20200915 165802

Recently I was making some phase noise (PN) measurements using John Miles PN programme in his GPIB Toolkit using my 8566B spectrum analyser (SA) and was finding that there were spurrs occuring at 50/100 Hz intervals at the low frequency ends of the plots. A closer examination of the spectrum showed a series of close in spurs at 50 HZ intervals.  (The line frequency in Europe). The spurs were at around 45 to 50 dB below the carrier which seemed to be too high. They are not normally obvious but at a 1kHz span and a 10 Hz bandwidth became quite apparent.

12575GHz sidebands

The HP8566B Installation and Verification Manual (HP Part No. 08566-90169) gave a detailed table of the sideband spur specification which showed conclusively something was wrong. The spurs were around 15 to 20 dB to high. This table is below.

power line sidebands

I was confident that the 12575.5 MHz Herley PLL I was using as a signal source was clean and was not the source of the spurs, so they were from the SA itself. I opened it up and checked the power rails for ripple and they were all perfectlty clean and set to the  correct voltages. At this point I was assuming that the sidebands were being caused by some spurious modulation of the first YTO (first YIG local oscillator) in the analyser. I checked out the capacitors in the YTO driver circuitry and all seemed fine for both ESR and capacitance. On consulting the HP 8566B Service and Repair manual pages on the A6A7 module I discovered that the YTX (YIG tuned mixer) was a known source of power line spurii. The manual suggested that noise on the tuning current fed to the coil of the YTO modulated the incoming signal which became visible on the display as spurs at resolution bandwidths of 100Hz or less. The solution adopted was to switch a 100uF (A6A7C1) electrolytic capacitor across the the coil to bypass the residual ripple noise as shown below.  C1 is switched into circuit across the YTX coil by the  2N6073 triac, Q8.

A6A7C1 sch 

As a crusty old EE it struck me that this was not a good way to treat an electrolytic capacitor and expect it to have a long life. I was aware of the history of this particular 8566B (serial number 2332A02879) and I knew that I was its first custodian after HP/Agilent and that it was unlikely to have been used on narrow bandwidths very much. It had probably come from a HPIB automated production line environment. It was highly likely that the capacitor had been seldom switched into circuit to keep its dielectric "formed"..

The coil bypass capacitor (A6A7C1) is on module A6A7 which is located at the right hand side of the rf/mixer area accessed below the bottom cover. The module is easily removed by undoing two countersunk screws and unplugging a PCB header connector and a ribbon cable.  On checking C1 looked in excellent physical condition but on checking with an ESR meter was open circuit. The pictures below show the original capacitor and its replacement along with the board layout.

20200926 203911

20200926 205203

A6A7 pcb

 

 

The location of the A6A7 board is here.

20200926 203936

After replacing A6A7C1 the spectral plot looked like this with the spurs largely below the PN sidebands of the Herley PLL.

repaired line spurs at 12575GHz

A close-up of the offending capacitor. It is in excellent physical shape but not there!  Now perhaps I can remove that yellow  "Out of band responses OOS (Out of Spec)" sticker on the front panel.

20200928 114936

GM8BJF 28/09/2020

img026 1img027 1img028 1img029 1img030 1img031 1img032 1img033 1img034 1img035 1img036 1img037 1img038 1img039 1img040 1img040 1img041 1img042 1img043 1IMG 0001scanIMG 0001IMG 0003IMG 0005IMG 0006IMG 0007IMG 0008IMG 0010IMG 0011IMG 0012IMG 0013IMG 0014IMG 0015IMG 0017IMG 0018IMG 0019IMG 0020IMG 0021IMG 0023IMG 0024IMG 0025IMG 0026IMG 0027

 

 

 

 

 


 

