Hi there, If I want to connect the X310 using 10 GbitE interface, do I have to connect the both cables in order to achieve 200 MSps on the both channels? What about PCI interface? Can I achieve 200 MSps on the both channels using this interface? BR, Vladica

Hi Vladica, yes, for 2x 200 MS/s you need the dual-10GE connection (quick calcution: 2*(16 b [I] + 16 b [Q])/S * 200e6 S/s = 12.8 Gb/s ). PCIe doesn't offer 2x200MS/s Best regards, Marcus On 19.10.2016 12:01, Vladica Sark via USRP-users wrote: > Hi there, > > If I want to connect the X310 using 10 GbitE interface, do I have to > connect the both cables in order to achieve 200 MSps on the both > channels? > > What about PCI interface? Can I achieve 200 MSps on the both channels > using this interface? > > BR, > Vladica > > _______________________________________________ > USRP-users mailing list > USRP-users@lists.ettus.com > http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

Thanks for the valuable info. I didn't know that PCIe does not support 2x200MS/s Best regards, Vladica On 19.10.2016 13:08, Marcus Müller via USRP-users wrote: > Hi Vladica, > > yes, for 2x 200 MS/s you need the dual-10GE connection (quick calcution: > 2*(16 b [I] + 16 b [Q])/S * 200e6 S/s = 12.8 Gb/s ). > > PCIe doesn't offer 2x200MS/s > > Best regards, > > Marcus > > On 19.10.2016 12:01, Vladica Sark via USRP-users wrote: >> Hi there, >> >> If I want to connect the X310 using 10 GbitE interface, do I have to >> connect the both cables in order to achieve 200 MSps on the both >> channels? >> >> What about PCI interface? Can I achieve 200 MSps on the both channels >> using this interface? >> >> BR, >> Vladica >> >> _______________________________________________ >> USRP-users mailing list >> USRP-users@lists.ettus.com >> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com > > > _______________________________________________ > USRP-users mailing list > USRP-users@lists.ettus.com > http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >

Hi Vladica, The PCIe interface we use can't deal with the raw data rates of 2x 200MS/s [1], and you can't have two of them. So, that won't work. Best regards, Marcus [1] https://kb.ettus.com/X300/X310#Choosing_a_Host_Interface On 19.10.2016 13:11, Vladica Sark wrote: > Thanks for the valuable info. I didn't know that PCIe does not support > 2x200MS/s > > Best regards, > Vladica > > > On 19.10.2016 13:08, Marcus Müller via USRP-users wrote: >> Hi Vladica, >> >> yes, for 2x 200 MS/s you need the dual-10GE connection (quick calcution: >> 2*(16 b [I] + 16 b [Q])/S * 200e6 S/s = 12.8 Gb/s ). >> >> PCIe doesn't offer 2x200MS/s >> >> Best regards, >> >> Marcus >> >> On 19.10.2016 12:01, Vladica Sark via USRP-users wrote: >>> Hi there, >>> >>> If I want to connect the X310 using 10 GbitE interface, do I have to >>> connect the both cables in order to achieve 200 MSps on the both >>> channels? >>> >>> What about PCI interface? Can I achieve 200 MSps on the both channels >>> using this interface? >>> >>> BR, >>> Vladica >>> >>> _______________________________________________ >>> USRP-users mailing list >>> USRP-users@lists.ettus.com >>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >> >> >> _______________________________________________ >> USRP-users mailing list >> USRP-users@lists.ettus.com >> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>

