaps more like parasitic LC at these frequencies),
2. conventional semi-rigid coax
3. DRO, or dielectric resonator
4. stripline

Version 1
Narrow tuning, layout needed more attention to get a higher frequency

Version 2

Despite my best efforts using a short piece of coax as a tuned circuit I had no real success with this approach. I was able to achieve oscillation but I didn't find this approach for me, a mere mortal, as being something I could reproduce when ever I wanted an oscillator at these frequencies.

Version 3
My DRO attempts were similarly mixed. I could achieve oscillation, but my junk box did run to a huge number of DRO's. I even tried modifying the DRO's found in old mobile phone filters.

I could get these oscillators to work. But just like crystals you never seem to have one on the frequency you want. I ended up achieving a maximum frequency of 1150MHz with this 1600MHz DRO. It seems from my testing of the DRO's I had that in general the maximum frequency when used in an oscillator is around 70% of the resonant frequency of the DRO alone.

What was interesting was how critical the board layout was. Putting a piece of copper foil under hte DRO, thereby extending the groundplane, lifted the frequency of this oscillator from 904MHz to 994MHz. That's about 10%. Placing a via next to this extension lifted the frequency still further to 1105MHz. I never managed any further significant increase in frequency. A raft of fiddling only achieved 1150MHz.

Version 4

Days had passed fiddling with these approaches when I tried a strip line version of something I had seen in "VHF Communications" once and at http://lea.hamradio.si/~s53mv/spectana/vco.html. Almost immediately I enjoyed success. I now had an oscillator I could  build over and over and I found that each time I built one it worked, albeit with some modest alteration to the stripline length if my guess was off the mark.

After my first attempt I decided to dispense with direct coupling the mmic to the transistor in the oscillator. Instead, I used a fourth stripline adjacent to the resonator striplines feeding a mmic to lift the output level. The circuit, shown below, has proved to be repeatable. The stripline length may need to be varied depending on your substrate and varactor properties. The stripline to change is the centre stripline which runs between the collector and base striplines. Shorten the stripline to raise the frequency, or lengthen the stripline to lower the frequency. Easy with a knife or a small piece of copper foil cut into a strip and soldered onto the existing stripline to lower the frequency. If you didn't have any copper foil, you could also use the shield from some thin coax or a piece of thick copper wire.

Apart from construction errors and defective parts this should always work. If you can't get it to oscillate try a transistor with a higher ft.

While I used some 6 pin mmics I recovered from a PCB, you could use just about any mmic that comes to hand if you modified the board layout.

Top layer of Stripline VCO, bottom layer is groundplane. Here the mmic is loosely coupled directly to the oscillator

I built one of these with a sweep generator on the pcb which I will post details of later. The pcb, which when compared to the direct coupled mmic version above, should help explain the principles of the layout.

Regards
Richard VK6TT

Parts Storage

If you're anything like me then you have a junkbox. The junkbox is a source of never ending distraction that can lead you astray from the project at hand. Over the last few years my approach to the junkbox has dramatically changed. To help you avoid some of the time consuming mistakes I made here is how I now do it.

The Now,3,Box,Bin mantra

Sometime in 2016 I realised I was spending too long recovering parts, sorting and finding containers. So I adopted some simple rules, which became the now,3,box,bin mantra, illustrated with the mythical part number abc123:

1. I would only recover abc123 if it was needed now
2. If I had more than one abc123, either on one board or on several boards, then all those abc123's would be recovered only if there was a reasonable chance I would need them in the next 3 months and the remainder put into a container. 
3. Every time I picked up a board I actually identified what I might recover for later use and put this into a spreadsheet. As each box filled up the sheet would be printed off, stuck on the lid and the box put away.
4. If I really had no foreseeable use for any of the parts on the pcb, put the pcb in a bigger box known as the bin. Before going to the verge I would try to give away as much of this as I could. I found that anything I considered rubbish was a view shared by everyone else. But I let each bin box sit there for 6 months before it went to the verge.

 I use this mantra for all parts, leaded or smd. I still do recover parts as needed and the question remains what to do with the other parts which get recovered at the same time.

