Analog, multichannel RF recorder?

What if you extract the modulation on the signals, record that, and regenerate the original signals by modulating oscillators? Why do you need a cycle-by-cycle record of the rf?

-- john

That isn't undersampling if the bandwidth of each IF is less than 30 MHz. If it's really undersampled, you have garbage; find another project. If the bandwidth is much less than 30 MHz -- which in a 75MHz IF seems likely -- you can sample at a lower rate, saving data. You probably need to know only the baseband signal. They can recreate the RF with appropriate modulators.

Jerry

Could you not write to a RAID rather than a single disk?

the

record

testing

I/O

for

You mention that the signal BW is just 5 MHz. Is the signal center frequency fixed within the IF? If so, you could digitally downconvert the signal, with sample rate reduction, and record the baseband.

I would think writing out to multiple disks simultaneously to reduce the required BW is the way to go.

John

Sorry to have not been more accurate, precise, and clear. First, I think I should have said "bandpass sampled" instead of undersampled. IF output is centered at 75 MHz, +/- about 15MHz of signal BW (spectrum from 60MHz to 90 MHz).

Sampling at 60MHz shifts that down to baseband (0-30MHz) where the DSP begins. The 5 MHz BW I quoted is after some of the filtering in our DSP.

I was hoping to capture analog on a multichannel recorder so I could do repeated playback thru the RF->IF analog chain, and iteratively tweak components (including my A/D converters), without having the artifacts of other (possibly inferior) A/D converters corrupting my "gold-standard" reference signals.

Digital capture also means I have to hook up 24 modulators to replay the RF. I wanted to avoid degrading my reference signals as much as possible. Replaying and modulating captured baseband means the next time I look at the signals in my DSP, they will have been thru my IF twice, plus somebody else's modulator, which munges things a bit. Thus, capturing original Rx antenna output would be ideal.

Also, capturing the data will be expensive, as it requires rental of an airplane for a couple of hours. Swapping disks every 2-5 minutes does not give me much continuous, uninterrupted data to work from. (I want to replay an entire flight segment thru certain maneuvers and across a good strectch of airspace.)

Having said that, if there is an "easy" (affordable, off-the shelf) RAID solution I can write to from my ADC converters (129.6 GB/minute for at least 10 minutes), I could use that for replay into my DSP datapath simulations. Not ideal, but far better than nothing.

Hope this better explains my needs and rationale. But as I mentioned before, all feedback is appreciated.

Thanks again,

mj

An off the shelf linux box with 4 400GB (~$300 ea) drives configured as one large disk with linux gets you 1.6 TB.

However, AFAIK the fastest disks around these days have a transfer rate of 1.5 GB/sec which only gets you 90 GB/minute.

There are some solid state disks around that can do better (but not much) than that with multiple fiber channel interfaces, but they aren't cheap.

I know of no computer system you can buy that will directly address over 1 TB of memory; direct to memory would get you the speed, but not the capacity.

I think you are screwed in storing this much data this quickly on any current storage device unless you can do real time compression on the data.

Just on general principles, you can beat down the RF signal to video using quadrature detectors. Record the in phase video and quadrature videos on two video channels of sufficient bandwidth. In fact, you would be best off using A/D converter and record digitally. That, and the frequency and phase of the carrier, should give you all the information you need. Actual waveforms could be reconstructed in a computer, if you really need that.

Bill

If you have 8 antenna's why do you have to record 24 channels.

Can't you record the 8 signals before they go into the 2 way converters (whatever those are)?

Saves you a factor of 3. Mark

Over 20 years ago, similar signals were recorded directly on tape; i believe the keyword was IRIG, and it was analog and there was no distortion, wow or flutter due to the signal processing done by the recorder. I think the signals could also be re-generated to be processed by a receiver as if the original signal had been fed into the receiver. These recorders allowed multiple channels, and one could record anything from almost DC to many megacycles (the term in those daze). It would not hurt to see what their full capabilities were..saves A->D then D->A conversions and all of the aliasing, etc that entails...

As you have "8 antennas receiving ~1GHz signals, ~5 MHz bandwidth each", what you might do is to mix the outputs the eight signals with a quasi-random, time-reference signal ( Perhaps from the GPS system.), and feed the outputs into nine video recorders. ( One for the quasi-random, time-reference signal.)

A digital computer or optical computer, could be used to sample any two of the channels and to determine auto-correlations, cross-correlations, perform Fourier transforms, power spectral density, etc.

-- Tom Potter

I almost posted previously about high-density digital tape recorders for this sort of thing about twenty years ago for airborne data gathering. However, those tape drives were big, expensive things that required custom interfaces, etc., etc. If you have the time to build some interfaces and dig some of these up (and your airplane is big enough to carry it (80lbs or so?)), and you can mount it safely, etc., it might be a good way to go.... ;)

IIRC the IRIG recorders (which we used to use as well) used a specific standardized format to capture things like time code, position, etc., and had limited bandwidth capability. My faulty memory seems to recall that we put the IRIG data multiplexed in with the captured data on the high-density tape.

I'm not even sure that those old Ampex HD tape drives could keep up, though. I seem to remember that the one we used had a max input clock rate of 10-20MHz and could take 20 tracks or so of data at that rate. Those numbers are only rough ballpark, it's been a long time...

I do think that you have a difficult problem, though. Even with the modern stuff data collection like this can easily challenge the capabilities of state-of-the-art systems.

Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions.

