Broadband spectral sweep of ni5660 to disk?

[cdschmit]

Hi,

I'd like to use the ni5660 to save broadband (100 MHz) spectrum to disk.  I understand the real-time bandwidth limitations (1.25 MHz DDC limitation, 20 MHz total bandwidth possible if first dumped to the onboard memory), but I don't know the sweep rate limitations (how quickly can I reset the niTuner to a new center frequency and is there any danger in damaging the tuner by sweeping too quickly over an extended period of time?).  My interest is the 2.4 - 2.5 GHz (rougly ISM band).  I'd like to imitate the sweep functionality of the 5660 Demo Panel, with a save to disk option and, perhaps, I-Q data in place of amplitude only.  I'll be saving a full day's worth of data to the PC's HD.

I'm looking for code/suggestions on to accomplish this with the highest resolution for a given sweep rate setting.

Thanks!

Chris

[Yardov]

Hi Chris,

The NI 5660 is not the best equipment to perform this operation. Mainly because the limitation you mention (1.25MHz DDC). This means that you will sample at max rate (64 MS/s) and then the software will perform the DDC operation which might take long time. Now, you do not need to worry in tunning faster than the hardware can, the driver will not let you damage the device and you will be able to get the data you need in chunks. Then you will have to concatenate this data.

The driver RFSA performs this operation automatically but not the NI 5660 driver. Saving all this data for a day would take massive amount of time and processing power. I would recommend to look at some of the newer devices with higher DDC bandwidths like the NI PXI 5661 and NI PXI 5663 which have 20 MHz and 50 MHz DDC bandwidth respectively.

Post back if you have further questions or thoughts about this.

[cdschmit · Accepted solution]

---

Yardov wrote:

Hi Chris,

The NI 5660 is not the best equipment to perform this operation. Mainly because the limitation you mention (1.25MHz DDC). This means that you will sample at max rate (64 MS/s) and then the software will perform the DDC operation which might take long time. Now, you do not need to worry in tunning faster than the hardware can, the driver will not let you damage the device and you will be able to get the data you need in chunks. Then you will have to concatenate this data.

The driver RFSA performs this operation automatically but not the NI 5660 driver. Saving all this data for a day would take massive amount of time and processing power. I would recommend to look at some of the newer devices with higher DDC bandwidths like the NI PXI 5661 and NI PXI 5663 which have 20 MHz and 50 MHz DDC bandwidth respectively.

Post back if you have further questions or thoughts about this.

---

Yardov,

Yeah, it's a simple matter of expense.  I have placed a proposal for the 5663, but will not have the funding until summer (if at all ) and the 5661 would require purchasing a new PXIe chassis anyway.  Thanks for the info regarding the tuning speed!

Chris

[qwwvrvwekvuw]

Thanks very much, these informations helped me a lot.

I have a quastion: can I analyse or record broadband signals with BW > 50 MHz (e.g. 100 MHz ) with PXIe-5663 ?

As I understand, the PXIe-5663 needs to retune its LO (or niTuner) to a new center frequency since BW > instantaneous BW.

Or may I use multiple of PXIe-5663's (two of them for 100 MHz BW), that what I understand from the specs of PXIe5663 page 2 in title "RF Record and Playback" .

[Yardov]

Hi qwwvrvwekvuw,

Welcome to the forums. You read the information correctly, you can use N PXI 5663 to record N * 50 MHz BW. Remember that you will need to be able to stream the data out fast enough (200 MB/s) for each analyzer. If you need help putting together this system, I would recommend to talk to your local NI guy so that he can help you out selecting every single part for this system. The RF equipment is important but also the PXI Chassis, Controller and RAID array.

As a correction, retuning the LO does not need niTuner. PXI 5663 driver support is only using niRFSG.

[cdschmit]

---

qwwvrvwekvuw wrote:

Thanks very much, these informations helped me a lot.

I have a quastion: can I analyse or record broadband signals with BW > 50 MHz (e.g. 100 MHz ) with PXIe-5663 ?

As I understand, the PXIe-5663 needs to retune its LO (or niTuner) to a new center frequency since BW > instantaneous BW.

Or may I use multiple of PXIe-5663's (two of them for 100 MHz BW), that what I understand from the specs of PXIe5663 page 2 in title "RF Record and Playback" .

---

Hi,

I talked to NI about this directly at one time.  From what I recall, placing two 5663 instruments in the same chassis would allow synchronous sampling across the entire 100 MHz band, but streaming to the RAID array would present a bottleneck and limit the acquistion to something closer to 70 MHz.  Hopefully, someone else can verify this for you.

Chris

[NeilF]

Streaming the data from two 5663s to disk should be no problem.  50 MHz of bandwidth corresponds to an IQ sampling rate of 62.5 MHz, which at 4 bytes per sample is 250 MB/s for the I and Q data combined.  Two units at 250 MB/s each is 500 MB/s, which is easily achievable with an NI HDD-8264 12-drive RAID array.  In fact, our system engineers have tested systems with 4 5663s and 2 HDD-8264s, all streaming to disk continuously for hours.  I think we can get to 6 5663s in a single chassis (PXIe-1075 chassis + PXIe-8130 controller), but that has not yet been tested.

Neil F.

Principal Engineer

Modular Instruments

[qwwvrvwekvuw]

so all I need to perform analysis of a broadband signal ( BW=120 MHz) is three PXIe-5663's in the same Chassis, the sampling and streming to disk requirs an RAID array.

then how can I know the required RAID for a specificr ecording time and  BW (or sampling rate) ?

[Yardov]

Neil, thank you for the correction (250 MB/s per 5663).

Qwwvrvwekvuw, basically, your complex sampling rate (IQ Rate) is 1.25 x BW. And each IQ sample pair is made out of 2 Bytes. Therefore:

Data Rate = IQ Rate [Samples/sec] x 4 [Bytes/Sample] = 1.25 x BW x 4

If you need 120 MHz, then 40 MHz each analyzer would give you and IQ rate of  50 MS/sec = 200 MB/sec x 3 = 600 MS/sec

This rate can be sustained with the 8264 for 2 TB.

One more thing to add to this conversation is that signals that are between devices (i.e. between two PXI 5663), will have phase differences since they are not downconverted with the same LO. In general, this is not a problem for spectrum analysis since phase information is not considered.

[Yardov]

A want to add something here. In general if you are working with data rates of 400-500 MB/s, the hardware selection is not difficult.If you are in the limits of the RAID (600 MB/s), then there are a few hardware and software things that come into play. I would recommend to talk to us to make sure you select each piece correctly.

Some of the hardware factors is that the PXI 1075 has PXIe switches in all the slots that will prevent you to go as high as 600 MB/s. You can solve this issue choosing a PXI 1065 chassis. This chassis has two slots that have direct access to the controller. (This chassis will require a different placement of the three 5663's).

On the software side, you should be able to stream 600 MB/s but this is assuming almost no other pieces of code are running.

Using two RAIDS would get around any of these factors and we would not need to think this so much. You just need to stream 300 MB/s to each of them. Hope this helps to scope a streaming application pieces in general.

Message Edited by Yardov on 02-25-2009 02:06 PM