# Why do we have a limited bandwidth in electronics?

I've been looking around a bit on optical communications and it is said that the main advantage in using optical communications, namely optical fibers, lies in the gigantic bandwidth available, when compared to the few GHz that electronics can manage. My question is: what is the physical phenomena that is limiting the electronics bandwidth? In light I can see it to be the optical properties of the waveguides such as absortion, and dispersive mediums, but what about in electronics? Specially in current nanoscale electronics like transistors.

• You asked about communications but then you said nanoscale...two different things entirely. Sending data around the globe is not nanoscale, and there's the rub. Mar 26 at 13:44
• Oh, you're right, I was thinking about communication between electronic components, in data processing. I'll correct it, thanks. Also, am I confusing things? Mar 26 at 13:46
• I guess it is the electronics that sends those signals into a fiber, that you call it gigantic bandwidth. Mar 26 at 14:02

I feel like there are almost two questions here

1. What limits the bandwidth of electronics? (Let's focus on solid-state electronics)

2. Why do optical waveguides provide such large bandwidth?

Side Note: Comparing the bandwidth of electronics and that of optical fiber is somewhat odd to me. I say this because electronics, such as LEDs and LASER Diodes, are frequently used to send signals down optical fibers. It would be more common to compare optical fiber to other options for carrying electromagnetic waves such as Coaxial cable, twisted pair, metallic waveguide, free-space etc. I touch on this somewhat in point 2 below.

For 1. you can spend a long time reading up on the semiconductor physics behind this but switching on and off a transistor, LED or LASER diode is not an instantaneous process. These structures have capacitance which means there are voltages that cannot change instantaneously. This limits the rise and fall time of signals. The slower the envelope of a signal changes, the smaller its bandwidth is.

You correctly noted that dispersion will limit the bandwidth in an optical fiber. Chromatic dispersion for instance means that you can not use too broad a range of wavelengths as the components will travel at different speeds and you will get Inter-symbol interference on your channel. It's similar with say a transistor driving a length of coax, the capacitance and inductance of the circuit will result in different phase shifts for different frequency components, and therefore transmitted symbols will broaden in time and interfere with one another.

For 2. Ultimately, it is all electromagnetic waves and governed by the same fundamental equations. However, there is substantially more bandwidth available at optical frequencies than say microwave frequencies. A given device structure may support a 10% 3db-bandwidth, but 10% of 1000THz is far larger than 10% of 10GHz.

• Laser may be an acronym, but it's never written in all caps like that. Mar 27 at 3:23
• @Hearth It used to be, particularly in the 1960s when they were a new thing, but it has certainly fallen out of favour since then.
– J...
Mar 27 at 10:57
• Chromatic dispersion is not a limitation to modern coherent optical communication, which is how all long-range optical comm in fibers is done nowadays­. Rather, it's optical nonlinearities, basic noise considerations, and the bandwidth of the underlying electronics processing the signals to be transmitted through fiber. Mar 27 at 17:13
• @Orhym It's not a limitation, but it is a problem that requires special optics to correct. Telecommunication lines do require dispersion compensating etalons to maintain signal integrity over long distances.
– J...
Mar 28 at 12:09
• @J... It's done on the DSP side in coherent links nowadays. Mar 31 at 23:45

I've been looking around a bit on optical communications and it is said that the main advantage in using optical communications, namely optical fibers, lies in the gigantic bandwidth available, when compared to the few GHz that electronics can manage. My question is: what is the physical phenomena that is limiting the electronics bandwidth?

The gigantic bandwidth of optical waveguides is simply due to their high center frequency (~250 THz), since even a 1% fractional bandwidth at THz frequencies gives an enormous absolute bandwidth. Waveguides actually scale continuously from GHz to TH to PHz bandwidths, so there is nothing specific to optical communications here (and you can get much higher frequency waveguides then those used for optical communications), so really you are asking about the difference the bandwidth of waveguides and the bandwidth of active circuits like transistors.

Waveguides can have gigantic frequencies simply because all they do is contain the wave. To do that you simply need a material with a reflection coefficient above 0 or a refractive index above 1. That isn't very hard, you can buy mirrors that work at PHz frequencies, so you can build waveguides (at least in theory) at any frequency up until the point where the photon energy becomes so high you can't build reflective or refractive materials. That happens well into the PHz with X-rays.

