16 Jun
2012
16 Jun
'12
8:59 p.m.
Hi Alexander,
I spent some time with UmTRX today and, after a number of tries, got
the FPGA and firmware code in the fairwaves/umtrx-dboard branch
running. I build the bootloader.rmi into the fpga image and loaded
through JTAG.
Some observations and questions so far:
The UmTRX identifies and comes up fine, but I'm having some tuning
difficulties on LMS 1 with the following command sequence. I don't get
the same output on LMS 2. Do I need to make any other changes?
Register dump attached.
=== LMS 1 ===
$ ./umtrx_lms.py --lms 1 --lms-init
$ ./umtrx_lms.py --lms 1 --lms-tx-enable 1
$ ./umtrx_lms.py --lms 1 --lms-tx-pll-tune 900e6
FREQSEL=62 VCO_X=8 NINT=276 NFRACK=7743330
VOVCO[0]=1
VOVCO[1]=1
VOVCO[2]=1
...
VOVCO[62]=1
VOVCO[63]=1
CAN'T TUNE
=== LMS 2 ===
./umtrx_lms.py --lms 2 --lms-tx-pll-tune 900e6
FREQSEL=62 VCO_X=8 NINT=276 NFRACK=7743330
VOVCO[0]=2
VOVCO[1]=2
VOVCO[2]=2
...
VOVCO[62]=1
VOVCO[63]=1
START=45 STOP=50 SET=47
Also, what is the status of storing to flash with
usrp_n2xx_net_burner.py? The firmware appears to write successfully,
but the FPGA image times out after 'Verifying data'. I suppose this
part is not essential right now, but it would be convenient because I
otherwise need to borrow a JTAG adapter.
Working with Xilinx tools in Linux is very frustrating, but I guess
that is old news. ISE 13.3 kept crashing until I installed it in a
virtual machine. That was the worst part.
Thomas
On Mon, Jun 11, 2012 at 7:12 AM, Alexander Chemeris
<alexander.chemeris(a)gmail.com> wrote:
On Mon, Jun 11, 2012 at 4:08 AM, Thomas Tsou
<thomastsou(a)gmail.com> wrote:
Signals look great!
Glad you like it :)
Yep, will do at Wed when I get back to Moscow.
Do you have a patchset to enable all needed debug settings for this?
On Sun, Jun 10, 2012 at 10:05 AM, Alexander
Chemeris
<alexander.chemeris(a)gmail.com> wrote:
Aah! Yes, we have some issue with setting the timestamp with UmTRX.
E.g. I had to modify UHD examples which set timestamp to 0 on startup
and rely on this later on. With my change they read the actual value
of the timestamp and start from that. Could we do the same for
OpenBTS?
We have to track down this bug in FPGA, as it's quite annoying. :( May
that's something Robin or Sylvain could take on?
OpenBTS starts with this change:
diff --git a/Transceiver52M/UHDDevice.cpp b/Transceiver52M/UHDDevice.cpp
index d4ba580..4baf824 100644
--- a/Transceiver52M/UHDDevice.cpp
+++ b/Transceiver52M/UHDDevice.cpp
@@ -48,7 +48,8 @@
tx_ampl - Transmit amplitude must be between 0 and 1.0
*/
-const double master_clk_rt = 52e6;
+//const double master_clk_rt = 52e6;
+const double master_clk_rt = 13e6;
const size_t smpl_buf_sz = (1 << 20);
const float tx_ampl = .3;
But the behavior is a bit strange - when I start OpenBTS transceiver
starts to consume 100% of CPU and doesn't transmit. After a while (a
minute or so) CPU usage drops and I start seeing a signal at the UmTRX
output. I had no time to look into this deeply.
If the load drops down and the signal transmits, I wouldn't worry
about it for now. The likely condition is that there is a offset
between the initial arriving timestamp and the expected timestamp.
Because the receive timestamps drive the transceiver, the receiver may
run at 100% until the timestamp and expected value align. If OpenBTS
transmits consistently, then the system is in sync. Can you transmit a beacon signal?
Yes. I could see the network with a phone. Signal hound pictures:
1) umtrx-openbts-spectrum.PNG - spectrum, generated by OpenBTS.
2) umtrx-openbts-timeslots.PNG - the same, but with zero span.
3) "GMSK VGA1 -5, VGA2 18, channel power 750kHz.PNG" and
"Tx_GMSK#6(hot)_240412.png" are spectrum of a pure continuous GMSK
signal, transmitted with UmTRX (captured by me and Andrey Sviyazov). Actually
I should just try to run it - may be it will work right away? :)
Yes, check for RACH bursts. The bursts will probably get rejected if
they arrive at GSM core due to timing offset, but that is an easy fix. Have you verified reception of a receive test
signal?
Andrew Karpenkov should have done that. I personally haven't tried yet.
--
Regards,
Alexander Chemeris.
CEO, Fairwaves LLC / ООО УмРадио
http://fairwaves.ru