Expert Pulse Amplifier

Hi Shyam, I've not come across the term "Channeltron" before, but reading your specs it looks quite simple....

Input voltage negative 10mV into 50 ohm.... 1ns rise time and 3 ns decay is easy....

1 Volt output, no problem!

There's no need for 10 GHz Ft transistors, a good design can use humble BSX20's etc...

Might have to use an emitter coupled pair, but no problem

John.

Dear John

BSX20 is a low gain 500MHz transistor. I am aware that it needs fater band wodth of 10GHz and also Hfe > 100 at about 10mA. In fact I need JFET, PNP and NPN transistors all that can work at high frequency due to 1ns rise time. I need proper pulse shape.

Pulse electronics requires some special consideration than RF electronics. Thanks for writing and if you have some better idea then write to me.

Have to disagree Shyam,

Just because you specify a 1 nS risetime doesn't mean you need expensive high Ft transistors....

ECL logic chips work with sub nanosecond rise / fall times and the chip transistors are fairly standard rf types....

The 'trick' is to make sure that the transistor stays out of saturation / that the Miller capacitance is kept to a minimum maybe by using a cascode circuit and hey presto!

You will have a super high speed edge.... My final year thesis was on this topic of super fast switching and high speed sample and hold techniques.

That was back in the days when a BSX20 was the high speed transistor of choice, anything faster was too expensive to use...

John.

Oh and also Shyam, you do not need high gain transistors, in fact the higher the Hfe the lower the Ft (generally speaking) for a transistor.

You only need a tiny gain of 100 and a single emitter coupled pair will give that without problem.

The speed of the edges will be around the 200 pSecond region depending on layout etc...

To drive 50 ohms you will then need a simple emitter follower with a pull down as output, operating in class A mode will be sufficient.

John.

Oohhhh just googled channeltron is a tradename ... its nothing more than a type of photomultipling tube or Gieger Muller tube....

For mass spectroscopy... interesting... how soon do you want this pulse amplifier Shyam?

I can email you a design within a few hours...

Dear John,

Channeltrons are channel electron multipliers. These are faster devices than Photomultipliers and work in vacuum as they are open devices and not vacuum packed like PM Tubes. Channel electron multipliers also come in an array called MCP or Micro-channel Plate.

Gain of the channel electron multiplier is of the order of 10^6 to 10^7 order and input is single electron only and gain depends on the channel biasing voltage similar to the PM Tube operating voltage or GM Tube as you have rightly said (but has no atmosphere or gas inside). Channel is a continuous multiplying tube of glass coated with resistive thin film kept at high potential difference and there are no dynodes in this device. Only anode or farady cup exists to collect electrons.

As the device generates charge with rise time of 100ps to 250ps and with decay time of about 1ns to 2ns, it is a very fast devices. Pulse charge is about 1.6x10^-13C range. Charge gets integrated in cable capacitance, and hence we use shortest cable of say 6" in length. This reduces peak height and also makes the pulse broader. Perhaps 20pF is the minimum capacitance comes from connectors. There is about 3000V on the Farady cup or anode where charge is collected.

I used 6" Teflon cable, 10nF isolation capacitor and 1M termination and then amplified using slow 2N3905 slow PNP transistor. I could get 800mV peak output but 1us long pulse tail. This is due to slow response of the transistor and perhaps capacitance as 2N3906 (On Semi) is only 300MHz but has high gain hfe of 100+.

This is a problem as my next pulse rides over the present pulse and then next pulse over the other. This is not what I want. The turn OFF time and Turn ON time of the transistor is very important and need to be very fast to about 1ns to 5ns range. I can't have this unless I use 10GHz switching transistor.

ECL logic uses transistors of this order BW and they are not slow. What you are telling me is to use the transistor in class A mode to let it work on faster slope. I understand that and that is the way they are used as signal is too small.

I actually have to amplify this fast signal and then have to use a comparator with different threshold levels to compare if signal is above the level and to obtain TTL level pulse for it.

I am sure your experience will be useful. Let us discuss more about the possible solution. I also have to worry about the safety of the transistors as there is high voltage which may generate transisnts on power ON and OFF conditions. Thanks for feedback.

Ahhhh are you saying the source impedance is not 50 ohms.... it sounds like it if you're measuring the output of one of these tubes...

In that case the design becomes a little more involved, but not excessively... with a high speed fet to buffer the input allowing a high impedance source that would be okay...

The rest of the circuit I must emphasise does not need to use such extremem transistors...

You mention you've tried it with the 2N3906 series, these are not very good for high frequencies, they have the gain but you don't want that.... The die geometry is designed for audio and medium power applications... Whereas the BSX20 (2N2969) is designed for high speed switching, they are two entirely different types of transistor.

In a straightforward class A common emmiter design both would be useless in your application, except for a small gain block perhaps... But when used in a circuit which cancels out the Miller capacitance and does not saturate the transistors the BSX20 will fly through your specifications without trouble.

So it seems with a high impedance source we would need a fast fet input, similar to a scope input, followed by a switching stage and then a power output stage to drive a 50 ohm load...

John.

Dear John,

Actually anode is biased using 1M or greater resistor to 3kV and hence you can say that it is high impedance. What we get is small change in voltage due to charge collected at anode. Hence, anode is not floating and can not be linked to GND directly.

We use 10nF capacitor of HV type to isolate the input and then can terminate it to GND using whatever resistor we want. 50 Ohms for 1.6 pC for 10 million electrons in say 10ns pulse will have only 8.0 mV pulse. same thing in 1ns will have 80.0 mV pulse. I am sure if pulse lasts only 1ns, it will go undetected by many transistors. As I have streched the pulse, it was detected easily but that is not what I want.