I have owned a FT736R for over 25 years and have been very happy with it. It has given me good service. It is fitted with the 6m and 23 cm modules. A widely recognised weakness with the radio is its switched mode power supply. It is an elderly design that is quite inefficient by present day standards and runs hot on account of four high wattage resistors which dissipate energy. One consequence of this is that several of the electrolytic capacitors “dry out” over time and lose their capacitance to a point that the power supply will not start up when switched on. Most other owners I know have re-capped their power supplies and I was no exception. Additionally, about a year ago my unit developed another fault whereby by the supply would make a ticking noise on transmission speech peaks. This was eventually traced down to a 10 mH inductor in the feedback circuitry which had gone open circuit. A replacement was sourced which seemed to cure the fault but I noticed that on transient loads and speech peaks the power supply kept tripping. I assume that it was probably due to the replacement inductor not having exactly the same characteristics as the original Yaesu part which looked to be a “special”. I decided it was time to think of replacing the whole PSU.

I remembered that I had seen an article some years ago about fitting a replacement by Harvey Laidman, W8DX in CQ Magazine but could no longer find a copy. (I subsequently found a copy).  It was in the June 2006 issue on page 28. I found mention of the article on the FT736R IO Group and found that this was the part recommended . It seemed to be available in the US but was on a long lead time so I started to search about for a more available part. For EMI and safety reasons I wanted to use the original perforated steel enclosure to encase/screen the replacement PSU it had to fit in a footprint of 80 by 140 mm and be less than 40 mm high. I also wanted the replacement to have a good EMI specification. Most of the standard modular supplies have a width of 90 mm and would not fit the available space. I looked at the catalogues of the usual distributors and settled on a unit manufactured by XP Power, the ECM100US12. This has a footprint of 64 by 115 mm and height of 31 mm so would fit comfortably inside the original casing. It also runs from either 110 V or 230 V line voltage without switching. It is available for around £70 - £80 from a number of different UK/European component suppliers such as RS, Farnell and CPC and sometimes appears on EBay. The data sheet states it has industrial and medical approvals and it uses a two section mains input filter with two cascaded bifilar (common mode) chokes. It can supply up to 8 amps which is more that the radio requires on FM modes. Although it is nominally a 12v supply it can be adjusted up to 13.4 V which runs the radio quite adequately.

 psu

To  fit a replacement first remove the original PSU from the radio.  To do rhis remove the three screws on the heatsink back plate, and the two self tappers securing the PSU base at the front of the unit.  Slide the PSU forward and raise it from the main chassis.  The cable clamp/bush in the back panel can now be released to take the PSU free of the chassis complete with the 6-way JST connector. The PSU can now be dismantled by removing the screws securing the perforated cover and the PCB to the base plate. Recover all of the metalwork including the heavy gauge aluminium end plate. Carefully store all the screws as these will be required for re-assembly. Cut a piece of 18 SWG sheet aluminium to the same size as the original Yaesu PCB to act as a mounting plate for the new PSU and drill six  fixing holes using the PCB as a template. (Two are for the aluminium end plate.) Also drill fixing holes for the new PSU and mount it on short pillars. Short leads with line input (3-way Molex) and DC output (12-way Molex) connectors were prepared. (Connector details are below).

20200525 151252

The power supply with connectors and short tails. The Molex connector shells and pins were as follows from RS. Links are to the RS catalogue. (A slight cock-up was made in ordering to 12-way output connector but a 6-way one sufficed!). Pins 9 -12 are unused.

 

SPOX 3.96mm female crimp housing, 3 way

Molex 43061-0003

896-7380

SPOX 3.96mm female crimp housing, 12 way

Molex 09-50-1121

670-5206

Crimp Term,18-24AWG,phosphor Bronze,reel

 Molex 39-00-0021

670-6303

 

 

The short tails were connected to the original Yaesu cabling with crimp connectors.

 20200525 164709

 20200525 164727

The PSU assembled into its original casing and back in the radio. The aluminium baseplate is visible where the PCB with the input and output connections and the voltage selector were.

