1.5" pipe where the gas is injected and a 2" coupler for the blower end and the furnace end. The pipe union houses the new jet assembly. The 1/2" nipple has a brass insert that couples 1/4" copper soldered in place.
 
 

Here you can see me removing the jet assembly. It's a long tube plugged up at the and and fastened in place with a compression fitting.
 
 

This gives a sense of scale to the jet relative to the 1/2" pipe and union.

Now this photo was the nozzle part of the assembly and I found that the flame would sit where the spark plug was due to both turbulence and the slowing down of the airflow as the opening increased in size.  Increasing airflow just blew the flame off the end.  So I tried smoothing it out by filling the inside with castable refractory.  I turned a piece of cedar on the wood lathe and used that as a form that I later burned out LWP (Lost Wood Process).  This didn't work any better.  Then Jens had the bright idea of make a square shoulder to create an abrupt change in size.  Next picture shows that.

You can just see the end of the spark plug.  This one worked really well but the flame,  which sat in the end of the nozzle was still actually outside the furnace boddy so I thought I'd try and do the same thing in steel and move the flame into the furnace.

Not much different than the flared fitting except that the spark plug has been moved back a bit but it sticks into the airflow a bit deeper.  However, I find a low velocities it doesn't light as easily.  The next photo shows how the transition happens.

From the edge of the 1.5" pipe the 2" flares out and then abruptly has an edge right at the opening.  The idea was to create a flame holder.  Now this worked but still a bit touchy outside the furnace body.  I also found,  confirmed by discolouration that the flame tended to sit at one side of the tube.  I did some bending of the jet assembly and suddenly the flame became quieter and more to the center.  Now it was also more reliable.  The next step was to test out the computer controlled airspeed and ignition circuit.

In this photo you can see the Gas Solenoid,  the flame detector tube and the high voltage driver for the tube.  The uComputer on the left has a small 28pin DIP circuit called a  CANDIP .  The carrier board has a K type TC input.  Driver outputs for the spark plug coil, the solenoid and the PWM controlled fan.  The two DB connectors on the let side are for RS232 and CAN bus.  Eventually I'll have the CAN bus connected to the PC and a graphical display of furnace temperature, motor speed etc. along with melt statistics.  The connector at the back is to be used for air flow sensing using the hot wire anemometer technique.