[TenTec] More IP3 Stuff

[Hare,Ed, W1RFI]

>First off:  an epiphany!  I had assumed all was being done
>in the linear region of the rcvr to audio transfer characteristic;
>that is,  NO AGC action at all!!  It had not dawned upon me that
>the AGC would be active during a test for a rcvr's linear
>characteristics and IP numbers....

AGC "on" is the ONLY way to make receiver measurments with a receiver
output that would be more than about 30 or 40 dB above the noise floor.  
 
But the effect is not quite what you are thinking.
 
In making an IP3 measurement with AGC active, the steps are as follows:
 
   Set a reference generator on channel at a level that gives an S5 recevier output
   Log its level
   Replace that generator with a two-tone signal spaced to give the desired IMD
   Adjust the level of the two-tone signal until the receiver gives an s5 output
   Log the level of the two tones and do the IP3 calculation from those levels
 
As can be seen in the sidebar, for some receivers, ACG has no significant effect on
IP3.  I will speculate that in these receivers, the onset of AGC is smooth and the 
filtering and AGC circuitry is such that the strong off-channel signals do not affect
the gain of the receiver in any significant way, at least at the levels used for the test
tones.
 
The effect of AGC on the tests is the same for the desired on-channel reference level 
and for the intermod product, so, in theory, the AGC effects should balance and what
is left is the IP3 response of the receiver. On the sidebar results, you will note that 
the levels of test signals varies by tens of dB, but even over the range where AGC
is active, the IP3 does not vary all that much, until we are getting at S9 and above, 
and even that for only some receivers.  In measuring IP3 with AGC active, we really 
are measuring the receiver the way it will be used in most cases.  Those that really
want to know the linear-range IP3 can calculate it from the noise floor and the
measured two-tone, third-order IMD dynamic range. When I started doing testing at
ARRL, that was the ONLY IP3 that ARRL calculated; we changed to higher test-tone
signals about 10 years ago, at the suggestion of Ulrich Rohde.
 
In the testing I did for the sidebar, I saw two effects. One was that "burble" seen on the 
third-order response just before AGC was sending the receiver into AGC compression.
I am not sure what aspect of the receiver design caused this, but it really happens and it
does indeed move up the intermod response by a few dB for intermod products that are 
just below the AGC threshold.   The other effect is that in some receivers, S9-ish level
test signals seem to also result in a higher IP3. My speculation is that the off-channel
signals, which are about 50 dB over S9 at that time, btw, are somehow pumping the AGC
a bit and reducing the gain of the receiver, moving up the apparent IP3.
 
To summarize, ARRL has chosen and stuck with the S5 level because it represents a
reasonable use of the receiver. Few listen to signals that are at the noise floor of the
receiver; even on VHF receivers, ambient noise is often ten dB above the receiver noise
floor, especially when preamps and high-gain antennas are used. So most folks would not
be affected by an intermod product appearing at the recevier's noise floor. An S9
intermod product is usually caused by undesided signals that are 50 or more dB higher
than S9, an interference situation not often encountered.  An S5 intermod output is
going to cause interference, and it is typically caused by S9+30 or +40 dB signals, at 
least something sometimes encountered in real-world situations, so an S5 reference
level for the IP2 and IP3 tests is a reasonable number that will give IP3 measurements 
of the receiver under conditions that approximate actual use.
 
> As Ed reports from his screen room experiment:
> "to measure IP3 in a receiver into full ACG compression...
> it looks like that at some point above S8 or so, the IP3 starts to
> rise."  And it certainly does,  because the AGC is causing the
> rcvr system to behave exactly as it would were ATTENUATION
> being added at the front end!  
 
No. This would be true for IP3 measured inside the receiver's ultimate
selectivity, but for test tones that are outside the selectivity, the intermod it taking
place *before* the part of the circuit that is controlled by AGC, generally,
so the only affect that AGC should have on the measurement is to affect the
receiver output level, the same amount for the on-channel reference signal
as for the intermod product in the same channel.
 
> In fact,  Rohde plots just such  behavior,  I now see,  in Fig. 45 of the 
> Jan/Feb QEX piece, where he illustrates with a curve,  and adds this text of
> explanation under the curve,  "A plot of intercept point behavior for the 
> receiver system (then he refers to a complete rcvr block diagram,  his Fig. 43).  
> The RF attenuation activated by the AGC voltage IMPROVES [increases] the 
> third order intercept point."  He shows that the IP3 of this particular
> rcvr is a constant +20 dBm with no AGC  from the MDS level
> up to the area of between S7 and S8 input signal pair level.

In receivers whose ACG also controls the RF amplifier, or stages before
where the intermod product is occurring, this can indeed be true. If ACC is
used to control the RF amp stage, it usually has a considerable threshold --
an area where the AGC does not affect the stage gain.  Not many receivers have this
design, though.  Also, note that the "S7 and S8" level signals Ulrich is
discussing are NOT the levels of "off-channel" signals, but the level of the
reference being used for the test.  If the reference level is an AGC voltage that
gives S7 to S8 on the receiver's output meter, then if the RF stage or attenuation
is affected by that level of AGC, then indeed, the IP3 level *for those test signals* 
will be increased. If one is dealing with S9 signals and inteferors greater than
S9+50, that would be a reasonable measurement.  
 
Do note, too, that the receiver architecture being described by Rohde is not common 
in amateur equipment. And it is does start being used, ARRL should probably start testing
receviers for their ultimate attainable IP3, although I am not convinced that testing a 
recevier's sucscptibility to S9+50 dB signals is a necessary test.
 
And again, in choosing an S5 reference level for the tests, most receiver's would be
below the threshold of this phenomenon anyway, and if not, would hams not want to know
something about the intermod performance of a receiver at a level that represents both
the signals apt to be encountered on the bands and an intermod product that represents
interference that would just begin to affect most amateur communication?
 
73,
Ed Hare, W1RFI
ARRL Lab