Firstly, I am working about AoA techniques such as Esprit and MUSIC. While I was searching the literature, I found an example given below as a link.
In this example, NI USRP 2920s were used and two RF outputs/inputs(TX1/RX1 and TX2/RX2) were used as receiver. However, I am using USRP 2921. My USRPs do not support using both RF outputs/inputs. So, how can I realize the reference TX signal input at USRPs?
The USRP-2921 is only half duplex (can not TX and Rx at the same time). so an additional device will be needed to provide the reference tx signal. Another alternative when using the USRP-2921 is to use a 4 way power splitter to transmit your reference signal.
Just for clarification, are all of the USRPs that you want to use 2921s?
Check out the block diagram below.
You are correct that the radio only has one receive chain. It looks like there are two on the front of the box, but you'll see that on the inside it's just an RF switch to switch between SMA inputs.
The switch however is not "perfect"...I think it's about 30-ish dB of isolation. So, assuming your two signals were in the same band, you'd see your active channel signal (say Rx1) as normal, and then the signal on the non-active channel (say Rx1) about 30dB down from the actual level.
Anyway...if I recall, that MUSIC example needs to measure both the signal of interest as well as a "reference tone" in order to phase align all the receivers. I believe the example actively measures the signal of interest on one Rx channel, and puts the reference tone into the other "non-active" input. If the reference and the signal of interest are close enough in frequency, they can be measured in the same band. Of course...the reference tone will be down 30dB...but who cares? It's a cute way to do measurement of two different signals even though you only have one Rx channel.
I haven't done MUSIC, but I have used this same trick to phase align multiple USRPs, and it does work. It's a bit of pain compared to just having two active Rx paths that can be phase aligned, or to be able to phase align using some other external reference, but it works!
Thanks for reply. I understood your aspect and I will try. However, I thought about another solution for beginning.
In this example, PPS In and Ref In are connected just receiver(x4). Also, ref tx is connected with cable to all receivers. If I do not use ref tx signal and if I connect PPS and Ref In to all USRPs like 1 Tx and 4 Rx, Does the system work? When the ref clk comes from PXI to all USRPs, the LO phase will be adjusted and will be normal. Then, when the receivers took the transmitted signal, the phase offset occurs because of just uniform antenna array properties.
Ideally, sharing an LO between Rx's would eliminate the need for the reference tone...unfortunately, this isn't what happens with the N210's. The way the LO and PPS on the N210's works is that sharing the LO will *phase lock* all the LO's, but it does not *phase align*.
In other words...sharing the LO will phase lock the USRPs such that they will have a constant, non-time-varying phase difference between them. Making matters worse...these phase differences will be different each time you run the radios. In order to phase align (which you need to calculate the actual distance/angle/etc), you need to measure each USRPs LO relative to some other signal (i.e. - the reference Tx) to calibrate out these constant (but always different) phase differences between radios.
You was talked about phase different between USRP.. and right now, I think I face this issue to my research.
Using the code from NI community about Angle-of-Arrival, this is the result from my measurement:
1) Receiver detected AoA from 90 (arriving signal = 1) and or 60 and 120 (arrivinf signal = 2) when transmitter antenna OFF
2) I got fine result from DoA 0-40 but after that, the result for DoA estimation error. System can't detected the right DoA
Why this behaviour occurs ?
Thank you for your consideration