Airspy architecture


Alberto I2PHD
 

A few naive questions, originated from my attempt to understand the architecture of the Airspy.

The tuner is an R820T2, which converts in the analog domain down to a low IF, just above 4 MHz.

First question : from the simplified block diagram of that chip, it doesn't seem to use a quadrature
mixer. That notwithstanding, the image rejection reported from its data sheet is 65 dBc... how is
that possible ?

A more detailed look shows that it has two IF outputs, pin 12 and pin 13, marked as IF N and IF P.
In the unconfirmed hypothesis that those outputs are actually I and Q, in other words the mixer
is as a matter of facts a quadrature mixer, then those two outputs must be digitized by the
12-bit, 80 MHz high speed ADC of the NXP LPC-4370 processor, which is a single ADC.
So it should work in a time interleaved fashion... is this correct ?

A bit of light on the Airspy architecture would be welcome, thanks.

--
73 Alberto I2PHD
Credo Ut Intelligam







This email has been checked for viruses by Avast antivirus software.
www.avast.com



Siegfried Jackstien
 

Hello alberto

I would guess the if n and if p means if negative and if positive (balanced
output) and NOT be I/Q signals

The r820t tuner does only have low if out and not 0hz if (i/q)

Hth

Dg9bfc

Sigi

-----Ursprüngliche Nachricht-----
Von: airspy@yahoogroups.co.uk [mailto:airspy@yahoogroups.co.uk]
Gesendet: Montag, 2. Februar 2015 11:26
An: airspy@yahoogroups.co.uk
Betreff: [airspy] Airspy architecture



A few naive questions, originated from my attempt to understand the
architecture of the Airspy.

The tuner is an R820T2, which converts in the analog domain down to a low
IF, just above 4 MHz.

First question : from the simplified block diagram of that chip, it
doesn't seem to use a quadrature
mixer. That notwithstanding, the image rejection reported from its data
sheet is 65 dBc... how is
that possible ?

A more detailed look shows that it has two IF outputs, pin 12 and pin 13,
marked as IF N and IF P.
In the unconfirmed hypothesis that those outputs are actually I and Q, in
other words the mixer
is as a matter of facts a quadrature mixer, then those two outputs must be
digitized by the
12-bit, 80 MHz high speed ADC of the NXP LPC-4370 processor, which is a
single ADC.
So it should work in a time interleaved fashion... is this correct ?

A bit of light on the Airspy architecture would be welcome, thanks.


--
73 Alberto I2PHD
Credo Ut Intelligam







________________________________

<http://www.avast.com/> This email has been checked for viruses by
Avast antivirus software.
www.avast.com <http://www.avast.com/>



Alberto I2PHD
 

On 2/2/2015 1:37 PM, 'siegfried jackstien' siegfried.jackstien@... [airspy] wrote:

I would guess the if n and if p means if negative and if positive (balanced
output) and NOT be I/Q signals

The r820t tuner does only have low if out and not 0hz if (i/q)
Yes, that was also my first guess, but the high image rejection made me wondering...
And in any case you can have I/Q outputs also if the IF is not at zero.

On 2/2/2015 3:45 PM, Leif Asbrink leif@... [airspy] wrote:

The R820T2 has I and Q internally and combines
them to a real valued signal with excellent image
suppression. 

The output is balanced to fit the balanced input 
of typical AD converters.

I have no idea how they can get such a good image
suppression...
Hi Leif,

� That explains a lot. From the simplified block diagram found on Internet the use
of a quadrature mixer is not apparent, and that made me scraping my head...

Yes, achieving 65 dB of image rejection working in the analog domain is really
a feat... kudos to them... everything must have been balanced with extraordinary
care...

Thanks Sigi and Leif.


--
73 Alberto I2PHD
Credo Ut Intelligam






This email has been checked for viruses by Avast antivirus software.
www.avast.com



Leif Asbrink
 

Hi Alberto,

The R820T2 has I and Q internally and combines
them to a real valued signal with excellent image
suppression.

The output is balanced to fit the balanced input
of typical AD converters.

