Re:SOLID STATE AMP(G4APV)

[HL2UOK]

G4APV's  EB104 Construction Page

EB104
This web page contains a sort running log about the construction of a MRF150 based 600W linear amp.  The basis of the amp is an old but reliable design by Motorola that came out as Engineering Bulletin 104, hence EB104.

Case
The case came from Maplin.



Parts
So what goes into the case?  Here is the kit of EB104 parts that I got from Communication Concepts for the amplifier itself:


Attenuator
As well as that I needed a few other parts, these are for the attenuator to drop 20W down to 6W.  This 6dB odd loss ensures that most HF transceivers with a typical low power output of about 10W can be used to drive the amplifier with a bit of a margin of safety. It also ensures the input impedance to the amplifier is more closely controlled.  The parts came from mainly from JAB components.


Hardware
I quickly came to the conclusion I would need quite a few nuts, bolts and other hardware.  I used M3 and M4 nuts and bolts as they fitted the devices..  This lot came from Farnell:


Thermal Issues
WIth 600W out and about 60% efficiency then there is as much as 400W to get rid of.  This means getting the heat away from the devices with a copper spreader and then getting rid of it with a heatsink.  Both the spreader and the heatsink also came from Communication Concepts:


I was not confident about pure convection and so went for forced air cooling using mains fans, again from Ebay.  

Power Supply
The amplifier runs of 50V and with 600W out will need about 1000W in, ie 50V at 20A.  To power this an ELTEK brand new power supply off Ebay but sold by Anchor Surplus in Nottingham is ideal.  It is nicely protected and can be started up by a 5V control line.  This allows it to be powered down on receive to avoid nasty noises from the switching, although in practice it seems to reasonably quiet anyway:


Auxiliary Power
I decided to use 12V for control systems to provide compatibility with other systems I had.  I got a 12V PSU, again off Ebay, shown here with the cooling fans.:

If you are really interested (or is it sad?) then here is the spreadsheet with all parts as ordered (except Ebay).

Main Board Assembly
Before the board can be assembled, there is a great deal of mechanical work to be done to mount the heat spreader onto the heatsink and then mount the unpopulated PCB onto the spreader. Only when all of this is done is it appropriate to put the components onto the amplifier board.  

After much planning, drilling and tapping here is the PCB on the spreader, with the spreader, on the heatsink.  I found it critical to have a pillar drill to do this.  I went and bought a £55 one, new, from Machine Mart.  Here's the trial of the board on the spreader and heatsink:



Population of Board
This a fairly long winded process with components on the top and the bottom of the board.  On the bottom of the board are some chip caps.  The board needs to be just clear of the spreader to ensure these don't short:


I next fitted the solder pins, and in retrospect the two big resistors that are part of the feedback loop (below) should have been on pins, it would make life much easier later on:



Putting the rest of components on gives:
 

 

Power Devices
After various trial fittings, marking out, drilling and tapping of the spreader and finally checking for shorts, the board was fitted to the heat spreader and the devices soldered in:
   
Notice the 1W resistors have got a bit bent to allow access to the device screws.  Next time they will go on solder posts.

Initial Testing and Bias Setting
At this point I decided to be brave and put the supply on and, hopefully, set the bias currents.  I bodged up a 3A fuse (just because I had one) and monitored both the voltage and the current.  I am fortunate to have a 0 - 50V, 40A variable PSU which makes life easier.  Here is the bodge up of the first test:


It proved to be pretty easy to set the bias current to 1A total, ie 250mA per device.  As I was taking it up from the initial supply voltage of 40V to 50V as a check there was a bang, a blue flash and a blown fuse - not good.  It turned out to be a small offcut strand of wire across one of the devices.  There is a lesson here about cleaning the board before testing!

