[evlatests] Setting levels
Robert Hayward
rhayward at aoc.nrao.edu
Thu Sep 1 09:54:47 EDT 2005
Comments on Power Level Setting and the T304:
From the Front-End Group's perspective, the purpose of the T304 is not
just to adjust the output level to the Samplers (i.e., the canonical -33
dBm) but to adjust the power coming from the receiver to ensure it is
optimized before presenting it to the subsequent post-amps and mixers in
the downconverter itself. For receivers with lots of gain, like the C
and K Bands, the signal will require higher RF attenuation. If not set
correctly, we could easily run into compression down stream (remember
our headroom spec is to be 20 dB below the 1% compression point of any
active device). For receivers with low gain, we need less RF
attenuation. If not set correctly, it might significantly degrade the
noise figure of the subsequent post-amps in the downconverter and
increase the overall system temperature (our goal is less than 2%
increase over the receiver's noise output).
So there is a delicate balance, which hasn't really been truly
characterized yet on the EVLA antennas. I did generate a noise model
which attempts to estimate the incremental noise contribution and
dynamic range of each device in the signal path all the way from the
input to the receiver’s feed right through to the output of the T304,
but it is far out of date (lots of integrated designs have arisen in the
interim). I will try to update it soon so that we can identify where the
critical points are for losing sensitivity or hitting compression.
All receivers are not created equal. The output power levels can vary by
several dB between polarization channels. Receivers of similar frequency
bands can also have gain differences of several dB. Thus we can easily
have 54 receiver channels all with different output power levels. Each
receiver band will have different average output power - K's have lots,
X's are wimpy. The frequency response of the broad band receivers will
also have a strong effect on the amount of power delivered to the T304.
The Q Band's high end will be at least 10 dB weaker than the low end
(due to the frequency response of the LNA's, post-amps and mixers). The
K-Band's gain rolls of strongly at the band edges, on the order of 5 to
10 dB lower than the center of the band. Throw in the roll off in the
cables and frequency response of the UX & LSC converters, there is no
way we can present a "perfect" input level that the T304. So we need the
RF Attenuator, and the Input Total Power Detector, in the T304 to
optimize the power level before the Output Attenuator gets around to
trying to make the Sampler happy.
In item 3 of Barry’s email, he asks about the need for the Output
Attenuator, especially in the 8-bit mode. Again, it is a delicate
balance to optimize sensitivity and headroom. If the Output Attenuator
were set to zero and the RF Attenuator has to provide all the gain back
off (for K & C-Band, it might behave to be set to 25 dB or more), we
might see the noise floor of the T304 start to rise. It makes me think
of the T108 frequency converter used in the VLBA, where its 500-1000 MHz
IF amp is rather noisy. If you disconnect the input coming from the
W-Band receiver, and you crank up its gain (it has a pot you can
adjust), the ALC in the Baseband Converters will see enough “signal” to
adequately servo to the appropriate level for the VLBA samplers. But
it’s all noise coming from the IF amplifier chain. Except for the fact
that there is no switched power, you wouldn’t know that the receiver had
been removed. So we should examine the entire receiver, frequency
converter and T304 signal path to come up with the ideal criteria for
setting both attenuators.
Another point. We have made a somewhat underwhelming attempt to provide
a Solar Mode on the EVLA that is comparable to the current VLA, at least
for the L-Band & X-Band receivers. Paul Harden has designed 20 dB pads
into the LSC Converter which can be switched in so that, in principle,
the “interim” L-Bands can be used with the rest of the VLA (new EVLA
L-Bands will sport the new Lilie “coupler-fed” Solar Mode, but that’s
another story). The “transition” X-Band receivers may be on the Array
for a long time should the new EVLA X-Band design get descoped. The
signal path in the X-band does not go thru the LSC and thus does not see
a switchable 20 dB solar pad. Thus the T304 RF Attenuator may have to
fill that roll. This was another reason for dropping back to the
low-gain VLA-style X-Band receiver yesterday (although having to buy 60
new higher gain amps and devoting the extra effort to retrofit them into
the receivers was really the deciding justification). We need to do some
further analysis to see if the T304 RF Attenuator will adequately allow
X-Band receivers to observe the sun without any compression problems.
So, as I see it, the RF Attenuator should be used to optimize the input
level to minimize sensitivity degradation and preserve headroom (i.e.,
to compensate for the wide range of input power coming from the myriad
of receivers). The Output Attenuator then sets the level into the sweet
spot of the Sampler (in 8-bit mode, this is the canonical rms of 5-10).
This nominally requires about -33 dBm from the T304. If we use the
Output Total Power Detector to find this spot, we must both be in the
linear region and ensure that all the subsequent amps that follow in the
Sampler module have identical gain. I suspect neither are likely to
happen with any great degree of accuracy. Ideally then, having the
T304's use the sampler rms in the feedback loop used to adjust the
Output Attenuator would give the most optimized Sampler levels. But this
scheme gives the software folks a fit. For a reason which I haven’t
understood yet, there is a general reluctance to have peer-to-peer
communications between the MIB’s. This sort of restricts the advantages
of all the distributed intelligence we have in the system. However, if
it was implemented, wouldn’t such communication only need to occur
during setup changes or source changes (and, as Barry points out, Solar
bursts would be a pain - is this the Solar tail wagging the VLA dog again?).
