G4APV's  EB104 Construction Page

EB104
This web page contains a whole load of stuff 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.

If you want the ShackCam then click here.

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 n 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.


This is still work in progress as I continue to build the amplifier.

Bob Harris, G4APV, 31st August 2009