I have no idea how they can get such a good image
suppression...

73

Leif




<airspy@yahoogroups.co.uk> wrote:



A few naive questions, originated from my attempt to understand the architecture of the Airspy.

The tuner is an R820T2, which converts in the analog domain down to a low IF, just above 4 MHz.

First question : from the simplified block diagram of that chip, it doesn't seem to use a quadrature
mixer. That notwithstanding, the image rejection reported from its data sheet is 65 dBc... how is
that possible ?

A more detailed look shows that it has two IF outputs, pin 12 and pin 13, marked as IF N and IF P.
In the unconfirmed hypothesis that those outputs are actually I and Q, in other words the mixer
is as a matter of facts a quadrature mixer, then those two outputs must be digitized by the
12-bit, 80 MHz high speed ADC of the NXP LPC-4370 processor, which is a single ADC.
So it should work in a time interleaved fashion... is this correct ?

A bit of light on the Airspy architecture would be welcome, thanks.

--
73 Alberto I2PHD
Credo Ut Intelligam






------------------------------------------------

This email has been checked for viruses by Avast antivirus software.
www.avast.com



Andrea IW0HDV
 

Hi Alberto, Leif,

  *am* iw0hdv


jdow
 

Using both I and Q outputs of a tuner into an SDR is a "Hail Mary" desperation ploy. Results with the E4000 show this. You get quadrature phase and amplitude errors that have to be balanced out, poorly in general. The AirSpy appears to need exceptionally clean in this regard using its sampler at twice the rate using every other sample as I or Q.

The mixer in the R820T is unknown. It is believed to be an image rejection mixer based on performance. Some versions of the rtlsdr.dll source code included code for fine tuning the balance. So far nobody seems to have managed to make it appear to do anything.

Call up the Rafael people and get a non-disclosure signed, probably concurrent with an order for a million chips, for more information.

{^_^}

On 2015-02-02 03:25, Alberto I2PHD i2phd@weaksignals.com [airspy] wrote:


A few naive questions, originated from my attempt to understand the architecture
of the Airspy.

The tuner is an R820T2, which converts in the analog domain down to a low IF,
just above 4 MHz.

First question : from the simplified block diagram of that chip, it doesn't seem
to use a quadrature
mixer. That notwithstanding, the image rejection reported from its data sheet is
65 dBc... how is
that possible ?

A more detailed look shows that it has two IF outputs, pin 12 and pin 13, marked
as IF N and IF P.
In the unconfirmed hypothesis that those outputs are actually I and Q, in other
words the mixer
is as a matter of facts a quadrature mixer, then those two outputs must be
digitized by the
12-bit, 80 MHz high speed ADC of the NXP LPC-4370 processor, which is a single ADC.
So it should work in a time interleaved fashion... is this correct ?

A bit of light on the Airspy architecture would be welcome, thanks.

--
/*73 Alberto I2PHD*
Credo Ut Intelligam




/


--------------------------------------------------------------------------------
<http://www.avast.com/>

This email has been checked for viruses by Avast antivirus software.
www.avast.com <http://www.avast.com/>





Alberto I2PHD
 

On 2/2/2015 10:45 PM, jdow jdow@... [airspy] wrote:

The mixer in the R820T is unknown. It is believed to be an image rejection mixer 
based on performance. Some versions of the rtlsdr.dll source code included code 
for fine tuning the balance. So far nobody seems to have managed to make it 
appear to do anything.

Call up the Rafael people and get a non-disclosure signed, probably concurrent 
with an order for a million chips, for more information.

No need for that... the link posted by Andrea is very interesting :

http://forums.radioreference.com/software-defined-radio/303443-airspy-comparisons-4.html

Reading that forum reveals many things that otherwise it would have been possible only to find
after having signed a NDA...


--
73 Alberto I2PHD
Credo Ut Intelligam







This email has been checked for viruses by Avast antivirus software.
www.avast.com



jdow
 

On 2015-02-02 06:45, Leif Asbrink leif@sm5bsz.com [airspy] wrote:
Hi Alberto,

The R820T2 has I and Q internally and combines
them to a real valued signal with excellent image
suppression.

