When we are taught about oscillators as newbie engineers, we are shown a variety of waveforms on an oscilloscope or in a textbook. This is a sine wave, they say, this is a sawtooth, this is a square wave, and so on. We’re taught to look at the lines on the screen as idealised, a square wave is truly square, and the transition from low to high voltage and back again is instantaneous.
In most cases this assumption is harmless. If we look into the subject a little deeper we learn that what seemed an instantaneous cliff-face is in fact a very steep slope, but when a circuit does its business in milliseconds there is usually no harm in ignoring a transition time measured in nanoseconds. The glue logic for your Arduino project can take its time.
Sometimes though, the rise time of a logic transition is important. The application that prompted this article was the measurement of oscilloscope bandwidth by looking at how quickly the ‘scope catches up with a pulse that exceeds its bandwidth, for example. When the instrument can happily measure the transition times of all your usual pulse generators, something out of the ordinary is called for. So it’s worth taking a look at the rise times you’d expect from everyday circuitry, examining a few techniques for generating rise times that are much faster.
That Was Considered Fast, Back In My Day
If you look at the data sheet for a typical transistor, you will find a section devoted to switching characteristics. Taking as an example the 2N3094 popular general purpose transistor, you’ll find it has a quoted maximum rise time of 35nS. Thus if you applied a perfect square transition to its base, the corresponding change at its collector would finish happening a maximum of 35 nS later. This might sound rather quick, but it corresponds to the rise time of a sine wave just over 7.14 MHz. Of course the 2N3904 is capable of working at much higher frequencies in small-signal mode, but if it has to traverse the entirety of its range you’re stuck at 7.14 MHz.
When it comes to faster transition times, you might expect our path to lead directly to components designed for square wave transitions, such as logic gates. But before we make that journey there is a surprising source of very fast rise times that’s not even electronic, it’s mechanical. The mercury-wetted relay is a type of reed relay in which the contacts are coated in mercury by capillary action. This produces an instantaneous contact, as the mechanism is that of liquid mercury droplets combining with each other rather than spring contacts touching. This contact time is well below a nanosecond, which means that the rest of the circuit around the relay and the voltage being switched governs the rise time, so extremely fast times can be achieved. We were fortunate enough to be able to borrow a mercury-wetted relay for this article, and when switching logic level into a 10 K resistor measured through an oscilloscope probe we were able to measure an impressive 4.6 nS rise time. This required some care with respect to lead lengths and ceramic decoupling capacitors to clean up and shorten the transition to this length, it is likely that further measures could shave some more time from this figure.
Logic gates are optimised for fast transitions, and should be correspondingly quicker than the 2N3904 we considered earlier. The archetypal logic gate family is of course the 74 series of TTL devices, of which there are many variants with ever-improving characteristics since the series first saw the light of day in the 1960s. We turned up the only original 74 series device we had to hand, a 7410 3-input NAND gate chip. Its data sheet quotes a typical low-to-high rise time of 11 nS, perhaps our device was one of the better ones as the ‘scope measured 7.1 nS. This corresponds to the rise time of a sine wave at about 35.2 MHz, but that figure is something of a theoretical upper maximum of the 7410’s performance envelope and the real usable figure would be rather less. Still, better than the 2N3904, but surely we can achieve more.
Casting around for higher speed 74 logic variants in a breadboard-friendly DIP package on the bench, the next up was a 74HC240 octal buffer. Around 20 years further up the technology ladder, and measured as having a rise time of 4.5 nS. Much better than the 7410, but still equal to the rise time of a sine wave at about 55 MHz.
More recent 74 series families offer improved rise times, but trawling through a lot of data sheets suggests that they still struggle to achieve significantly below 2 nS. As we move into the realm of picoseconds it’s obvious that we need something a little more special.
Speed Doesn’t Come Cheap… Or Does It?
Other devices optimised for very fast voltage transition aren’t hard to find, we’re used to using comparators to produce a quick logic level change based on the ratio of two analogue voltages. Of course, not all comparators are even in the class of the components above, the ubiquitous general-purpose LM139 and its derivatives for example have an almost leisurely 300 nS quoted transition time between TTL logic levels. But just as 74 logic has seen successive generations of technological improvement, so have comparators, and some of the more exotic devices leave the fastest 74 logic rise times in the dust. The ADCMP580 from Analog Devices for example is a SiGe emitter-coupled logic device that has a rise time of an astonishingly low 35 pS. There is a catch though: each chip will cost you around $18 and the evaluation board is just short of an eye-watering $300. That’s $8.30 per picosecond.
If you are seeking a picosecond-class rise time for the transition itself, as in our ‘scope bandwidth application, rather than for the timing while conveying some information, happily there is a much cheaper alternative. Avalanche breakdown is a phenomenon in which an insulator under an electric field can become conductive very rapidly indeed due to a chain reaction of accelerated free electrons dislodging more electrons. When applied to a transistor it can turn the device on much more rapidly than it would be when used in a conventional fashion, and it is this property that can be used to create a relaxation oscillator with an extremely fast pulse rise time. It is claimed a 2N3904 can achieve 500 pS rise times in this manner, something of an improvement on the 2N3904’s stock 35 nS mentioned above.
Of course, there is a catch with an avalanche pulse generator. The avalanche breakdown voltage of a bipolar transistor is quite hight, well over 100 V in the case of a 2N3904. Thus while the oscillator is simplicity itself, you are likely to need to also build some form of step-up power supply. If pushing boost converters beyond their usual boundaries is not yet your area of expertise, expect to plug that gap in your knowledge. It is, however, unusual to find a design that pushes the limits of what is possible in an area of electronics without resorting to exotic devices or special techniques, and as our rather messy prototype in the picture above shows this is a project that should be well within the abilities of many Hackaday readers. We’ll cover it in more detail in a future article as we examine its application, but meanwhile this piece would not be complete without a picture of it.
There is a satisfaction in achieving the fastest rise time, not dissimilar to that of achieving the most accurate frequency reference or atomic timepiece. We hope we’ve given you something of an introduction to some of the issues surrounding rise times in logic transition, and with the surprise that one of the fastest transitions can be achieved with components you are likely to have to hand then perhaps you’d like to have a go yourself.