EEVblog #306 - Jim Williams Pulse Generator

[free_electron]

I've built up my kit - and it is not working


Having got through two batteries I've connected it to my lab power supply. It draws about 10mA. The voltage at the end of the diode chain is a healthy 106V.

The voltage at the transistor across C2 is 35V. The Voltage across the 50 ohms is 0.33mV which corresponds to 6.6 microAmps. The voltage drop across the 10M R1A is about 71V
so the current through this is approx 7.1 microAmps.

This means, I guess, that Q1 is leaking too much current to reach avalanche at 40V?

The voltage across R5 is 252mV where as it perhaps should be 106V*20k/(10M+20k) =209mV. In the circuit diagram R5 is only 10k whereas on the board layout
and as supplied in the kit it is 20k?

Any suggestions as to what Voltages and currents I should be seeing would be most welcome. Is 7 microAmps very leaky? (I've no idea.)

Should I be connecting a 10M R1? (No R1 came in the kit, I thought R1A replaced it?)

If I added a 10M as R1 then the voltage drop would be halved and the Voltage across Q1 would rise to about 53V potentially (though presumably it would
do what it is supposed to do and breakdown at 40V).

It's nearly midnight so I'm off to bed but any help would be much appreciated.


EDIT (following morning) :

I've added a 10M as R1 and it is now working! At least it is avalanching at a rate of just under 40kHz or every 25 microsecs. Q1 looks like around 5M when it is not avalanching, so the circuit is
two 10M in parallel then a 2pF in parallel with 5M so it is not quite as simple as just RC for the time constant - the effective R is 12.5M - I suppose the last bit of charging is stretched out. I've just looked at it with a probe but will properly terminate it and get some plots when I get a bit of time later on today.

The current draw is still about 10mA so most of this is probably the current of the led (around 8mA or so) - so if you want to save battery life disconnect the led.

Looks like you have leakage around the transistor. Did you clean off the flux ? That area needs to be clean !

[jpb]

Looks like you have leakage around the transistor. Did you clean off the flux ? That area needs to be clean !

I didn't because the flux described itself as "no-clean flux" and claims to meet SF-818 for surface insulation resistance what ever that is.
Probably I should have (though it all looks very clean anyway). This was my first ever bit of smd soldering and use of a flux pen so it
has been a bit of a learning curve. Initially no flux came out of the pen at all, but then I actually read the instructions and found you
need to press it down on the point/knib to start the flow!

I'll clean it up with a bit of IPA, though now I have the extra 10M in there it won't make any difference to the behaviour except perhaps
to increase the frequency from 40kHz if it gets rid of the leakage.

The extra leakage mattered when it was preventing the capacitor charging to above 35V, but once avalanche is reached having approximately
5M in parallel presumably is neither here nor there. With the 5M resistance, even with the leakage it should be able to reach avalanche voltages up to
around 55V.

Thanks for the advice, and all the work you put into this project - I am having a lot of fun with it.

[jpb]

Following Free_electron's advice, I've thoroughly cleaned everywhere with IPA including the connector. The results are a small but noticeable improvement as can
be seen in the curve.

A slight downside is the varnish on the circuit board goes sticky with IPA but hopefully it will dry out again and anyway it is safely sealed in its box.

[jpb]

Here is the output curves before and after cleaning - the difference is generally very small but there is visible improvement on the leading edge where it matters.

[van-c]

Are the kits still available?  If so, I'm very interested in buying one; if not, I'd like to get in on the next backlog.  Please post or PM me with details.
Thanks,

--Van

[BravoV]

A bit related to this, just want to thanks Vincent for selling me bunch of these gold doped transistor, and those gold pins are still really shiny after all these years (batch code 87xx).   

[sdscotto]

I wanted to thank Vincent publicly for the work put into this project.  Also a thank you to those of you who have taken the time to advance the project and teach dummies like me how things work.

scotto

[free_electron]

A bit related to this, just want to thanks Vincent for selling me bunch of these gold doped transistor, and those gold pins are still really shiny after all these years (batch code 87xx).   

That's real SGS (Sociedad Generale Semiconduttore ) quality ! we be using real gold at 15 microns instead of flimsy 3 micron flash (if its not rhodium )

SGS was a Spinoff from olivetti
ATES (Aquila Tubi E Semiconduttori ) licensed the transistor directly from Shockley ! The fab was in Catania ( which is today still making IC's)
Thomson : borged Secosem , Mostek and others.
SGS-ATES + Thomson  = SGS-Thomson Microelectronics or STmicroelectronics for short.

