Re: LF: Re: Class D current spikes

[Andy Talbot <andy.g4jnt@gmail.com>]

IN that case some sort of protection needs to be incorporated - and few have come up with really properly designed and foolproof circuitry.

I use the technique adopted in the hold Decca transmitters, detailed here http://g4jnt.com/137tx.pdf  but it assumes a tank circuit is in use.    For transformer and traditional PI/Tee LPF arrangements it's not applicable

I don't know of anyone who has used this technique outside designs based around that article that I've been told about- which seems amazing considering how absolutely robust and foolproof it is.   Complete overload protection against any mismatch.   

I've seen all sorts of overload protection circuitry in use, current monitors and trips, Hall sensors etc etc  But if there is going to be a current spike then it has to be designed-out in the first place.   Choke fed PAs could perhaps have a bigger choke that can absorb a few 100us .   Put voltage clamps in if overvoltage is the problem.  Just Schottky diodes up and down to the rails.

Andy

www.g4jnt.com

On Sun, 24 Feb 2019 at 18:08, Eric NO3M <no3m@no3m.net> wrote:

Andy

This recommendation is all well and good, but seems applicable to only CW where either the operator is manually keying or using computer controlled keying with a constant carrier input.

With digi modes (WSPR, JT9, etc.) that do not present a constant RF input, the traditional PTT signaling is of no use as it keys active prior to the RF envelope start and goes inactive after the RF envelope ends.  The spikes would still be present and problematic.

Other than Rik's suggestion of killing the lingering active high driver output, it seems that only using a supply control scheme (whether driver or FET drain supply) would require sampling the original RF input to detect the end of the envelope and apply the key up (inactive) signal.  Prior to reaching the driver subsystem, the input RF would need to be delayed by some amount.

73 Eric NO3M

On 2/24/19 9:41 AM, Andy Talbot wrote:

For on-off keying of a switch mode PA you really should be using supply line control.  Switching the drive then trying to reduce rise /fall times for keyclick elimination by controlling drive amplitude(something that seems to be done in some designs) is really asking for trouble.  Even hard-switching drive gives its problems , as we see here repeatedly

Instead, why not a simple emitter follower in the positive supply, then you can apply waveform shaping  to the base drive at low current, where you've got full control.   Even a 20A supply could be controlled this way with a suitably rated transistor.  Quite likely have to be a Darlington.   There will be a slight fixed drop across it, say 1V at saturation / full power, so a few watts of dissipation, but it's only operating non saturated for a short period during the rise / fall times.

A power FET could be used, but you'd need a drive voltage Vgs higher than the supply, albeit only a few mA capability; nothing a small switcher module couldn't manage.

 

Andy

www.g4jnt.com