Hi Markus, But if I get the 2x 10G Ethernet, it has enough bandwidth for the RAW 2x200MS/s rate. Is this supported by the radio? Do I see them as 2 separate radios, since they would have separate IP addresses? Any idea about the latency when using 10G Ethernet? Best regards, Vladica On 19.10.2016 13:15, Marcus Müller wrote: > Hi Vladica, > > The PCIe interface we use can't deal with the raw data rates of 2x > 200MS/s [1], and you can't have two of them. So, that won't work. > > Best regards, > > Marcus > > [1] https://kb.ettus.com/X300/X310#Choosing_a_Host_Interface > > > On 19.10.2016 13:11, Vladica Sark wrote: >> Thanks for the valuable info. I didn't know that PCIe does not support >> 2x200MS/s >> >> Best regards, >> Vladica >> >> >> On 19.10.2016 13:08, Marcus Müller via USRP-users wrote: >>> Hi Vladica, >>> >>> yes, for 2x 200 MS/s you need the dual-10GE connection (quick calcution: >>> 2*(16 b [I] + 16 b [Q])/S * 200e6 S/s = 12.8 Gb/s ). >>> >>> PCIe doesn't offer 2x200MS/s >>> >>> Best regards, >>> >>> Marcus >>> >>> On 19.10.2016 12:01, Vladica Sark via USRP-users wrote: >>>> Hi there, >>>> >>>> If I want to connect the X310 using 10 GbitE interface, do I have to >>>> connect the both cables in order to achieve 200 MSps on the both >>>> channels? >>>> >>>> What about PCI interface? Can I achieve 200 MSps on the both channels >>>> using this interface? >>>> >>>> BR, >>>> Vladica >>>> >>>> _______________________________________________ >>>> USRP-users mailing list >>>> USRP-users@lists.ettus.com >>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>> >>> >>> _______________________________________________ >>> USRP-users mailing list >>> USRP-users@lists.ettus.com >>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>> >

Hi Vladica, On 19.10.2016 13:18, Vladica Sark wrote: > Hi Markus, > > But if I get the 2x 10G Ethernet, it has enough bandwidth for the RAW > 2x200MS/s rate. > Is this supported by the radio? Yes! > Do I see them as 2 separate radios, since they would have separate IP > addresses? No; you'd just specify a second IP address; see the "Dual 10 Gigabit Ethernet" section in the manual [1]. > Any idea about the latency when using 10G Ethernet? Not really much higher than PCIe in most usage scenarios. Your mileage might vary, but think of 100µs up. It really depends mostly on your host computer and its OS, so general numbers are really impossible to give. I haven't done any latency testing with dual 10GE, and it probably doubles the latency. In many applications, latency can be "hidden" by usage of *timed commands*; are you familiar with those? what's your application? Best regards, Marcus > > Best regards, > Vladica [1] http://files.ettus.com/manual/page_usrp_x3x0.html#x3x0_hw_10gige > > > On 19.10.2016 13:15, Marcus Müller wrote: >> Hi Vladica, >> >> The PCIe interface we use can't deal with the raw data rates of 2x >> 200MS/s [1], and you can't have two of them. So, that won't work. >> >> Best regards, >> >> Marcus >> >> [1] https://kb.ettus.com/X300/X310#Choosing_a_Host_Interface >> >> >> On 19.10.2016 13:11, Vladica Sark wrote: >>> Thanks for the valuable info. I didn't know that PCIe does not support >>> 2x200MS/s >>> >>> Best regards, >>> Vladica >>> >>> >>> On 19.10.2016 13:08, Marcus Müller via USRP-users wrote: >>>> Hi Vladica, >>>> >>>> yes, for 2x 200 MS/s you need the dual-10GE connection (quick >>>> calcution: >>>> 2*(16 b [I] + 16 b [Q])/S * 200e6 S/s = 12.8 Gb/s ). >>>> >>>> PCIe doesn't offer 2x200MS/s >>>> >>>> Best regards, >>>> >>>> Marcus >>>> >>>> On 19.10.2016 12:01, Vladica Sark via USRP-users wrote: >>>>> Hi there, >>>>> >>>>> If I want to connect the X310 using 10 GbitE interface, do I have to >>>>> connect the both cables in order to achieve 200 MSps on the both >>>>> channels? >>>>> >>>>> What about PCI interface? Can I achieve 200 MSps on the both channels >>>>> using this interface? >>>>> >>>>> BR, >>>>> Vladica >>>>> >>>>> _______________________________________________ >>>>> USRP-users mailing list >>>>> USRP-users@lists.ettus.com >>>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>>> >>>> >>>> _______________________________________________ >>>> USRP-users mailing list >>>> USRP-users@lists.ettus.com >>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>>> >>

