RE: [Discuss-gnuradio] IF ADC, and the radio itself

[Brian_Whitaker]

Dave and everyone,

We're worked pretty hard over here at Maxim, but I agree that it would make
an excellent show to have put together a slew of receiver front-ends for the
various SDR applications this team is working on. Its only been a day since
I asked about who here is working on the RF stuff, and I've heard from one
fellow here (Jamie Morken), but no one else. I can't make a career out of
working on this, and don't want to step on the toes of people who might be,
but I certainly want to get my hands dirty! Putting Maxim's name on at least
one of the radio designs will allow me to justify the use of my RF lab here
to support my efforts. Most of my experience is with cellular standards, so
I'd like to start there.

Dave's right in that we need to start planning what receivers to start
designing first, and make them as broad-ranging in application as possible.
So, I've got a few questions for the leaders of this pack:

(1) What SDR implementations are going to be ready first?
  It sounds like HDTV is on the way soon... but what cellular standards will
be available first?  AMPS seems like an obvious choice, because its just a
simple FM modulation, low BW, etc.

(2) Have people here already come up with some standard receiver
archiecture?  By this I mean, "everything will be direct-sampled at a 70MHz
IF" or the like. In particular, would I have the opportunity to use a
direct-conversion receiver (RF directly down to baseband)?

(3) Dave mentioned requiring a bandpass IF filter to knock down the images
in my proposed undersampling scheme... that's what the IF channel-selection
filter is all about. If we decide that we want to provide for a 50MHz
channel centered at 200MHz, then we set up the channel selection filter to
be just this (fc=200MHz, BW=50MHz), and sample the resulting If at no less
than 100MHz. Now if the channel was only 30kHz wide (AMPS, TDMA), then
nothing has to change, because the ADCs images are still outside the IF
filter BW. Does this sound reasonable?

Of key concern here is to model our architecture after widely-used
commercially available stuff. For analog TV, I think that's a 6MHz channel
on a 70MHz IF... and have a good enough ADC to under-sample this, say at
20MHz or so (2x6MHz plus lots of margin), so we can use  this for as many
different systems as possible -- anything that we can bring down from RF and
stuff in this 6MHz filter.  ...then we have another, more expensive radio,
that has a higher IF with channel select filters and ADCs set up for a much
wider channel -- whatever HDTV is, probably, and re-use that for 802.11b
(22MHz), 802.11a/g (110MHz or so), etc.

I know it seems like this RF stuff should be simple, but since the goal here
is to have a radio that is defined by what demod routine you run, we should
have as few different radio front-ends as possible.  It seems silly to have
to have a different USB/firewire box for every system you want to listen to.

Brian Whitaker
Maxim RF Applications

> ----------
> From:         David  Bengtson
> Reply To:     address@hidden
> Sent:         Monday, March 10, 2003 7:21 PM
> To:   address@hidden
> Cc:   address@hidden
> Subject:      Re: [Discuss-gnuradio] IF ADC, and the radio itself
> 
> On Mon, 10 Mar 2003 10:15:11 -0800, you wrote:
> 
> >All,
> >
> >Akshay mentioned Maxim ADCs for IF sampling -- here's a link to the
> >front-end of Maxim's parametric search page for high-speed ADCs. You can
> >either 'show all products' or drill-down to filter out only the ones with
> >full-power BW, ENOB, etc that you want. 
> >
> >http://para.maxim-ic.com:80/ss.asp?Fam=Fast_ADC&Tree=ADConverters&HP=ADCD
> ACR
> >ef.cfm&ln=
> >
> >Of particular interest are the MAX104/106/108 family... probably
> over-kill
> >for any real IF sampling, but they offer 600 to 1500 Msps (yes,
> >1.5giga-sample per second) with the analog front end to match this
> >performance (2.2GHz FullPower BW).
> >
> >Have any comm buffs out there laid down minimum required specs for things
> >like Spurious-Free-Dynamic-Range (SFDR), Signal-to-Noise-ratio+distortion
> >(SINAD), or effective-number-of-bits (ENOB) ?  The IF itself will tell us
> >what the input BW needs to be, and Nyquist  will tell us what the sample
> >rate needs to be.
> 
> 
> Part of the problem is that the front end spec's are undefined since
> modes and protocols are a bit undefined. it would probably be possible
> to come up with some band level specs, for a "broadband" (Think CDMA)
> and a "narrowband" (think GSM and AMPS, iDEN, LMRS). I'm not
> conversant with the various video modes. Coming up with a list of
> potential target protocols/modes is a starting point. 
> >
> >Just to clarify one point -- hardware folks here are expecting to
> >under-sample the IF, right, not try to grab _everything_ from DC to 500+
> >MHz?  There was a comment a while back that sounded like people were
> talking
> >about needing to sample a 300MHz IF at more than 600Msps -- remember that
> >Nyquist under-sampling will effectively mix down for us, so the sample
> rate
> >only needs to be 2x the actual BW used at IF, since that's the only info
> >we're interested in.  ...and this is assuming we're going to provide
> >ourselve's the opportunity to grab every channel in the band. Sorry if
> I'm
> >just stating the obvious.
> 
> Undersampling makes some of this easier. However, you still have to
> have a bandpass filter in from of the ADC to protect against sampling
> images. Grabbing everything in a band is a probably too
> computationally expensive for a 3 to 5 years. 
> 
> >
> >In this light, the MAX 1180 and 1190-series (there are about a dozen in
> >these families) all have input track/hold stages worth about 400MHz, with
> >sample rates anywhere from 10Msps to about 100Msps, all 10-bit res... see
> >that parametric search page for more details.  Most cellular handsets use
> >IFs in the range of 100MHz to 300MHz, and the widest receive band is only
> >60MHz (UMTS, US PCS), so just in these respects, these ADCs should be a
> good
> >match. Frankly, the target application for most of Maxim's high-speed
> ADCs
> >are as basestation IF sampling, so they ought to do well for cellular
> >applications.  
> >
> >For HDTV, the specs might be similar, since we're only grabbing a single
> >channel... can someone post what they know about HDTV's modulation
> scheme?
> >
> >A quick question about the radio itself -- most of the conversation here
> is
> >about the software implementation of the SDR, and very little about the
> >radio front-end (how we get from RF to IF).  What are people's plans here
> >for desiging the radio?  I'm a junior RF guy (only 5 years out of
> school),
> >but I'd like to contribute to this part of the effort if I can. 
> >
> >Brian Whitaker
> >Maxim RF Applications
> >
> 
> Brian, I think this would make a hella app not for Maxim.........
> 
> 
> 
> 
> Dave
> 
> 
> 
>