EEVblog #306 - Jim Williams Pulse Generator

[alm]

However we must remember that the pulse generator itself has a ~ 350 ps rise time, and so the total system rise time is defined as :

t_sys ^ 2 = t_scope ^ 2 + t_pulsegen ^ 2

and so if t_sys = 0.73 ns  and  t_pulsegen = 0.35 ns  we get that the scope rise time must be 0.64 ns which correspond to an approximate bandwidth of 625 MHz which is what was expected.

Note that this formula only applies if all parts of the system have a Gaussian response. For maximally flat response, I believe Agilent's suggestion in an appnote is 'ask the manufacturer', since it's quite rare for scope vendors to publish bode plots of the scope response.

[free_electron]

Right. Here's a promised update:


Final design.
Deviation from the JW schematic : i use a CR2032 coincell and have two additionla 1-uf Ceramic caps around the LT1073.


The @n2369 is mounted Dead-bug style in a hole drilled in the PCB. The cae of this transsitor is electrically connected to the Collector. so the hole has the 1 Megaohm and 2pF caps connected to it. The collector wire is simply snipped of , or can be tacket to the rim of the case.
The base wire is bent outwards and down to land on an SMD pad to the 10k base resistor.


the 50 ohm resistor is made with two 100 ohms 1206 body resistors soldered back to back and vertically. A standard through hole SMA connector is put on edge and soldered on the board.

The emitter wire is bent out form the to-18 can and lies on top of the resistors electrode and contacts the center pin of the SMA. Two dabs of solder do the rest. The goal wa ot minimize stray capacitance as much as possible. I also opened the soldermask around the critical area's to avoid leakage there.

Copper features with soldermask removed :




The loop on the back is intentional. This is the feedback point of the voltage divider. I do not want that running underneath the switching inductor as it would pick up the field , so it 'detours'. that point is high impedant.

Everything fits in a Hammond case model 1553AABK or 1553AAGY. All the cutouts and mounting holes are present in the PCB.
All parts are available from digikey apart from the 2n2369.

I will order the PCB tomorrow from IteadStudio. I'll have a few extra's. If anyone is interested. 5$ gets you a blank PCB (US shipping included). I order the 10x5 cm as this board is 6.5 by 5. And go for the color version (i don't like green boards. I'm colorblind and don't see green. They are grey to me ... dull and drab. I may go with yellow soldermask and black text ... or Blue and white ). Shipping will be done to me from itead using UPS. I don't want to wait an unknown amount of weeks to get this.

I will probably order parts from digikey in quantity a well ( no point in buying 3 diodes.... ) So if anyone is interested in the full kit.
I have to calculate what the final would be. I have 2n2369 plenty. I'll throw in 2 or 3 per kit.

[krivx]

Looks like a really nice layout. I think I would like a full kit 

[chavotronic]

One thing i don't get with the rise time / BW formula.
How can this only be dependent on the rise time, and not the rise time AND the amplitude (the slope).
I would guess that with a higher amplitude, your rise time is higher even with the same rising slope.
Would be nice to have an answer to that.
David

[G7PSK]

Hi Free Electron, I would like to buy a full kit please. Are you willing to ship to the UK though, if you are you have sold one, many thanks. G7PSK

[BravoV]

FE, count me in for one kit.

PS : Suggesting you to create new thread for it in the "Buy/Sell/Wanted" sub-forum.

[vk6zgo]

One thing i don't get with the rise time / BW formula.
How can this only be dependent on the rise time, and not the rise time AND the amplitude (the slope).
I would guess that with a higher amplitude, your rise time is higher even with the same rising slope.
Would be nice to have an answer to that.
David

No,the rise time is defined as the time between 10% & 90% of the final amplitude of the signal at any signal amplitude.
It is the percentage change,which defines the slope,not the absolute value.
Think about ordinary swept frequency response--this is not affected by the signal amplitude,so why should rise time be any different?

Rise time is unaffected by changes in amplitude,if a perfect attenuator or amplifier is used.
Real attenuators & amplifiers have a response of their own,which will affect the  rise time of the signal after passing through them.
The rise time is worse after passing through these devices.
You can improve the rise time of a rectangular or square wave by"clipping" the signal at some level which is reached earlier on the leading edge of the signal.