Hi Marcus, I use the radios for ranging and localization applications mainly. I schedule basically the transmissions and receptions, using timed samples. I have tried timed commands on N210 and B205mini, but on N210 I get relatively large latency, which is variable and not consistent. Situation with B204mini is slightly better. BTW, PCIe x4 should be able to support the rate of 12.8 Gb/s. Is the bottleneck in the radio itself? BR, Vladica On 19.10.2016 13:24, Marcus Müller wrote: > Hi Vladica, > > > On 19.10.2016 13:18, Vladica Sark wrote: >> Hi Markus, >> >> But if I get the 2x 10G Ethernet, it has enough bandwidth for the RAW >> 2x200MS/s rate. >> Is this supported by the radio? > Yes! >> Do I see them as 2 separate radios, since they would have separate IP >> addresses? > No; you'd just specify a second IP address; see the "Dual 10 Gigabit > Ethernet" section in the manual [1]. >> Any idea about the latency when using 10G Ethernet? > Not really much higher than PCIe in most usage scenarios. Your mileage > might vary, but think of 100µs up. It really depends mostly on your host > computer and its OS, so general numbers are really impossible to give. I > haven't done any latency testing with dual 10GE, and it probably doubles > the latency. > In many applications, latency can be "hidden" by usage of *timed > commands*; are you familiar with those? what's your application? > > Best regards, > Marcus >> >> Best regards, >> Vladica > [1] http://files.ettus.com/manual/page_usrp_x3x0.html#x3x0_hw_10gige >> >> >> On 19.10.2016 13:15, Marcus Müller wrote: >>> Hi Vladica, >>> >>> The PCIe interface we use can't deal with the raw data rates of 2x >>> 200MS/s [1], and you can't have two of them. So, that won't work. >>> >>> Best regards, >>> >>> Marcus >>> >>> [1] https://kb.ettus.com/X300/X310#Choosing_a_Host_Interface >>> >>> >>> On 19.10.2016 13:11, Vladica Sark wrote: >>>> Thanks for the valuable info. I didn't know that PCIe does not support >>>> 2x200MS/s >>>> >>>> Best regards, >>>> Vladica >>>> >>>> >>>> On 19.10.2016 13:08, Marcus Müller via USRP-users wrote: >>>>> Hi Vladica, >>>>> >>>>> yes, for 2x 200 MS/s you need the dual-10GE connection (quick >>>>> calcution: >>>>> 2*(16 b [I] + 16 b [Q])/S * 200e6 S/s = 12.8 Gb/s ). >>>>> >>>>> PCIe doesn't offer 2x200MS/s >>>>> >>>>> Best regards, >>>>> >>>>> Marcus >>>>> >>>>> On 19.10.2016 12:01, Vladica Sark via USRP-users wrote: >>>>>> Hi there, >>>>>> >>>>>> If I want to connect the X310 using 10 GbitE interface, do I have to >>>>>> connect the both cables in order to achieve 200 MSps on the both >>>>>> channels? >>>>>> >>>>>> What about PCI interface? Can I achieve 200 MSps on the both channels >>>>>> using this interface? >>>>>> >>>>>> BR, >>>>>> Vladica >>>>>> >>>>>> _______________________________________________ >>>>>> USRP-users mailing list >>>>>> USRP-users@lists.ettus.com >>>>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>>>> >>>>> >>>>> _______________________________________________ >>>>> USRP-users mailing list >>>>> USRP-users@lists.ettus.com >>>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>>>> >>> >

Hi Vladica, yes, the latency over network and USB is random. So that's why you'd need the absolute times you get by issuing stream commands – you don't really need the estimate within microseconds of signal reception, you just need to know the delay between the RX and TX stream's first sample, right? > BTW, PCIe x4 should be able to support the rate of 12.8 Gb/s. Is the bottleneck in the radio itself? No, the radio internally doesn't make a difference which interface you use for the most part. MXI, the external PCIe bus used[1], is based on PCIe 1.1. A PCIe gen1 four-lane link has a link capacity of 8Gb/s, which is enough for the 6.4 Gb/s of 1x 200MS/s SC16, but not enough for twice that rate. The real bottleneck here will be the PC: The PCIe driver the way we use it