Hi guys
I havent working with RF for ages, but am using it now. I know about "classic" radios, with coils for filters and OSCs. Now, with SOC (system on chip) I dont need to know about it, but I still wonder - I can make an oscillator or PLL, not problem, but say the IF and filters on chip? What replaces the coil-filter in an SOC?
Maybe they use a surface acoustic wave filter. The ISSCC papers tell those trade secrets. International Solid State Circuits Conference.
Direct conversion with DSP reception and demodulation at the receive frequency. There is no IF. A/D converter digitizes the RF input and DSP does the filtering and demodulation. The only gotcha is an added analog AGC stage in the front end to maximize the A/D dynamic range. If the demodulator is programmable, you get an SDR (software defined radio).
Take a look at the block diagrams and descriptions at:
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
Pretty sure the Silabs parts are not quite direct-conversion - I think they convert down to a low IF around 45KHz, and digitise I/Q to a DSP from there. The DSP is programmable (I have an SSB patch for the Si4735, under NDA) but the interface is private and the code is encrypted, so only Silabs can write code for it.
I want to do an SDR in an ARM7 Cortex, I think that would be fun. Working at 50KSPS, you don't need too much CPU horsepower.
The idea is generally that one uses a minimal amount of analog electronics, digitizes the signal, and then does DSP. An AM radio could be a bandpass filter and an ADC and an FPGA and a DAC, something like that.
There are cost and quantization issues, so an AM radio is not literally a good candidate for DSP. But lots of signal processing works great inside an FPGA or SoC [1]
You can make a great DDS signal generator with an SoC and a DAC, for a superhet or I/Q zero-IF front-end.
[1] If you can actually buy an SoC. Our Altera distributor A**** seems reluctant to sell us Cyclone V chips. Our Xilinx distributor A**** is actually selling us away from the ZYNQ chips towards some weird Igloo things.-- John Larkin Highland Technology, Inc picosecond timing laser drivers and controllers jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Let me take a guess: "Arnet"? ;-)
Tim
-- Seven Transistor Labs Electrical Engineering Consultation Website: http://seventransistorlabs.com
Not quite. 2.45GHz+ chips have tuning and stuff. They do it with the help of many metal layers.
900MHz might be direct conversion, or need some support hardware (which might be as simple as a resonant antenna).Think multilayer PCB, but with thin aluminum, SiO2 insulator (hey, way better than FR-4), and a goopy semiconductor base always underneath everything. You make a micro stripline with one trace on the top layer, surrounded by a box formed by traces on each layer stacked vertically, firmly joined with vias. And probably a flood on the bottom layer, or heavy doping in the semi.
So, you can make modest transmission lines, and poor inductors (usually of the planar spiral type, with Q factors of 5-10, peaking at 5-15GHz). But it's good enough for tuning power amps and doing IF -- as far as I know.
As a previous reply suggests, ISSCC probably has some interesting stuff on current tech or research level designs. Maybe they do direct conversion a lot more than I imagine.
I've read articles about monolithic amps (tuned or wideband), where when you're pushing 200GHz on a slab of InP, you don't have much choice. But you also have more freedom in insulating vs. heavily doped semi, and you can do bizarre things like nonlinear schottky junction transmission lines for shock wave generation, which (AFAIK) is how Tek and Lecroy do their fastest sampling scopes. In addition to a little RF magic, shuffling around bands and sampling them, and running everything into a pile of massive FPGAs for heavy processing and buffering.
Tim
-- Seven Transistor Labs Electrical Engineering Consultation Website: http://seventransistorlabs.com
Den mandag den 13. juli 2015 kl. 15.59.53 UTC+2 skrev Sonnich Jensen:
-Lasse
The SiLabs chips I mentioned are made for AM/FM/SW, which have a maximum RF frequency of about 110 MHz. Certainly 2.4/5 GHz chips will benefit from some front end filtering, probably done in the manner you describe. At lower frequencies, it's usually an external inductor and internal varicap tracking the receive frequency for AM/SW and no LC filter for FM. See typical block diagram at:
The various wi-fi chips haven't had an LO/Mixer/IF since the original Lucent/Orinoco and Raylink 802.11b chips around 1995. Everything has been direct conversion since then. Pick a block diagram, any block diagram: For example, here's some Broadcom chip: which even says "direct conversion radio" for both 2.4 and 5Ghz.
