Sunday, 20 September 2015

Cavity resonators; size does matter and preview of my HP 8591A spectrum analyzer

Cavity resonators; size does matter and preview of my HP 8591A spectrum analyzer

The Q or quality of a cavity resonator or filter is critical. The simplest way of increasing Q is to increase the diameter.

The club has been lent what seemed a pretty ordinary 2 m cavity resonator; even has UHF rather than N connectors and is light. I wasn't expecting too much. But it is 150 mm diameter compared to the 100 mm ex-government cavity filters the club has.

The two cavity filters. Note that the filters are not tuned to the same frequency; but they are not too far apart to make a difference to Q.

However, what a difference size makes! The Q is visibly much better. Note, none of the instruments were calibrated for this exercise, it is just to show the qualitative difference.

First, the quick and dirty antenna analyzer and a 50 Ohm terminator.

Then with the club's Rigol spectrum analyzer.

Then with my recently acquired HP 8591A spectrum analyzer; 1990s' technology, 10+ kg but half the price of the Rigol. Even had to read the manual on how to use the tracking generator! All the same, it works quite well.

Then with my Chinese KC901H "RF multi-meter". These are are a neat device, with a new improved version out (and twice the $600 approx that I paid.)

The KC901H as an antenna analyzer with 50 Ohm termination on one side of each cavity.

Both filters are presumably made of silver-plated brass sheet. The diameter of the larger one is only 50 percent bigger, as is the surface area, but what a difference that seems to make to Q. Without chasing the relationship of Q to diameter, it seems more than linear. I shall try to find out, as I thought it was linear.

There may be another explanation due to the construction of the filter, but I can't see inside the bigger filter. It may have a larger diameter antenna or probe, as the ratio of the two diameters affects Q and impedance.

Thursday, 10 September 2015

HP 437B power meter: new toy; traps for beginners

HP 437B power meter: new toy; traps for beginners

Introduction and traps for beginners

I have recently turned 60 and decided to get some new toys, in the form of stand-alone test equipment. For some odd reason, mainly curiosity, I decided a power meter would be useful, so I bought a 20 year old HP 437B power meter, with sensor and attenuators. Modern power meters of similar capability are about $10K or more!

As well as being expensive, the meters are not particularly intuitive to use, primarily because they use a range of power sensors that have very different features, discussed in this post, as it is a trap for beginners. In my case, the sensor was too sensitive for the unit's calibration signal; requiring attenuators just to get the device working.

Power measurement is not a simple task, other than for a unmodulated carrier; much less, a 7 MHz wide digital TV signal, especially DVB-T with 8000 carriers across its bandwidth.  Power measurement is not covered well in the literature. One of the best sources is a series of Agilent application notes: "Fundamental of RF and Microwave Power Measurements" (Part 1-4; Part 1: application notes 1449-1, literature number 5988-9213EN)

Recommissioning the parts

I bought the four main parts of the power meter from four sources to try and save money; "poor man pays twice". The parts are the meter, the cable, a sensor and an attenuator. After a quick glance at the operating manual, plugging it in and turning it on went ok, so it seemed to be working. However, when I went to zero the meter, it showed overload or would not zero.

The sensor I had was a common HP8484A thermistor sensor. It is useful from 10 MHz to 18 GHz, but has an input range of -70 to -20 dBm or 100 pW to 10 uW; a very low range. The problem is that the meter's calibration output, at the back of mine, but usually on the front, is 1 mW at 50 MHz; higher than the maximum for my sensor!

The solution is fairly easy, just add 40 dB or more of calibrated attenuators (not cheap) and carefully read the operating manual (especially Section 3-10: Simplified operation). Essentially, plug sensor (via attenuators) into calibration output and push "Zero" etc. Disconnect from the calibration port, then the instrument is ready to use, indicating a couple of thousandths nW (a very tiny amount of power and close enough to zero!). Each sensor has a table of calibration coefficients. These need to be keyed into the meter to get field calibration of the device.

Using the meter 

The power meter after zeroing and displaying signal generator output via 40 dBm of attenuators.

The HP 8484A thermistor type sensor (vs diode), a cheap 10 dBm attenuator, and a HP 1234 reference attenuator; 50 W 30 dBm, connected to the signal generator. The sensor's calibration table can be seen in the photo.

The HP 1234 RF signal generator generating a 1 GHz signal at 10 dBm (with some modulation), which is attenuated by 40 dBm before entering the sensor.

