calibration error

[Max]

Hi.

I've installed CT (non-GPS version) into USRP and trying to calibrate
it using http://thre.at/kalibrate/
Unfortunately it doesn't work: whatever optios I choose I got error:
usrp_source::tune

I'm running it as follows:
LD_LIBRARY_PATH=/home/user/bin/gnuradio52/lib ./kal -v -D -s 850

Any advice on how to fix this error and perform calibration properly?

best regards,
Max.

[Alexander Chemeris]

Hi,

Most likely it means that whether (1) ClockTamer has no power (check
that you're able to connect to its control console with miniUSB) or
(2) it has loose connection to USRP, or (3) you did something wrong
during USRP modification.

In a nutshell, it most-likely means that USRP doesn't have clock on the FPGA.

PS Here I assume that you use kal v0.4.1

--
Regards,
Alexander Chemeris.
http://www.fairwaves.ru

[☠]

Thanks for prompt answer.

You're right about kalibrator version.

I double-checked and it seems weird:
- when I connect CT to usb-only it works just fine e. g. tb.py shows the version and
paameters.
- when I connect it to usrp (power and ext. port) AND usb - I do not see device even
via lsusb but usrp fan is spinning e. g. powr goes through
- when I connect it to usrp regardless of usb I got tune error mentioned before

Does it mke any difference how and to which ext. port I conncet CT on usrp?

I'm stuck right now, maybe some pictures of install CT on usrp - just to make ure
that I hadn't missed something obvious.

best regards,
Max.

[Alexander Chemeris]

Max,

On Mon, Apr 11, 2011 at 19:22, ☠ <ex5...@gmail.com> wrote:
> I double-checked and it seems weird:
> - when I connect CT to usb-only it works just fine e. g. tb.py shows the version and
> paameters.
> - when I connect it to usrp (power and ext. port) AND usb - I do not see device even
> via lsusb but usrp fan is spinning e. g. powr goes through
> - when I connect it to usrp regardless of usb I got tune error mentioned before

What if you connect to USRP power without connecting to USRP extension port?

> Does it mke any difference how and to which ext. port I conncet CT on usrp?
>
> I'm stuck right now, maybe some pictures of install CT on usrp - just to make ure
> that I hadn't missed something obvious.

Look at this picture to see how to connect ClockTamer to USRP extension port:
http://wiki.clock-tamer.googlecode.com/hg/images/clocktamer-v1.0/with_usrp_2.jpg

You may connect to any of USRP extension ports, but it's critical to
connect with right polarity - e.g. like on the picture above. USRP
extension port don't have a key, so it's easy to connect ClockTamer in
wrong polarity.

[Max X]

I connected CT exactly as it's shown in the picture.
Now it's visible to tb.py when I connect both usb & usrp ext so it looks like the polarity indeed was wrong.
However it doesn't fix main problem - I still can't tune in 52Hz mode.

2011/4/11 Alexander Chemeris <alexander...@gmail.com>

What if you connect to USRP power without connecting to USRP extension port?

it works with tb.py (clock-tamer GUI) just fine.

You may connect to any of USRP extension ports, but it's critical to
connect with right polarity - e.g. like on the picture above. USRP
extension port don't have a key, so it's easy to connect ClockTamer in
wrong polarity.

This is fixed now but it seems that somehng else is wrong.

What else can I try to diagnose why usrp is unable to obtain reference signal from CT?

best regards,
Max.

[Max X]

Just in case - attached is screenshot of CT-GUI and output of test commands.

best regards,
Max.

[Alexander Chemeris]

Max,

Just in case - what daughterboards do you have installed on your USRP?
If you use RFX1800, then you should re-flash it to RFX900 to work in
GSM850 band,

--

[Max X]

No, I use rfx 900, besides tuning without ct works and I can see
neighbour cell signal.

вторник, 12 апреля 2011 г. пользователь Alexander Chemeris
<alexander...@gmail.com> писал:

[Alexander Chemeris]

Could you, please, provide a high-res picture of your USRP modifications?

Do you re-solder everything back to test without CT?

--
Regards,
Alexander Chemeris.

[☠]

18.04.2011 08:51, Alexander Chemeris пишет:

> Could you, please, provide a high-res picture of your USRP modifications?

Sure. Attached are results of my ugly attempts at soldering and overall connection view.

> Do you re-solder everything back to test without CT?

No, I just disconnect ct and it work fine.

cheers,
Max.

[sergey kostanbaev]

On Sun, Apr 24, 2011 at 3:05 AM, ☠ <ex5...@gmail.com> wrote:

18.04.2011 08:51, Alexander Chemeris пишет:

> Could you, please, provide a high-res picture of your USRP modifications?

Could you provide pictures from the front side, near SMA

Sure. Attached are results of my ugly attempts at soldering and overall connection view.

> Do you re-solder everything back to test without CT?

No, I just disconnect ct and it work fine.

It shouldn't be happened, because if you'd done it right it wouldn't have been working. Check http://code.google.com/p/clock-tamer/wiki/ClockTamerUSRPInstallation with your modifications Moving capacitor removes path from internal oscillator.



 

cheers,
Max.

[Max X]

*facepalm*!

Resoldering fixed all the hardware issues.

Now calibration works fine except that I miscalculate correction values for oscillator.

Here http://code.google.com/p/clock-tamer/wiki/ClockTamerCalibration following formula used:
calibration offset = freq_error * LO_freq / ARFCN_freq
where freq_error is "average" value obtained from kal, in Hz
for example:

./kal -c 234
kal: Calculating clock frequency offset.
Using GSM-850 channel 234 (890.4MHz)
average        [min, max]    (range, stddev)
+ 32.284kHz        [32188, 32358]    (169, 35.854992)
overruns: 0
not found: 17913

freq_error = 32284 Hz

LO_freq is CT's oscillator frequency in MHz (by default is 20, could be obtained using tb.py for example)

ARFCN_freq is channel frequency.

In the example above for ARFCN 234 I can obtain following via http://www.aubraux.com/design/arfcn-calculator.php

Start Frequency = 890 MHz
Channel Width = 200 KHz
ARFCN=234

Channel Frequency = 936.8 MHz

What shall I use to calculate correction value 890 or 936.8 in this case?

thanks,
Max.

2011/4/24 sergey kostanbaev <sergey.k...@gmail.com>

[Alexander Chemeris]

Max,

32kHz offset doesn't look real. Offset should be no more then 250Hz.
So you're probably synchronize to some pure tone noise :)

--
Regards,
Alexander Chemeris.

[Max X]

Ok, that just weird: I definitely have a gsm signal on my android phone.
But judging from offset values all I can see is only "pure noise" signals:

./kal -v -s 850
kal: Scanning for GSM-850 base stations.
channel detect threshold: 866.156701
GSM-850:
    chan: 142 (872.0MHz - 15.277kHz)    power: 1291.67
    chan: 143 (872.2MHz - 9.884kHz)    power: 1287.06
    chan: 172 (878.0MHz - 32.270kHz)    power: 1191.07
    chan: 235 (890.6MHz + 32.299kHz)    power: 3046.15

How can it be?

And just to double-check: in correction calculation I use "channel frequency", not "start frequency"?

cheers,
Max.

[Alexander Chemeris]

Because we have only GSM-900 and GSM-1800 in Russia, no GSM-850 here.

--
Regards,
Alexander Chemeris.

[Max X]

Ok, that's it - I think I have collected all the possible errors now :-)

I think I've finished calibration (correction offset was calculated using channel frequency provided by kal itself).
Currently 30 Hz which seems fine.

thanks a lot everyone for patience and helpful advices,
Max.