Leaded parts

I still put leaded parts into containers, but since I don't use that many I only use broad groupings eg all npn transistors into one container. Sometimes, if the board is a real mixture of semiconductors, everything just goes into a semiconductor container. If I ever need to find a diode, transistor or led I know where to look. Clear containers are invaluable for finding things later on. The same though process applies to capacitors and high wattage resistors. IC's tend to be sorted then and there. 74ACxx into one container, 74HC into another, op amps into another.

Once, when leaded parts was all I used I would group these into AF/GP, Switching and RF. Perhaps the reason I sound so dismissive now about the sorting is that I still have too many transistors in those bins to worry about any further leaded parts I might recover.

The only reason I recover a leaded part nowadays is that I must use a leaded part. Otherwise I leave them on the board, catalogue and box.

Surface mount parts

As I moved into surface mount I started sticking recovered smd parts onto old business cards, grouped by value. Then, as my source of containers grew, the recovered and increasingly "new" proportion of parts went into containers.

 One of the great discoveries of 2016 was these containers on eBay:





10 of these are currently selling for around A$4, or US$3. If you look closely there are 12 compartments in each "wheel". And there are 12 values in the E12 series we commonly use for parts values. You can find these with an eBay search string of "10pcs empty round storage", available worldwide option checked. I use the 6cm versions though I sometimes see them in a larger size which might be useful for my miscellaneous board.

I did keep these in a separate container but I now have glued them onto a few boards. So on a piece of board some 25cm x 25cm I have 7 wheels. That allows me to find any smd resistor from 1 ohm to 8.6 mega ohms just by looking at the board.

The board holding my smd resistors.
Get the orientation the same on all wheels, I overlooked this on the first two wheels.

It also tells me what values I don't have. If I am going to accidentally recover several 1206 resistors of a value I don't have then I know where they will go. And if the compartment is well populated then it's into the bin with them.

The same thing works for 1206 capacitors. Where the capacitors are labelled then I know where they will go if the compartment is empty. I don't bother with unlabelled capacitors unless I'm really stuck for a part value and waiting 4-6 weeks is unpalatable.

On my third board I keep a miscellaneous collection of surface mount parts.

Ic's tend to go into empty metal mint containers now that I found a large container that could house the mint containers standing up.

Catalogue

Definitely worth doing. Since the box is in the workshop and I'm often in the house this is a great way of keeping track of the more exotic parts. Think rf capacitors found on transmitters. I know which values I have and which board they are on. So I can design, for example, a transmitting filter or diplexer, knowing what values I have without walking up to the workshop and rummaging around.

I'm wondering if a photo of each board, or section of the board, pasted into the spreadsheet might be useful too.

Summary

I spend more time on building things now that I have reduced the amount of time I touch boards. Hopefully there is something in this that helps you become a more productive home brewer too! Comments on alternative approaches most welcome.

Regards
Richard VK6TT

Unilab KL70 10m FM Converion - PA Matching schematic

Please find below the schematic showing what was changed. I ended up achieving 50W across the FM segment of 10m. Now I just need to set the deviation and the radio is ready for use!

Regards
Richard VK6TT

Unilab 10m FM Output Matching Network - Error in original radio corrected

You may recall from the post here where I first discussed the modifications needed to the PA that I modelled the output matching network as shown below:

With only 40W of RF out I removed one of the trim caps to see if I could learn anything that might explain why. What this revealed was that due to the pin-out of the trim cap it was not connected to the right point in the output network. Instead of the network above we had this:

Actual output network before error spotted

A combination of the trim cap being connected on the wrong side of the LC trap and using ideal parts when real might have been more appropriate mean the output network was matching to perhaps 32ohms, not 50 ohms.

So I connected the trim cap to the correct side of the LC tank. Now I was getting 50W and to be fair that is not a bad result. The 50W was flat across the 10m segment and the second harmonic was 50dB down. The third harmonic was over 60db down and I felt no need to measure it with more precision.

But as I was writing this I realised that an extra 39pF padding the very last trim cap might improve the matching. I tried this but the only benefit was the second harmonic content fell even further.

