X5-RX Input Coupling

[mrodby]

The data sheet for the X5-RX says that its inputs are DC coupled. We assumed that this meant that its response would go down to DC.

In our testing, when we use a low frequency signal (e.g. 5 kHz), the resulting peak amplitude is a small fraction of the peak amplitude when we use a high frequency signal (e.g. 1 MHz). When we feed these signals to an oscilloscope, the low frequency amplitude matches the high frequency amplitude, so the problem is not in our signal generator. We have not done extensive characterization of the X5-RX frequency response, but could do so if necessary.

Is the X5-RX input actually DC coupled? If so, does this mean that response should go down to DC?

[jhenderson]

The X5-RX actually employs two A/Ds for each analog input connection: An AC-coupled high-speed A/D and a DC-coupled low-speed A/D. This design allows the design to achieve the best possible signal quality and bandwidth on the high-speed channel, while preserving the ability to measure low-frequency or DC on the low-speed A/D.

Channels 0..3 correspond to the high-speed inputs, 4..7 the low-speed. 0 and 4, correspond to input channel 0, etc.

[mrodby]

The X5-RX data sheet makes no mention of this aspect of the architecture. It mentions that the channels are DC coupled several times, and includes several graphs that imply that the response extends from 0 Hz up past 100 MHz. A diagram shows only 4 channels of A/D, and the text only mentions 4 channels.

In our application, we need to use the channels in pairs to digitize I and Q for quadrature demodulated signals. The demodulated signals can appear anywhere in the spectrum, often at 0 Hz.

Is there any way to get a true DC coupled digitization at a high speed (we currently use 130 MSPS) by combining the high and low speed channels on the X5-RX?

[jhenderson]

It is possible to design firmware that would meld the data received from the high and low speed A/D channels into a coherent whole. However, this is non-trivial and is accompanied by some technical challenges such as handling phase discontinuities to achieve seamless performance across the coverage band.

We elected to sidestep those problems on the stock firmware by maintaining separate data streams from the LS and HS A/Ds. For most customers of the X5-RX and the X5-G12, which uses a similar architecture, the ability to process the DC component independently is sufficient and often desirable.

There is no provision on the design to achieve DC coupling via more traditional means. Classic DC coupling of the HS A/D would require a complete board redesign and qual.

[mrodby]

Upon further reflection, we don't actually need DC, since our quadrature demodulator does not go all the way down to DC. But we do need to go down to probably a few kHz, which is well below the minimum frequency we appear to be getting now.

What is the frequency response of the high speed channels?

[jhenderson]

Why can't you use the data from the LS channel to support analysis of data in the <= 1 MHz band?

[mrodby]

We are not looking at a single frequency, or even a small range of frequencies, at a time. Our analysis algorithms treat the digitized quadrature signal as a continuous frequency band from -X to +X Hz. The algorithms look for LFM chirps, BPSK modulated pulses, and similar signals, anywhere in this whole range. We use quadrature demodulation so that we can analyze twice as much of the frequency band in a single capture. With a fairly large hole in the middle of that continuous band, our analysis algorithms don't work right, at least for signals with components in that hole.

Since the manual for the X5-RX is not yet available, and the data sheet does not say, could you tell me what the response of the both the high speed and low speed channels should be, and the maximum sample rate for the low speed channels?

Speaking of the manual, back in February I thought the manual was in final review, to be released shortly after the board was shipped. Do you know its current status?