[alank2]

I wanted to thank Vincent publicly for the work put into this project. 

+1, I can't wait for them to be available again so I can get my hands on one!

[JuanPC]

can someone explain, ¿why make such a complex device?


isn´t easy to make a short circuit, ?
9v or 12v, or 18v, or 24v.

555
relay,
50ohm resistors.
done.
A fuse just in case 

[c4757p]

There is no way in hell you will get sub-nanosecond rise time from that. The vast inductance of the system will limit it to significantly longer edges.

[JuanPC]

There is no way in hell you will get sub-nanosecond rise time from that. The vast inductance of the system will limit it to significantly longer edges.

what kind of inductance has a relay and a couple of resistors?

[c4757p]

At least a few tens of nH, if they are small and you have connected them very carefully. If you want 500ps rise time, that translates to a bandwidth requirement of about 0.35/500ps = 700MHz. 6.28*10nH*700MHz = 44 ohms per 10nH, making this quite significant.

[JuanPC]

At least a few tens of nH, if they are small and you have connected them very carefully. If you want 500ps rise time, that translates to a bandwidth requirement of about 0.35/500ps = 700MHz. 6.28*10nH*700MHz = 44 ohms per 10nH, making this quite significant.

the Relays i have here has: 0.000mH
QIANJI JQC-3F(T73)
10A 28VDC / 7A 240VAC
5VDC
Relays have a <10ms Operating Time, and <5ms Release time. = 15ms max period = 555 must not exceed/should operate at  66.667Hz.* i´ve pushed those relays up to 120Hz.

even better...
dual cascaded Relays!!!

18vdc--->555--->Relay1--->Relay2--->5v 50ohms output.
Relays work at 5vDC each.
just a couple of voltage dividers.
done.

[c4757p]

Here's my best attempt at measuring the rise time of a mechanical switch. It ain't pretty. The fall time was so bad as to not merit the 60 seconds or so it takes my scope to record a screenshot.

[c4757p]

the Relays i have here has: 0.000mH
QIANJI JQC-3F(T73)

Bullshit it has 0.000mH inductance! Show me.

[JuanPC]

the Relays i have here has: 0.000mH
QIANJI JQC-3F(T73)

Bullshit it has 0.000mH inductance! Show me.

The Uni-T UT70A is calibrated for the long probes, thats why measures -0.002 with the short alligator clips.
Short Probes -0.002 = 0.000mH
Just for fun added an EMI filter.

[c4757p]

Datasheet claims 2% + 10 counts on inductance. That means it could read all the way down to 0 for something that's actually 10.2 nH. At the low end, that's not unlikely.

And mechanical switches don't behave as well as you would think, as I just showed. See how the voltage mostly ramps up over a few microseconds, and only speeds up at the end? That's pretty useless for any real measurement.

Also... try zooming into a 1ns edge that only happens at 10 Hz or so on an analog oscilloscope. You could crank the brightness to maximum and still not see it!

[free_electron]

uni-t 'professional multimeter' 
2millihenry range trying to measure nanohenry's ... 

you don't stand a chance with a relay. even a normal transistor can't switch that fast. we are dabbling in the picosecond range with this pulse generator. spanning 8 volts in 200 picoseconds needs something very fast. fast as in atomic physics fast. which is why we use the avalanche effect of driving the transistor in controlled breakdown.
the energy stored in the stripline is released over time to give the top plateau. The stripline is tuned for that so it releases energy correctly.

[BravoV]

Datasheet claims 2% + 10 counts on inductance. That means it could read all the way down to 0 for something that's actually 10.2 nH. At the low end, that's not unlikely.

And mechanical switches don't behave as well as you would think, as I just showed. See how the voltage mostly ramps up over a few microseconds, and only speeds up at the end? That's pretty useless for any real measurement.

Also... try zooming into a 1ns edge that only happens at 10 Hz or so on an analog oscilloscope. You could crank the brightness to maximum and still not see it!

uni-t 'professional multimeter' 
2millihenry range trying to measure nanohenry's ... 

you don't stand a chance with a relay. even a normal transistor can't switch that fast. we are dabbling in the picosecond range with this pulse generator. spanning 8 volts in 200 picoseconds needs something very fast. fast as in atomic physics fast. which is why we use the avalanche effect of driving the transistor in controlled breakdown.
the energy stored in the stripline is released over time to give the top plateau. The stripline is tuned for that so it releases energy correctly.

Guys, some hints -> HERE or HERE, I guess he has trouble with scales.   