It appears I have to be satisfied with 50W unless I change the MRF247 to a transistor better suited to this application. If anyone manages to coax more than 50W from a MRF247 at 10m I would dearly like to hear how you achieved this.

Next time I'll publish a circuit for all the PA stage changes I made.

Regards,
Richard VK6TT

Unilab 10m FM conversion Transmitter PA mods - Low Pass Filter

Well it's been hot here, with two days over 40degrees Celsius now. I've used that time incorporating a pre-driver into the radio and considering the harmonic issue. It turns out that in Australia there is no allocation of the spectrum where the second harmonic would be an immediate problem. With that in mind I have stopped worrying about the exciter harmonic level for now.

The existing 7 pole LPF can be rebuilt. I have settled on a 7 pole Chebyshev filter as follows:

The suggested attenuation of the second and third harmonics is 37dB and 67dB. I would like more attenuation of the second harmonic but I will wait and see what I achieve in practise before redesigning the filter. If I achieve my targeted 50dB below the carrier for the second harmonic that will be less than 1mW of RF power being radiated which should keep me out of trouble.

I have some 110pF surface mount RF caps which has shaped the filter I have selected. I only need remove the two 33pF outside capacitors on the filter, to be replaced by 110pF capacitors, and pad the inside capacitors with 110pF.

LPF Board before modification

This morning I quickly wound the inductors, measured them to ensure they were close and put everything back together. And I removed that disc ceramic capacitor shown in the photo above.

Regards
Richard VK6TT

Unilab 10m FM conversion Transmitter PA mods - Revised Driver Stage

After a few cooler days I have been able to finish the driver stage. In place of the BLY32 there is now a two stage amplifier. It was all a bit of a squeeze but we managed to fabricate a PCB that would fit in the cut-out .

Schematic of replacement for BLY32 module

PCB of BLY32 replacement module before preparing to etch.

BLY32 replacement module installed

I made an error on the pcb layout which is why the collector of the 2SC1971 is floating. If you would like to see a revised circuit board as a pdf so you can build you're own please ask!

Using the 2SC1971 as a driver means we are limited to driving the final stage with around 6W. Which is enough for 100W from the radio if we have 13db of gain in the final transistor which I suspect we do have.

At present I am getting 40W out of the radio, current draw is around 5 amps. I think I have discovered why I am not getting 100W. It was to do with output matching. As built, the radio's output matching network is not as per the circuit diagram. It turns out the variable capacitors on my radio had a different pin-out with the result the matching network is different than I has assumed. I will alter my radio and let you know what happens.

Regards
Richard VK6TT

Measuring inductors from 10nH to 10uH

Update - Even though this approach is valid, I now simply use the NanoVNA with a test fixture that is included in the calibration. It allows me to measure the effective inductance at the intended frequency of use.  Covered in the early posts with the Test label.

Here is a test fixture I use to measure small, down to nH size, inductors. I got the idea from the book Solid State Design.

Provided your power meter shows a loss of at least 20dB with no inductor connected then the capacitor between the bnc connectors and the floating pcb are small enough in value. When Cg is small then the peak detected is more pronounced and the measurement of Q is more accurate.

In my implementation I use 2 bnc sockets from a panel which are connected to a small piece of blank PCB supported above the ground plane.I glued the floating PCB down with some plastic spacers between it and the PCB underneath. My reasoning was this would reduce the impact of temperature variations on the capacitance by putting an air dielectric between the substrate and the groundplane. I haven't determined if it really makes any practical difference so you might find this unnecessary.

Each bnc connector is connected to this supported pcb via a gimmick capacitor. I used a 2cm or so length of miniature coax. I should have built this with the inner to the bnc connector, the shield to the floating pcb. You can see from the photo below I got one of the small pieces of coax transposed. You could use a gimmick capacitor from twisted wire. If you were using a larger known capacitor then you might find a really small value disc ceramic is suitable.

From this pcb to ground I have a nominal 100pf capacitor. The value isn't critical since you can measure the actual capacitance with a capacitance bridge or meter. I use my digital LC meter which appears suitable for measuring C but I can't rely on it when measuring L since it works at low frequencies. If you can't measure capacitance then use a tight tolerance capacitor and assume a few pf extra.

