Hi all, I spent some time yesterday testing UmTRX with MCBTS. Aside from some trivial fixes (inverted sample rate), the code ran fine with no issues. The default bandwidth of the LPF was insufficient, so that was widened. I tested with 9 handsets between 3 and 7 carriers on a Core2 Duo (P7570) laptop. git://github.com/ttsou/openbts-multi-arfcn.git umtrx Configuration is through sqlite database, which is unchanged from mainline; just set the number of ARFCN (7 max) and C1 channels. Because of the bandwidth involved, be sure to increase the maximum buffer sizes for UHD. The uhd_usrp_probe utility will warn if the target size cannot be set. http://files.ettus.com/uhd_docs/manual/html/transport.html#linux-specific-n… Design ===== There are three resampling stages. 1. Outer resampler converts from the device rate to a multiple of the 400 kHz channel spacing. 2. Channelizer demultiplexes input stream to M channels 3. Inner resampler converts from 400 kHz to a multiple of the GSM symbol rate. UmTRX rates: Channelizer Rate Device Rate Decimation 1.6 Msps 1.625 Msps 8 3.2 Msps 3.250 Msps 4 Options ====== To broadcast dummy bursts on all carriers for spectrum testing, uncomment the following preprocessor declaration in radioParams.h. #define ENABLE_ALL_CHANS With dummy bursts on all carriers: http://filebox.vt.edu/users/ttsou/http/7carrier_umtrx.PNG Calibration ======== Calibration remains a concern. With single carrier, the carrier leakage and and IQ imbalance appear as in-band distortion. For MCBTS, OpenBTS offsets C0 on to the lowest carrier. This makes the carrier and image quite visible, as shown in the following uncalibrated capture. I checked with a swept tone to verify that this issue was not a result of baseband DSP. The USRP also shows similar effects. http://filebox.vt.edu/users/ttsou/http/carrier_image.PNG Any idea how much carrier and image suppression we should expect with calibration on UmTRXv2? Thomas

Hi Alexander, It is IQ imbalance indeed. It has to be corrected in BB. I believe you have access to the document: "LMS6002D Improving transceiver performance using digital techniques -1.0r1.pdf" It may be useful. Also, we recommend to implement RX DC auto cancellation as explained in section 3.3 of the same document. This should solve RX LO leakage issue. We usually measure < -60dBc unwanted image level when IQ imbalance is corrected. Regards, Srdjan On 29/10/2012 12:33, Alexander Chemeris wrote: > > Hi Srdjan, > > Have you seen a question at the end of this e-mail, regarding the LMS > calibration? > To me it looks like a result of the I/Q imbalance which we should > compensate at the baseband level. Is that correct? > > On Thu, Oct 25, 2012 at 10:16 PM, Thomas Tsou &lt;ttsou(a)vt.edu&gt; wrote: >> >> Hi all, >> >> I spent some time yesterday testing UmTRX with MCBTS. Aside from some >> trivial fixes (inverted sample rate), the code ran fine with no >> issues. The default bandwidth of the LPF was insufficient, so that was >> widened. I tested with 9 handsets between 3 and 7 carriers on a Core2 >> Duo (P7570) laptop. >> >> git://github.com/ttsou/openbts-multi-arfcn.git umtrx >> >> Configuration is through sqlite database, which is unchanged from >> mainline; just set the number of ARFCN (7 max) and C1 channels. >> Because of the bandwidth involved, be sure to increase the maximum >> buffer sizes for UHD. The uhd_usrp_probe utility will warn if the >> target size cannot be set. >> >> >> >> http://files.ettus.com/uhd_docs/manual/html/transport.html#linux-specific-n… >> >> Design >> ===== >> There are three resampling stages. >> >> 1. Outer resampler converts from the device rate to a multiple of the >> 400 kHz channel spacing. >> 2. Channelizer demultiplexes input stream to M channels >> 3. Inner resampler converts from 400 kHz to a multiple of the GSM >> symbol >> rate. >> >> UmTRX rates: >> >> Channelizer Rate Device Rate Decimation >> >> 1.6 Msps 1.625 Msps 8 >> 3.2 Msps 3.250 Msps 4 >> >> Options >> ====== >> >> To broadcast dummy bursts on all carriers for spectrum testing, >> uncomment the following preprocessor declaration in radioParams.h. >> >> #define ENABLE_ALL_CHANS >> >> With dummy bursts on all carriers: >> >> http://filebox.vt.edu/users/ttsou/http/7carrier_umtrx.PNG >> >> Calibration >> ======== >> >> Calibration remains a concern. With single carrier, the carrier >> leakage and and IQ imbalance appear as in-band distortion. For MCBTS, >> OpenBTS offsets C0 on to the lowest carrier. This makes the carrier >> and image quite visible, as shown in the following uncalibrated >> capture. I checked with a swept tone to verify that this issue was not >> a result of baseband DSP. The USRP also shows similar effects. >> >> http://filebox.vt.edu/users/ttsou/http/carrier_image.PNG >> >> Any idea how much carrier and image suppression we should expect with >> calibration on UmTRXv2? >> >> Thomas > > >