The power meter indicates -34.41 dBm, which is approximately correct. The sensor's correction factor had not been entered and the accuracy of the 10 dBm attenuator is not known. But it is working for the first time since obtaining the power meter.

The instruments are quite stable, a photo of the power meter about three hours latter, indicating 34.44 dBm, a tiny difference.


Power meters are not particularly simple devices to operate and are designed to give much more accurate measurements than the average amateur radio operator would want. Power measurement, for anything other than CW, is complex, especially for wide-band digital modes like TV, especially DVB-T, my interest.

Power measurement can be done to some extent with a spectrum analyzer, which is needed anyway, to ensure the amplifier is operating in its linear mode.

Saturday, 7 March 2015

Problems of different metals in a cavity resonator: Galvanic series

Problems of different metals in a cavity resonator: Galvanic series

A problem building a cavity resonator is the potential of electrolytic corrosion between different metals, such as copper and aluminium. This can be overcome by keeping the cavities dry and using a sacrificial anode, such as zinc, already part of the design.

The Galvanic series for metals in sea water is shown below.
From the table mixing copper and aluminium, is potentially not a good idea. However...
A cavity usually resides in the same room as the repeaters that is usually dry, hence no electrolyte to enable galvanic corrosion. Putting the cavity outside is a different story.
Another sneaky trick is to use a sacrificial anode to protect the other metals. Zinc, in the form of galvanised steel can do this. An integral part of my cavity design is the use of galvanised threaded rod to hold everything in place. A secondary function is the galvanic protection of the device with the sacrificial zinc, if there does happen to be an electrolyte, such as water on the cavity.  

Friday, 27 February 2015

IDSG101/GK101 10MHz touchscreen Function/Arbitrary Waveform Generator

IDSG101/GK101 10MHz touchscreen Function/Arbitrary Waveform Generator

Arbitrary waveform capability is compatible with

The GK101 10MHz touchscreen Function/Arbitrary Waveform Generator is a indigenous-Chinese designed instrument with significant capabilities at low cost. It is part of the emerging Chinese-designed equipment industry where the design is in-house rather than a copy or licenced design from overseas. Historically, this pattern follows Japan and Korea in indigenous design. China educates more engineers each year as the total number of engineers in the UK, so it is perhaps not surprising.

At the moment, there is not a lot of English language support for the GK101. I bought my GK101 for US$99 from InstradStudio, the distributor for the AirSpy SDR (Software-Defined Radio),: It is available from eBay at a higher price.

There is a wiki for the device: It provides most relevant information for the device. However, the software and firmware sites are in Chinese. The Chrome browser does an easy translation from Chinese to English, or other language as appropriate.

The firmware and software sites are protected by a password. This is given in the Chinese sites, but for the moment is


altera max II epm240m100c5n complex programmable logic device (CPLD) is a programmable logic device with complexity between that of PALs and FPGAs, and architectural features of both
The STM32F103xC, STM32F103xD and STM32F103xE performance line family incorporates the high-performance ARM®Cortex™-M3 32-bit RISC core operating at a 72 MHz frequency, high-speed embedded memories (Flash memory up to 512 Kbytes and SRAM up to 64 Kbytes), and an extensive range of enhanced I/Os and peripherals connected to two APB buses. All devices offer three 12-bit ADCs, four general-purpose 16-bit timers plus two PWM timers, as well as standard and advanced communication interfaces: up to two I2Cs, three SPIs, two I2Ss, one SDIO, five USARTs, an USB and a CAN.
The DAC904 is a high-speed, digital-to-analog converter (DAC)
offering a 14-bit resolution option within the SpeedPlus family
of high-performance converters. Featuring pin compatibility
among family members, the DAC908, DAC900, and DAC902
provide a component selection option to an 8-, 10-, and 12-bit
resolution, respectively. All models within this family of D/A
converters support update rates in excess of 165MSPS with
excellent dynamic performance, and are especially suited to
fulfill the demands of a variety of applications.
The ISSI IS61/64WV6416BLL is a high-speed, 1,048,576-
bit static RAM organized as 65,536 words by 16 bits

Tuesday, 24 February 2015

Tuning a commercial UHF duplexer with RigExpert antenna analyser

Tuning a commercial UHF duplexer with RigExpert antenna analyser

The club is assembling a 70 cm repeater with some equipment kindly loaned to us to use; a Kenwood TKR-820 repeater and a good quality commercial six cavity band-pass duplexer. The tuning was done with a RigExpert antenna analyser that gives a plot of SWR swept across a frequency range. The analyser was able to tune each cavity, but the overall response, while good for separation was compromised, possibly by incorrect inter-connecting cable lengths.

