Direct Conversion Receiver

"Randy Yates" <yates@digitalsignallabs.com> wrote:


That must be a learning exercise with FPGAs.
Direct conversion or direct sampling receivers are bound to poor 
performance.

> I'm considering the TI ADC08D502. The SFDR (about -60 dB) actually looks
> to be well below the quantization noise floor.

1. SFDR of -60dB is way above 8 bit quantization floor.
2. Quantization noise is not a noise but nonlinear distortion.
3. SFDR of -60dB is extremely poor performance figure for wideband receiver.

Vladimir Vassilevsky
DSP and Mixed Signal Consultant
www.abvolt.com

>> anti-aliasing requirements

that would be the least of my concerns, run a fixed-frequency NCO to shift
the whole band near 0 Hz still at the high rate, then lowpass -/ decimate
in multiple stages (that said, I've never actually implemented this, and
there may be better ways)

no, I think getting the RF signal through the ADC is the hard part, unless
the radio requirements are very easy.
The main difficulty in receiver design is that I have to be prepared for
any likely combination of unknown interferers. 

-markus

>> I didn't mean just stick the antenna into the ADC. 
>> I presumed you would have some sort of "sufficient" preselector filter.
..
>>Then possibly a signal conditioning amp prior to the ADC (after the
preselector).
OK, sounds like a plan

>>Did you mean ADC?
yes, that's what I meant...

>> and the preamplifier makes it even worse. With a wide-band frontend, 
that can cause inter- or crossmodulation products.
>> Isn't that true for either DC or superhet topology? One needs a preamp
either way.
yes, it should be no different. Maybe I created the problem here myself by
assuming you want to do also the preselect filter digitally.

>> Your clock and timing uncertainty need to be as clean as a local
oscillator
>> in a conventional receiver. This alone often rules out the idea.
>Why? Why not make the ADC clock as clean as an LO?

Maybe it's no problem at 200 MHz. Just keep an eye on it that it doesn't
become one, especially if strong blockers are possible at high frequencies
(the derivative of the waveform is proportional to frequency, that means so
is the waveform error from a given sample timing deviation => error can
scale 20 dB / decade with frequency).

Randy Yates <yates@digitalsignallabs.com> wrote:
> I am contemplating a direct conversion receiver architecture
> (versus a single-conversion superhet) but have heard of issues
> in such designs.
 
(snip)

> My contemplated architecture will use a fast ADC at the front end to
> capture the RF spectrum from DC to my highest frequency of interest,
> then filter and decimate in an FPGA down to a usable baseband bandwidth.
> In other words, no mixers! So I won't have to worry about images.

I had wondered in this group some time ago about a radio receiver
that would receive all stations at once.

For AM, the band ks 540kHz to 1600kHz, so the whole thing is
less than 2MHz. It should easily be possible to filter and sample
to get the whole band into an ADC, then digitally filter and
demodulate the 100 or so possible signals. 

The problem, though, is that some will be at a higher power
than others, and you have to AGC to the highest one. You need
enough ADC bits to get even the lowest one out.

For FM, it is a little different. From 88MHz to 108MHz, so 20MHz wide.
Might be better to mix down first, but there might be affordable
ADCs that could do it. You still want to filter out some of the
lower frequencies that might be very high power, though.

Then filter and demodulate the 100 FM signals. Well, you probably
don't have all 100 anywhere, so you can get by with fewer filters
and demodulators, but all in the digital side. 

Then, still in the digital domain, convert to MP3 and write the
data stream out, maybe to a disk drive. 

Is that at all what you were thinking about?