[tecman]

I will take a full kit as well.

paul

[HackedFridgeMagnet]

Dave you mentioned RG59 at just before 7 minutes, you must've meant rg58.

I'll take one Free Electron and postage to Oz if possible.
What case are you using? I'll go in for a case too if possible.

[all_repair]

Count me one for a full kit if possible with case to Singapore.  Thanks !!!

[dbinokc]

I just signed up for the forum just so I could say count me in as well for one.

[Alexei.Polkhanov]

You could try these from ebay:

1I308J tunnel diodes 2pcs

They're supposedly russian Germanium tunnel diodes with 20mA peak point current and 1...4 pF valley point terminal capacitance. Less than 100 ps could be possible with these. Fast tunnel diodes are very rare, and very expensive.

I got 6 of those from same seller on ebay 2 months ago for laser distance measuring project that I was planning. I checked one diode for basic diode action when I got the package - it is fine, but I have a problem however - I need a holder for these tiny devices. Germanium ... most likely not solderable at all. No clue where I can get nice holders. I soldered a needle to a small pad and tried to use it - looks like I need something better than that and preferably gold plated. Also I am scared to touch them even with wrist band on - they must be uber sensitive to ESD.

Woods metal. It melts at little over blood heat  70 deg C and was often used for mounting germanium in the past. There are alloy metal that melts even lower.   http://en.wikipedia.org/wiki/Wood's_metal

Fact that they have lo leads probably means they never meant to be soldered. Kind of like those gunn diodes that go into special fixture inside microwave horn I guess.

[robrenz]

Count me in for a kit also.

[Alexei.Polkhanov]

I will probably order parts from digikey in quantity a well ( no point in buying 3 diodes.... ) So if anyone is interested in the full kit.
I have to calculate what the final would be. I have 2n2369 plenty. I'll throw in 2 or 3 per kit.

I don't understand what is necessity of taking gamble with these 2N2369 devices and "hand-picking" them when choice of specialized avalanche action transistors available like FMMT413 from Zetex. 50V breakdown voltage - should work in this design as well and it is in SMD package.
DigiKey: http://www.digikey.ca/product-detail/en/FMMT413TD/FMMT413TDCT-ND/2248551

[free_electron]

There's a couple of things at play

First of all the 2n2369 is NOT a true avalanche transistor. It is just being used in avalanche mode ( which is actually a bad idea ... unless the current is critically limited this would normally fry a transistor )

Avalanche mode operation is actually a controlled breakdown of the Vceo voltage. the collector voltage is so high that it it succeeds in pulling electrons from the emitter through the base region. The transistor enters conduction and you get a snowball effect. the high collector voltage causes the transistor to start conducting all by itself.

Real avalance transistors, in real avalanching applications such as laser drivers are held just under this point . The little extra energy that is required is created by pulling the base high. this gets the electrons shot in the emitter and they race towards the collector en-masse. there is suc h a thing as secondary avalanche but i'd have to look it up.. can't remember the exact physics.

Any transistor can be made to avalanche. It's just a matter of finding the right voltage and limiting the current so you don't fry it.
I did some poking around and there are other transisotrs that will perform very well in this pulse generator like the BFG541 and the ZTX415.

The problem with true avalanche transistors is that they are pricey .... 7 to 15 $ ... a pop... while a 2n2369 costs 0.5$ ...
Anyway , this circuit has enough room for experimentation. if anyone wants to have a go at true avalanche transistors be my guest.

[BravoV]

Btw the BFG541 cost just $0.5 and its in SOT223 form, I guess we could even solder it with the cap + resistors straight at the at back of the bnc connector.

[Alexei.Polkhanov]

BFG541  is not designed for avalanche action. It is a same story as with 2N2369 - avalanche action is a side effect of what otherwise sold as regular transistor. I think BFG541 has to be "hand-picked" in same way as 2N2369. I see people raising concerns about how long these transistors will survive if used for avalanche action continuously like in case of laser distance meter for example.
 

[free_electron]

True . those transistors are not designed for avalanche operation, but then a true Avalanche transisotr is not meant to be used in this fashion either.
A true avalanche transistor is 'ignited' by a base pulse that starts conduction by pulling electrons from the emitter to the base and opening the channel.