is less capable of distributing workload across multiple CPU cores (it's rather optimized for low-latency and LabVIEW usage), and hence, you will have a hard time configuring a system to even handle 1x 200MS/s via PCIe. Best regards, Marcus [1] http://www.pxisa.org/userfiles/files/Specifications/PXIEXPRESS_HW_SPEC_R1.PDF On 19.10.2016 19:14, Vladica Sark wrote: > Hi Marcus, > > I use the radios for ranging and localization applications mainly. I > schedule basically the transmissions and receptions, using timed > samples. I have tried timed commands on N210 and B205mini, but on N210 > I get relatively large latency, which is variable and not consistent. > Situation with B204mini is slightly better. > > BTW, PCIe x4 should be able to support the rate of 12.8 Gb/s. Is the > bottleneck in the radio itself? > > BR, > Vladica > > > On 19.10.2016 13:24, Marcus Müller wrote: >> Hi Vladica, >> >> >> On 19.10.2016 13:18, Vladica Sark wrote: >>> Hi Markus, >>> >>> But if I get the 2x 10G Ethernet, it has enough bandwidth for the RAW >>> 2x200MS/s rate. >>> Is this supported by the radio? >> Yes! >>> Do I see them as 2 separate radios, since they would have separate IP >>> addresses? >> No; you'd just specify a second IP address; see the "Dual 10 Gigabit >> Ethernet" section in the manual [1]. >>> Any idea about the latency when using 10G Ethernet? >> Not really much higher than PCIe in most usage scenarios. Your mileage >> might vary, but think of 100µs up. It really depends mostly on your host >> computer and its OS, so general numbers are really impossible to give. I >> haven't done any latency testing with dual 10GE, and it probably doubles >> the latency. >> In many applications, latency can be "hidden" by usage of *timed >> commands*; are you familiar with those? what's your application? >> >> Best regards, >> Marcus >>> >>> Best regards, >>> Vladica >> [1] http://files.ettus.com/manual/page_usrp_x3x0.html#x3x0_hw_10gige >>> >>> >>> On 19.10.2016 13:15, Marcus Müller wrote: >>>> Hi Vladica, >>>> >>>> The PCIe interface we use can't deal with the raw data rates of 2x >>>> 200MS/s [1], and you can't have two of them. So, that won't work. >>>> >>>> Best regards, >>>> >>>> Marcus >>>> >>>> [1] https://kb.ettus.com/X300/X310#Choosing_a_Host_Interface >>>> >>>> >>>> On 19.10.2016 13:11, Vladica Sark wrote: >>>>> Thanks for the valuable info. I didn't know that PCIe does not >>>>> support >>>>> 2x200MS/s >>>>> >>>>> Best regards, >>>>> Vladica >>>>> >>>>> >>>>> On 19.10.2016 13:08, Marcus Müller via USRP-users wrote: >>>>>> Hi Vladica, >>>>>> >>>>>> yes, for 2x 200 MS/s you need the dual-10GE connection (quick >>>>>> calcution: >>>>>> 2*(16 b [I] + 16 b [Q])/S * 200e6 S/s = 12.8 Gb/s ). >>>>>> >>>>>> PCIe doesn't offer 2x200MS/s >>>>>> >>>>>> Best regards, >>>>>> >>>>>> Marcus >>>>>> >>>>>> On 19.10.2016 12:01, Vladica Sark via USRP-users wrote: >>>>>>> Hi there, >>>>>>> >>>>>>> If I want to connect the X310 using 10 GbitE interface, do I >>>>>>> have to >>>>>>> connect the both cables in order to achieve 200 MSps on the both >>>>>>> channels? >>>>>>> >>>>>>> What about PCI interface? Can I achieve 200 MSps on the both >>>>>>> channels >>>>>>> using this interface? >>>>>>> >>>>>>> BR, >>>>>>> Vladica >>>>>>> >>>>>>> _______________________________________________ >>>>>>> USRP-users mailing list >>>>>>> USRP-users@lists.ettus.com >>>>>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>>>>> >>>>>> >>>>>> _______________________________________________ >>>>>> USRP-users mailing list >>>>>> USRP-users@lists.ettus.com >>>>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>>>>> >>>> >>