The output is balanced to fit the balanced input
of typical AD converters.

I have no idea how they can get such a good image
suppression...

73

Leif




<airspy@yahoogroups.co.uk> wrote:



A few naive questions, originated from my attempt to understand the architecture of the Airspy.

The tuner is an R820T2, which converts in the analog domain down to a low IF, just above 4 MHz.

First question : from the simplified block diagram of that chip, it doesn't seem to use a quadrature
mixer. That notwithstanding, the image rejection reported from its data sheet is 65 dBc... how is
that possible ?

A more detailed look shows that it has two IF outputs, pin 12 and pin 13, marked as IF N and IF P.
In the unconfirmed hypothesis that those outputs are actually I and Q, in other words the mixer
is as a matter of facts a quadrature mixer, then those two outputs must be digitized by the
12-bit, 80 MHz high speed ADC of the NXP LPC-4370 processor, which is a single ADC.
So it should work in a time interleaved fashion... is this correct ?

A bit of light on the Airspy architecture would be welcome, thanks.

--
73 Alberto I2PHD
Credo Ut Intelligam






------------------------------------------------

This email has been checked for viruses by Avast antivirus software.
www.avast.com



------------------------------------
Posted by: Leif Asbrink <leif@sm5bsz.com>
------------------------------------


------------------------------------

Yahoo Groups Links



Marcus D. Leech <mleech@...>
 

On 02/02/2015 07:02 PM, jdow jdow@... [airspy] wrote:
 

Of course, I'm forced to point out that everything you can do in the analog domain in a weaver mixer you can do in the digital domain with an I/Q
  signal that may not be perfect.  Once you have an estimate of phase/amplitude imbalance, you can easily correct it digitally--any residual imbalance
  would be due to quantization error in both the estimates and the corrections.

Really, the only difference between the image properties in a "real" IF and a complex-direct-conversion is exactly where the images fall.  And with low-IF,
  they also tend to fall in-band.  TANSTAAFL :)



{^_^}

On 2015-02-02 06:45, Leif Asbrink leif@... [airspy] wrote:
> Hi Alberto,
>
> The R820T2 has I and Q internally and combines
> them to a real valued signal with excellent image
> suppression.
>
> The output is balanced to fit the balanced input
> of typical AD converters.
>
> I have no idea how they can get such a good image
> suppression...
>
> 73
>
> Leif
>
>
>
>
> wrote:
>
>>
>>
>> A few naive questions, originated from my attempt to understand the architecture of the Airspy.
>>
>> The tuner is an R820T2, which converts in the analog domain down to a low IF, just above 4 MHz.
>>
>> First question : from the simplified block diagram of that chip, it doesn't seem to use a quadrature
>> mixer. That notwithstanding, the image rejection reported from its data sheet is 65 dBc... how is
>> that possible ?
>>
>> A more detailed look shows that it has two IF outputs, pin 12 and pin 13, marked as IF N and IF P.
>> In the unconfirmed hypothesis that those outputs are actually I and Q, in other words the mixer
>> is as a matter of facts a quadrature mixer, then those two outputs must be digitized by the
>> 12-bit, 80 MHz high speed ADC of the NXP LPC-4370 processor, which is a single ADC.
>> So it should work in a time interleaved fashion... is this correct ?
>>
>> A bit of light on the Airspy architecture would be welcome, thanks.
>>
>> --
>> 73 Alberto I2PHD
>> Credo Ut Intelligam
>>
>>
>>
>>
>>
>>
>> ------------------------------------------------
>>
>> This email has been checked for viruses by Avast antivirus software.
>> www.avast.com
>>
>>
>>
>
>
> ------------------------------------
> Posted by: Leif Asbrink
> ------------------------------------
>
>
> ------------------------------------
>
> Yahoo Groups Links
>
>
>



-- 
Marcus Leech
Principal Investigator
Shirleys Bay Radio Astronomy Consortium
http://www.sbrac.org


jdow
 

Nope. The digital domain is a little limited in its ability to sample at 1.7 GHz in any meaningful manner. So you're correct in theory; but, practice lags theory in this regard.