20200527 150513

 

While I was working on the radio and had it apart I decided to rewire the internal mains switch to improve safety. The Power switch on the front panel it  is a 2-pole mains voltage rated part but one side is used to switch the 13.8 V DC line while the other side is used as the 230V mains switch. The problem is that the switch is in the neutral side of the power feed to the PSU. This is very bad practice as the live is permanently connected to the PSU. The Y-Capacitors in the PSU are subject to full mains voltage as long as the radio is connected to the mains. NOT GOOD!!! I decided that as I had only ever run it off a DC supply to test it I could dispense with the low voltage switching and use the double pole switch to properly isolate the radio from the mains when switched off.

I was able to do it by running new correctly colour coded line and neutral wires to the front panel switch and I was able to utilise the existing Yaesu (white) wires that were already in the harness as the returns to the fuse and backplate. I also had to rearrange the fuse wiring. I now have it arranged so the L and N come off the pcb on the back of the IEC socket directly to the switch. The fuse has been rewired so that it is in the L return from the switch before going to the 6-pin JST connector on the back. The L and N pins on the JST socket have been swapped as I was trying to preserve the original wiring as much as possible, but I see no harm in that. I have cut a piece of clear plastic so that it clips over the back of the mains socket to prevent the ingress of passing fingers as had happened! I also enclosed the front panel mains switch in heat shrink.

Overall the radio runs much cooler with its new more efficient power supply and will hopefully continue to give good service for some years.

31 May 2020.

 

Addendum

Notes on Mutek replacement front-end kits

Recently the has been some discussion on the kits that were available from Chris Bartram's, G(W)4DGU's, company Mutek back in the mid/late 1990s, These were to improve the sensitivity and the strong signal handling performance of these radios. There were two versions of the kit. The original kits consisted of four PCBs.

1. A 2m replacement front-end

2. A 70cm replacement front-end

3 A mixer and downconverter for the 70cm board (47.43 Mhz IF)

4  A board with a 13.690 MHz crystal "roofing" filter to improve adjacent channel rejection

Mutek kit

A somewhat messy picture of a four board kit on Ebay!

A set of fitting instructions for the four board kit is here.

 

Later production kits consisted of the two front-end PCBs and a combined single IF board with the roofing filter. This combined  board was housed in its own screened housing and fitted above the central 2m main unit PCB in the upper deck of the radio. I believe these kits were fairly rare and only appeared towards the end in the late 1990s. The fitting instructions that I have seen on various websites are for the earlier four board kits and I have never seen the instructions for the later kit.  I only became aware of the later three board kit when a friend of mine, Briain Wilson GM8PKL mentioned he had a Mutek kit that he had bought way back when they were current and had never fitted it. As he had got interested in VHF/UHF operation again he decided he would fit it and I was curious to see what it consisted of as I had never seen one in the flesh. At first I thought that a board had been lost but then on examination of the PCB the penny dropped that there had been a re-design of the kit. Presumably this was to reduce costs and possibly improve shielding of the circuitry as the original boards were out in the "open", unshielded and quite large for what they contain. They had also gone over largely to surface mount construction. Briain kindly took photos of the board and these are below, and also OCRed a copy of the somewhat poor photocopied manual. This is largely as per the original but has some clearer instructions for setting up the ALC. The original had some circuit diagrams as well but the block diagram below tells the story more clearly.

Mutek block diagram

 

 This is a block diagram of the replacement board(s) traced out from examination of the PCB.

IMG 0712

 

 The replacement IF PCB with the screening enclosure.

IMG 0709

 IMG 0710

 

Close-ups of the IF PCB.

 

 

Front-end improvement

I do not have pictures of the two RF front-end PCBs but these are simply to replace the existing front-end circuitry with a low-noise pre-amp and a mechanical relay. The original Yaesu boards used PIN diode switching which was somewhat lossy, degrading the noise performance. In my 736 I replaced the PIN switching with an Omron G6Z-1F 9V SMD RF relay on 2m and 70 cm. This improved the sensitivity, NF and power output. The relay PCB I used is below. This NF on 70 cm went from around 10dB down to around 2dB. The original NEC GaAs FET is quite a respectable device.

relay pcbrelay pcb 007

 Ant relay PCB

relay pcb layout

relay pinout

Below is a (rather poor) schematic showing the circuitry for the relay replacement on one of the VHF/UHF front-end PCBs The 144 and 432 MHz front-ends use the same PCB layout.

