ARRL

Forum Home - Rules - Help - Login - Forgot Password
Members can access, post and reply to the forums below. Before you do, please first read the RULES.

Two Tone 3rd Order IMD Testing

Sep 1st 2011, 19:54

W1RFIAdmin

Joined: Jul 25th 2011, 14:25
Total Topics: 0
Total Posts: 0
Two Tone 3rd Order IMD Testing; Rationale for ARRL's Test Method Reasoning.

ARRL staff have seen some discussions of its test methodology on a number of email lists. I have put together a series of documents that explain why ARRL uses the test methods it uses and how those test methods are derived from industry standards and, in some cases, even international Recommendations through the ITU-R.

As with our Blocking Gain Compression test, there are times in which received noise is present (from reciprocal mixing) that makes it difficult to hear the desired and IMD signal by ear, or by an audio meter at the MDS level; as was the case with our old test method using two signal generators used a few years ago.

Today, due to improved methods*, we use two signal generators (two tone) to create an IMD signal at or very close to the desired signal and combine this IMD signal with a third signal generator representing the desired signal. The combined IMD and desired signal is then fed into the antenna jack of a receiver. The audio from the external speaker jack is then fed into a signal analyzer so we can visually observe both the IMD and desired audio signals and match the IMD signal level with the desired signal to get the IMD level needed to calculate the 3 IMD DR figure and Third Intercept Point figure. Due to the noise created by reciprocal mixing, narrow bandwidths are use with our signal analyzer to assist in measuring both test signals with noise present, down to 20 Hz, if needed.
I’ve noticed on several forums on the web than some question our dynamic range figures and that they are corrupted by noise and do not represent the usable dynamic range of the receiver and we should report this dynamic range as, “noise limited”. This erroneous conclusion can be made if reciprocal mixing is left out of the picture. None-the-less, the 3 IMD level measurement must be done in order to obtain the dynamic range measurement and a Third Order Intercept Point at the MDS level.

The use of a signal analyzer set to a narrow bandwidth is, at times, needed. While third order intermodulation products occur at a lower level than gain compression, noise created by reciprocal mixing is still a problem. However, it must be remembered that reciprocal mixing is not a third order phenomenon, while third order intermodulation is and one cannot use a noise limited value in this case.

A third-order dynamic range test or intercept point test that is noise limited is not a valid test of a third-order phenomenon at all
.
To the extent possible, the ARRL bases its tests on the best possible industry standards and consensus. Although specific to a measurement of intermodulation, the International Telecommunications Union, Radio Communications Sector, has released a report, SM.1837, describing how to do receiver intermodulation testing in the presence of noise:


"Phase noise

If the levels of the IM products at f3 and f4 are near the noise floor or phase noise skirts, the measured levels will represent the IM product level plus noise. The actual IM product level can be obtained by subtracting out the noise contribution."

Although one could measure the noise separately and actually subtract it, this is less accurate than simply discarding the noise by using narrow test instrumentation (our signal analyzer).
Considering the collective opinion of the international scientific community that published a report which advises noise should be subtracted to obtain a 'pure' measurement, I am afraid that the learned collective knowledge expressed through the ITU-RCS trumps the naysayer.

I also read on the web that some have virtually implied that the reason that ARRL Lab uses the newer IMD test methods is to make (ARRL) advertiser's receiver performance look better than they actually are. That may sound good from someone high upon a soapbox, but he should realize when the ARRL Lab changed its methodology to make the dynamic range measurements in the presence of noise, as well as obtain a more accurate and useful result that applies to all signal levels, it also simultaneously started to include in the Product Review test-result tables, a measurement of reciprocal mixing noise.

In reality, a poor receiver will look no better than it did (using our newer method) if blocking gain compression, intermodulation dynamic range and reciprocal-mixing dynamic range are all considered. In fact, because reciprocal mixing is done at the noise floor, it tends to make a receiver look worse than the noise-limited level at the much higher levels of the gain-compression test. We clearly addressed out test method change of IMD testing and the addition of the reciprocal mixing test in QST and we stressed the fact that reciprocal mixing is a stringent noise test. The ARRL Lab strives to use the best test methods possible with the purpose of reporting what is fair, accurate and impartial. While our test data does help the manufacturers, who strive to improve their product, our test data is presented entirely as a service to our members in order to help them decide what equipment to purchase. As a reader of this forum, please consider all three key tests: Blocking, 3IMD and Reciprocal Mixing.

