Intel builds 'can't be built' working digital RF transceiver chip
Analog out as radios enter SOCs
IDF 2012 Intel has developed a truly digital radio chip, a part that replaces the analog elements in today's radio frequency transmission and reception circuitry with digital equivalents.
The result of a ten-year research project, the digital transceiver is a 32nm part capable of delivering Wi-Fi. It's still at the experimental stage: it can only handle 2.4GHz 802.11g, not 802.11n or 5GHz operation, but as a proof of concept it shows what can be achieved.
"A lot of engineers and managers Intel said this couldn't be done, that components like digital frequency synthesisers will not work," said Intel CTO Justin Rattner.
The next stage is to integrate the radio directly into a processor die, and Intel Labs researchers have already working on Rosepoint, an experimental system-on-a-chip that combines an earlier version of the digital radio demo'd today with a pair of Atom CPU cores, memory controller, PCI Express control and other IO, have begun punching out 32nm test wafers.
Analog hasn't been totally eliminated - a few components still need to be rendered in analog circuitry. But by replacing the vast majority of analog elements with digital versions, the radio is able to take advantage of silicon process technology in a way analog can't. Intel is working on converting as many of the remaining analog elements as feasible, it said.
Digital circuitry scales neatly as chip fabrication processes shrink. That's not the case with analog. In fact, some analog components cease to operate efficiently - or even at all - when they are reduced in size.
Elements such a digital frequency synthesisers, phase modulators and ADCs, can shrink from 0.3mm² at 32nm to 0.04mm² at 14nm.
That might not matter if devices remain large. Analog radio systems aren't expensive, either. But as more hardware needs to be connected to networks - think of tiny sensors that run for years on a battery charge, streaming data throughout that time, or of wearable devices - that puts a premium on shrinking the radios they must contain.
Put radio into the CPU, and any device that processes data can be easily and cheaply internet connected.
And existing devices such as smartphones contain multiple radios. Digital RF is key to delivering Bluetooth, Wi-Fi, GSM, 3G HSPA and 4G LTE with a single, tunable radio unit. That will make for better battery life and more compact devices.
They will be better radios too, said Rattner. "We think this radio will out-perform best-in-class analog radios because it can automatically adjust for interference and self-callibrate in real time."
Intel didn't comment on power consumption and other potential disadvantages emerging when the digital radio is compared to analog kit, but Rattner said the efforts behind today's demo chip was focused on "getting it to work", not optimising it for power consumption and such.
However, he claimed the radio already offers full spectral purity for Wi-Fi and cellular communications. ®
They forgot to update Clark's first law .......
"When a distinguished but elderly scientist [/engineer/manager] states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong."
Re: They forgot to update Clark's first law .......
Including engineers I'd agree with. Not managers though, they are often wrong about what is possible.
Shooting at a target that moves faster than your bullit.
Intel has been working at making digital radios for more than the mentioned decade.
From a certain point of view, this makes perfect sense, but not so much for WiFi.
The strength of digital radios is that they can shift between radically different radio protocols simply by changing the SW settings.
This would be a great boon to mobile phones since they can then reconfigure to using whatever network and frequencies are available in a given area without the need to include a physical RF circuit for each and every protocol and frequency out there.
So in principle an extremely good idea from Intel.
The problem and the reason they have continuously failed to deliver (and the current implementation seems to be a bit short of a total success) is that they are aiming at a moving target.
The reason RF circuitry is (partially) implemented in analog circuitry is that analog chips are by definition faster than digital chips working on the same chipset technology (so the transistors are identical).
Digital chips need a generation or more of chipset technology development in order to do what can now be done in analog chips.
Radio frequencies are a scarse commodity, so everybody tries to use them as well as possible and to cram as much data into the available space as posssible.
This means that the RF standards are continuously being upgraded for higher speed and bandwidth as soon as there is an available technology than can handle the needed processing.
So digital radios are playing catchup to standards that are continously being upgraded to a level, where only analog radios can follow (yet).
What Intel needs in order to succeed is for RF development to stagnate and reach a plateau, where speeds can no longer be increased by using faster or more powerfull chipsets or where the speeds are simply good enough for most practical purposes.
When that happens, they have a chance of success
Re: Flexibility vs Consumption
Sampling speed is one thing, bit depth is the other.
For maximum SDR flexibility, you need to have a wide open front end (no pesky hardware filtering). This is required to meet the promises made by the SDR folks. So how many bits depth do you need to pull Radio Wingwong out of the noise while parked three blocks away from Radio Luxembourg? Without bothering to do the math, probably about 32 bits. Maybe more. I know that 16 or 24 isn't going to cut the mustard.
Humans will get there. Just not until about 2030.
Re: Some work left to be done...
You're getting "integration" conflated with "software defined". Of course virtually anything can be integrated. My point was that these technologies (the ones I've listed) are not yet ready to be replaced with software. In contrast, the transceiver's IF strip can now be pure computer software to great benefit.
Listening to some of the idiot promises made by the 'Software Defined Radio' folks over the past decade is enough to make me reach for the nearest barf bag. I could tell you some horror stories of certain SDR projects, but I'm not going to.
Free advice to decision makers: SDR technology is applicable to the IF strip (not the entire communications system). This imposes practical limitations on the flexibility that can actually be offered by SDR. The NSA policies may impose additional limitations. End result is that the promises made are not realistic.