ALSO, you're always hostage to the analog components in front.

{^_-}

On 2015-02-02 16:52, 'Marcus D. Leech' mleech@ripnet.com [airspy] wrote:


On 02/02/2015 07:02 PM, jdow jdow@earthlink.net [airspy] wrote:

Look up self calibrating Weaver mixer.

http://www.eecg.utoronto.ca/~kphang/papers/2002/jchow_imagereject.pdf
Of course, I'm forced to point out that everything you can do in the analog
domain in a weaver mixer you can do in the digital domain with an I/Q
signal that may not be perfect. Once you have an estimate of phase/amplitude
imbalance, you can easily correct it digitally--any residual imbalance
would be due to quantization error in both the estimates and the corrections.

Really, the only difference between the image properties in a "real" IF and a
complex-direct-conversion is exactly where the images fall. And with low-IF,
they also tend to fall in-band. TANSTAAFL :)



{^_^}

On 2015-02-02 06:45, Leif Asbrink leif@sm5bsz.com [airspy] wrote:
Hi Alberto,

The R820T2 has I and Q internally and combines
them to a real valued signal with excellent image
suppression.

The output is balanced to fit the balanced input
of typical AD converters.

I have no idea how they can get such a good image
suppression...

73

Leif




<airspy@yahoogroups.co.uk> wrote:



A few naive questions, originated from my attempt to understand the
architecture of the Airspy.

The tuner is an R820T2, which converts in the analog domain down to a low
IF, just above 4 MHz.

First question : from the simplified block diagram of that chip, it doesn't
seem to use a quadrature
mixer. That notwithstanding, the image rejection reported from its data
sheet is 65 dBc... how is
that possible ?

A more detailed look shows that it has two IF outputs, pin 12 and pin 13,
marked as IF N and IF P.
In the unconfirmed hypothesis that those outputs are actually I and Q, in
other words the mixer
is as a matter of facts a quadrature mixer, then those two outputs must be
digitized by the
12-bit, 80 MHz high speed ADC of the NXP LPC-4370 processor, which is a
single ADC.
So it should work in a time interleaved fashion... is this correct ?

A bit of light on the Airspy architecture would be welcome, thanks.

--
73 Alberto I2PHD
Credo Ut Intelligam






------------------------------------------------

This email has been checked for viruses by Avast antivirus software.
www.avast.com



------------------------------------
Posted by: Leif Asbrink <leif@sm5bsz.com>
------------------------------------


------------------------------------

Yahoo Groups Links



--
Marcus Leech
Principal Investigator
Shirleys Bay Radio Astronomy Consortium
http://www.sbrac.org




Marcus D. Leech <mleech@...>
 

On 02/02/2015 08:27 PM, jdow jdow@... [airspy] wrote:
 

Nope. The digital domain is a little limited in its ability to sample at 1.7 GHz
in any meaningful manner. So you're correct in theory; but, practice lags theory
in this regard.

ALSO, you're always hostage to the analog components in front.

I/Q amplitude and phase imbalance appears in the output of the mixer, which is neither high-bandwidth, nor high absolute
  signal frequencies.    Digital methods of compensation are well known.  Ask Youssef, he's invented at least one of them...

While I agree that you're hostage to analog imperfections "up front", *some* of those can be corrected downstream in
  the digital domain.  And they routinely *are*.