By the way, some forum goers have taken exception to our use of 0 dBm by the ARRL Lab as one of its test tone levels for intercept-point measurements. The ITU RCS in SM.1837 also addresses this issue, specifically for IP3 measurements:


"The permitted level range of test signals f1 and f2 at the antenna input of the monitoring receiver is −30 dBm to 10 dBm."

The international scientific community finds 0 dBm IMD level testing appropriate, so ARRL intends to continue to do testing at this level, in addition to the testing it does at the noise floor and at a level that represents S5 signals and intermodulation products.

One last point, there is a misconception by some, as seen on a few list postings, that the ARRL Lab measures the 3IMD DR of receivers with AGC off. That is incorrect. For ARRL's measurements of two-tone, third-order dynamic range, intercept point, the receiver AGC is left on. Technically, the AGC can be left off at the MDS level, but for higher IMD levels and for most IP3 measurements, AGC must be on, as the measurement is being made well into the region where the receiver would not be working correctly without AGC.

Bob Allison, WB1GCM
ARRL Test Engineer

*First reported with the Yaesu FT-2000 in February, 2007 QST
Sep 2nd 2011, 20:39

AB4OJ

Joined: Apr 4th 1998, 00:00
Total Topics: 0
Total Posts: 0
The ITU-R have made many of their Recommendations freely downloadable from their website, so here is the link to SM.1837:

http://www.itu.int/dms_pubrec/itu-r/rec/sm/R-REC-SM.1837-0-200712-I!!PDF-E.pdf

Section 3, 3rd paragraph (p. 6) calls for a "realistic choice" of RBW:
--------------------------------------------------------------------------------
The IF filter bandwidth or resolution bandwidth (RBW) for the measurement can be chosen by the manufacturer but must be a realistic choice for the type of receiver and intended application. The chosen bandwidth (BW) for the frequency range 9 kHz to 30 MHz must be ≤ 5 kHz and for the frequency range 20 MHz to 3 000 MHz it must be ≤ 30 kHz.
--------------------------------------------------------------------------------
However, Section 2, penultimate paragraph (p. 5) speaks to the case where phase noise is dominant:
--------------------------------------------------------------------------------
Phase noise

If the levels of the IM products at f3 and f4 are near the noise floor or phase noise skirts, the measured levels will represent the IM product level plus noise. The actual IM product level can be obtained by subtracting out the noise contribution.

--------------------------------------------------------------------------------

This appears to validate the ARRL's current DR3 test method (at 3 Hz RBW) within the narrow context of the test engineer wishing to determine DR3 and thence IP3, with a view to comparing the usable sensitivity attributable to IMD3 products with that traceable to phase noise (reciprocal mixing noise, RM). One could say that the ARRL test method would be more universally applicable if the DR3 test results were placed side-by-side with those of the RM test, with a clear statement as to which of the two degrading factors, IMD3 or phase noise, is dominant at each spacing tested. The RM data should be presented as ranking equally in importance to the DR3 data.

I believe that one could conceive of presenting the two sets of data -
DR3 or IMD-free dynamic range measured at 3 Hz RBW, and reciprocal mixing noise - side-by-side in a table, with a clear and unequivocal statement as to which of the two parameters determines the receiver's usable sensitivity.
_____________________________________________

Example: IC-7410. 14.1 MHz, 500 Hz CW, 3 kHz roofing filter.

2 kHz spacing/offset, DR3* = 88 dB, RM = 78 dB, RM dominant

5 kHz spacing/offset, DR3* = 98 dB, RM = 88 dB, RM dominant

20 kHz spacing/offset, DR3* = 103 dB, RM = 101 dB, RM dominant

* DR3 is measured at 3 Hz resolution bandwidth (RBW) to subtract out the noise contribution.