Hi Marcus, On 19.10.2016 19:24, Marcus Müller wrote: > Hi Vladica, > > yes, the latency over network and USB is random. So that's why you'd > need the absolute times you get by issuing stream commands – you don't > really need the estimate within microseconds of signal reception, you > just need to know the delay between the RX and TX stream's first sample, > right? Yes basically. For estimation of time when the signal arrived, I use timestamps, and this works quite well. The problem is when I want to change the gain (I use xcvr2450 + n210, and also b205) between two time slots, in which I receive signals from 2 different stations. For this operation a hundred to few hundred useconds are required plus some random delay. Since my time slots are ca. 200 us, this is a bit of an issue. Anyway, having a lot of bits on the A/D, solves the problem. I do not have the optimal gain for each time slot, but that is not a huge issue. I make only coarse adjustment of the gain, just to avoid clipping of the signal in any time slot. And of course this is performed in not used time slots. In semi-static scenarios, this works well. > >> BTW, PCIe x4 should be able to support the rate of 12.8 Gb/s. Is the > bottleneck in the radio itself? > > No, the radio internally doesn't make a difference which interface you > use for the most part. > > MXI, the external PCIe bus used[1], is based on PCIe 1.1. A PCIe gen1 > four-lane link has a link capacity of 8Gb/s, which is enough for the 6.4 > Gb/s of 1x 200MS/s SC16, but not enough for twice that rate. I thought it was at least Gen 2.0 or 3.0. Best regards, Vladica > > The real bottleneck here will be the PC: The PCIe driver the way we use > it is less capable of distributing workload across multiple CPU cores > (it's rather optimized for low-latency and LabVIEW usage), and hence, > you will have a hard time configuring a system to even handle 1x 200MS/s > via PCIe. > > Best regards, > Marcus > > [1] > http://www.pxisa.org/userfiles/files/Specifications/PXIEXPRESS_HW_SPEC_R1.PDF > On 19.10.2016 19:14, Vladica Sark wrote: >> Hi Marcus, >> >> I use the radios for ranging and localization applications mainly. I >> schedule basically the transmissions and receptions, using timed >> samples. I have tried timed commands on N210 and B205mini, but on N210 >> I get relatively large latency, which is variable and not consistent. >> Situation with B204mini is slightly better. >> >> BTW, PCIe x4 should be able to support the rate of 12.8 Gb/s. Is the >> bottleneck in the radio itself? >> >> BR, >> Vladica >> >> >> On 19.10.2016 13:24, Marcus Müller wrote: >>> Hi Vladica, >>> >>> >>> On 19.10.2016 13:18, Vladica Sark wrote: >>>> Hi Markus, >>>> >>>> But if I get the 2x 10G Ethernet, it has enough bandwidth for the RAW >>>> 2x200MS/s rate. >>>> Is this supported by the radio? >>> Yes! >>>> Do I see them as 2 separate radios, since they would have separate IP >>>> addresses? >>> No; you'd just specify a second IP address; see the "Dual 10 Gigabit >>> Ethernet" section in the manual [1]. >>>> Any idea about the latency when using 10G Ethernet? >>> Not really much higher than PCIe in most usage scenarios. Your mileage >>> might vary, but think of 100µs up. It really depends mostly on your host >>> computer and its OS, so general numbers are really impossible to give. I >>> haven't done any latency testing with dual 10GE, and it probably doubles >>> the latency. >>> In many applications, latency can be "hidden" by usage of *timed >>> commands*; are you familiar with those? what's your application? >>> >>> Best regards, >>> Marcus >>>> >>>> Best regards, >>>> Vladica >>> [1] http://files.ettus.com/manual/page_usrp_x3x0.html#x3x0_hw_10gige >>>> >>>> >>>> On 19.10.2016 13:15, Marcus Müller wrote: >>>>> Hi Vladica, >>>>> >>>>> The PCIe interface we use can't deal with the raw data rates of 2x >>>>> 200MS/s [1], and you can't have two of them. So, that won't work. >>>>> >>>>> Best regards, >>>>> >>>>> Marcus >>>>> >>>>> [1] https://kb.ettus.com/X300/X310#Choosing_a_Host_Interface >>>>> >>>>> >>>>> On 19.10.2016 