Gain Test
The next check was to see if it amplified.  I used a signal generator which has a maximum output of 19dBm (ie about 100mW) at 14MHz to see what would happen.   Thsi produced about 5 or 6W into a 50 Ohm dummy load which seemed ok to me.   The next check was to use an FT817 (out of picture) to provide some RF as this can have it's output reduced down to about 100mW.

     

Attenuator Board
Once I had established the amplifier board itself seemed to be working, I designed the attenuator.  The actual design of the component values was done using WinAtt from GM4PMK on G3SEK's webpages.   I used TinyCAD for the schematic capture, FreePCB for routing the board and ViewMate to print to pcb out.  

 

I got the board made at work using a prototyping system that uses a routing technique to remove copper from those areas where it is not wanted:
 
       
Another idea I investigated but did not really got to work properly is the "iron on" technique.   The finished, routed boards come out ok, but are obviously not plated, but that I can live with:
   

Here it is populated, fitted but not yet wired in:


I then bodged on an output transmit/receive relay and the 12V power supply to give a complete working system:



Output Switching and SWR Board
To protect the devices it is clearly necessary to shutdown the amplifiier in the event of high SWR, for example a flashover in an ATU.  Also a relay for transmit/receive swithching ant the output is needed.  I then designed a board to do this, it removes the 50V bias supply if it detects excessive SWR at the output:.  The PCB layout starts as a schematic capture and is then imported into EasyPCB to give "ratlines":
           

The final result looks like:


Mounting in the Case
Much of the time was spent puzzling out how to mount all the bits in the 19" rack mount case.  Here is the stage with the amplifier, input and output boards and power supplies mounted.  The cooling fan, output filters and front and rear panels have not yet been dealt with.


Here are  the major components installed and wired up.  The amplifier is now usable but is minus the output filters.  Some low level instability in the amplifier disappeared with the input and output SO239 and fitted and the power supplies properly wired in
   


At this stage the lack of  filters means the output was not too clean, this was on 80m, some harmonics are only some 25dB down on the fundamental:


The other lesson is not to push things to hard.  The bypass caps on the drains supply don't like too much RF current.  While trying to see how much power I could get out of it I discovered they go bang and burst into flames.  Here is the aftermath of this happening, the remaining leads in the plated through board were very difficult to remove.
 

It seems as if too many harmonics are the problem which leads nicely into the design of the output filters.

Filters
It seems to be fairly well established that a 5th order Chebyshev design is appropriate for removal of harmonics from power amplifiers.  The design of the filters was based on the ARRL Standard Value Capacitor (SVC) tables.   Once the values had been calculated they were modelled using ELSIE.  Using information I already had (not sure where it came from) it became clear that a T-130-2 ferrite would handle the power and provide the necessary inductance.  An example of the design process is given in this document.  

  

I put together a prototype of the 3.5MHz filter as well as simulating it:  I then used a Panasonic VP-8191A signal generator plus a HP8590A spectrum analyser set to peak hold to find the filter's response:  



I decided to put 3 filters on a PCB and then use 2 boards to cover all bands.  This was in an attempt to keep the track lengths acceptable.  It seemed a bad idea to have lots of RF going down long tracks.  The higher the frequency, the shorter the track ought to be.  I used an earth plane approach in attempt to cope with the potentially high RF currents.  The unpopulated boards look like:
 

And once populated and the topband filter tested:
 

The second board was built and tested.  I then stacked them using some M3 studding and wired it all up:
 
These photos also show the tray made out of bend and painted steel that carries the boards and acts as a duct for the cooling air.  The left hand side where the space is is where the control board will eventually go.

Using the amplifier on 160m and 80m at about 350 - 400W  it developed a fault where it would become erratic on transmit.  The power output would disappear and then reappear.  This did not appear to be related to band, drive level, output power level etc.  At this stage I bodged on some meters to watch the supply voltage and current (see below).  Of course it did not do it again although I did get it to smoke while giving it a long wwwaaaahhhhh.  I suspected that the PSU was going into some sort of shutdown, but not due to over current as the drive level did not affect the problem.  I suspected RF getting into the control circuitry of the PSU.  
 