It will be seen from the above comparison that for the frequency spacings tested, synthesiser phase noise is the dominant factor in determining the receiver's usable sensitivity. To avoid confusion on the part of the reader/radio shopper, it will be necessary to present and explain this data comparison very clearly. This will also increase the incentive to the OEM's to improve the phase-noise performance of their synthesisers.

73, Adam Farson VA7OJ/AB4OJ

Sep 5th 2011, 00:31

W1RFIAdmin

Joined: Jul 25th 2011, 14:25
Total Topics: 0
Total Posts: 0
Hi, Adam,

Thanks for the input.

One of the reasons that ARRL added the reporting of reciprocal-mixing noise at the same time it used a narrower bandwidth to determine IMD and gain compression in the presence of noise is that in many cases, noise, not intermod or gain compression, IS the limiting factor. From the perspective of we engineers, we tested, it was reported in the review, so it was a "done deal."

However, from the perspective of the discussions I have seen about the testing, it does appear that was seemed obvious to we test folks may not be as obvious to all readers, so suggestions on how to present the data in a more useful way are always welcome. I can't speak for the editors, but I do get to speak with them often.

I personally like the idea of ensuring that the reciprocal mixing information is presented in a way that puts it 100% on a par with the IMD and gain-compresion dynamic range. Great minds do think a like, right? :-)

73,
Ed Hare, W1RFI
ARRL Lab
Technical forum moderator
Sep 5th 2011, 00:39

W1RFIAdmin

Joined: Jul 25th 2011, 14:25
Total Topics: 0
Total Posts: 0
Quote by AB4OJ

--------------------------------------------------------------------------------
However, Section 2, penultimate paragraph (p. 5) speaks to the case where phase noise is dominant:
--------------------------------------------------------------------------------
Phase noise

If the levels of the IM products at f3 and f4 are near the noise floor or phase noise skirts, the measured levels will represent the IM product level plus noise. The actual IM product level can be obtained by subtracting out the noise contribution.

--------------------------------------------------------------------------------

This appears to validate the ARRL's current DR3 test method (at 3 Hz RBW) within the narrow context of the test engineer wishing to determine DR3 and thence IP3, with a view to comparing the usable sensitivity attributable to IMD3 products with that traceable to phase noise (reciprocal mixing noise, RM).

That was my interpretation of the clause, as well. One way to "subtract the noise" would be to measure the noise separately, then do an actual subtraction. This can be done pretty well when the signal and noise are within about 6 dB of each other, but to get down further into the noise requires more and more precision on the measurement of both the noise level and of the signal in the presence of the noise level. Unfortunately, noise is, well... noisy, and that precision is almost impossible to obtain,

ARRL chose instead a more accurate way to subtract the noise, and that is to simply use a narrower measurement bandwidth. That, coupled with a separate measurement of reciprocal mixing, does give a rather complete picture of the the receiver's performance.

It is, IMHO, important to report intermodulation and noise separately, as the intermodulation result, although lost in the noise at lower signal levels, will increase rapidly, and the reporting of actual intermodulation dynamic range and intercept point (IP3) allows some extrapolation between the intermodulation at the noise floor and the intermodulation separately measured at about S5 levels of intermodulation product. Even ten or twenty dB above the noise floor, reciprocal mixing is usually less important and, in real use on HF receivers, most band noise is at least ten or twenty dB above the receiver's noise floor.

73,
Ed Hare, W1RFI
ARRL Lab
Technical forum moderator
Nov 10th 2011, 14:11

KF5GTX

Joined: May 12th 2010, 11:05
Total Topics: 0
Total Posts: 0
Is there a list of current transceivers 3rd order IMD specs and what frequencies were used during testing?

73,
Bobby Duncan KF5GTX
Nov 10th 2011, 14:46

WB1GCM

Joined: Apr 4th 1998, 00:00
Total Topics: 0
Total Posts: 0
Bobby,

Our Product Review summary is a bit dated. However, one of our valued Lab consultants, Rob Sherwood, has an excellent chart to refer to :http://www.sherweng.com/table.html. His test results always agree with our results.

It is hoped that I will have the time someday to built an brand new summary chart. I have to start from scratch due to the methods used with creating pages in our web site.

Bob Allison

Back to Top