One of Barry’s questions (in his item 2) was whether we need a
synchronous detector in the T304 at all. I’d say yes, regardless of
whether it is a useful function astronomically. It is a valuable tool
for receiver checkout. We use the same function in the F4's on the VLA
antennas all the time for monitoring the health of the receiver. And we
do the same with the VLBA antennas using the Screens program. Having a
means to measure Tsys in real-time is an extremely desirable diagnostic
feature.
Of course one of the problems is that the current Total Power Detectors
used in the T304 aren’t quite matched for the new Sampler input level
(-33 dBm). If I have my information correct from Travis, the typical low
end is -35 dBm, but some TPD’s work fine down to -37 dBm but others give
up at -32 dBm. This range was adequate when the desired Sampler input
level was about -27 dBm. But Mike’s mod to the digital processing system
has dropped our signal levels by several dB so the TPD’s are forced to
work at the bottom of their range. What I’d like to see is a re-spin of
the TPD circuits that have a 10 dB of gain or so in front of each of the
detectors so that the signal gets into a more desirable range. Travis
says these mods would not be overly expensive (perhaps about $75 per
module).
Lastly, what about our Gain Slope Equalization scheme. As pointed out in
EVLA Memo 80, the frequency responses of some of the receivers will lead
to severe gain slope, perhaps as high as 10 dB or more across the 2 GHz
bandwidth of the 3-bit samplers. And this was just the receiver itself
and excludes all the other converters and cables. I believe Rick has
relaxed the original Project Book slope specifications (which we could
never have met anyway, I think) but the Gain Slope Equalizer will still
be needed (we have not been told otherwise). So then how do we determine
which slope setting will be required for all the various combinations of
antenna vs. receiver band vs. frequency setting vs. receiver/T303/T304
serial number vs. God knows what else that we will encounter. I think
any look up table we would come up with would be so overwhelming to
build and maintain that we really need some sort of real-time system. I
have always hoped that WIDAR could calculate the autocorrelation
function so that the mean bandpass slope could be determined. This would
allow the appropriate gain slope correction setting in the T304 to be
chosen to provide the ideal compensation. Is this a too unrealistic a
feature to be implemented?
-Bob
Barry Clark wrote:
> I have been operating under the assumption that the way we would run
> the IF system of the EVLA antennas is that we would use the detectors
> in the T304 to set the levels to the digitizers, which, because they
> are multibit devices, have a reasonable latitude for what they are
> fed. From the the discussion at Monday's test meeting, it appears
> that the T304 detectors may not have enough accuracy for the purpose.
> This note is to ask a few questions leading to a redesign of that
> concept.
>
> In any event, it appears to me that we have had an entirely misplaced
> emphasis on the precision, as opposed to the accuracy, of the T304
> detectors.
>
> I ask a few questions below. I also answer them to some degree,
> rather tentatively, incompletely, and subject to revision.
>
> 1. To what accuracy need the level be set going into the digitizer?
> By actual measurement on the eight bit sampler transition system
> (adjusting the level going into the digitizer and making compensating
> adjustments on the gain before the intra-FIR resampler in the
> deformatter), the answer appears to be that an error of 2db in the
> level setting results in a loss of SNR of about 4%. (The optimum
> level is determined by noise in the digitizer on the low end
> (and possibly some clipping loss), and by aliasing of clipped
> highs on the high end.) One db accuracy would seem to be a reasonable
> goal. The corresponding factor is surely known by calculation for
> 3 bit samplers, but I do not have that number handy.
>
> 2. Is there ever a need to have a synchronous detector in the T304?
> I think that in the usual cases, synchronous detection should
> be done in the T5 for the transition system, and in the WIDAR itself
> for the final system. However, I am very unsure whether such a thing
> might be needed for solar mode.
>
> 3. What is the function of the T304 output attenuator?
> In three bit mode, it serves to equalize the power levels between
> the 1 and 2 outputs. In eight bit mode I do not believe it has
> a function, and should always be set to zero. Is the best strategy
> in three bit mode to always set the output attenuator in the channel
> with the lower throughput to zero?
>
> 4. What is the function of the T304 input detector?
> None.
>
> 5. If, as seem likely, the detectors in the T304, elegant as they are,
> are inadequate to the task of level setting, should they be replaced
> by something simpler - a diode, a 1 sec RC, and a SPI A/D?
>
> 6. Given that the digitizer itself must serve as the main level detector,
> what is the best strategy for doing that?
> Clearly the only acceptable mechanism is through the data capture
> buffer in the D30x. There may need to be a tradeoff between
> responsiveness (which I believe is important for solar observations),
> complexity of software, and saturating various system elements with
> traffic.
> Alternative designs are:
> a) Moving the T304 output attenuator into the D30x module, under
> control of the D30x MIB.
> b) MIB to MIB communication.
> c) A special ALC program running in the CB that just acts like a
> detector, turning around the messages from the D30x and sending
> to the T304.
> d) Closing the the ALC loop within the AntennaPhysical software
> element.
> e) Recording, in the system parameter database, the relation between
> the T304 output detector and the actual power level presented to
> the digitizer.
> As a firm believer that hardware problems should be solved as close as
> possible to the hardware in question, the above is ordered by desirability
> in my eyes.
> _______________________________________________
> evlatests mailing list
> evlatests at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>
More information about the evlatests
mailing list