MOSFET gate to drain unwanted coupling

[charliecoutas]

Thanks David. I'm now clamping the feed to the feed to the reference. The reference itself will only come into action when the voltage exceeds its rated voltage (2.5V). So a leading spike could still get through and cause trouble.

Our Colossus takes about 12KW from one phase. Switching on power to that many heaters, which are much lower resistance when cold, caused a few surprises as various protection devices operated. So we slowly wind the heaters up automatically with a huge motorised Variac, generously given to us by Claud Lyons who make said items.

The "up" time varies but it's generally quite good. Problems can be corroded terminal blocks, bad soldering, valves going low emission etc. Our machine was started about 15 years ago. It is a rebuild. You only need five original parts to qualify as a rebuild. The paper tape reader, or Bedstead to give it its proper name, used selenium proximity valves to sense the light through the holes in the tape. These were used on guided missiles: when the area in front of the flying missile went dark, it must be the target, so bang! We have had to disguise modern photodiodes in the original glass housing because you simply can't get the originals any longer.

https://www.tnmoc.org/colossus

Hiya Les.

Charlie

[trastikata]

trastikata: I need to clamp the voltage right down to zero (well a millivolt of two). A BJT would have some Vce surely?

Ok I see what you need.

Just a question - do you need sharp rise and fall times? A 2k-4.7k resistor at the FET gate usually solves this spike issue if indeed the rise and fall times allow it.

[david]

Charlie - many thanks for the link.  12kW - jeez.  I bet the lights flicker a bit when that's running. Fascinating stuff!

Les  - a digital attenuator (I think) attached.  It seems to model ok and I seem to remember back about early 80s that ETI magazine did a filter design using PWM of both capacitors and resistors.  Perhaps someone else could comment on this approach.  Obviously maximum attenuation is when the duty cycle approaches 100%.  There's no distortion (with ideal parts) and it is high level.  Of course the voltage controlled switch would be a FET like a 2N7000 in which case you'll end up with the same spikes Charlie has had but they should get filtered out.

Cheers,
David

[charliecoutas]

Yes, sharp rise and fall times are needed. Hopefully better than 5uS.

The lights don't flicker but we've had to install fans for warm days. The main culprits which cause a hotspot in the room are the 501 thyratrons which hold the 501 "bits" of the Lorenz cipher. There were not many ways of providing "memory" in 1943 so gas filled valves which could be either on or off were used. But 12KW is 12KW however you chop it up!

I think I have the spike problem solved, thanks to you guys!

Charlie

[JonW]

A good old 74LVC3157 for shunting and the audio application, its Fast, easy to use and low RDS on with respect to your line Z of 4k7 so should clamp very close to 0V and has great isolation in off state to boot. 

Les have you considered the series RF pin diode  it may work? Never used one at Audio though and may need a rather high bias current! , DC applied through resistors and L&C.  Looks like a VERY LOW capacitance when off and a variable resistor Vs Current Vs bias current.  You can drive with PWM or using PIC peripherals opamps, Dacs , PWM, R2R etc.  For better linearity ( current vs attenuation)  you can use in a quad configuration but will be more costly.   
Could also use the analog switches as a switched attenuator (passive)? 

Or a resistive R2R type and a 4051 multiplexer or I/O to Gnd forming a 2 resistor pad?


J

[top204]

Thanks John, but I did the job in software using Fixed Point for the audio, so the decrements in the audio data are in small steps and produce very clean sounds. It also takes a few components from the final build.