Transistors are different. A transistor is an active device that must couple into a wave and then undergo an electronic transition. There are limits to how fast that can happen. Unlike photons, electrons have mass and so cannot move at the speed of light. If your transistor works by moving charge to or from a gate, there will be a delay while those charge carriers move. That delay limits bandwidth.

• I think the dielectric constant means electrons bump in waves always at the speed of light (yet crawl very slow) for the medium and for a given diffractive index photons also move at the speed of light for that medium, but the only speed difference is parasitic Miller Capacitance and diode RsCp time constant in semi's.... which is the basis of my answer Mar 27 at 14:04
• or the tunneling -ve ESR Cj time constant in "Back Diodes" referring to another answer up to 40 GHz but yes ECL aka CML is the best for logic Mar 27 at 14:11

The biggest physical effect that that hampers frequency is usually capacitance. On a PCB pico farads exist between planes, or in wires there is a small amount of capacitance. This coupled with resistance those found naturally in metals forms and RC circuit, which is a low-pass filter. Another parasitic that also limits frequency is inductance which blocks high frequencies, and is also found in all materials.

What designers have to do is design wave guides/ transmission lines that are matched to convey high frequency signals, this is done to design high speed communication lines and antennas. These transmission lines balance inductance capacitance and resistance be in wires and in PCBs

For example it's hard to get a trace to have a bandwidth with more than DC to 50 megahertz on a PCB without a transmission line.

In 1st order approximations -3dB BW is stated with a spectral density and related to the step response slew rate. When they do not correlate , it is due to either active current limits or higher order RC effects.

$$f_{-3dB}=0.35/t_{R} ~~\text{for 10 to 90%}$$

The C value is inversely related to RdsOn and size of MOS chips and Vds ratings while BJT’s have much lower Miller capacitance they also have an inverse Rce to Cm relationship while the RceCm or RdsOnCoss product is a figure of merit (FoM)but both depend on structural designs and substrates e.g. GaAs , Saphire, Diamond are better than Si for dielectric constant.

In put impedance of the configuration and GBW also play a role in signal bandwidth as trade offs as Common Base (CB) have better frequency response but much lower Rin= (Rb+rBE)/hFE

The most significant improvement in CMOS has been due to the laser lithography resolutions from xxxx nm to x nm demanding lower RdsOn, which was standardized for 4000 series CMOS at 300 Ohms @ 12V And 1k for low Vdd as I recall. Then 74HC00 family became 5.5V max and 50 Ohms +/-50% then 25%. Later 3.3 or 3.6V max technology improved speeds again with lithography to design standard RdsOn at 25 Ohms +/- xx % such as ARM chips and 74ALCVxx chips. Thus the lower the voltage rating and lower threshold voltages of CPU chips also reduced capacitance at the same RdsOn value and greater CPU speeds until it reach a limit on the process capability.

Then high energy processes for sub 10nm processes demanded lower impurity levels for ESD which also brought about lower yields but Taiwan Semi is still the leader in this field to achieve yet faster slew rates from the reduced capacitance.

Thus in high speed communication electronics, GaN, GaAs and other semiconductor materials are used to improve BW. There are far more exotic materials in physical electronics that are being researched to achieve terabyte speeds but not in production AFAIK.

Perhaps you can edit your question to illuminate us on optical nanoscale state of the art.

• Well, for example polysilicon emitter w/ SiGe-base BJTs (so not really exotic materials) have been around for a while now, and if you couple BJTs being inherently faster than MOSFETs and these optimizations, you get easily around hundreds of GHz and maybe even THz. And you can get extremely fast logics with ECL and these devices. The problem is of course scaling and large power consumption, making the Silicon CMOS process tough to beat.
– edmz
Mar 27 at 11:50
• "ohms" (not "Ohms") Mar 27 at 13:32
• Subscript (<sub></sub>) is the poor man's MathJax (though it does not work in comments). E.g. "V DS" and V DD. Mar 27 at 13:34
• TY @PeterMortensen $I_{am~still ~learning~~~~~~~~~~~~~~~~~~~~~~~~}$ feel free to edit Mar 27 at 13:59

Electronics still drives the optical devices so, asking about the electronics might be a bit of a red herring. The electrical limitation is in the copper cable (the predecessor of optical fibre). Most electrical cables start to have worsening transmission characteristics above (say) 100 MHz and, it is this limitation that you probably need to consider more than the actual silicon parts.