Monitoring the voltage and current soon showed the problem was not the PSU.  Putting an old SWR bridge onto the input showed a very poor match and that the drive from the IC735 was disappearing when the problem occurred.  My conclusion is that rig was shutting down as it got warm due to a poor mismatch (or is that a double negative?).  The solution is to run the IC735 via the 6dB input attenuator so that it always sees a good match.  The problem has not happened again since keeping the attenuator in all the time.

However the filters do seem to be working, the second harmonic is now some 45dB down on the fundamental on 80m.  Compare this to the earlier plot of no filter which had the fundamental only 25dB down.  The filter should, at about 7.5MHz provide about 20dB of attenuation so, surprisingly, it seems to do wha the theory says it should!  The third harmonic now seems to have gone - good.

Front Panel
At this point I started to look at the front panel design and the associated control circuits to select the filters, attenuator etc.  I decided to go for "push to select" approach rather than rotary controls.  Rotary controls and toggle switches could have been connected directly to the relays etc but but aesthetically this would have look amateur and I want to have a reasonably professional end result.  This will imply some form of logic to interface the buttons to the relays etc.  

 The front panel is made from mainly off the shelf parts from Farnell  but the meters are from a rally.
 
The front panel layout was done using free software from Front Panel Express.  I used this as it is intended for front panel work and so the dimensions and positions can be easily controlled.  I am fortunate to have an office that is directly opposite the Graphic Design student's printing workshop, so I went and talked nicely to the technician and got the screen printing done......  The printing is water based so I lacquered the blank panel to protect it.


Mains Board
I wanted a hard on/off with complete isolation of the live from the psu when the amp was off.  I do not like the psu to but in some sort of shut down mode.  It is too easy to end up with the supply on when you don't want it.  The on/off buttons feed a  pcb with a mains relay that latches via the front panel buttons:



I've put this onto a pcb for neatness:

 
 

Here it is in place on the base plate next the 50V and 12V PSUs.



Control Logic
I thought I had a few options for the control logic each of which had pros and cons:

• hard logic - significant design effort, hard to modify later but potentially simple and no EMC problems
• microprocessor - significant hardware and software design effort but flexible and I have programmed in assembler before.  Potential EMC problems from the clock
• Raspberry Pi - this was a late consideration.  This met all my connectivity, processing power and  cost requirements BUT is hard to mount (no mounting holes), lots of cabling in strange positions to cater for and, most importantly, it would need programming in a high level language, probably C.  I have no competence in any high level language and accessing I/O ports is more complex - this was the killer for me.

 I went for option 2 and chose a PIC as all the tools for generating the code and loading the code onto the PIC is cheap and well documented.  After a lot of thinking I went for a two chip solution.  The idea was to separate the control of the purely digital interface of the front panel to the relays from the analogue  functions (over temperature mainly)  and the comms function.  I intend to include a CI-V interface to interface to my Icom radios so that the amplifier is automatically on the correct band.  It might add Yaesu CAT as well if I can be bothered later on.  For the digital interface I chose a 16F57 as this had the necessary number of digital I/O lines.  To date I have been happy with this choice, with one exception.  I thought it could use ICSP (In Circuit Serial Programming), in fact it must use this method.  This had the slightly annoying  effect of needing an extra socket attached to my programmer.  

I designed a board using Tinycad and FreePCB as before.  FreePCB has some problems in terms of producing the Excellon drilling file.  It seems to generate non compliant code.  I checked the code using gerbv and hacked it (it's a text file) by hand until I cleared all the errors.


This is a double side board, but the top side only has a few links on it.  It was produced on the LPKF system as per the other boards:
 
Having populated it and used it I would, in retrospect have done it differently.  I would have use SIL resistors to improve the packing density, put ICSP on the board and put the digital controller and analogue/comms PICs on the same board.  The only problem is the sheer number of bus lines takes up a great deal of real estate and I am loathe to use tracks that are too narrow as this is, effectively, a prototype so hacks may be necessary - much easier on big tracks.