[JuanPC]

uni-t 'professional multimeter' 
2millihenry range trying to measure nanohenry's ... 

you don't stand a chance with a relay. even a normal transistor can't switch that fast. we are dabbling in the picosecond range with this pulse generator. spanning 8 volts in 200 picoseconds needs something very fast. fast as in atomic physics fast. which is why we use the avalanche effect of driving the transistor in controlled breakdown.
the energy stored in the stripline is released over time to give the top plateau. The stripline is tuned for that so it releases energy correctly.

#1. The UNI-T works fine for me.
it measures the probe inductance: 1.2mts vs. 20cm
0.002mH = 2uH,

#2. The 2n2369 is just an old 123AP !!! used for Audio.-
#3.
the word you are looking for is: "krytron, used for extremely precise rapid high-voltage switching. Krytrons with certain specifications are suitable to initiate the precise sequence of detonations used to set off a nuclear weapon, and are heavily controlled at an international level ... This design, dating from the late 1940s, is still capable of pulse-power performance which even the most advanced semiconductors (even IGBTs) cannot match easily. Commutation time: less than 1 nanosecond can be achieved, with a delay between the application of the trigger pulse and switching which can be as low as about 30 nanoseconds. The achievable jitter may be below 5 nanoseconds. "

Wrong, faster than Atomic Bombs.
http://en.wikipedia.org/wiki/Krytron

[c4757p]

Oh dear god, we both missed it.

0.002mH = 2nH,

No. 0.002mH = 2uH. Micro. One thousand nH. You weren't even looking at nanohenries, you were looking at thousands of nanohenries.

#2. The 2n2369 is just an old 123AP !!! used for Audio.-

No it's not. It's a "high speed saturating switch". It's made for switching very fast. It will give 1-2ns rise time even just being switched in standard common-emitter configuration. Absolutely ridiculous for audio purposes. Read the datasheet. Spend some time with the one for your multimeter too.

#3.
the word you are looking for is: "krytron, used for extremely precise rapid high-voltage switching. Krytrons with certain specifications are suitable to initiate the precise sequence of detonations used to set off a nuclear weapon, and are heavily controlled at an international level ... This design, dating from the late 1940s, is still capable of pulse-power performance which even the most advanced semiconductors (even IGBTs) cannot match easily. Commutation time: less than 1 nanosecond can be achieved, with a delay between the application of the trigger pulse and switching which can be as low as about 30 nanoseconds. The achievable jitter may be below 5 nanoseconds. "

Er.... What does this have to do with fast rise time pulses, exactly?

[BravoV]

#2. The 2n2369 is just an old 123AP !!! used for Audio.-

Sigh .. I guess I must be conned by free_electron into buying couple of those 2N2369A "audio" transistors from him. 

[JuanPC]

Oh dear god, we both missed it.

0.002mH = 2uH,

No. 0.002mH = 2uH. Micro. One thousand nH. You weren't even looking at nanohenries, you were looking at thousands of nanohenries.

#2. The 2n2369 is just an old 123AP !!! used for Audio.-

No it's not. It's a "high speed saturating switch". It's made for switching very fast. It will give 1-2ns rise time even just being switched in standard common-emitter configuration. Absolutely ridiculous for audio purposes. Read the datasheet. Spend some time with the one for your multimeter too.

#3.
the word you are looking for is: "krytron, used for extremely precise rapid high-voltage switching. Krytrons with certain specifications are suitable to initiate the precise sequence of detonations used to set off a nuclear weapon, and are heavily controlled at an international level ... This design, dating from the late 1940s, is still capable of pulse-power performance which even the most advanced semiconductors (even IGBTs) cannot match easily. Commutation time: less than 1 nanosecond can be achieved, with a delay between the application of the trigger pulse and switching which can be as low as about 30 nanoseconds. The achievable jitter may be below 5 nanoseconds. "

Er.... What does this have to do with fast rise time pulses, exactly?

Sorry Typo Mistake. Corrected.

milli m       0.001    thousandth    1795
micro u     0.000001    millionth    1960
nano n     0.000000001    billionth    1960
pico p       0.000000000001    trillionth    1960
femto f     0.000000000000001    quadrillionth    1964
http://en.wikipedia.org/wiki/Nano-

1H = 1000mH
1mH = 1000uH
1uH = 1000nH
1nH = 1000pH

#2. 

[c4757p]

Yes. Exactly. So that relay could have been all the way up to 10 microhenries and your meter could have read the same. So nanohenries? Forget about it.

Don't even get me started on the utter uselessness of an inductance meter with an error of 10 uH...