432 front end mods

 

I should add that my 736R still does not have a Mutek upgrade. Back when I bought mine the kits were around £600-700  and having spent £950 on a  second hand radio with 50 MHz and 23 cm I considered it expensive.  I have considered trying to "homebrew" one. The components are all readily available apart from a suitable 13.69 MHz crystal roofing filter. Anyone know of a good source of filters ?? Eliminating the PIN switching improved the sensitivity on both the VHF and UHF sections. The 23cms optional module uses relay switching so that was left alone. I have not invesitgated the 6 m front-end. Locally the noise level is so high it probably wont make a difference.

 Brian GM8BJF

5 Jan 2021.

 

I have always been spurred on in my radio activities by the desire to get onto higher and higher frequencies so when I saw that kits for puting together a 122 GHz transceiver were becoming available I could not resist getting one. I ordered the 122 GHz populated PCB designed by Andrew Anderson VK3CV / WQ1S and produced by Tim Tuck VK2XAX. It duly arrived about six weeks ago. I decided to make the antennas myself as I have a small lathe and enjoy a bit of machining now and again.  I produced two antennas as per the original drawings supplied as part of the project. The board is pictured below.

PCB

I boxed up the board and mounted my circular horn on the PCB and arranged for the horn to protrude through the box.as shown below.

IMG 20200721 WA0000

Locally to me Peter bates GM4BYF had also bought a set of bits to build up a system so there was the possibility of having some QSOs! Earlier this week I visited Peter and we had a preliminary test acoss his garden and then went out to do a test over a longer path. Fortunately my home town of Edinburgh is endowed with plenty of hills. Peter went onto the Blackford hill and I set up shop on the Braid hill and we had a QSO over a path of 0.68 km.

1stGM GM122GHz

 

We believe this is the first GM to GM contact on the 122 GHz band. Once we got the horns lined up the signals were S5 to S7 and good copy on FM. There was some QSB on the signals which we attributed to a tree which was close into the path and it was quite windy. This test showed up a few deficiencies in my setup.and a couple of days later I improved the mounting and antenna pointing arrangement for my transceiver. It also was apparent that I was overdeviating and I corrected that.

Flushed with success from our first QSO we decided that we would try to have a cross border QSO GM to G!  We considered a number of locations around Coldstream on the A697 and Carter Bar on the A68. Finally Peter spotted a disused rail way line that crossed the border just east of the town of Kelso. As the Carter Bar site was likely to be busy with people visiting the border viewpoint the disused railway sounded better proposition! 

 

england Scotland border

Introduction

20200503 142102

Recently a small vector network analyser (VNA) has become available on the usual auction sites from China. It started life as a kit aimed at hobbyists in China but has evolved into a product that has been widely “cloned” and made available for sale. There are a large number of sellers and apparently several different manufacturers. However at a cost in the £30 to £40 they seemed too good a bargain not to get one! Some of the clones are better than others and it is hard to judge which one you will get when you order one. I bought one from a Chinese seller on Ebay. It was version 3.1 of the PCB and broadly conformed to the boards designed by Hugen79.

The design makes ingenious use of low cost consumer ICs to produce a VNA that is capable of giving useful results up to 900MHz. The block diagram is below:

nanovna blockdiagram

The heart of the unit is an audio codec chip that takes as its inputs the outputs from three SA612 mixers. The mixers are fed with a common local oscillator signal and their outputs are the reflected signal from the DUT, a reference signal and the transmitted signal through the DUT. The mixer inputs are the common local oscillator and:

  1. The reflected signal sampled by the bridge
  2. The reference signal
  3. The transmitted signal from the DUT output

The test signal and the local oscillator are generated by a Si5351A clock generator chip and are spaced by a constant frequency difference in the audio range of 6kHz (or 10kHz with some firmware versions). The three mixer outputs are I and Q signals at a 6kHz (or 10kHz) intermediate frequency, which are digitised in the audio codec chip to provide three data streams to the STM32F Microcontroller which computes the display data to present on the touch screen.

pcb
 

The Si5351A data sheet specifies its maximum output frequency as 200MHz but the designers have found that in practice most of the chips are able to operate to 300 MHz. As the outputs from the clock generator chip are square waves they are rich in odd harmonics and by generating LO and test signals whose third harmonics are spaced by 6 kHz it is possible to extend the frequency range to 900 MHz.

Extending the Frequency Range

Over the past year newer versions of the firmware have appeared on Github and the Nanovna IO group that extend the frequency  range of the unit up to 1500 MHz by using the fifth harmonic of the Si5351A output. I was curious to see how well it would work and so I re-flashed the firmware in my unit with the DFU files uploaded to Github by Dislord, [1].

There are extensive guides to flashing firmware to STM32F microcontrollers on the internet and Youtube so I will not cover it here.

On calibrating my unit set to a span of 100 MHz and a centre frequency of 1300 MHz the results were not encouraging.  A 50 ohm load on a Smith chart plot looked a bit like a sizable hedgehog jumping around on the middle of the chart which was not good.

Looking at the schematic revealed that there is a switch mode charge control IC (IP5303) to control the charging of the single cell LiPo battery and also provide the 5 Volt supply when running off the battery. This was a prime suspect as a source of noise. A clue was that the noise was worse when running on battery.  Looking at the signal at the CH0 port with a spectrum analyser revealed that it was noisy. Taking the back off the unit and probing with a ‘scope showed that there was noise on the 3V3, VDD rail powering the microcontroller and the Si5153A. Adding a 33uF 16V tantalum SMD capacitor to the 3V3 rail, across C4 on the schematic(link) Vdd rail reduced the noise on the RF signal at the test ports suggesting that the source was switching noise from the IP5303 modulating the Si5153A output. This did not have an immediate effect on the noise on the trace and attention turned to the 5 V rail powering the three SA602/612 mixers. This showed several 10s of mV of noise and two further capacitors were added across this rail. Firstly a 33 uF SMD tantalum type was placed on the track leading to the 3V3 regulator and a 220 uF 10 V electrolytic type was placed across C49 on the 5 V output of the IP5303. The positions of the three capacitors are as shown in the two pictures below:

 

V3.1 PCBmods

mods2

This was found to reduce the noise on the traces considerably suggesting that it was getting in via the power rails to the mixers. I also added tinplate screens over each of the three mixer channels to improve isolation but overall I did not find this made a significant difference.  Overall after the addition of the screens and the extra decoupling the unit I have gave the following performance:

 

50-150 MHz

400-500MHz

800-900MHz

1.2-1.3GHz

S21 Dynamic Range

   70dB

  55-60dB

  50dB

   35dB

S11 Dynamic Range

   70dB

    40dB

   30dB

    25dB

 

Some Results of Calibration at 1300 MHz

open

Open

short

Short

50 ohm load

50 Ohm load

S21 thru

S21 through

S21 open

S21 open

 

A Simple Reflection Measurement of a 23cm Antenna

The following two pictures show an S11 measurment of a WA5VJB 1290 MHz "Big Wheel" PCB antenna over a frequency range of 1250 MHz to 1350 MHz.

big wheel

The "Big Wheel"

Measurement setup

 

Conclusion

Overall the unit has become useful as an indicator of performance at frequencies in the 23cm amateur band and the additional capacitors have reduced the noise on the traces at frequencies in its original operating range improving the dynamic range there as well. There may still be some scope for further noise reduction.

Brian Flynn GM8BJF 

04 May 2020

References

  1. https://groups.io/g/nanovna-users/files/Dislord%27s%20Nanovna%20-H%20Firmware
  2. Hugen79 schematic