Here is board in the amplifier fully wired up:
 

In terms of writing the code for the PIC I used MPLAB IDE v8.86 available from Microchip as a download.  In terms of how program, the manual for the 16F57 is good, but you need more than this unless you are really experienced.   I used some ideas from here and most usefully from Gooligum Electronics who provide very good tutorials which should get most people going.  To load the code onto the PIC I used a Quasar Electronics 3128 PIC programmer but with a Vero board extra to allow me to use ICSP on the chip.

Version one of the software is now working, this provides control of the band, attenuator and  amplifier on button.  The code is not sophisticated, at this stage I just wanted to get it working.   This document details the how the PIC interfaces to the board.  The PIC is now programmed such that the board works correctly as the front panel interface for manual control.  It also now has the code to allow the second controller board to override the manual settings when it is automatic mode where the band is set from the CI-V input from an Icom radio.

Communications Boards
This is the second processor board and it provides control of the cooling fan based on the temperature and also will ensure the amplifier automatically on the correct band.  It will do this for Icom radios initially.  This is because I understand the Icom CI-V bus, have several Icom radios and finally I can't be arsed to do it for Kenwood or Yaesu at this stage.  The slow bit was writing the code for it.  I am conscious that this will be time critical code so I will need to be a bit more careful to avoid timing issues when programming the PIC.  The PIC I chose is the 16F627A as it provides the analogue comparators needed for the temperature control as well as having the UART needed for the CI-V interface.

Here's the design:
     

Currently I have got the 16F627A configured to use a comparator to turn the fan on and off based on the temperature as measured by the thermistor.  This sounds simple but the different internal architecture and the sheer number of configuration options meant it took me a day to get this code working correctly.  The nice thing is that now that this code is done I have a completely working amplifier, albeit in manual mode only.

The auto band switching code is now written and working!  It proved to be easier than I expected the only the thing that caused any problems was failing to set the global interrupt enable - that tends to stuff things.  If you really are interested then here is the working code for control board #1 and control board #2. If you want to use my code that is fine but I would appreciate acknowledgment of this if you in any way publish your code.  To write the code a good understanding of Icom's CI-V code is needed, I used a very old paper document that describes it for the IC735 but you can find suitable information from Icom or with a users perspective here.

Completed Hardware
 

.
Above are pictures of the current hardware which is more less final - at least until I change it.

Software Updates
21 April 2013

I have modified the software to ensure if band is changed automatically then the amplifier is turned off.  Otherwise it would be too easy to change band on the radio and go to transmit with the amplifier in but without matching the aerial - not good.

The software will now work with any Icom radio not just the IC735.  The IC735 was the first radio to implement the CI-V and used a 4 byte data string for the frequency rather than 5 used on later models.  The software can now cope with either, I've tested with an IC706 and it works fine - only after 4 hours of  messing about just to discover I had to enable the Transceive function on the IC706 for it to produce data,  oh well......

28 April 2013

I recently bought a Kenwood TS480 which I am really enjoying playing with, particularly the remote control abilities.  Obviously I want it to able to control the amplifier so I have been looking at the Kenwood Network System (KNS).  The key document has been this one.  So far I have gleaned a few things:

• it has a separate txd and rxd line (unlike CI-V which has a single line)
• standard RS232C voltage levels are used (+15V and -15V)
  
• it seems to only be suitable for one radio controlled by one PC (unlike CI-V which is a CSMA-CD system)
• to get the rig to output it's parameters as they are changed on the front panel (as opposed to interrogating it for its status) it has to have the Auto Information set to on.  This is done by sending it the command AI2; .  Unfortunately this is cleared back to off when the rig is powered off.

