Today’s topic is about ”emerging” technologies… and yes, you will probably find that some of these topics have been around for ages. My point is however that they have got a kind of revival in today’s research community and if you want to publish something – why not aim for one of these ten. You might also find some of them even to be outdated and will never return. My opinion though is that they still leave room for new ideas when it comes to mixed-signal design.
A few of the bullets are closely related to eachother and you could probably argue that some of them are subtopics of eachother, but anyhow … I need to divide them somehow. And once again: these are supposed to be teasers for you and you have to do some googling on your own.
Sorry for a lengthy top ten list today – hope you won’t be too bored. And as usual, these are only ten of many and further on. Please comment and add your own list…
The list is not ordered, please enjoy:
- #10: Smart dust
Smart dust is back (again). Smart dust, autonomous sensor networks, internet-of-things, etc., etc., got its popularity in the early 1990s, and just a few days ago, even the Swedish newspapers highlighted a story of a one-square mm “microcomputer”. It was an implantable device to be deployed inside (!) the human eye. However, we have completely different applications to think of: surveillance, environment monitoring, smart buildings, trafic control, logistics, etc. Each little smart dust mote would be equipped with a sensor, a microcontroller, a radio interface to communicate and means to scavenge energy from its surroundings.
The main idea with smart dust is to distribute a very large number of sensor units over a certain area. Quite a few of these systems are event-driven and thus the sensors do not have to operate continuously, but could instead wait for a certain activity and once that happens the come to life.
The telecom company Ericsson and others, have proposed the 50 billion devices prohecy to happen quite soon. See for example:
How do we handle these immense amounts of units? How is data handled and communicated through the back-bone? Can all these units be handled in an environmentally friendly manner? Are they disposable as such, i.e., biologically decomposable? This leads us in to the next few bullets…
- #9: Bio-/Organic/Printed/Paper electronics
Ok, so sorry for bundling so many topics into one… I am probably not being fair to the researchers in these fields, but yet they are closely related.
In Norrköping, Sweden, Professor Magnus Berggren and his team at Linköping University have established a well-known research group within the field of organic and bioelectronics. Further on, surf into the Printed Electronics Arena (PEA) for more information on how e.g. Acreo and Linköping University, et al., have teamed up.
So, imagine, for example, massive volumes RFID tags where we want some some simple, mostly passive interaction. One could think of groceries of any kind: imagine very simple protocols such as the opportunity to program information about the grocery (weight, name, price, last date of consumption, etc., etc.) Not only could this information be scanned at the counter of the super market, but also at home by the fridge, etc.
Similar to bullet 2 the bioelectronics fields is getting more and more attention. Not only how to create electrical devices but also how to interface with biological circuits and signals. We have of course had the pacemaker for quite some year now: a low-power, implantable devices that “interacts” with the human body.
Anyway: these areas calls for new paradigms when it comes to design, different tools, different voltage levels, different characteristics.
- #8: Ultra low power/voltage
Lowest possible energy is a must to be able to even think of smart dust (bullet 1). Of course, a goal has always been to lower the power consumption since it improves battery life, reduces the nead for e.g. heat slugs, etc.
The theoretical mainframe is already there obviously. We are quite aware what kind of energy levels that are required to represent certain levels of information. For digital circuits it has for long been known that the kT noise will set a limit on the minimum energy required to represent a ‘1’ or a ‘0’ with high probability. Remember to also always talk about minimum energy rather than minimum power/voltage in your design. You always have a trade-off between speed and accuracy…
Also, as related to another bullet, with lower dimensions, the gain of the transistor decreases, other capacitors dominate in the transistor model. How do you design a wide-swing, 4000x gain amplifier? How much out of the box do you need to think?
In addition to saving power/energy/voltage you could make sure that you can recharge your batteries, see next bullet…
- #7: Energy harvesting
Alternatively, or more correctly in addition, to the previous bullet you could make sure that your circuits have a very efficient scheme to scavenge energy from the surrounding environment. One well-known technique is off course solar panels or in its simplest form, a photo diode. Other techniques would involve chemical harvesters, wind mills, or extracting the energy in the radio signals emitted from some kind of base station.
On the site http://www.energyharvesting.net/ you can find lots of more information on this.
Anyways, there are plenty of advances that can be done in conjunction with all the other bullets on this list. For standard CMOS and larger components/systems, the energy harvesting is ”pretty straight-forward”, but what happens in a handheld, miniaturized environment? Here, for example, the temperatures could vary between the extremes and large volumes of units are distributed and potentially never recovered (unlike mobile phones which – fingers crossed – ends up recycled properly). How do we design efficient harvesters for these purposes?
- #6: Memristors
Has anyone missed the HP Labs by-now famous article from Nature, 2008? The paper was titled “The missing memristor found” and referred to the memristor component ”introduced” by Leon Chua in 1971. Memristance was the ”missing component/link” in the mesh of electrical flux, current, voltage, and charge. To fill the gap, the memristor had to be added to the set of inductor, resistor, and capacitor.
Already in 1971 did demonstrate memristance in an active circuit by hooking up several OPs and transistors on a bread board. This large, bulky ”component” was then in 2008 (or so) successfully implemented by HP labs, but then in a 50x50x50 nm (!) large device.
The memristor could become a quite huge success. Due to its quite nonlinear behavior, it can be biased/operated such that it has a hystereses with respect to voltage and current. This hysteresis can be utilized to implement a small memory element. The read and write operations then become quite simple in classical crossbar configurations of any kind. Source current in one direction = write, sink in the other = read.
