SDR#'s Automatic IQ correction now works with RTL-SDR (xpost from yahoo groups)

[Youssef]

Hi all,

After a day of investigation and testing, Steve M. from the Osmocom team has
solved an issue that was preventing SDR#'s IQ correction from working properly.
The spectrum was shifted by about 2kHz and thus image correction wasn't working
at all. This has changed with the latest release of librtlsdr (rtlsdr.dll) which
is included in the development package of SDR#. Now even these cheap DVB-T
dongle have over 80dB image rejection. Isn't that cool?
Here are some screenshots of the FM broadcast band with IQ correction in action:
http://sdrsharp.com/downloads/rtl_without_iq_correction.png
http://sdrsharp.com/downloads/rtl_with_iq_correction.png
http://sdrsharp.com/downloads/rtl_without_iq_correction2.png
http://sdrsharp.com/downloads/rtl_with_iq_correction2.png
The download location is as usual:
http://sdrsharp.com/index.php/downloads (rev 330+)

Feedbacks are welcome.

Youssef

[Alan Moffet]

Awesome. Youssef - thanks for your work. Steve M, also. We are grateful.

Alan, KE7IJZ

> --
> You received this message because you are subscribed to the Google Groups
> "Ultra Cheap SDR" group.
> To view this discussion on the web visit
> https://groups.google.com/d/msg/ultra-cheap-sdr/-/H4M4lAFgnCsJ.
> To post to this group, send email to ultra-c...@googlegroups.com.
> To unsubscribe from this group, send email to
> ultra-cheap-s...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/ultra-cheap-sdr?hl=en.

[G7Tony]

Great work again..
It just gets better and better!
rgds
Tony

[imo]

Hi, I tried r330

1. more CPU needed (60->65%)

2. two crashes (XP SP3 freezed fully) within few minutes - one when tuning, one when setting manual gain

 

No such issues with r327.

imo

[Youssef]

Only rtlsdr.dll has changed between these versions. I'll investigate.

[imo]

.. the buffer size has been changed in the r330 (1024*16 -> 1024*8) as well..

i.

[Youssef]

Hmm. Should I revert it? :-p Well, download in a few minutes :-)

[Chris Stake]

R330 is working very well here (W7 32  home premium). No crashes, no stutters. Nice job.

When gain is set to Auto, could the Front End perhaps display its actual setting?

Chris Stake

 

 

--

You received this message because you are subscribed to the Google Groups "Ultra Cheap SDR" group.

To view this discussion on the web visit https://groups.google.com/d/msg/ultra-cheap-sdr/-/4LIsSfkUHnsJ.

[G7Tony]

330 is working fine for me on XP SP3 (fully updated)..
No crashes or freezes with gain auto or manually adjusted
rgds
Tony

[DJ0MY]

Hi Youssef,

outstanding achievement....

Let me be a bit picky however ;-)

How can you achieve 80dB image rejection if the 8 bit dymanic range of the dongle is only about "40ish" dB wise....?
;-)

Well done,

Oscar DJ0MY

[Miguel A. Vallejo]

Because samples first are converted to 16 bit values and then the IQ correction applied, then corrected samples are 16 bits and 80dB values for image correction is possible :-)

--
You received this message because you are subscribed to the Google Groups "Ultra Cheap SDR" group.

To view this discussion on the web visit https://groups.google.com/d/msg/ultra-cheap-sdr/-/Wm7WTw0NiOAJ.

[Youssef]

Good question. The algorithm itself was designed for sound card based SDRs with a 16 and 24 bit-widths and the measured image rejection was over 80dBc. Now that the spectrum is perfectly centered on the new librtlsdr release, the correction algorithm seems to work in the exact same fashion as in narrow band SDRs. So here, we speak about the capability of the algorithm. Note however that with decimation and filtering, even 8 bit width can achieve higher dynamic range than the theoretical "40ish" dB (without averaging or filtering.) The algorithm will be able to kill the eventual images to over 80dB, which means images will be nulled out even if under the noise floor :-) Overkill for RTLSDR, indeed. I hope this clarifies things.

[Youssef]

Hi Miguel,

It's more because of oversampling/decimation/filtering than the conversion. In fact, the 8 bit samples are converted directly to single precision floats (32bit) "without loss", then decimated to the detection sample rate (>250kHz for WFM, > 24kHz otherwise.)
Question: What's the sample rate of the ADCs in the RTL chip? ;-)

Youssef

On Tuesday, June 5, 2012 11:48:21 AM UTC+2, Miguel A. Vallejo wrote:

Because samples first are converted to 16 bit values and then the IQ correction applied, then corrected samples are 16 bits and 80dB values for image correction is possible :-)

El 05/06/2012 11:28, "DJ0MY" <oscar...@googlemail.com> escribió:

Hi Youssef,

outstanding achievement....

Let me be a bit picky however ;-)

How can you achieve 80dB image rejection if the 8 bit dymanic range of the dongle is only about "40ish" dB wise....?
;-)

Well done,

Oscar DJ0MY

--
You received this message because you are subscribed to the Google Groups "Ultra Cheap SDR" group.
To view this discussion on the web visit https://groups.google.com/d/msg/ultra-cheap-sdr/-/Wm7WTw0NiOAJ.

To post to this group, send email to ultra-cheap-sdr@googlegroups.com.
To unsubscribe from this group, send email to ultra-cheap-sdr+unsubscribe@googlegroups.com.

[Miguel A. Vallejo]

Youssef wrote:

>>> Hi Miguel,
>>>
>>> It's more because of oversampling/decimation/filtering than
>>> the conversion. In fact, the 8 bit samples are converted directly
>>> to single precision floats (32bit) "without loss", then decimated
>>> to the detection sample rate (>250kHz for WFM, > 24kHz otherwise.)

I don't know why I thought SDR# internals worked at 16 bits... obviously 32 bit is needed for those SDR working at 24 bits...

>>> Question: What's the sample rate of the ADCs in the RTL chip? ;-)

I'll bet for the user selected samplerate: 0.9 up to 3.2 Msps :-)

[mrzfrn]

On Tuesday, June 5, 2012 11:59:38 AM UTC+2, Youssef wrote:

Question: What's the sample rate of the ADCs in the RTL chip? ;-)

Youssef

O

... two 8 bit ADC,  up to 4 MS/s
fM