Would something like this maybe work for you>

best of luck - Mike

MJ, My first reaction to your apparent (rx testing) problem is "WHy isnt simulation sufficient?" or inother words, "What physical phenomenom (affecting the signal ) is so complex, that it can not be statistically re-created".??

Vibration and movement of and in an airplane (assuming this is part of your system) can be fairly easily measured, modelled, and use to 'modulate' an "ideal" signal into a pseudo-actual one. Recording the relationship between the acutal phenomenon and the affect on signal ( just use CW) is fairly easy, and can be used to derive a statisitcal model. (MATLAB is very helpful here)

Noise, wow flutter etc due to ionospherics is also able to be modelled. Im sure there are lots of papers on how this phenomenom affects signals. In short, are you really sure recording the signal is a) the easiest method to acheive what you need? b) Is able to achieve the depth and dynamics of testing youd like, when compared to simulation?

Before jumping into a recording-solution, I would justify mathematically that I could not acheive the same testing requirements using an easier-to-built simulation (which also probably has a far broader potential for testing / use than a one-of recording).

Having said all that,....................

My first guess at actually doing what you want, with easily available COTS, is to use a number of fast customized linux boxs, with maximum RAM, A/D cards, and fast disk, then setup them as a set of rotating buffers, You will need N = data-rate/disk-write-rate boxs, so the 1st just completes writing all its data to disk before the last triggers it for reading to RAM again. How you multiplex the various signals (each of 5MHz?) is another issue, but it seems wise that you'd shift each of your individual signals to baseband (just use mixer and LO) .

Even with this system, there is a limit to how accurate it will be for suitable recreation of the actual signal. Even after locking the local osciallators, accurately recording phase noise, especially for long term-intergration, may be a problem.

Happy working! David Merrett

Hi,

I am unclear what you are trying to evaluate. Are you gathering data for research on the world outside the aircraft, or are you trying to evaluate how well your own kit on the aircraft performs? It makes a big difference to the best approach. Flights are expensive, and you need to minimise the chance of a wasted flight. The less things between you and your data source, the less things you might find out later were not behaving properly. :-) In other words, record as close to the raw data as data rates and volumes will permit.

I haven't been involved in this kind of work for some years. It used to be that disk recording on aircraft was problematic for a number of reasons - vibration, acceleration, pressure, etc. I guess these days it would probably work out OK, as notebook disks seem to behave well on aircraft. A quick Google seems to show Enertec and others still make tape recorders for the kind of applications you are looking at. I used to use Enertec and Ampex machines in the 80s. The Ampex ground playback machines were good, but I had better results recording with Enertec machines. That was a long time ago, though, and newer machines may bear little resemblance to the ones I used. It seems these machines haven't speeded up that much. We used to write at about 120M bits/second to Enertec machines 20 years ago. They are talking about less than 1G bits/second now :-) Those solutions certainly work solidly in aircraft of various types, when used to suitable shock and vibration mounts - recording defense system flight data is their core market.

Exactly what you should try to record is another matter entirely. This comes back to my first question. A modern fighter gives surprisingly modest levels of vibration, but massive acceleration. A C-130 gives little acceleration, but vibrates hard at about 68Hz. The microphonic effects of these things on RF kit can be interesting, to say the least. If the aircraft operators will let you do it, tethering the aircraft and running the RF systems with the engines flat out can be educational. They tend not to like extended tethered tests, though. Whether done on the ground or in the air, I would recommend trying to get a proper understanding of how well the RF equipment works in its operational conditions before putting too much credance in any information you gather from the outside world.

Regards, Steve

snipped-for-privacy@cs.ucf.edu wrote:

Thanks VERY much for all the informative posts. Please keep em' coming if you have other info; I'm very grateful.

I'm old enough to remember the big multi-channel analog recorders, but not quite old enough to have ever used one... :-) Will do more digging on IRIG, Enertec, Ampex, etc..., as well as the Linux/RAID configurations, and post what I find.

To answer some of the other questions:

Yes, recording analog close to the antenna gives me the easiest way to replay a scenario, provided the tape stability and dynamic range are sufficient. 8 antennas would only require recording 8 channels, but if I have to record IF instead of RF, it's 24 channels, because we're actually pulling 3 different signal bands out of each antenna.

Simulation is not feasible *yet*. I need real data from some really busy airspace first (hundreds of planes, with their tcas and transponder systems squittering away) in order to get realistic statistics (signal levels, overlaps/collisions, etc...) Then I can create all kinds of models with controlled deviations from a known, realistic environment. But the "real world" sanity check is essential. Analog recording is preferred, to capture the effects of ultra-close transmitters saturating my receiver.

Aircraft will either be a slow jet or a small turboprop, depending on which is available. No high-rate maneuvers. I don't know what to expect in terms of microphonic effects from aircraft vibration, but as long as it's realistic (on the antennas, not on my recorder!), I'll be glad to have it. (Ideal scenario would be jet *and* turboprop data thru same airspace at same time, to see any differences.)

Not yet sure how I would interface my A/D converters to a fast disk array, but since I have an FPGA deserializing and processing the A/D input, some compression inside the FPGA (simple deltas?) is an obvious option that I didn't think of; thanks for pointing it out!!!

Regarding the 90GB/min disk arrays: I didn't know you could get 'em that fast! (Any recommended suppliers?) 12 channels at 90MB/sec = 64.8 GB/minute; at that rate I could put about 16 minutes of uncompressed data on a 1TB drive. That would certainly get me out of the starting gate.

Thanks again, very, very much!

mj