VEGAS IF Target Levels Discussion

Introduction

In an effort to update the Observer's Guide, Amanda asked for iftarget values for VEGAS (in units of Volts - see table below and Table 5.4 in the OG) to include in an updated email. The values in this table is what config-tool is setting in the IF Rack (by default). If set too low, the signal will be too weak. If these values are set too high, it can cause saturation in the optical fibers leading from the IF Rack downstream, potentially leading to non-linear signal sensitivity. This value depends on the bandwidth of the signal leaving the IF rack, with larger bandwidths generally having a larger IF target level.

http://www.gb.nrao.edu/gbtprops/obsman/GBTog/node48_ct.html#table:iftarget

Potential for Saturated Fibers

The default values "for all other backends" could potentially lead to fiber saturation with VEGAS if set to 1.0 V, as noted by Ron:

From Ron August 11

While reviewing this with Adam I came to the realization that we probably
are having an issue with any VEGAS projects that do observations like
Galactic HI.  And, we probably have issues with multi-frequency work with
widely spaced centers for some receivers.

1) For a single, narrow-band line, config-tool will pick a narrow filter in
the IF Rack (so as to keep RFI from overwhelming the optical drivers).
Since VEGAS is not an explicit column in table 6.4 of the user's guide, I
think we can assume config tool will be using the 'All Other" column to
determine the IF Target level.  Config tool should then be setting the
balance point in the IF Rack to 1 V.  But, for narrow band work, the optimum
target level for the optical drivers is more like 0.1 V.  In this common
observing case, it's likely we're slightly overdriving the optical drivers.
In addition to changing the noise characteristics, overdriving may also make
the system nonlinear, thereby systematically skewing the calibration.

Fixing this may not be as simple as getting config tool to use 0.1 V for
narrow IF Rack filters.  If we use such a target for narrow work, then the
balancing of the converter rack will remove 10 dB of CR attenuation from
what you've been traditionally getting with VEGAS.  I don't know the typical
attenuation setting one has been getting in the CR.  If the CR attenuators
have been balancing to below 10 dB, you'll end up needing amplification in
order the balance the input to VEGAS.

2) For a single wide bandwidth where the IF Rack filter will be 320-1280
MHz, the current 1V IFR targets are probably OK for healthy driving of the
optical fibers.  

3) Multi-frequency work with widely spaced centers (either wide or narrow
VEGAS bandwidths) that require "All Pass" we'll need IF Targets of ~3 V.
Since config-tool currently sets 1V, we're probably underdriving the fiber.


Next, the balancing of the CR should produce attenuator values that straddle
midrange (i.e., 16 dB, plus or minus a few).  We need the CR attenuators to
have plenty of head room and floor room since many of our receivers (those
labelled class B in table 6.4, plus W-band) have > 10 dB of gain slope
across their instantaneous bandwidth.  

If the CR attenuators don't straddle ~16 dB, then some banks may have maxed
out or zero-valued CR attenuators and some VEGAs banks will either be under
or overfeed.  If we're drastically off, then the only way to get the CR
attenuators near midrange is by adding amplifiers or attenuators after the
CR and before VEGAS (which is what we had to do for the Spectrometer).  If
the straddle point differs by few dB from ~16, then some nudging of the IF
Target levels might be possible but for only those receivers with an
appreciable power slope.

I've hit three use cases, two of which probably have some issues at some
level.  Are there any others that differ significantly from these?

Ron

Discussions of Values

Steven White and Melinda Mello clarified this point. First from Steven:

The linearity of the components downstream are likely impacted more from the
settings of the ODM than the fiber optic links.  The spectral density is the
limiting factor where the target is set at 0.1 V lowers this by 10 dB.
Balancing at 1 V regardless of the bandwidth selected sets the power level
into the optic links at the optimum value (for interference suppression) and
maximizes the signal to noise.  This lower target value may also be to allow
more head room for interference, since the signal to noise ratio is higher
at the narrow bandwidths.  With no interference considerations, there is at
least 10 dB of head room, when balanced at 1.0 V, above the 1% compression
point.

→  Steve  →

and also from Melinda:

For VEGAS the iftarget is NOT being set to 1 for all cases.  The
dependencies are based on receiver first, and then backend. So the target
level are:

For all kfpa observations the target level is set to 5
For all wband observations the target level is set to 1.5
For all other receivers with VEGAS, the target level is set to 1.


If you care for more details for the other backends, read on:


For all kfpa observations the target level is set to 5
For all wband observations the target level is set to 1.5
For all receiver-backend combination beside guppi and the spectrometer the
target level is set to 1.

For GUPPI:
 for backend bandwidths >= 100 the target level is set to 1
 for backend bandwidths < 100 the target level is determined by the
receivers nominal if and the filter in use. I assume it corresponds to
all the values in the table in the users guide, but I only spot checked
it.

Forthe Spectrometer:
  for backend bandwidths >= 200 the target level is set to 1
  for backend bandwidths < 200 the target level is determined by the
receivers nominal if and the filter in use. I assume it corresponds to
all the values in the table in the users guide, but I only spot checked
it.

For the Holography receiver/backend combination the target level is set to 1.

We thus conclude that the values as set in the Table should be sufficient, as the Receiver is the dominant factor in setting these values.