What this 'pulser' system is doing is causing avalanche by pulling the collector above the breakdown voltage and ignite the avalanche that way. No transistor is meant to be used that way. The energy that is dissipated is small due to the fact we only have a 2pF reservoir and we can only feed current through a 1 Megaohm resistor.
Short term it is no more destructive to the transistor than trying to find what is Vceo using a curve tracer. Long term .. your guess is as good as mine.

[GK]

Datasheets for transistors marketed for avalanche operation actually spec max pulse peak collector current under such conditions - for example 50A (25A secondary breakdown) for the FMMT413.

Shouldn't that imply that these transistors are specifically designed to be robust and long term reliable in this regard?

The FMMT413 datasheet also gives nano farads of external C-E capacitance as a test condition, which is a lot more than 2pF. With only 2pF and "ordinary" small signal transistor in this application is probably more than safe.

[free_electron]

Datasheets for transistors marketed for avalanche operation actually spec max pulse peak collector current under such conditions - for example 50A (25A secondary breakdown) for the FMMT413.

Shouldn't that imply that these transistors are specifically designed to be robust and long term reliable in this regard?

The FMMT413 datasheet also gives nano farads of external C-E capacitance as a test condition, which is a lot more than 2pF. With only 2pF and "ordinary" small signal transistor in this application is probably more than safe.

True , avalanche transsitors are made fur such behavior. Note that they are intended to be driven into avalanching by pulling the base up , not by overloading the Vceo breakdown voltage ! No transistor is happy in that.

witht he few picofarads the the non avalnche transistors will survivie. Avalannche breakdown is one of the tests done using curve tracers. Then again , you do that during the design of the transisotr on a few thousand  samples to get a spread and set the safe operating margins.

I'd have to talk to our transistor people but i beleive the physical construction of an avalanche transisotr differs from a regular in the sense that the CE channel is much wider, for a same base doping, than in a regular transistor. Like you said. in avalanche the current can be quite high.

Strange beasties..

anyway. It's a Jim williams design. He knew what he was doing. And the idea is to have a pulse generator to compare scopes. You can always install the BFG or the FMMT on the board.
the BFG will fit as its big tab is collector. poke it in the hole , tack it down with solder and connect emitter and base in floating style.
the fmmt is trickier as that is a tiny sot23. but its doable.

Boards are in production since this morning...

[notsob]

Count me in for a kit also.

[Alexei.Polkhanov]

Here is the link to a Masters thesis from Oulu university. There on page 33 (chapter 3.4.3) there is along description of many experiments  with laser and avalanche pulse generator, including current measurements etc. Guy definitely had access to good equipment. http://herkules.oulu.fi/isbn9514272625/isbn9514272625.pdf.

[carpin]

Count me one for a full kit to the Netherlands.
Thanks

[GK]

True , avalanche transsitors are made fur such behavior. Note that they are intended to be driven into avalanching by pulling the base up , not by overloading the Vceo breakdown voltage ! No transistor is happy in that.

Hmm, I'm not sure about that. See the Zetex app note here:

http://www.diodes.com/_files/products_appnote_pdfs/zetex/an8.pdf

The series stack configuration for high voltage pulses, as detailed in that app note, rely on Vceo breakdown of all the transistors with the exception of the one fired.

An interesting note in the 4kV stack design paper I linked to a few post back, on the 2n5551, is that they found the Vceo breakdown subject to significant drift in operation, and had to subject the transistors to months of cooking in the test jig to stabilise them.
I wonder if these dedicated avalanche transistors like the FMMT's are pre cooked at the factory - might explain in part the price premium.

 

[free_electron]

@gk. interesting reading material.
So it does look like they trigger one, that one avalanches thus the voltage on ce. Becomes next to nothing which means the other guys actually see their vceo go up beyond Bvces so they avalance.

Hmmm. Interesting. Cool, i learned something.   Very oddball parts apparently.
I wonder if there is a typo in that document. They claim trsting to 4x 1011.. Strange number unless the 11 need to be an exponent. 4x 10^11 or 4E11.. It would be odd to do a lifetime testing with 4044 pulses....