Now here again, Ed has responded to my last post. He
has also requested that I repost to the reflector, so here
goes (this one came here about only 10 minutes ago):
"
> I am especially pleased that you plan to review, and
> probably do some updates to the procedures, Ed. I like the
> idea of a standard input signal level for the pair of tones;
> the Collins Radio idea of using -109 dBm plus the rcvr noise
> figure will be somewhere near a particular radios S5 meter reading I
> would bet, hi.
Actually, by "Collins" standard, I was referring to the Collins S-meter
standard, where S9 is 50 microvolts (-73 dBm at 50 ohms) and each S unit is
6 dB. I don't know what levels Collins used/uses for their IP2 and IP3
measurements. My guess is that they would have derived their IP3 from the
noise floor and dynamic range measurements. Many of the older radio
references *define* IP3 as a relationship between sensitivity and dynamic
range. This is a reasonable assumption for most receivers, although subject
to all the factors in the sidebar I wrote for Doug's article. Other texts,
especially dealing with high level mixers, where there may NOT be a noise
floor to measure (a passive mixer, for example) use a more generic
definition of IP3. Considering the way that the "real" IP3 varies with
signal levels in some receivers, a signal level for IP3 measurements above
the noise floor is quite justified -- it just becomes a matter of selecting
an appropriate signal level. I won't repeat the material in my earlier
post, but if you repost it as I had asked, folks can read all about my views
there. :-) As a quickie here, I will point out that any refinements ARRL
makes will NOT make the difference between night and day on the results; the
measurements as made are reasonable and accurate and are measuring the
equipment under test, NOT measuring the test instrumentation. I have been in
test engineering for 25 years now, so I know where to draw the lines!
> Or maybe just settle on the -97 dBm level for all radios;
> haven't quite gotten all the subtleties you have outlined
> digested just yet!
That could work for me, except that there can be a 20 dB difference in
sensitivity between different radio designs. An 80-meter receiver with
a -120 dBm noise floor would be QUITE usable on 80 meters, but -97 dBm would
really be a different level inside the receiver than it would for a receiver
with a -144 dBm noise floor. And I could imagine an insensitive receiver
with a high AGC threshold that might not be showing any AGC action at all
at -97 dBm. It bears more thought and some investigation.
W1RFI wrote:
>> The only sticking point appears to be dynamic range. Most agree that
>> testing IP3 at a level somewhere in the AGC range of the receiver is
>>the best test, although I don't think that most hams understand how
>>and why IP3 varies in real receivers based on signal levels. If
>>nothing else comes of this, understanding that may help hams
>>to stop quibbling over a dB or two.
>> Those who work with high-level mixers and/or visualize receivers whose
>> non-linearity's follow a classic response tend to think in terms that
>> noise floor, dynamic range and IP3 are all precisely related.
>>Ulrich, for example, whose opinion I greatly respect, has stated
>>that dynamic range can be calculated from the noise floor
>> and an IP3 measurement made at a high level. When you consider that
>>the formula ends up assuming that the 1st-order and 3rd-order
>>responses are 1:1 and 3:1 sloped and makes a calculation for
>>the levels at the noise floor of the receiver, it is, IMHO, more
>>accurate to make a measurement at the noise floor because it can
>> be done. No one has yet convinced me that making a measurement
>>at a higher level and assuming ideal slopes of lines that are
>>probably not ideal is a better way to determine dynamic range
>>than actually making a measurement that can be made
>> accurately.
I will clarify that in the above paragraph, I am discussing making IP3
measurements at at S5 receiver output level (approximately -109 dBm, if
memory serves) and making dynamic range measurements at the noise
floor, the very definition of dynamic range.
>> Of course, one must always remember that Product Review is
>>dealing with a sample of one. IP is really a very nebulous number
>>for a number of reasons, and I have just scratched the surface in my
>>explanations. When we are dealing wtih receiver intermod with
>>receivers that are linear over 90-110 dB, look at how small
>> a deviation from perfection a linear range of 100 dB would be.
>>Differences of fractions of fractions a percent in the non
>>linearity of a mixer can make tens of dB
>> of difference in the dynamic range. And if a manufacturer derives
>>his IP3 from the measured linearity and the specified receiver
>> sensitivity and makes a receiver more sensitive than the spec,
>>but whose linearity is as specified, the IP3 calculated from
>> the actual measurements is lower than the spec. So if they
>>make a better receiver, it doesn't meet spec."
> Somewhere, Ed, in your above two paragraphs must lie the
> reason for the disparate numbers that Ten Tec is listing for the
> IP2 and IP3 numbers of their new Argo V, vs. the numbers your
> lab has reported in the April 2003 QST. I went to the
> Ten Tec web site this morning, and yes, they do list IP2
> of +66 dBm, and they also add that the ARRL Lab test method
>was used to determine both it and their listed IP3 number of
>+4 dBm, both tested using 20 kHz signal pair spacing.
> However, they make no claims about 5 kHz signal spacing.
> I thought it was especially interesting that they specifically
> mention and list use of your ARRL Lab procedure,
> that is they used a power level for the two input signals such
> that the S meter on the Argo V read S = 5. And who knows how
> accurately that was read by them, or, as you suggest how
> accurately the S 5 setting was read by you folks during the test.
> The S meter does not have a mirror behind the needle as the old
> time Triplett meters did to aid in the elimination of parallax, hi.
Parallax is not going to make that much difference. If the reading were
S4, S5 or S6, if those lines of first- and higher-order responses were
straight, the calculated IP3 would be the same, as long as the parallax
error were the same in both cases. If the lines are not ideal, and they
probably aren't, the IP3 calculation difference would be different only by
the amount that the first- and higher-order responses deviated from ideal
over that small portion of the curves.