-- glen

On Tue, 06 Nov 2012 12:23:30 -0500, Randy Yates
<yates@digitalsignallabs.com> wrote:

>eric.jacobsen@ieee.org (Eric Jacobsen) writes:
>
>> On Tue, 06 Nov 2012 11:46:11 -0500, Randy Yates
>> <yates@digitalsignallabs.com> wrote:
>>
>>>I am contemplating a direct conversion receiver architecture
>>>(versus a single-conversion superhet) but have heard of issues
>>>in such designs.
>>>
>>>  1. They are susceptible to microphonics 
>>>
>>>     http://www.arrl.org/files/file/Technology/tis/info/pdf/9208019.pdf
>>>
>>>  Why?
>>
>> That depends on the rf implementation.   Designs I've worked with
>> didn't have this problem.
>>
>>>  2. They are susceptible to DC offset. 
>>>
>>>  Why?
>>
>> Because any LO bleed-through winds up at DC as a static DC offset in a
>> coherent system.
>>
>>>My contemplated architecture will use a fast ADC at the front end to
>>>capture the RF spectrum from DC to my highest frequency of interest,
>>>then filter and decimate in an FPGA down to a usable baseband bandwidth.
>>>In other words, no mixers! So I won't have to worry about images.
>>>
>>>Any thoughts regarding this? 
>>>-- 
>>>Randy Yates
>>>Digital Signal Labs
>>>http://www.digitalsignallabs.com
>>
>> What you just described isn't a direct-conversion system, but a
>> sampled-IF or super-Nyquist sampled-IF system.  
>
>Hi Eric,
>
>Thanks for the feedback. 
>
>What IF are you talking about? There is no IF in what I described. 

If there's no AFE/ODU or whatever, then perhaps it's sampled RF.

>> If you do a mix to baseband digitally, or there is no mix to baseband
>> and you demodulate a bandpass signal, you won't have the DC offset
>> problem.
>
>The latter. I was planning on complex bandpass filtering at the high
>sample rate, then decimating. I'm choosing the initial (high) sample
>rate so that my band lies centered at DC after decimation. So, no mixer!
>-- 
>Randy Yates
>Digital Signal Labs
>http://www.digitalsignallabs.com

Not sure how you wind up with a real-valued input centered at DC
strictly by sampling, unless it's done coherently and the spectrum is
already symmetric.   If it's already complex-valued then it either
already had some processing done to it or it didn't come out of an
antenna.

If the real-valued input spectrum is not symmetric and you want it
centered at DC you need some way to generate the quadrature component.
This is usually done with a mixing operation of some kind or, rarely,
with a Hilbert transform.


Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

On 6.11.12 6:46 , Randy Yates wrote:
> I am contemplating a direct conversion receiver architecture
> (versus a single-conversion superhet) but have heard of issues
> in such designs.
>
>    1. They are susceptible to microphonics
>
>       http://www.arrl.org/files/file/Technology/tis/info/pdf/9208019.pdf
>
>    Why?
>
>    2. They are susceptible to DC offset.
>
>    Why?
>
> My contemplated architecture will use a fast ADC at the front end to
> capture the RF spectrum from DC to my highest frequency of interest,
> then filter and decimate in an FPGA down to a usable baseband bandwidth.
> In other words, no mixers! So I won't have to worry about images.
>
> Any thoughts regarding this?


The strong AM stations in the MF band quite probably will use
all the dynamic range your A/D has, and the desired signals
get buried in the AM spurs and birdies.

IMHO, you cannot (yet) make a wide-band receiver without
some kind of band-limiting ahead of the active components.

Some more points in an analog direct-conversion receiver:

  - In an image-reject system, it is very difficult to keep
    the channel balance in a wide frequency range,

  - The dirct-conversion system is basically a receiver
    with very low IF. The 1/f noise kills the sensitivity
    compared to a conventional superheterodyne.

-- 

Tauno Voipio

On Tuesday, November 6, 2012 11:28:01 AM UTC-8, Randy Yates wrote:
...
> 
> > ... it is much more efficient to
> > mix and then desample with a multistage low-pass implementation ...
> 
> Dale,
> 
> Thanks for your input. I hear your point here, but I have a question on
> one detail: did you mean "mix" in analog?

I meant 'mix' in digital: multiply the real samples from the ADC by [cos(wt) - i * sin(wt)] to produce a complex signal with the center frequency at DC.

> 
> > Also, when the center frequency and bandwidth change, your
> > architecture may require a new sampling frequency, desampling ratio
> > and filter design. That means reloading the FPGAs and supplying a
> > variable frequency clock instead of a fixed frequency clock to the
> > required accuracy and jitter specs.
> 
> That's not going to happen. Sorta like expecting GPS satellite
> frequencies to change.