When I checked with my capacitance bridge I had 109pf from the supported pcb to ground. That forms the C in the well known f = 1 / (2PI sqrt(LC)) equation. I then sweep with my signal generator on one port, and my power meter on the other. At the peak response I solve for L. This allows me to measure nH inductors with ease. Did you know that capacitance is essentially constant with frequency whereas inductance changes with frequency? Measuring an inductor at 1kHz  gets you close, but I prefer measurements at RF frequencies.

By sweeping the frequency to find the points 3dB down I can estimate the Q! The unloaded Q is equal to the resonant frequency divided by the difference between the two frequencies where the response is 3dB down. I will post a separate blog soon about this aspect.

For inductors of 800nH and upwards I have an alternative instrument I can use which measures the capacitance at several megahertz using a clock module. It works well despite only providing a subjective measurement of Q.

The fixture described does work really well and since I strive to build things that do not need tweaking it reduces the uncertainty around measured inductance.

Regards
Richard

Unilab 10m FM conversion - Transmitter VCO Board

In the quest to get the harmonic content of the transmitter under control I returned my attention to the VCO board. Sure enough, when I measured the output of the mmic I found it was around 900mVpp and the flat topping of the waveform was clearly seen on the CRO. This equates to 3dBm for a 50 ohm load and is right on the limit for the mmic used. You may recall that the coupling capacitor from the VCO to the mmic was a parallel combination of 1p5 and 2p5. I removed the 2p5 and replaced it with a 1p5, giving 3pF in total. Did it make a difference?

Measured at the exciter output into my spectrum analyser (all readings in dBm):

	Before	After	Difference
Fundamental	22	20	-2
2nd Harmonic	13	3	-10
3rd Harmonic	-6	-7	-1
4th Harmonic	-7	-30	-23

So yes, it made a difference. While there was little change to the output or the third harmonic, the second and fourth harmonics were significantly reduced.

While the original mmic was fine when the VCO was running at 80MHz and the pre-scaler had plenty of sensitivity, the reduced pre-scaler sensitivity at 30MHz means the mmic has to be run at the limit to get enough signal into the pre-scaler for reliable counting to occur.

I see two alternative courses of action: substitute a mmic with a higher output rating and similar gain, or put a small amplifier stage between the mmic and the pre-scaler so the output level from the mmic can be reduced still further. A different mmic is perhaps the most expedient solution.

The existing mmic has a gain of 18dB at 30MHz and an output level of 3dBm. The suitable alternatives I know of are a MAR8 or a BGA616. Neither is a simple drop in replacement though.

Type Po -1dB Freq Range Gain at 30MHz uPC1651G (existing) 3 10MHz-1GHz 18 MAR8 12.5 DC-1Ghz 32 BGA616 18 DC-2.7Ghz 20

The BGA616 would be my first choice. It is capable of a higher output level then the MAR8 and this should result in even less distortion, hence harmonic output, than the MAR8 will give. The biasing of the BGA616 would also be easier given the 8V rail available on the VCO board. A 100ohm resistor and inductor will suffice. The MAR8 runs at a nominal 7.8V and I suspect the 5.6ohm biasing resistor it will need is too low in the event the temperature rises.

In some ways a small amplifier stage between the mmic and the pre-scaler is less critical. A voltage gain of say +4 would be ample and there is no need to re-jig the board for accepting a different mmic. And just about any small signal transistor could be pressed into service.

So I built this onto a piece of vero-board and mounted it as shown in the photo below:

And after removing one of the two 1p5 caps that coupled the oscillator to the mmic the results were:

Which confirms the mmic was definitely being over-driven. The relationship of fundamental to harmonic output levels now "appears" better. However, when I consider that these are measured after the tuned tank circuit I added on the main exciter board I think the results should be better.

I'm going to ponder this over the next few days as we go through a heat wave. I am going to rebuild the driver stage onto a small pcb which will fit in the cut-out where the BGY32 used to go. It will include a pre-driver stage to lift the transmit power from 20W to something closer to 100W. And since this involves etching I will also rebuild the buffer onto a small pcb which will look a lot neater. If you'd like either board just let me know!

Regards
Richard