I am not sure calibration precision is affected by sample rate. Theoretically, we are rotating I vector by alpha/2 and Q vector by -alpha/2 and that is done by multiplication. Hence, the resolution of quadrature phase error (alpha) which we can correct is defined by the with of multipliers. Any way, we clocked ADCs at 15.36MHz (30.72MS/s digital IQ interface) in the experiment I mentioned. On 29/10/2012 13:26, Alexander Chemeris wrote: > > Hi Srdjan, > > If I understand correctly, I/Q imbalance calibration precision depends > on the sampling rate. What sampling rate do you use to achieve <-60dBc > image level? > > On Mon, Oct 29, 2012 at 5:19 PM, Srdjan Milenkovic > &lt;s.milenkovic(a)limemicro.com&gt; wrote: >> >> Hi Alexander, >> >> It is IQ imbalance indeed. It has to be corrected in BB. I believe you >> have >> access to the document: >> >> "LMS6002D Improving transceiver performance using digital techniques >> -1.0r1.pdf" >> >> It may be useful. Also, we recommend to implement RX DC auto >> cancellation >> as >> explained in section 3.3 of the same document. This should solve RX LO >> leakage issue. >> >> We usually measure < -60dBc unwanted image level when IQ imbalance is >> corrected. >> >> Regards, Srdjan >> >> >> On 29/10/2012 12:33, Alexander Chemeris wrote: >>> >>> Hi Srdjan, >>> >>> Have you seen a question at the end of this e-mail, regarding the LMS >>> calibration? >>> To me it looks like a result of the I/Q imbalance which we should >>> compensate at the baseband level. Is that correct? >>> >>> On Thu, Oct 25, 2012 at 10:16 PM, Thomas Tsou &lt;ttsou(a)vt.edu&gt; wrote: >>>> >>>> Hi all, >>>> >>>> I spent some time yesterday testing UmTRX with MCBTS. Aside from some >>>> trivial fixes (inverted sample rate), the code ran fine with no >>>> issues. The default bandwidth of the LPF was insufficient, so that >>>> was >>>> widened. I tested with 9 handsets between 3 and 7 carriers on a Core2 >>>> Duo (P7570) laptop. >>>> >>>> git://github.com/ttsou/openbts-multi-arfcn.git umtrx >>>> >>>> Configuration is through sqlite database, which is unchanged from >>>> mainline; just set the number of ARFCN (7 max) and C1 channels. >>>> Because of the bandwidth involved, be sure to increase the maximum >>>> buffer sizes for UHD. The uhd_usrp_probe utility will warn if the >>>> target size cannot be set. >>>> >>>> >>>> >>>> >>>> http://files.ettus.com/uhd_docs/manual/html/transport.html#linux-specific-n… >>>> >>>> Design >>>> ===== >>>> There are three resampling stages. >>>> >>>> 1. Outer resampler converts from the device rate to a multiple of the >>>> 400 kHz channel spacing. >>>> 2. Channelizer demultiplexes input stream to M channels >>>> 3. Inner resampler converts from 400 kHz to a multiple of the GSM >>>> symbol >>>> rate. >>>> >>>> UmTRX rates: >>>> >>>> Channelizer Rate Device Rate Decimation >>>> >>>> 1.6 Msps 1.625 Msps 8 >>>> 3.2 Msps 3.250 Msps 4 >>>> >>>> Options >>>> ====== >>>> >>>> To broadcast dummy bursts on all carriers for spectrum testing, >>>> uncomment the following preprocessor declaration in radioParams.h. >>>> >>>> #define ENABLE_ALL_CHANS >>>> >>>> With dummy bursts on all carriers: >>>> >>>> http://filebox.vt.edu/users/ttsou/http/7carrier_umtrx.PNG >>>> >>>> Calibration >>>> ======== >>>> >>>> Calibration remains a concern. With single carrier, the carrier >>>> leakage and and IQ imbalance appear as in-band distortion. For MCBTS, >>>> OpenBTS offsets C0 on to the lowest carrier. This makes the carrier >>>> and image quite visible, as shown in the following uncalibrated >>>> capture. I checked with a swept tone to verify that this issue was >>>> not >>>> a result of baseband DSP. The USRP also shows similar effects. >>>> >>>> http://filebox.vt.edu/users/ttsou/http/carrier_image.PNG >>>> >>>> Any idea how much carrier and image suppression we should expect with >>>> calibration on UmTRXv2? >>>> >>>> Thomas >>> >>> >>> >