{^_-}

On 2015-02-02 16:52, 'Marcus D. Leech' mleech@... [airspy] wrote:
>
>
> On 02/02/2015 07:02 PM, jdow jdow@... [airspy] wrote:
>>
>> Look up self calibrating Weaver mixer.
>>
>> http://www.eecg.utoronto.ca/~kphang/papers/2002/jchow_imagereject.pdf
>>
>
> Of course, I'm forced to point out that everything you can do in the analog
> domain in a weaver mixer you can do in the digital domain with an I/Q
> signal that may not be perfect. Once you have an estimate of phase/amplitude
> imbalance, you can easily correct it digitally--any residual imbalance
> would be due to quantization error in both the estimates and the corrections.
>
> Really, the only difference between the image properties in a "real" IF and a
> complex-direct-conversion is exactly where the images fall. And with low-IF,
> they also tend to fall in-band. TANSTAAFL :)
>
>
>>
>> {^_^}
>>
>> On 2015-02-02 06:45, Leif Asbrink leif@... [airspy] wrote:
>> > Hi Alberto,
>> >
>> > The R820T2 has I and Q internally and combines
>> > them to a real valued signal with excellent image
>> > suppression.
>> >
>> > The output is balanced to fit the balanced input
>> > of typical AD converters.
>> >
>> > I have no idea how they can get such a good image
>> > suppression...
>> >
>> > 73
>> >
>> > Leif
>> >
>> >
>> >
>> >
>> > wrote:
>> >
>> >>
>> >>
>> >> A few naive questions, originated from my attempt to understand the
>> architecture of the Airspy.
>> >>
>> >> The tuner is an R820T2, which converts in the analog domain down to a low
>> IF, just above 4 MHz.
>> >>
>> >> First question : from the simplified block diagram of that chip, it doesn't
>> seem to use a quadrature
>> >> mixer. That notwithstanding, the image rejection reported from its data
>> sheet is 65 dBc... how is
>> >> that possible ?
>> >>
>> >> A more detailed look shows that it has two IF outputs, pin 12 and pin 13,
>> marked as IF N and IF P.
>> >> In the unconfirmed hypothesis that those outputs are actually I and Q, in
>> other words the mixer
>> >> is as a matter of facts a quadrature mixer, then those two outputs must be
>> digitized by the
>> >> 12-bit, 80 MHz high speed ADC of the NXP LPC-4370 processor, which is a
>> single ADC.
>> >> So it should work in a time interleaved fashion... is this correct ?
>> >>
>> >> A bit of light on the Airspy architecture would be welcome, thanks.
>> >>
>> >> --
>> >> 73 Alberto I2PHD
>> >> Credo Ut Intelligam
>> >>
>> >>
>> >>
>> >>
>> >>
>> >>
>> >> ------------------------------------------------
>> >>
>> >> This email has been checked for viruses by Avast antivirus software.
>> >> www.avast.com
>> >>
>> >>
>> >>
>> >
>> >
>> > ------------------------------------
>> > Posted by: Leif Asbrink
>> > ------------------------------------
>> >
>> >
>> > ------------------------------------
>> >
>> > Yahoo Groups Links
>> >
>> >
>> >
>>
>
>
> --
> Marcus Leech
> Principal Investigator
> Shirleys Bay Radio Astronomy Consortium
> http://www.sbrac.org
>
>
>
>



jdow
 

On 2015-02-02 17:31, 'Marcus D. Leech' mleech@ripnet.com [airspy] wrote:


On 02/02/2015 08:27 PM, jdow jdow@earthlink.net [airspy] wrote:

Nope. The digital domain is a little limited in its ability to sample at 1.7 GHz
in any meaningful manner. So you're correct in theory; but, practice lags theory
in this regard.

ALSO, you're always hostage to the analog components in front.
I/Q amplitude and phase imbalance appears in the output of the mixer, which is
neither high-bandwidth, nor high absolute
signal frequencies. Digital methods of compensation are well known. Ask
Youssef, he's invented at least one of them...

While I agree that you're hostage to analog imperfections "up front", *some* of
those can be corrected downstream in
the digital domain. And they routinely *are*.
The only image that can be reduced in digital is the one the other side of the spectrum center. You have this feature in a zero IF tuner configuration as with the E4000 tuner. Experience says this is a poor tradeoff.

With the R82x tuner family you cannot suppress the image the other side of the Dongle's VCO. That has to be done with an image reject mixer of one kind or another in the dongle in analog. The dongle itself would have to deliver samples at about 14 Msps to achieve the effect.

{^_^}