I sent it a few commands by hand to check I understood this.  To do this I used the terminal program Tera Term.  This is very useful for checking things out.  I have reached a few conclusions based on the above:

• the amplifier cannot just passively monitor the data coming out from the radio as there won't be any unless you send AI2; to the radio
• when the amp is turned on it will need to send AI2; as part of its start up sequence
• to have PC control of the radio via HRD and the amp connected, then the amp will have to take over the rxd and txd lines to the radio, send AI2; to the rig, and then connect the lines back to the pc and HRD.  This will probably mean looping the radio to the amp and then the amp to the pc
• to ensure when the amp is off that HRD will still work then I may have to use relays to switch txd and rxd
• the current comms board will need to be modified to bring out rxd and txd separately, currently they tied together to drive the CI-V interface - I should have thought about keeping them separate earlier.....................!

My current thinking is to have an extra small board on the back of the radio which feeds two 9 way D types, one from PC, one from radio.  The board will then have some relays for txd and rxd switching and some sort of control line to switch it for Icom, Kenwood, or even Yaesu (if I can be arsed to do it).  The only thing that really worries me is having to drill the chassis as I will have to do it with all the electronics in the case.  I'll see if I can borrow a 9 way D type hole punch from work.

Switching Noise.
The switched mode 50V psu was causing some switching noise on some bands, particularly 15m.  However it does have a control input to turn it on, although checking showed the manual to be wrong on the pin number; it is 12b not 12d.   I am using the input to power it down on receive, and then power up on transmit.  The only slight snag is that it takes a short time for the 50V to come up once the control line enables it.  I could measure how long it takes but I can't be bothered as there is nothing I can do about it.

Bugs fixed:

• Output power varies band to band - this turned out to be the bias switching transistor not going on fully.  Reducing the base resistor fixed it.
• RF getting into the processor making it switch bands etc when on transmit - not good!.  The copper earth plane on top of the board (clearly visible above RF board in the pic above) was not earthed.  Obvious really, but it took me a while to find......

Bug to still fix - RF output varies band to band.
I've yet to figure this one out.  I suspect the filters are maybe presenting a poor match to the RF board.  I've taken some Smith charts to see if there is a clue there.  Here's some results:

Top band in-band and at second harmonic which looks ok:
 

18MHz in-band and at second harmonic which does not look ok:
 

Trouble is, they both seem to work all right so this is still work in progress..........

These plots were taken using a RigExpert AA30


This is overtly an antenna analyser but is much more powerful than that as it can determine Z, R and X, including the sign of X.  Most analysers don't seem able to do the sign of  X.  The user and software manual tell you more if you are interested.

SWR protection Board
The original SWR protection board just did not work :-(  I don't seem able to measure the SWR effectively so this board needed a redesign.  Marc van Stralen, DK4DDS has sent me some helpful information by Bill Leonard, N0CU.  He also gave me a link to his All in one SDR TRX.  This is a great project and includes a high power SWR bridge design which I have partly replicated.

I have taken the SWR bridge parts and combined them with buffering and latching circuits to provide protection.  Here is Marc's bridge design:


My implementation of this looks like:


The circuit to go with this contains buffers, a latch, a piezo buffer to warn of excessive SWR plus some 50V bits to remove the bias.  If it works as I hope (!) then putting the amp back to receive should reset the latch.  I haven't yet built it so we will see.......

 
I have designed the PCB for this, the main problem is that the bridge parts are rather big and so has to sit above the omp amp etc pcb. The PCB is:

 

The populated board before fitting into the amp:
    

This is now fitted into the amp and works as intended.  The only thing I had to do was to reduce R3 from 10k to 1k as the voltage at pin 3 of U3 was too small.  I'll take a photo of it in place when I next take the lid off.

Next thing to do is to redesign the radio interface to cope with Kenwood as well as Icom rigs.

Bob Harris, G4APV, 17 January 2014

[나무와 샘]

오호~~ 멋지네요 ^^