• But are the silicon parts limited in bandwidth? Mar 26 at 13:50
• Everything has a limit but, in comms, it is usually the cable that is the first to give up the ghost. Mar 26 at 13:54
• Yes, but silicon parts are also used to transmit and receive data over optical fibers, so silicon clearly is not the limiting factor. Or it is but it applies to both optical and electrical transmissions. Mar 26 at 13:54
• Precisely as I said above. Mar 26 at 13:55
• You're right that makes sense. But suppose now this scenario (and I know it's a bit far from the initial question perhaps) but say you replace the copper cable by an optical fiber network and you have to process the data coming from that channel. Now the silicon will be a limiting factor, right? If so by how much? Mar 26 at 13:59

The main advantage in using optical technology over semiconductor circuits is that it is independent from having any interference (cross talk and ground noise) which can easily pollute signals that are of electrical components using traditional circuitry.

So in many cases it relates to a quality-of-signal issue, and not a bandwidth issue. Fiber optics are considered optical electronics because signals usually come from a copper source – converted to optical, and then converted back to copper. These extra conversion steps can slow a signal down in terms of latency. This means that you maintain large bandwidth, but the transmission arrives in time at a delay. Therefore latency sensitive applications (such as real-time data communications) will stick with semiconductor components and not use the more expensive higher fidelity optical gear.

Also, bandwidth in terms of performance is going to be brokered at a cost. This means your devices that have limited bandwidth are throttled by the provider because they expect consumers to pay for performance. The limitations on bandwidth are a function of cost and not the equipment. Traditional semiconductor components might appear to have slower bandwidth than optical alternatives, but they actually can achieve better performance than one would expect.

Fiber optics are also used for applications such as long hauls (long distance connections), seen as the technology doesn't allow for signal degradation or attenuation. Fiber also won’t necessitate the need for so many repeaters and signal boosters.

Based on created market value, vendors distribute products at a certain price keeping bandwidth commodity cost prohibitive. So how much bang for the buck do costumers get? Many times high performance technology gets sold and at a cost that’s much less what it truly worth.

Typically consumer markets get the best deals, 50-100 dollars a month for a cell phone with an unlimited “high speed” Internet connection. This merchandise should honestly sell closer to \$1000 a month, for example. But in this ever so present communications era, the phone and data line is capable of transmitting data hundreds of times faster than what we’re used to. They only sell you a small piece of the pie.

Business class circuits are priced much higher and more closely to what these products actually cost. They usually provide a superior level of service (specialized switching and more security features). Telecommunications companies usually pay more to deploy and integrate these technologies than they would retrieve in sales. They end up selling a very under-valued product so that there is cash flow. This makes the economics work out in the long run.

These points made hold true to all markets of technology and electronics such as: audio equipment, computer technology, medical equipment, television, video games, and many other areas of interest.

I don’t know; you’ll need to speak with an economics expert to get more light shined on this subject matter. I’m just so happy and blessed that everything works as well as it does and I get to use it.

• Are you sure that optical has more delay than something like UTP? This reads more like a rant about economics than anything actually about bandwidth... Mar 27 at 3:21
• Yes this is typically the case. Recently optical technology have been able to fight latency issues by using proprietary technologies such as fiber channel, and fully optically wired components (not just the connection). Mar 27 at 3:30
• Tried to edit out the rants but I wasn't sure when to stop. Giving it a downvote instead. Nowhere does OP ask about the economics or internet speed.
– pipe
Mar 27 at 3:31
• I'm pretty sure propagation velocity in optical fiber is far faster than any common copper cables. And still, none of this is what the asker asked about--they wanted to know why things have limited bandwidth, which this answer doesn't seem to address at all. Mar 27 at 4:03
• I don't think we should feel sorry for companies no longer being able to take advantage of monopolies. They just need to learn to become more efficient. That is what competition is for. Mar 27 at 13:49