Also, at 900 MHz, nobody uses an antenna with sufficiently high Q to be considered useful for resonant. Mostly, they're coaxial sleeve antennas, which have a very wide bandwidth. Here's the insides of a
2.4GHz rubber ducky antenna: The 900 MHz equivalent is more of the same, just larger.I don't know of any that are built like that. The aluminum or ceramic hybrids are usually reserved for circuits that need to get rid of lots of heat. If you look inside a commodity wi-fi router, there's usually a shielded radio "module". Looking at a few of these in my junk pile, they all seem to use FR4/G10 PCB material. Cheap is the driving force. Looking at my box of assorted mini-PCI and assorted laptop wireless cards, they're also on FR4/G10.
Yep. That's commonly done in transmitters. Here's a typical VHF 25 watt power amplifier module: However, you've drifted well off the topic. The OP wanted to know where the mixer and IF amplifier went with an SOC (which I assumed meant an all digital direct conversion receiver), not technology more appropriate to transmitters.
One is more likely to find examples of all digital direct conversion receivers in the IEEE Transactions on Communications (TCOM):
Y'er way over my head on all that. I'll pass.
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
Yeah, their identical disinterest in selling parts suggests that they recently merged.
I can buy a 7010 ZINQ on a board with Ethernet, DRAM, Flash, USB, power supplies, connectors, Linux, all that, for not a lot more than Arnet wants for the chip alone. Their sales guy mumbled about $15 each but the quote came in over 4x that.
The Cyclone SoCs are apparently made to order, with long lead times. There seem to be no parts that will be distributor stock.
We'll do something else for now and revisit the SoC situation in a year or so, when things might have settled down.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
The inductors that can be made on silicon chips usually have a Q less than 20, so they are ok for oscillators (better phase noise than a RC or ring oscillator) and useful in RF filtering and matching, but not good enough for narrow IF filters.
Therefore they usually use direct conversion or low-if. In direct conversion, they mix the incoming RF down to baseband using two mixers, driven with LO signals 90 degrees apart in phase. A 5MHz wide channel centred at 2.15GHz would be mixed down produce I and Q baseband signals each extending up to 2.5MHz, since the centre of the RF channel ends up at DC. Then an active RC filter with maybe 5 to 9 poles is used to remove the strong adjacent channel signals (which might otherwise be
40dB stronger than the wanted channel) and then the signal is digitised and digitally filtered.Some people think that the incoming RF signal is directly digitised right at the antenna. Whilst this is a simple idea that appeals to programmers and academics, it is not a very good approach where good performance and economical manufacture are of interest. For example, a cellphone will happily receive signals of roughly -110dBm at the same time as unwanted signals of maybe -15dBm, meaning that one would require >16 bits of linearity in the ADC to capture this dynamic range without clipping or distortion products that swamp the wanted signal, and at channel frequencies of 2.2GHz, to avoid aliasing that would require a
4.4GHz+ sample rate. Nobody makes such a converter, and if they did, it would cost as much as a car and would probably need a car battery to get a few hours battery life. Even if the ADC did not use a lot of power, the DSP to process several gigasamples per second with a narrow-band channel filter would use a lot of power. Using mixers and analogue filtering ahead of the ADC allows vastly superior performance at a vastly more acceptable price.=_NextPart_000_0029_01D0BDCA.2794EFE0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable
Ah, interesting.
Those amp triangles on the RF end there are probably the tuned things, = but beyond that, they just dive right into it, eh?
Nonono =E2=80=93 I don=E2=80=99t mean using PCB, or hybrid. I mean on = the chip itself. The SiO2 and metal layers over semiconductor. Microns = worth. Like this:
Ah, that=E2=80=99s neat too. Hmm, interesting, that coupling capacitor = in the middle there, it=E2=80=99s got a =E2=80=9Ctuning fork=E2=80=9D = extending below it. Just another tweak I=E2=80=99m sure.
Anyway, those are certainly an option for higher power =E2=80=93 point = being, it works monolithically (usually at >...wave generation, which (AFAIK) is how Tek and Lecroy do their = fastest=20
bands=20
for=20
There=E2=80=99s a whole video about it on the Signal Path Blog, if you = don=E2=80=99t mind a little YouTubing and have a passing interest. The = insides are something to gawk at, if nothing else! (They tear down a = Lecroy something or other.)
Tim
Direct conversion means that the receiver converts the RF frequency USING A MIXER, directly to baseband, and after some analogue filtering, the I and Q baseband signals then go into the ADCs.
Tom Lee's book "The Design of CMOS Radio-Frequency Integrated Circuits" is a good reference on this stuff. Very readable and still mostly relevant.
Here is my ISSCC paper on a direct conversion transmitter:
You still have the problem with the I and Q mixer IP3 (third order intercept point). To have a decent IP3, say +10 to +30 dBm, you are going to need a lot of DC power, which you do not have in a portable device.