There are quite some complexities concerning the memristor. One of them is the lack of models (there are not too many design houses offering memristors in their PDKs yet…) and the other is that it is quite trick to implement, they cannot be too big, for example.
Also, the interfacing circuits may need some attention, since when reading the memristor you actually also destroy the information in it. How do we handle this?
- #5: Process scaling/quantum electronics/graphene
The 40-nm CMOS nodes are fairly well established these days. Whole analog front-ends for video can successfully be designed and operated at sub-volt supply voltages. Can it be done in 32 nm? Yes, probably…
There has been a constant “oohh, at the hmmm-hmmm-hmmm nm node it is impossible to do analog design”, etc., etc. Back in the old days (i.e., the 1990’s) they talked about micrometers, we could never go below 1 micrometer channel length. Well, we could… Of course there are limitations imposing new advances needed.
Leakage is very high in sub-90nm nodes and therefore the LP options are often found: thicker oxide – less leakage – but more voltage needed to recapture the decrease in speed due to larger gate-source capacitance.
Graphene has been put forward as a new promising material and the 2010 Nobel prize in Physics was awarded to research within this field. At Linköping University, Prof. Lars Hultman is heading the successful group research on this topic. Could this be the future?
Something more modest perhaps: how do we handle the fact that core transistors have really poor gain in the small geometries? In 65nm you have to be happy if you get more than 10 times gain in one single transistor. How can this be combined with high swing.
- #4: Putting the data converter directly on the antenna
Direct sampling ADCs and digital to RF converter or RF DACs are not really new topics either. The idea is straight forward: throw away all discrete components and throw away all analog filters (well, give or take, of course, …) then connect the antenna directly to the ADC/DAC.
The holy graal offers multistandard devices, or so called software-defined radios, where e.g. a mobile terminal should be able to switch between any communication of band of interest, say 2G (GSM), 3G (WCDMA), and 4G (LTE) or even bluetooth/WLAN or similar radio standards. Quite often you find in litterature direct-sampling ADCs, a common popular choice is the use of a subsampling sigma-delta ADC.
A direct-sampling ADC/DAC would require very high clock frequencies and a large dynamic range. This combination also impliest very low jitter on the sample and update clocks.
Hmm, so what is ADC ”directly on the antenna”? Surprisingly often they have an LNA there and eventhough passive band-select filters should be avoided they often have a filter anyway – wide enough to cover as many bands as possible. Maybe Dr. Bengt Jonsson could shed some more light on this – what do you say at Converter Passion? Anyone else?
Further on, for the DAC we have quite often a similar case: the power amplifier needs to be there to drive the signal. A brick-wall pulse-amplitude modulated DAC would give rise to quite high-powered out-of-band spuriouses – thus the filters are required… An all-ideal, sinc PAM DAC would do the trick, but they are kind of hard to get your hands on … Anyway, there is still plenty of room for advances also on the DAC side.
- #3: All-digital analog components
The all-digital ADC is already there – the all-digital DAC is almost already there too. However, their resolution times speed is moderate. With an all-digital DAC I am a thinking of some of the sigma-delta modulators with very high oversampling ratios, whereas all-digital ADCs are a bit more complicated, most of them being time-based, and some of the components are ”pseudo-digital”.
Let’s get a bit philosophical, or perhaps even naïve:
My short questions are, relating to the previous bullet too, how far are we from producing a system which interacts with the analog world, but does not contain any analog components? Can it be done? What are the theoretical limits? When can we use digital and when do we have to use analog? Are all decisions based on case-to-case or can we be more pragmatic and develop something that can be used for several things, like we talk about in the next bullet…
- #2: Reconfigurable microsystems/Cognitive radios
Multi-standard applications of any kind, as hinted in the previous bullets, open up for a couple of another requirements: reconfigurability. Area is expensive and so is time to market, it could sometimes be beneficial to have one system developed to handle many different tasks. This idea is of course also far from new, a computer is such a device – and a human.
A bit more technical: a recent dissertation from Linköping University illustrates how mobile phones could communicate directly with eachother in case of a failing base station. A point-to-point connection could be created using intermediate phones/terminals as relay stations.
Another popular field of research is cognitive radios. In short – dependent on load on the network, trafic can be redirected for optimum quality of service. The stations should also individually be able to make as clever decisions as possible to accomplish this.
In this bullet, I am thinking of more humble units, such as e.g. reconfigurable ADCs, reconfigurable filters, reconfigurable networks, etc., that would fit a similar purpose (or fit in the components needed for the above tasks) but on a smaller, microsystem level.
- #1: Cognitive electronics and biologically inspired architectures (BIAs)
Ok, so I am putting ”my own” topic here: cognitive electronics and biologically inspired architectures (BIAs). I should mention here that the term ”Cognitive Electronics” was used by Dr. Staffan Gustafsson first.
Remember the good old neural networks? Roughly the same, but a little bit different… The things we are referring to, when it comes to cognitive electronics and BIAs, are components that have more redundance, are not intended for high-speed, nor high-accuracy resolutions. Systems that are tolerant towards errors, both in terms of decisions but also due to failing components, etc. The systems also have a much higher degree of feedback connections than a traditional neural network does.
A couple of years ago we outlined a system for hearing aid devices where we at Cognicatus AB, suggested to model the phonemes and then use that information to improve the digital signal processing in the hearing aids. Now, at Linköping University, a group of researchers is using this approach in their research.