Even if I can get a charge integrator that can sense 1pC and hold the charge and decay in 20ns then also OK.

Ooowww 1nsecond can be a long time

Just think of the timebase in a counter or oscilloscope - yes I've designed these as well! The trigger can be set along a 1 nS edge quite easily and with great stability...

I'm wondering if a simpler solution would be to use a buffer amp followed by a tunnel diode to give a 100pSec pulse to feed a transistor buffer output....

There are many ways of desinging this type of circuit and most of them are very cost effective.... afterall a great deal of design work has already been carried out into designing high speed triggers for scopes using only millivolts of signal...

John.

Dear John,

Now there are two things here.

1. Amplifier (x100) for 1ns pulse which is random with maximum one million pulses per second. 10MHz to 1GHz frequency band with fast settling within 1ns.

2. Voltage comparator to get TTL output as level detector after amplification

If you have done such thing and can find time then you can make small PCB for me for this and see if it works fine. I can pay for it. Alternatively, you can give me scheme and I can try here and will ask students to wire up on test PCB.

I will like to buy JFET, NPN and PNP transistors for this purpose. I have 2SK152 which is good for input stage charge amplifier but need fast PNP and NPN also as 2N3906 and 2N3904 are slower. Do you have any fast PNP known for this purpose.

I have Tektronix TDS2024 2GSPS 4-channel oscilloscope and hence can't see any better. Perhaps 10GSPS should have been much better.

How about BFR93 and BFT93 transistors? Is BSX20 better in performance in comparison to these transistors? I need both NPN and PNP of similar specs.

Dear Shyam, I'm a bit puzzled, why would you want to have a TTL output?

TTL will not 'see' a pulse of 10nS width - well maybe but unlikely. The rise and fall times wouldn't need to be as fast as 1 nano second?

I thought from what you have said you want to catch the next electron burst as fast as possible, so you need a short a pulse as possible to avoid missing the next one... If you're going to feed the output into a TTL gate then that isn't going to respond fast enough surely?

I will have to check on the spec for those 2 transistors, be back soon!

John.

Dear John

While amplifier remains fast, it undergoes small pulse shaping to decrease the rise time and fed into a fast comparator which often requires 10ns pulse width. The output also becomes same and then it is all right in TTL level with about 3ns rise and fall time domain. TTL output is used for pulse rate counting. I can also use ECL but Fast TTL is OK as it gives me about 120MHz count rate (74F161A or 74LVC161A for CMOS use). I am already using that part.

I will use the scheme as follows.

Fast high gm low noise JFET at the input - IF1801 or IFN152 - I Drain 10mA

Fast PNP in Common base - IC 2mA

Fast NPN driver - IC 10mA

Power +/-6V or +/-12V

In second stage I want to use some pulse shaper R-C network and Fast OPAMP drive and fast ECL output comparator followed by ECL to TTL converter.

Dear John,

I am not able to get PDF for 2N2969 or supplier for it. I did find BSX20 catalog and have to find supplier. I will also try 2N4957 PNP and 2N2857. I am not sure if I can get those Motorola 2N3960 I used earlier were 1.6GHz type.

The BFR93 you mention is a nice rf transistor, ideal for rf front ends and i.f. strips etc.... But its not a switcher and it would be wasted as one...

I am still a little confused at your description though... with all the comparator, ECL and TTL plus a fast op amp etc...

The original spec you asked for can still be completed without most of those, a few transistors would do the job nicely...

I wouldn't even think of using a high speed op amp, maybe a single schmidt high speed TTL gate for the output, but you said you wanted a 50 ohm output.... so I'm still convinced that a few transistors is all thats required....

John

Dear John

The pulse looks like this one. Fast negative pulse. Then charge amplified and filtered for noise using fast pulse shaper amplifer and finally fast comparator output ECL or TTL. There is about 10-bit information in the pulse height which is extracted using comparator which is programmed using a 10-bit DAC (10-bits of 12-bit DAC).

Without pulse shaper, lots of HF noise comes due to fast signal. Spaping time is restricted to 10ns to 20ns else will interfere with next pulse. MAX9012 comparator to be used.

Dear John

There is another very important factor that plays the role in charge amplifier is the isolation capacitor. MCP anode signal filter capacitor C1 and isolation capacitor in the charge amplifier C2 generate high voltage in the BNC cable. This voltage causes corona discharges in the BNC connectors. Capacitor always does not mean DC isolation. I need to place the C2 in the charge amplifier as some MCP may not have that C1 installed and all 3kV will appear in the charge amplifier input and damage it. Yes, I do use input diodes for protection but they can't take 3kV.

Selection of the capacitor value is based on the all frequency components in the signal and perhaps from 1Mhz to 1GHz range.

Dear John,

Recently I made two such amplifiers. I found two serious problems creeping in these.

1. Ground current noise from 50Hz power sources

2. RF Noise from 30MHz high power RF source installed in same room

Machine in which Channeltron is mounted has common earth with lots of power driven pumps etc and is having very poor quality body earth even though entire system has metal body and rests on floor. User provided different power ground and then there was ground loop current. Typically 200mV in 5V signal range was the noise at amplifier output.

In second problem, user was running RF source of 30MHz and through large optical viewing windows, RF was directly entering into Channeltron detector sensing points. 1V noise was seen at the amplifier output due to RF pick up.

I am planning power isolation, common ground and RF shielding for reducing the effect of these two types of noises. I am also going to decrease the gain of the amplifier as Channeltron has high gain capability in itself.

Amplifier tested in our lab worked fine but all these problems creeped out at user's place.