In addition, if you feed the mixer with square wave LO, you should watch for the mixing products with the LO odd harmonics (mainly the
3rd). So you should use a bandpass filter in front of th I/Q mixer (no wider than an octave) and preferably much narrower, if strong signals (such as cell phone or TV-transmissions) to attenuate them and hence avoid mixer overload.For direct conversion, two strong signals with distance less than your base band width but possibly at a relatively long distance from your wanted signal, will also create 2nd order intermodulation.
If the mixers are saturated by strong unwanted signals or intermodulation products fall into your base band, no post mixer base band low pass filters ahead of the ADC will help anymore.
These days it is possible to build mixers with good enough linearity and modest power consumption. They are used in every cellphone. (The direct conversion cellular radio is not the reason for the poor battery life of smartphones, many of the older phones that run for a week on one charge were also direct conversion.) A colleague made an LNA with some on-chip selectivity and and mixers with good enough linearity that it was possible to eliminate the receiver SAW filters in UMTS receivers around
2008:True. Square-wave LO is needed in order to achieve reasonable noise and linearity performance in the mixer. It is therefore necessary to thoroughly filter out interferers at the third harmonic of the wanted channel, before they can get to the mixer. This is not too hard to do off-chip because the frequencies are well-separated (by several gigahertz!) so no SAW filter is needed for that purpose. Often there is a duplexer between the antenna and the receiver, which can perform some filtering. Other interferers closer to the wanted frequency may need to be removed using a SAW filter, but with modern processes and good circuit design it is often possible to make the mixer so linear that SAW filters are not needed.
A more challenging frequency range is presented to TV tuners, as they must cover a range of wanted frequencies spanning more than a 3:1 ratio. I suspect that many of these are not really direct conversion, though I have not worked on them myself. Perhaps they up-convert to a very high IF and then convert back down again, like a spectrum analyser.
Yes, the mixers need to be very linear and well balanced, and you also need to prevent incoming RF signals from getting coupled into the LO path and then mixing with other unwanted signals to produce products that fall within the frequency range of wanted baseband signals. For this reason it is unwise to run the VCO at the RF centre frequency. It is better to at least run the VCO at a multiple and divide down. Sometimes even more elaborate schemes are needed to prevent interactions.
Yes, good circuit design and good LO chain frequency planning is necessary. Those problems are difficult, but at least they can be solved well enough on-chip.
My post was mostly to point out that "direct conversion" doesn't mean connecting the antenna directly to an ADC, as it seemed Jeff was suggesting: >>> The various wi-fi chips haven't had an LO/Mixer/IF since >>> the original Lucent/Orinoco and Raylink 802.11b chips around >>> 1995. Everything has been direct conversion since then.
Chris
Direct conversion for the 87.5 - 108 MHz FM band could be implemented as
There are multiple ways that this functionality can be implemented.
My apologies for screwing up my explanation. It's wrong. I'm not sure what happened, but somehow I got the erroneous idea that the wi-fi chips did not downconvert to baseband (despite having screens full of block diagrams, all showing the mixer and baseband I/Q demodulator. Also, thank you for being diplomatic with your comments, even though I don't think I deserve it for this bad a mistake.
I'm not really into designing the IC's but more into using them in consumer products. However, it looks like I'll probably need to do some reading to prevent further mistakes. Most of what I do in RF are design reviews, antennas, and oddly, cleaning up products that don't quite work or can't be easily built.
Thanks. I'll give it a read after I'm done sulking and beating my head against the wall.
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
Actually I thought I was being a bit snarky but thanks anyway!
Tom Lee's book is really quite entertaining and it starts off with the history of radio. I recommend it even if you never expect to do any chip design. It is the required text for a lot of degree courses so you sometimes find old editions cheaply second hand, in which case it is well worth it.
It's not all that relevant any more. It was the only (very complicated) way I could find to get rid of the SAW filter after the transmitter, and the tricks became unnecessary a couple of process generations later when the jitter of the dividers and LO buffers improved such that a much simpler scheme works now. It was fun getting it to work though.
Chris
Abebooks.com has several copies of the second edition for about $17. I bought one because I'm a sucker for early radio. For years, I've wanted to build an optical superregen out of a diode laser, but it's never been quite the right solution to the problem at hand.
Superregens are magic too.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
On 16/07/2015 15:35, Phil Hobbs wrote:>
Amen to that! Thanks to something you posted some time ago I bought myself a copy of the Whitehead book on Super-regens.
Did you notice how I squeezed a reference to super-regens into that thread about square-rooters a week or two back?
piglet
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.