[hpsdr] Level calculation error in Two Tone Test generation routine

[john_eckert at keysight.com]

Hello All,

First of all I want to thank Warren for hanging in here and taking the time to respond to my post.  I also want to apologize my suggestion in the original post that the 'author made a mistake'  in calculating the two tone power.

I have since seen the light.  That 'light' has also lead me to the solution of my original problem.  I'll get to that in a bit.

After connecting an oscilloscope across my TX load. I measurement the peak voltage in CW(tune) , then with the same power setting I measured it again with the two tone.  And as I claimed in an earlier post the two need to be the same to drive the amp to peak power.  And guess what,  they are.  So why did I think there was an issue?  My power sensor.  Simply put the peak/average power ratio for CW is one(linear) and 0.5 for two tone.  My lab grade sensor is an average reading sensor.  I had to switch to this sensor because I am making measurements at less than two watts.

When I get my amp back together I'll switch back to my peak power wattmeter and hopefully it will read the same power for cases.  If not I'll have another night of head scratching.

So, how did I get here?  I have a somewhat home brew radio/amp all running under control of OPENHPSDR.  The amp is liquid cooled solid state that has eight combined PA sections and up to 45 dB of total gain.  IMD is typically -40db.  Being excited about the performance gains from 'Pure Signal' I decided to build a HPSDR radio.  Unfortunately,  firing up 'Pure Signal' most often made my IMD worse or at best reduced it by 5dB.  I have always been confident that the problem was in my hardware.  I upped my amp 50VDC filtering to over 1 farad.  It didn't make a bit of difference to 'Pure Signal'.  Next I carefully adjusted each PA bias for lowest IMD.  Still no love.

A couple days ago I decided to get this working.  I was snowed in and finally had some time.  I have three stages of class A driving an AB stage that in turn drives the eight final PAs.  Since I had recently measured the PAs the problem had to be in these lower level stages.  So that is why I started this thread.  I thought the two tone drive power was incorrect.

In my design I decided to minimize the number of relays that switch on RX/TX transitions.  The board tx goes directly to the high power amp.  A single vacuum relay switches the antenna between the amp and rx adc path.  The PA sample comes from a high power coupler and is feed to my radios FWD spigot.   Now the interesting part.  The forward power sample feeds a two way splitter.  One path towards the RX adc and the other to a forward power detector for the radios internal software metering.  Instead of using a switch, I feed the forward power internally through the -20dB coupled arm of an internal reversed directional coupler.  This adds the forward power signal to the couplers through path to the receiver.  No relays!  There is also additional padding to get the level correct.

EUREKA!!!!
Today I connected a scope and a spectrum analyzer to my tx load.  Quite to my surprise, the IMD from my driver on the external spectrum analyzer was 40 dB lower than what HPSDR was indicating on the panadapter!!!  As far as pure signal is concerned,  garbage in = garbage out.

I quite sure the solution is at hand.  The culprit, I think, is my power detection circuit.  The forward power flows through the splitter hits the nonlinear detector(diodes), generates IMD, flows back through the splitter and adds with the forward sample.

Happy Thanksgiving,
K2ox


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openhpsdr.org/pipermail/hpsdr-openhpsdr.org/attachments/20161124/9a22f895/attachment-0001.htm>