Tuning the duplexer

The duplexer is a band-pass type, with six cavities, three each for RX and TX with a "T" combiner. The cavities use N connectors and thick LMR400 type cable, and are about twenty years old.

The probes are set for maximum coupling, which is desirable for selectivity but for some transmission loss.

As a band-pass cavity filter, each cavity has a N connector in and out, making tuning each cavity quite easy; put a 50 Ohm termination on one side and the RigExpert analyse on the other. A typical response for a single cavity is a nice smooth curve, as you would expect.

However, once each cavity is tuned and all connected again, with a terminations on the RX and TX ports and the analyser on the antenna port, the high level of separation of the RX and TX can be seen.

The depth of each pass-band is less than that for a single cavity, plus the pass-band is not nice and smooth. This is presumably an effect of the interconnecting cables and the "T" type combiner.

The individual RX and TX pass-bands can be observed. The pass-band is sharper than for a single cavity, shown in the first photo, but the response is not as deep or as smooth.


The most likely cause are the interconnecting cables between cavities, which are about 270 mm. My understanding is that cables should be an odd number of quarter wavelengths. At 433.6 MHz, a quarter wavelength is about 173 mm. The velocity factor of RG-213 coax is about 0.66, so the cable length should odd multiple of  about 114 mm. In this case, the connecting cables are close to a half wavelength, not good for duplexer's performance.

The duplexer was factory tuned for TX 466.95 and RX 457.45 MHz. For TX, the cable quarter wavelength is about 106 mm, with the cables being closer to three times a quarter wavelength. The length of the cable is a bit of a black art, often done by substituting cables of different lengths. I think the length of the cable plus some of the length of the probe inside the cavity may be the correct length.

The type of coaxial cable is very important as the velocity factor for RG-213 is 0.66, whereas for more modern and better shielded LMR400 coax, it is 0.85, a very significant difference.

Making new cables is a considerable task, especially the "T" combiner, as it is custom made. For LMR400 cable, 147 mm is too short for thick cable, thus requiring three times a cable quarter wavelength or about 441 mm. Thinner cable may allow a quarter wavelength.


For the moment, I think we shall use the duplexer as-is and see if it has any practical issues. The separation is good, but the transmission losses may be too high, especially for an aged repeater with  5 W output and aged RX frontend.

The RigExpert is a very convenient instrument for tuning the cavities, however a tracking spectrum analyser may be needed to measure the overall duplexer's performance, particularly transmission. losses. I have a portable one and the club has a desktop spectrum analyser, so we may use them another day.

Sunday, 8 February 2015

A $300 2m repeater duplexer and cavity resonators-hardware shop special

A $300 2m repeater duplexer and cavity resonators-hardware shop special


I have been experimenting with cavity resonators for a 2m repeater for some months now. I had been using aluminium tube and brass fittings. However, with 150mm tube, the optimal antenna should be about 50 mm diameter, with a 3:1 ratio of diameters for impedance and performance. However, the mechanical arrangements and conduction between them became problematic. If I wrapped all the joins with aluminium foil, it worked well, but is not practical.

In this post I will outline a new technique that uses shim copper (0.1mm) for all the RF parts. The aluminium tube is lined with copper and soldered to copper shim top and bottom plates, in turn attached to the main aluminium plates. The antenna goes back to a conventional fixed lower section, with a movable section to adjust tuning.

The basic premise of the design is still that it can be made from readily available parts at low cost, and with basic workshop facilities. As such, a six cavity 2m duplexer can be made for about $300 over a weekend or two.

The basic design

The basic design of each half the duplexer has not changed much from my earlier attempts, shown below.

The aluminium tube is 150mm diameter and 600mm long. The top and bottom plates are 150 x 6mm, 480mm long. Threaded rod is used to hold the assembly together. Plumber 25mm threaded pipe is used to adjust length. In this prototype, only one cavity is working with a single input to the probe. In this prototype, the threaded rod moved the whole antenna to adjust length. That arrangement became problematic when the diameter of the antenna was increased to 50mm, the optimum for this size cavity.

The main cost is the aluminium tube at about $150 for 6m, enough to make ten cavities.
Aluminium is used as it is much cheaper than copper, as well as lighter. Further, it is difficult to get large diameter copper tube, over $1000 for a 6m length.