To my knowledge, the rig never did make claims wrt 5 kHz spacing. What you
saw on the TT Web site was exactly what we were told the specs were when we
bought the rig. I think that some may have confused this rig with the Orion,
which DOES make claims for 5 kHz spacing.
I went into the screen room today and set up our Rohde and Schwarz ESH-2
EMC receiver and made some IP2 measurement. I measured the IP2 of
the R&S at +86 dBm or so at a number of different levels, indicating
that the test-fixture intermod is at least that good. In the past, with our
older
HP8540Bs, we had determined that the IP3 setup was good to at least +33 dBm
and that IP2 measurements could be made to at least +75 dBm. Our newer
Marconi generators are even better, and using a separate test fixture
donated by Ulrich Rohde, we had verified the IP3 capabilty to at least +40
dBm, and possibly higher. I presume that the IP2 capability is similarly
increased between the HPs and the Marconi, and I believe by the way that the
step attenuators were behaving that I was measuring the ESH-2 directly
(every dB of attenuator changed the product by 2 dB for IP2).
I may have a bit more to add after I do some testing of the rig again, but
I want to coordinate this with Ten Tec before adding anything to what was
said in the review. The review process contains checks, balances, retests of
anything at all out of line, consultations with multiple engineers in the
ARRL Lab and full communication with manufacturers before it all goes to
press. I want to give what I say here about any specific product the same
full consideration. We have, btw, made mistakes in the lab testing, but
this system of checks and balances has, to the best of my knowledge, caught
them all, though a few have been caught by the manufacturers at the 11th
hour. :-) And I thank them for it, because manufacturers know their products
better than the ARRL Lab dudes could.
I will also point out that MOST of the time, though, when a manufacturer
questions a result, the original result has held up. A few manufacturers
have even sent their engineers to HQ to work with us on a problem and when
all was said and done, a Feedback in QST was not necessary. :-) I believe
we had to run one Feedback because one of the Lab guys made a calculation
error from raw data. That one broke a 15+ year record of no Feedbacks due to
a Lab error, thanks in part to careful cross checking and good manufacturer
communication. :-)
> On the other hand, Ten Tec has listed some outstanding numbers for IP2,
> IP3, and phase noise performance of their soon to be released Orion
> rig. BUT, their beautiful IP3 curve of +25 dBm typical for all signals
> spacings down to about 5 kHz spacing, followed by a gradual roll
> down to about +21 dBm at 1 kHz spacing was performed using two
> input signals of 0 dBm, 1 mW, 0.22volts, or S = 9 plus 73 dB, if
> I did my arithmetic correctly, hi. See their Fig. 1 on down this
> page about the Orion from the TT web site:
> http://www.tentec.com/TT565.htm
>
> They have certainly performed their IP3 test to produce that curve well
> up into "the AGC range of the receiver as the best test" to quote
> from one of your paragraphs above, hi. But that is way above where
> the Orion S meter would read S = 5 !! Bet there will be a difference,
> based upon your past results, between the IP3 as measured with
> 0 dBm signals and when with something like - 97 dBm signals.
That is quite a bit higher than the S5ish levels we would use. I don't
want to cross that bridge until we come to it, but that is a lot higher a
level than I would use to make the test. I have not measured a lot of rigs
at an S9 level, but one of the rigs we measured for my sidebar showed quite
an increase in IP3 for S9 level recevier output. The Orion is, however, a
quite robust SDR type rig, where much of the performance is in the DSP used
for signal generation and detection, so some differences in the way it is
tested -- or even should be tested -- may be in order. Naturally, Mike will
work closely with Ten Tec. If nothing else, it will be a very interesting
expanded test-result report. :-)
For most receivers with final selectivity done at IF with filters, one
could possibly not really measure IP3 with 0 dBm signals. Those signals
would pin the S meter, so it couldn't be used as an indication of received
signal level. The filtering would not allow the test signals at 5 kHz
spacing and the resultant IMD products to all pass through to the receiver
output amplified exactly the same. But with DSP filtering, it may be quite
possible to gain access to the IF before the final DSP and look at the test
tones and resultant IMD products. If so, one can use those levels to
calculate IP3, at least before the DSP. To know how ARRL would want to
apply all this, I really would want to look at the radio, do some
preliminary testing and talk to the designers (and possibly a few Technical
Advisors!).
But this DSP/SDR stuff that Ten Tec is doing is personally very exciting
to me, so I can't wait to see this radio show up in the Lab. I may kick
Mike Tracy out of the screen room, or at least spend a lot of time hanging
over his shoulder. :-)
>> Did you get a Certifi-Cat for working the TT2?
> No I did not, but probably my fault, as I did not send Dean, KH6B a
> QSL for the contact, so he didn't bother to send something, hi. Anyway,
> I have too many QSL's in drawers here already as well as a few hundred
> hanging on my shack wall!
Send me the contact info, including both RSTs, and I will send you a
Certif-Cat! Trust me, it is WELL worth asking for it. It is a one-of-a-kind
sorta' thing. :-)
> Enjoy these discussions, Ed
As am I. I hope that when all is said and done, folks will realize that
the ARRL Lab *is* a world-class test facility. I saw another post on that
subject that is a better place to address that, so I am waiting for my sub
to be approved and my posts to show up.
Please post this one to the reflector, too, and we can let people catch up
on where it started later. :-)
73,
Ed Hare, W1RFI
ARRL Lab "
|