Help designing a 10MHz square wave oscillator with very short Tr/Tf

[narshben]

I want to generate a 10MHz square wave with Tr/Tf less than 2ns(or is this demand seems very odd? But it's really in need). The solutions all i could imagine now is:
1. use of a TCXO or the like, i checked several oscillator supplier's data sheet, the Tr/Tf is all around 10ns with a oscillating frequency of 10MHz.
2. use of a oscillator without tr/tf considered, then let the output go through a Schmidt Trigger or logic gate/Inverter ? I have little confidence about this method as i know little about the mechanism a signal passes the digital gate.
3. use a ring oscillator comprising crystal, fast inverters？
Could man have such expertise give me some tips/considerations in designing or other superior method？

 

[FvM]

You can achieve the intended rise time by connecting a fast logic gate after the oscillator. trf will be however strongly affected by capacitive load. What's the intended logic standard, what's the load impedance?

 

[narshben]

Actually, it will drive a capacitive load, a MOSFET. The capacitor value would be around 20pF, and i could accept a rising time adding up to 4ns.

 

[biff44]

pass it thru one of these: PO74G38072A

Potato Semiconductor / The GHz TTL/ CMOS IO Interface Logic/ Potato IC

 

[FvM]

If you assume e.g. 20 pF, 1 ns and 5V level, you get 100 mA required drive strength (respectively 80 mA, because 1 ns is for 10/90 percent, strictly spoken). Referring to standard high speed logic ICs, this ends up in paralleling multiple drivers. Don't use hex or quad ICs with only one supply terminal per package but single gate drivers or special bus drivers with multiple VCC and GND pins and effective low inductance supply bypassing.

 

[poorchava]

10 MHz is very high frequency for a mosfet. You need to consider the fact, that MOSFETs are minority carries devices, and speaking in general terms, the minority carriers are far less mobile than majority carriers. I think that you would need a special transistor for such a frequency.

Also, the power rating of the MOSFET is important. High power ones might have gate capacitances in the order of some nF. That kind of capacitance is hard to charge/discharge rapidly. As FvM pointed out, for a 1nF capacitance you would need 0.1A of drive capability, which is a really large value with so high frequency.

Other matter is that lots of medium and high power mosfets need voltage higher than 5V (around 7-8V) to completly turn on the transistor. So that depends on the transistor you choose.

And the last thing: if you succeed in creating a square wave of 10 MHz frequency and capable of 100 mA driuve, that kind of signal would generate HUGE EMI. Generally the higher frequency and higher currents result in higher EMI. If you're aiming at using that circuit in a place with other electronic devices, you may want to consider some shielding system to reduce EMI.

You may want to consider replacing MOSFET with a BJT (unless your design explicitly requires that transistor to be a MOSFET of course). You may use some king of darlington pair to minimize the current drive requirements. Also base-emitter junction capacitance of BJT's is usually an order of magnitude lower than gate capacitance of a MOSFET

 

[FvM]

10 MHz is very high frequency for a mosfet.

Narshben didn't characterize the MOSFET device further, except for telling about 20 pF gate capcitance. That's quite low for a MOSFET and apparently referring to a small area signal transistor. In so far, high speed won't be necessarily an issue. You should also consider, that some MOSFET types perform well at microwave frequencies.

 

[poorchava]

I know of FETs capable of microwave frequency operations, but these are mostly JFETs. And they they are usually made with different technologies, like GaAs and GaN which have far higher carrier mobility.

To be honest, I've not seen a FET with 20 pF gate capacitance. I've done my masters thesis on the topic of semiconductor junction and mosfet capacitance. I've emasured C-V curves of many semiconductor devices and i don;t remember a single mosfet that would go under 70 pF of gate capacitance. For BJT's and diodes it could even reach as low as single picofarads, but not in mosfets.

 

[mtwieg]

Small FETs can operate easily at microwave frequencies. A friend of mine recently made a circuit that drives a 2n7000 with a Tr of <1ns, which was driving (through a cascode) a 400V pulse.

It does generate incredible amounts of EMI though.

 

[poorchava]

Well, i can't argue with results. It's always nice to learn new things.

As for EMI i generally use an enclosure made of solid steel. You can buy those as shielding cases for mains transformers. I found that placing a noisy ciurcuit in one of those reduces EMI by a great amount. At prototyping stage a metal food can is also sufficient (like the one in which u can buy tomato soup or something). At higher frequencies the aluminum foil also helps.

 

[mtwieg]

I know of FETs capable of microwave frequency operations, but these are mostly JFETs. And they they are usually made with different technologies, like GaAs and GaN which have far higher carrier mobility.

To be honest, I've not seen a FET with 20 pF gate capacitance. I've done my masters thesis on the topic of semiconductor junction and mosfet capacitance. I've emasured C-V curves of many semiconductor devices and i don;t remember a single mosfet that would go under 70 pF of gate capacitance. For BJT's and diodes it could even reach as low as single picofarads, but not in mosfets.

FETs can certainly have gate capacitances below 20pF. Ones that small aren't generally useful in power or switching applications but rather in small signal amplifiers. For example I've used the BF998 in low noise amplifiers and I can verify that its input capacitance is no more than a couple pF.

 

[narshben]

Thanks your guys. As i have little knowledge about microwave devices, i don't know if the point "The BJTs have an input capacitance of one magnitude lower than MOSFETs" is true. In my case, as well as a low input capacitance, the Cgd or Crss is another important term to be considered. So as mtwieg mentioned, a cascode configuration should help.

 

[FvM]

i don't know if the point "The BJTs have an input capacitance of one magnitude lower than MOSFETs" is true.

In my view, it's not true for devices with similar output current.

As far as you specified the problem, fast standard IC buffers should be able to fulfill the requirements, as previously discussed. Custom designed circuits or exotic ICs come into play, where standard devices don't work.