I was looking for alternatives and was browsing a non-ferrous metal supplies website, particularly sheet cooper; I saw that 0.55mm copper sheet was available and seemed quite thin, thus bendable ($136 for an 1800x900 sheet, $17/kg, not bad as I had been paying $7/kg for scrap copper)). I bought a sheet, but noticed they had 0.1mm, 600mm wide in a roll on the shop desk. I bought one metre thinking I might use it for something. At home, I noticed the price, $35 per metre or about $75/kg! However, I have found a cheaper supply in Melbourne:, with the shim price about $25/kg.

I have built one cavity with copper shim lining the outer aluminium pipe. While this is possible, it is difficult for one person to do. To make it a little easier, I ran some thin lines of non-corrosive silicone to hold things in place while I solder the join. The ends have turned-over copper shim to give a copper end and to solder to the copper top and bottom plates. A special-purpose soldering iron bit would help with the inside seam. I am not sure copper-lining the cavity wall is worthwhile compared to aluminium. I am currently experimenting with this.

The copper shim on the the antenna/probe was more success. I used PVC plumbing parts and tube, with the brass screw, to construct the basic antenna. The two parts are show below.

The assembled antenna/probe soldered to base plate of 0.5 mm copper. Contact between the fixed and movable section, just cuts in fixed section, over-laying each other, with zip tie holding them tight and two fingers turned back to hold zip tie,

The copper-lined aluminium tube is in the background.

 The Q of each component is important, however size matters! The smallest bits may warrant silver-plating, whereas there is little benefit in plating the larger pieces.

With that in mind, I thought I would try silver-plating the probes, as they are tiny, by area compared to the antenna, tube or top and bottom plates. I could get them silver-plated, but that would take time and money. With a bit of lateral thinking, a second-hand silver tray seemed a good source. I had a quick look on eBay and bought a reasonably sized silver-plated tray for $6 plus $10 post. The tray arrived a couple of days later but was a bit heavier/thicker than I envisaged, making cutting and  working harder. However I fashioned a probe, on the right in the photo, as best I could to match the other one, made from copper sheet. The probes are about 5 mm from the antenna per the QST/ARRL design.

The assembled filter and the impressive SWR, much better! I fixed a few things, so all the improvement may not be from the one silver-plated probe.

For ball-park, I ran the analyse on a old commecial 2m cavity, 100 mm diameter, and I am not too far out; most pronising.

In my tests with two probes, I have had to put a 50 Ohm terminator on the second probe. Without it, just a dimple. When I started, I only had one probe and was able to get usable SWR. Not sure what is happening, other than the probes work well!


Thermal stability

While I have not mentioned it so far, I recognize issues such as temperature and mechanical stability. PVC has a thermal expansion double or triple that of brass or copper, which in turn is worse than Invar. I simply can't afford Invar, plus it is not easy to get; old off-frequency filters being a possible source. For the moment I will continue with PVC fittings for the antenna adjustment rod, but will be looking for better ways of doing it. I want to build 6 m band cavity filters where I will need something other than PVC, given the longer length.

However, the duplexers are to be used in my club's repeater room on the Gold Coast, Queensland, Australia, where the temperature is fairly stable (maybe 15 to 40 degrees); they don't need to do the +/- 40 degrees of snowline North America.

Aluminium flashing rather than copper shim

I have contemplated using aluminium flashing instead of copper shim, 0.3 mm thick at a number of widths, cheap and easy to get at hardware stores. The aluminium flashing could be used for the probe/antenna. It could even be used to line 150 mm or 200 mm PVC drainage pipe, half the price of aluminium pipe at about $80 for 150 mm 6 m length. However, that price difference is probably not enough for the hassle involved. It may be different for 200 mm, as aluminium is not readily available at that diameter, but I have not checked further. I think 150 mm will be ok for 6m cavity filters.

An advantage of all aluminium is less electrolytic corrosion. The Gold Coast can get pretty humid and marine, but again, I think the repeater room is dry and stable enough for it not to be a problem. In a shed on a mountain, that may be a different issue.

Bandpass vs notch

The filter is a bandpass filter. I could make it into other types of filter, such as the various types of notch filter using an inductor or capacitor between the probes. I will do that later; a good bandpass is sufficient or proof of concept for now.

Similarly, I have not been concerned with losses through the filter. Losses and selectivity are a compromise with the degree of coupling, along with the Q of the filter (ref??). On RX, I am less concerned, just use a pre-amp; selectivity is crucial compared to losses. On TX, I am not sure I even need the sharpness of a three cavity filter.


Close enough for me for now. I have a design that seems to work quite well. I won't do any more major changes on this filter, rather call design closure and make another one, I have enough copper shim to do that, plus I can make a tidier new one.