I was surprised at the size of the capacitors
  so that added a bit of square inches to the
  PC board. Back in the old days (60's) we used
  large metal can capacitors that were oil filled
  for our CD systems and vac tubes of course.

  Those caps store the spark energy instead of
  being stored in the coils. CD coils
  run cool as they are simply used as pulse
  transformers.

  Boards cost about 65 cents per square inch
  Plus a fixed cost of $55 for tooling.

  The caps are the 4 big boxes in the center of the PC
  Board. I'll use my favorite heat sink and
  TV connector for the output to the coil.
  This is for one spark plug on one rotor.
  The trailing plug does not need it and
  and any back up ignition can be used on the
  trailing rotor.

  This requires 4 Hall Effect devices
  per plug to trigger every 10 degrees for
  40 degrees after 22 degrees BTDC.
  In other words 20 BTDC, 10 BTDC, TDC and
  10 degrees after TDC with no RPM limit
  such as MSD's have problems with.
  One cap for each spark.

  The whole idea here is to see if it will
  cut the fuel burn.

  I'll see if I can get Mazdatrix to test it on
  the dyno. I'll loan it to them for use on
  what ever engine they are fooling with at
  the moment.

  It uses one 400 watt 220 Volt AC off the shelf 12 V power
  supply for up to 6 PC boards if you want to implement
  that many on a 3 rotor.

  Paul Lamar

  It would appear that isolation devices are needed between the four
  2mfd caps
  and the coil.  When trigger is low, there is a low impedance to ground
  thru
  the lower FET of the half-bridge, essentially putting the caps of the
  untriggered drivers in parallel with the coil.

  Victor Roberts,

  Yes I see what you are saying. It is helpful I have other engineers
  looking
  over my shoulder.

  I need some more switches in there.
  Thanks Victor. Do you have any suggested circuits?

  Paul Lamar

  The current configuration still leaves the 3 untriggered caps with one
  lead
  open. This does not put them in parallel with the coil or am I missing
  something?
  Dick Handverger


  Thanks Dick.
  First condition.  All caps are fully charged through the coil.
  This group is in series with the coil. The problem Victor pointed
out was
  when any lower gate closed all caps now in parallel would be discharged
  through that gate bypassing the coil as that would be the path of
  least resistance. The coil would see no current flow.

  I think I need Triac's in series with all caps to prevent that from
  happening so only the coil sees the sudden increase in current.

  Triacs are bidirectional switches unlike most semiconductors.
  They look like a relay contacts. Relays contacts don't care
  which way the current flows.

  All Triacs need to be on when no trigger pulse is occurring so
  the caps can charge through the coils. Only the Triac that
  is associated with the trigger pulse stays on when the trigger
  occurs.

  I have yet to figure out the logic on how to do that
  short of a computer. Any suggestions?

  Here is an incremented schematic.

  ---------------------------------------------

  Here is the truth table. Output refers to the triac.
  0 is off and a 1 is on.

  Looks like I will need a programmable logic chip
  with 4 inputs and 4 outputs unless somebody can
  come up with a better idea.

  The all triggers on condition never happens.
  In that case there is an error so all triacs
  should be off.

  Paul Lamar
  I believe that if you are using a positive trigger that all outputs
  are low until trigger, then output goes high, charging the cap between
  supply and ground through the coil. This probably means it is a
  capacitive-charge rather than discharge system but since the caps are
  not electrolytic charge and discharge should be symmetrical.

  In order for trigger 2 to deliver full potential, trigger 1 should
  be short enough that the voltage at the coil has returned to near zero
  before trigger 2 begins.  Some experimentation with the cap and coil
  should reveal the duration of the spark. As soon as the spark ends, the
  trigger should go low to "clear the deck" for the next pulse as
there is
  no benefit from continuing  the connection once the current has decayed
  below the level required to sustain spark  If I recall correctly, I
  think one of your earlier posts showed the typical duration of a CD
  spark as being around 100 microseconds. At 8 ms per rev, 10 degrees=220
  microseconds. So, a 100 microsecond trigger gives 120 microseconds to
  recover.

  Victor


  The caps are supposed to charge during the other 320 degrees of E-shaft
  rotation. That means the upper gate should be on during that time.

  The output impedance of the high voltage power supply is finite
  so even though the cap is grounded on one side by the coil
  and connected to 220 volts on the other side by the upper gate
  the internal resistance of the power supply, the coil DC resistance
  and the upper gate resistance restricts the speed at which the
  caps can recharge.

  If I could be sure the cap could recharge in 10 degrees at 7500
  RPM I could use only one section and one cap.  I would use
  4 hall effect sensors and a 4 input nand gate to trigger
  the system ten degrees apart.

  7500 RPM is 125 pps. The time it takes between pulses
  is .008 seconds or 8 ms. 10 degrees at 7500 RPM is .00022221 seconds
  I think or 222 micro seconds if my arithmetic is correct. Hopefully
  the capacitor can recharge in 222 micro seconds.

  That is why I am using  separate caps for each spark.

  I am going to bread board this shortly with only one section
  to keep it simple at first. Then  I'll pulse it 36 times per rev
  or 4500 PPS and see if the system can keep up.

  Here is another idea for isolating the coil.

  Here too are a bunch of pictures with Craig's function
  generator pumping 12 volts  into a Mazda 2nd gen
  coil through a 1 Mfd cap. The screen readouts
  will tell you what is going on frequency wise. The upper trace
  is out of the function generator and the lower
  trace is across the coil. This is only about 12 volts
  so heaven elp us  at 250 Volts.

  Paul Lamar



 Certainly, the larger the cap, and the higher the impedance of the
 charging circuit, the slower the charging time of the cap.  The
 charge/discharge curve is exponential, so worrying about whether the
 cap get's "fully" charged probably doesn't make sense.  To get a sense
 of the charge/discharge timing, you can look at the "time constant" of
 the circuit.  That says the time for the voltage to get to Vi/e.  "e"
 is about 2.8.  It's about 63% of the final value.  That's given by the
 product RC..

 If R=100ohm and C = 1mF, the time constant is 100ms.

 If you can get the R down to 10ohm, the time constant is 10ms

 Do you have a datasheet for the power supply?  It's probably pretty
 low - for efficiency.  It would be interesting to see what the short
 circuit current is.  I'd guess it has current limiting circuitry for
 protection.  Charging the cap may likely bump into that.


 Regards,

 Matt-

 No info on the power supply as it is an off the shelf 220 volt inverter.
 400 watts so it must be pretty low.
 10 ms is a long time at 7500 RPM.

 Paul Lamar



I know 10ms is a long time..  Do you know that you need a cap that large?


Regards,

Matt-

Actually we are a bit shy of high voltage so I was thinking of trying
a 2 mf.

Paul Lamar

You could use a voltage doubler instead of a full-wave bridge to boost
the voltage.

Victor

Good idea Victor. That might work great. I might be able to
get 300 to 350 volts out. Some CD systems are 500 volts or
more. Thanks for thinking of that.

Paul Lamar