About the AmQRP Club
QRP Forums

Technical Discussion Topic

"Effects of Diode Linearity on Measurement Accuracy"

FROM: George Silver KD4WSI ...

The 1N5711 diodes used in the AA908 Reflectometer, and as positive amplitude clippers on U3 will have slightly different forward Vf curves one to another.. The lower the current thru them the closer to the steep knee and greatest nonliniarity experienced.
  When 50 ohms is used, the internal impedence of the reflectometer is close to 50 ohms so the currents in the legs of the reflectometer is nearly evenly balanced.
However as the resistive or reactive loads are increased the currents thru the 1N5711 diodes will also decrease for some and then the lesser current, and the greater nonlinearity experienced will create a mismatch error with  sensativity to small self heating changes and changes in Vf of the diodes.

It is some of this that is effecting your readings as a function of frequency and time from turn on

 This has been a never ending problem with full wave diode multiplexers and demuxers for years.

  Good ones have all diodes of  the same build lot. then selected to match the Vf curves as well potting them all tightly together so that the temperatrue of all are nearly the same.  Still some nonlinearity match effects sneak in at low or very low signal and current levels.

In some High accuracy needs -applications, a exteranal adjustabe bios is established for each of the 4 diodes of a Detector or Demuxer and that is used to set the operation currents thru the diodes at points along the flatest part of the Vf curves where it is easier to match the set of diodes.

FROM: Joe Everhart, N2CX ...

Indeed there may be some non-matching issues with the diodes we use in our detectors.  That's one of the things we are investigating.

I'm not sure what you mean by "positive amplitude clippers".  You may be describing the prebiased diodes in the feedback loops on the amplifiers fed by the diode detectors.  This compensation method has been used for quite a while in RF power meters and have been described in some detail in the ham literature:

1. Jan 1987 QST p. 18, The Tandem Match--An Accurate Directional Wattmeter, John Grebenkemper, KA3BLO

2. Aug 1990 QEX p. 3, Calibrating Diode Detectors, John Grebenkemper, KI6WX

3. Feb 1990 QST p. 19, A  Simple and Accurate QRP Directional Wattmeter,  Roy Lewallen, W7EL

We chose the diode current to give reasonable performance at low signal levels where the diodes are indeed quite non-linear.

Our original prototype testing and circuit simulation seemed to indicate that matching of the diodes was not needed for reasonably good  performance. We are undergoing comparative testing to identify the mechanism for the observed errors in measurements at other than 50-ohm situations.  None of the testing we have done to date indicates any frequency effects on the detectors in the HF region. The major cause for error to date has been input-level related errors in the detectors with our simple circuitry.

The interesting thing is that the multiple prototypes and breadboards that we have set up with the current circuit configuration and software do not seem to exhibit the problems to the degree noted by some in the field. We are attempting to duplicate the same effects to study them.

The bottom line is "stay tuned!"

FROM: George Neron, N2APB ...

Indeed this whole area of low-level detection, and the techniques we used to implement it in the Micro908 reflectometer and buffers, was strongly considered during the 2-year design cycle for the product.  In fact, we provide similarly-detailed analysis and data in the Technical Reference Manual located online at the project web page http://www.amqrp.org/kits/micro908.  Sure hope all kit owners have read through that document, as it provides some great insight to the design solutions.    

That all said, George Silver is right on the mark regarding the challenges confronting designers looking to reliably detect and correlate multiple-channel, low-level RF signals.  The technique that Joe implemented in his design for the reflectometer, as cited in the Grebenkemper/Lewallen references below, utilizes a diode in the first buffer amplifier of each channel to compensate for the nonlinearity presented by the diode upstream in the reflectometer.  In other words, the nonlinear current-voltage response of the reflectometer diode is counteracted by the slight change in the near-unity gain of the first amplification stage caused by the use of the same diode in the feedback loop.  If the reflectometer diode is a bit higher than it should be (because of the nonlinear response near its low-end "knee"), the gain of the first buffer will be slightly lower by the same amount because that feedback diode is also operating at that same low-end "knee".    

This type of compensation would produce perfectly linear response through the whole signal chain *only* when the diode characteristics are absolutely identical.  Of course we all know this "identical diode characteristics" condition doesn't happen in the real world, so we get as close to it as possible by using good diodes with low threshold "knees" (like the 1N5711 Shottky diodes), and providing all diodes in each kit from the same batch as received from the vendor.  It's a reasonable assumption that components along the SMT reel tape are provided from the same manufacturing batch from the original manufacturer, so the characteristic of the diodes are likely as close as possible to each other.  Further, we purchased all diodes for 1000 kits at the same time, thus ensuring a good likelihood for very close characteristics for this all-important component.    

BTW, we did a detailed experiment upfront in the first kitting round in order to determine the sensitivity of instrument measurement accuracy vs. diode response. We measured a large batch of diodes and sorted them according to their low-level voltage characteristic, then we used those diodes with near-identical response in a Micro908.  The measurements obtained from this test controlled-kitted unit were nearly identical to a standard-kitted Micro908, so we felt that the "compensated design" approach was indeed valid.     

Now, after going through all this, the situation is that some kit owners are experiencing some measurement variability at higher-resistive loads (like when using the 274-ohm resistor).  Several factors could contribute to this condition ...    

1) diode nonlinearities at low signal levels -- discussed above.    

2) math inaccuracies & round-off -- studied before but revisiting.    

3) calibration software (mis)assumptions -- studying again.    

4) calibration approach -- studying again.    

So you can see that our current focus is on the math and calibration implementations.  Joe and I are currently revisiting this whole area of calibration to see if we can indeed provide some improvements to the instrument's upper-end accuracy so that *all* units experience the same measurement quality that most of them have already.  Most of you will recall several different calibration approaches that I had previously implemented - the most notable being the "multi-multi" technique that required many adjustments for each of several different load resistors.  There may be an adaptation of this calibration technique that will work better/easier, now that we've solved the "spectral purity" issues by adding the DDS Amp.  

So as Joe said "Stay Tuned" ... the pun was intended of course, by pun-master N2CX.  

And thanks again to George Silver (et al) for his insight on the topic ... this is an "open design" and we certainly welcome all input, suggestions and experience on the technical issues of the Micro908.  This is what will enable the product to continually evolve and continue to be better than anything else on the market in this price range.  

Back to Micro908 Project Page

Material and concepts presented on the AmQRP website is Copyright 2005 by the American QRP Club, Inc. 
These pages are designed and maintained by George Heron, N2APB (n2apb_at_amqrp.org)

Page last updated: 
March 19, 2005