13:11, Vladica Sark wrote: >>>>>> Thanks for the valuable info. I didn't know that PCIe does not >>>>>> support >>>>>> 2x200MS/s >>>>>> >>>>>> Best regards, >>>>>> Vladica >>>>>> >>>>>> >>>>>> On 19.10.2016 13:08, Marcus Müller via USRP-users wrote: >>>>>>> Hi Vladica, >>>>>>> >>>>>>> yes, for 2x 200 MS/s you need the dual-10GE connection (quick >>>>>>> calcution: >>>>>>> 2*(16 b [I] + 16 b [Q])/S * 200e6 S/s = 12.8 Gb/s ). >>>>>>> >>>>>>> PCIe doesn't offer 2x200MS/s >>>>>>> >>>>>>> Best regards, >>>>>>> >>>>>>> Marcus >>>>>>> >>>>>>> On 19.10.2016 12:01, Vladica Sark via USRP-users wrote: >>>>>>>> Hi there, >>>>>>>> >>>>>>>> If I want to connect the X310 using 10 GbitE interface, do I >>>>>>>> have to >>>>>>>> connect the both cables in order to achieve 200 MSps on the both >>>>>>>> channels? >>>>>>>> >>>>>>>> What about PCI interface? Can I achieve 200 MSps on the both >>>>>>>> channels >>>>>>>> using this interface? >>>>>>>> >>>>>>>> BR, >>>>>>>> Vladica >>>>>>>> >>>>>>>> _______________________________________________ >>>>>>>> USRP-users mailing list >>>>>>>> USRP-users@lists.ettus.com >>>>>>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>>>>>> >>>>>>> >>>>>>> _______________________________________________ >>>>>>> USRP-users mailing list >>>>>>> USRP-users@lists.ettus.com >>>>>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >>>>>>> >>>>> >>> >

Hi Vladica, On 19.10.2016 20:07, Vladica Sark wrote: > Hi Marcus, > > On 19.10.2016 19:24, Marcus Müller wrote: >> Hi Vladica, >> >> yes, the latency over network and USB is random. So that's why you'd >> need the absolute times you get by issuing stream commands – you don't >> really need the estimate within microseconds of signal reception, you >> just need to know the delay between the RX and TX stream's first sample, >> right? > Yes basically. > > For estimation of time when the signal arrived, I use timestamps, and > this works quite well. Combine that with specifying when samples will be sent, and you've got your deterministic RX/TX delay. By doing so, the variable transport delay becomes irrelevant. > > The problem is when I want to change the gain (I use xcvr2450 + n210, > and also b205) between two time slots, in which I receive signals from > 2 different stations. For this operation a hundred to few hundred > useconds are required plus some random delay. Since my time slots are > ca. 200 us, this is a bit of an issue. Indeed. That should be possible by using a timed command on the N210, but it won't be possible on the B205, I guess. > Anyway, having a lot of bits on the A/D, solves the problem. I do not > have the optimal gain for each time slot, but that is not a huge > issue. I make only coarse adjustment of the gain, just to avoid > clipping of the signal in any time slot. And of course this is > performed in not used time slots. In semi-static scenarios, this works > well. Indeed! What are your bandwidths? The ADC of the N210 has nominally 14bits, but of those, roughly 12 are effectively available, but oversampling usually saves the day and gives you something close to the full 16 bit of dynamic range the samples going to your PC have. On the AD936x of the B2xx series, the ADC is less potent, but with appropriate preselection filtering, the noise figure is probably better than that on the (a bit dated) XCVR2450's LNA. > >> >>> BTW, PCIe x4 should be able to support the rate of 12.8 Gb/s. Is the >> bottleneck in the radio itself? >> >> No, the radio internally doesn't make a difference which interface you >> use for the most part. >> >> MXI, the external PCIe bus used[1], is based on PCIe 1.1. A PCIe gen1 >> four-lane link has a link capacity of 8Gb/s, which is enough for the 6.4 >> Gb/s of 1x 200MS/s SC16, but not enough for twice that rate. > > I thought it was at least Gen 2.0 or 3.0. > Nope, sorry. Best regards, Marcus