They don't have to change, new frequencies can be added to enhance capabilities. There are a number of reasons to have systems that don't need a new precision clock source, but only a different value written to a register, if the requirements change.

> 
> > People have been designing mixer/desampler architectures for good
> > reasons.
> 
> Perhaps so, but I'm not going to hang on to coat tails, i.e., I want to
> see the issues and make the tradeoffs myself.
> 
> Randy Yates

That's why we are talking about them here.

Dale B. Dalrymple

"glen herrmannsfeldt" <gah@ugcs.caltech.edu> wrote:


> I had wondered in this group some time ago about a radio receiver
> that would receive all stations at once.
>
> For AM, the band ks 540kHz to 1600kHz, so the whole thing is
> less than 2MHz. It should easily be possible to filter and sample
> to get the whole band into an ADC, then digitally filter and
> demodulate the 100 or so possible signals.
> The problem, though, is that some will be at a higher power
> than others, and you have to AGC to the highest one. You need
> enough ADC bits to get even the lowest one out.

This is first order blocking problem. It won't be too bad as the lowest 
signal would be sufficiently dithered, so it could be pulled out with 
processing gain. What is worse, the nonlinear products of the stronger 
signals appear in band of the weaker signals; so there is no way to filter 
them out.

> For FM, it is a little different. From 88MHz to 108MHz, so 20MHz wide.
> Might be better to mix down first, but there might be affordable
> ADCs that could do it. You still want to filter out some of the
> lower frequencies that might be very high power, though.
>
> Then filter and demodulate the 100 FM signals. Well, you probably
> don't have all 100 anywhere, so you can get by with fewer filters
> and demodulators, but all in the digital side.
>
> Then, still in the digital domain, convert to MP3 and write the
> data stream out, maybe to a disk drive.

What is the point?  The 100 of single IC analog receivers would beat the 
100-channel digital monster by every parameter; except, perhaps, size.

Vladimir Vassilevsky
DSP and Mixed Signal Consultant
www.abvolt.com

"Tauno Voipio" <tauno.voipio@notused.fi.invalid> wrote:

> Some more points in an analog direct-conversion receiver:
>
>  - In an image-reject system, it is very difficult to keep
>    the channel balance in a wide frequency range,
>
>  - The dirct-conversion system is basically a receiver
>    with very low IF. The 1/f noise kills the sensitivity
>    compared to a conventional superheterodyne.

Agreed.

- The direct conversion receiver is hit by the 2nd order nonlinear products; 
whereas superhet is hit in the 3rd order. For everything else equal, dynamic 
range is going to be worse.
- Typical problem of direct conversion is receiving back your own LO 
modulated by power frequency.
- Typical problem: two direct conversion receivers nearby jamming each other 
with  LO leakage.

Vladimir Vassilevsky
DSP and Mixed Signal Consultant
www.abvolt.com

Vladimir Vassilevsky <nospam@nowhere.com> wrote:

(snip, I wrote)

>> For FM, it is a little different. From 88MHz to 108MHz, so 20MHz wide.
>> Might be better to mix down first, but there might be affordable
>> ADCs that could do it. You still want to filter out some of the
>> lower frequencies that might be very high power, though.

>> Then filter and demodulate the 100 FM signals. Well, you probably
>> don't have all 100 anywhere, so you can get by with fewer filters
>> and demodulators, but all in the digital side.

>> Then, still in the digital domain, convert to MP3 and write the
>> data stream out, maybe to a disk drive.
 
> What is the point?  The 100 of single IC analog receivers would beat the 
> 100-channel digital monster by every parameter; except, perhaps, size.

Well, size is sometimes important.

But OK, 100 analog receiver ICs, 100 ADCs, 

no, maybe 100 analog receivers, 100 to 1 analog MUX, 1 ADC, then
digital stream into 100 way MP3 encoder.

I haven't looked lately, what is available as single IC receivers?
All the way from FM RF input to stereo audio output?

Is there a data sheet PDF available for one?

But otherwise, it is size and the challenge of doing all 
the processing digitally, and in one IC.

-- glen