From Qwiki

We build a lot of electronics in MabuchiLab. Most of these only need to operate at fairly low frequencies, and so their design and construction are fairly forgiving. We buy most of our high-frequency electronics from companies like Mini-Circuits. However, it has never been clear exactly where the transition between low and high frequencies lies --- just what speeds are we capable of reaching with the basic construction skills and tools available in MabuchiLab? One day, with lots of more important work to do, we decided to find out.

Contents 1 Factors affecting high-frequency performance 1.1 Stray capacitance 1.2 Line inductance 2 Construction techniques 2.1 Breadboard 2.2 Dead-bug 2.3 Circuit mill, leaded components 2.4 Circuit mill, surface mount components 2.5 Linear Tech style 3 Testing 4 Conclusion 5 References

Factors affecting high-frequency performance

First of all, why should it be hard to build high-frequency circuits? Operational amplifiers with unity gain bandwidths in the GHz range can be purchased for a few dollars from companies like Analog Devices and Linear Technology. Can't we just substitute one of them in place of our old favorite OP27 and instantly get huge bandwidths? No.

Stray capacitance

Every little piece of metal on your circuit board forms a capacitor with every other one. Therefore each resistor in your circuit forms a low-pass filter with each of these tiny capacitances. If you need to squeeze every last bit of bandwidth out of your circuit, you should do the following:

• Use small resistors, because the low-pass cutoff is as ω = (RC) ^{− 1}
• Avoid big conductors on your circuit board. This means cutting component leads as short as possible, using small pads and short leads, avoiding IC sockets, etc.

Very high-speed ICs will often come in small packages, with small leads, to minimize capacitances. Similarly, high-frequency chip resistors and capacitors are very small and do not have leads. At what frequencies do you need to use such specialized components? Read on...

Line inductance

Any current-carrying wire has some inductance because of the magnetic field induced around it. As a result, it resists high-frequency changes in the current it carries. This is another effect that is usually tiny, but it can become a major source of bandwidth limitation. These techniques may improve your circuit's performance by reducing line inductances:

• Use short leads! The inductance is proportional to the length of the lead.
• Use a ground plane. Current flowing from the power lines into the circuit travels along a similar path, but in the opposite direction, as current flowing from the circuit to ground. These two currents create magnetic fields that cancel each other to a certain degree, thus reducing the line inductance. This reduction improves as the ground plane gets closer to your power lines.
• Use bypass capacitors on amplifier power pins.

The degree to which you get serious about your ground plane can determine a lot about the way you construct your circuit, as you will see below. Is it worth it? You'll need to continue reading to find out.

Construction techniques

In MabuchiLab we typically use one of four different construction techniques, with our choice determined by balancing the demands and complexity of the circuit with our own laziness. Those are detailed here, along with a fifth technique introduced to us by the boys down at Linear Technology in their high-speed amplifier applications note (Reference 1).

Breadboard

The standard for simple circuits. A fiberglass board with a bunch of holes punched in it and metallic pads surrounding each hole, you can buy these at Radio Shack and make decent circuits with them. Note that when we say "breadboard" we never refer to the solderless kind. Those should only ever be used for testing!

Pros

• Easy
• Fast to build

Cons

• Not easily reproducible
• Not easy to modify
• Scales poorly: big circuits with many ICs and connections are annoying to build this way
• Big pads (stray capacitance...)
• No easy way to incorporate ground plane

Dead-bug

So-named because the op-amps look like dead bugs when their pins are bent outward. This technique is typically chosen by only the laziest MabuchiLab members, because of the general impression that it is kind of junky.

Pros

• Fast to build
• Looks crazy --- impress your friends!
• You can make extremely short leads by soldering directly to the op amp pins

Cons

• Extremely hard to modify --- you'd better make sure you choose the right amplifier!
• Not easily reproducible
• Scales even more poorly than breadboard
• No ground plane
• Looks crazy --- people might not take you seriously

Circuit mill, leaded components

Sometimes we use our rapid-prototyping machine to make circuit boards. The machine drills holes and cuts isolation channels out of a fiberglass board with copper surface layers, creating a custom circuit board that simulates a professional PCB. This method is usually chosen for complicated circuits with many components (because it cuts all the wires for you) or when higher-frequencies (say, 100kHz to 10MHz) are needed (because it is expected to outperform the above methods).

Pros

• Easy to make complicated circuits
• Easy to change components or modify connections
• Easily reproducible: can cut additional boards or send layout files to have PCBs printed
• Built-in ground plane

Cons

• Slow due to large overhead (layout files must be created, board must be cut)
• Big pads (stray capacitance...)

Circuit mill, surface mount components

For extreme high-bandwidth circuits, the rapid-prototyping machine can be used to create boards that use surface-mount ICs and chip resistors and capacitors. In MabuchiLab, this is almost only used for circuits that need to be very fast (over 10MHz).

Pros

• Stray capacitances kept very small
• Built-in ground plane
• Easily reproducible

Cons

• Extremely annoying to solder. These components were meant to be installed with a reflow oven, and we don't have one of those. Instead we use a hot air soldering station, which is unpleasant at best.
• Large overhead due to construction of layout files

Linear Tech style

Basically a dead bug circuit with a ground plane, taken to the extreme. All components are placed as close as possible to the ground plane, everything is soldered directly to the op-amp pins, BNC cable connectors are soldered to the board, etc. Nobody in MabuchiLab constructs any circuits using this method because it is impractical for circuits that will be used regularly (although I guess you could put it in a box to improve this).

Pros

• Fast to build
• Short leads
• Ground plane
• Looks crazy --- impress your friends!

Cons

• Extremely hard to modify --- you'd better make sure you choose the right amplifier!
• Not easily reproducible
• Scales even more poorly than breadboard
• Looks crazy --- no-one will take you seriously
• Completely impractical

Testing

To quantify the differences between these construction techniques, we built a very simple amplifer using each method and measured their frequency responses.

The circuit was a simple non-inverting AC-coupled unity-gain amplifier. This circuit is basically useless, but its simplicity should allow us to reach the full bandwidth of the IC we used, the AD9631 (-3dB at 320MHz).

Note that the input coupling resistor network was chosen to give approximately 50 ohm termination. Here are the results, measured with our 500MHz Agilent network analyzer:

As you can see, there are big differences between the construction techniques!

• As expected, the breadboard gives the worst results; however, it still has useable bandwidth out to 10MHz! This was much faster than we expected.

The remaining amplifiers all extend out to 100MHz, but there is peaking in the responses of most of them. This is probably due to phase lag in the feedback loop due to stray capacitance and inductance.

• In fourth place is the amplifier with the largest resonant peak, the one made with the circuit mill and axial components. I guess the big pads contribute a little too much capacitance.

• In third place is the dead-bug circuit. So much for looking junky! This method produced a decent, fast amplifier.

• Second and first place come down to the judges. Junkiest-looking of all, the Linear Tech style amplifier produces slightly flatter response out to 200MHz due to a very slight peak peak at about 150MHz. However, it falls off faster after that than the surface mount amplifier, and the two ultimately have the same -3dB bandwidth of about 400MHz. Due to such similarity in responses, I'm going to have to give the first place award to the Linear Tech style amplifier, simply for its good looks. You might decide otherwise, but either way you'll be choosing a very impressive amplifier.

Conclusion

We now have real numbers to use as guidelines for the frequency ranges of different construction methods. When it comes to producing high-frequency amplifiers, we recommend the following guidelines:

• For anything less than 1MHz, or maybe even 10MHz, use whatever technique you feel like using.
• For less than 50MHz, use either dead bug or circuit mill. Don't bother with the faster stuff because it's harder to build.
• For anything faster than 50MHz, use Linear Tech style (modified to make it useable) or surface mount.

Note: we did not compare the noise properties of any of these methods, and it's quite possible that some will be better than others in this respect.

References

1. High-speed amplifier techniques, Jim Williams, Linear Technology Application Note 47, August 1991.
2. Op Amps for Everyone Design Guide, Ron Mancini, Texas Instruments, August 2002.