Real IMD Results for the New RF Linear Amplifier Chain - Pictures

[Steve - K4HX]

Amazing. Looks like the output of a signal generator it's so clean!

[w4bfs]

yeah Steve .... I believe you could call Tom's rig a signal generator .... nice work Mr. Vu !

[K1JJ]

Thanks for the kudos, guys.  It really makes it all worthwhile.  Other than using the project, half the fun is showing the results off to your peers. Who else would give a crap about it anyway, the dog or your XYL?  :-)


Last night I got half way thru the 20M tuneup.  The 4CX-350J had a small glitch that required re-neutralizing it. It runs at 30 dB of stage gain.   Finally, the output of the 8877 driver showed good IMD, like the other bands.
Other than that, today I will try the whole chain on 20M. And go from there.


All of the drivers also work on 160M, but I think the final amp (pair of 8877's) need some mods to cover 160M.


I also changed the final's input circuit from a L network to a pi-network. Seems to work better that way and give me more versatility to compromise loading  vs: best IMD.

It really a juggling act now, but I hope to carve these settings into stone and easily go from band to band later on.


T

This 1.5 KW "yallo" voice print from 40M still blows me away as I look at it this morning. I still wonder if the slopes down -70dB (starting at -90 dB on the graph) are from the 1000D's synthesizer noise -  as the real signal drops into the noise floor.  :

[ke7trp]

Tom,  Which lower marker are you using?   The little right side pip at about 91db or the "mass" below that around 80?  I have the same thing here.   Not sure which point to use. 

Maybe you are using the left slab down to the first point?

I got some work to do here with my test setup.. I am seeing more signal on the left side harmonic than the right... That makes no sense to me right now...  It should be close to equal on both sides??

C

[K1JJ]

Clark,

I usually do not see a perfectly symetrical SSB signal. Remember that we are attenuating the upper side on LSB and it shows.. the signal is centered on 7135 and it drops like a rock above 7135, but it extends down ~ 3 KHz into the lower side.  So some of this non-symmetry is the ssb filter - and could also show poor opposite sideband suppression. It depends how far down it is and what the ssb driver TX specs are..

In a 2-tone test, we use the first significant peak after the signal drops off. This is 3rd order. This freq can be easily calculated using the tone freqs.  The next peak is 5th order, etc.   Tuning can affect the symmetry of the tones too. I often see the upper side tone vary from the lower side by 5 dB, especially the 3rd order.  The two main tones should alway be set up equally, however.

For voice tests, we just eyeball it and pretend we are listening up or down away from center freq and hear the crud down so many dB. The 2-tone is scientific while the voice test is more subjective, but still a real inducation.

We want the amplifier output signal to be as close to the original input as possible. As a check, look at the input signal and see how close it looks to the output.

Also, some cud be artifacts or "humps" as Frank calls it, generated by who knows what in the test setup - overload, signal leakage, etc.

The point is to get a reasonable comparison and baseline so you can make improvements and know when you are making headway.  We will never get perfect lab-type results, but we don't really need it at the extreme low levels near the noise floor.

As long as you are not overloading the SDR's ADC, you should get very useable graphs.

Post some of your results once you get a handle on it via screen captures.


****  BTW, with a 2-tone test an easy way to indentify the 3rd, 5th, etc and weaker tones is to intentionally overdrive the amplifier. Those higher order tones will jump right out.  The higher order ones can sometimes hide, but not when overdriven.

Voice audio analysis will not be as clear since your voice is always changing.  Also, using a white noise or pulse pecker into the audio can help to standardize your subjective "voice" results for comparison.

T

[ke7trp]

I am getting false results here.  Here are some pictures of the radio alone at 100 watts.  It shows peak at -38 to -40 DB.  The left shelf is -105 DB.  This is not possible.  The rig is on 17 meters.  I just worked TX5K on 5 bands and 17 was the last one I used so I left it there. Rig is into bird 500 watt dummy load.

I dont understand what I am doing wrong.  The rig is rated at -45 to -55 depending on whos numbers you want to go with. But still, This is not possible at 65..  I cannot trust the SDR in tests like this.  I need to get Spec AN running in shack to confirm.  The data seems good but the PEAK is way to high on the screen. So maybe SDR has flaw in software. There is no way radio is a full 10 to 20 DB better than the published and private tests. 

This is the latest board (they had six major revisions over the years) which has not been tested anywhere, with the latest software which has not been tested. However, I think its a false test.

C

[K1JJ]

Is it possible you are keying the SDR and seeing it's own low level internal TX signal? Or is this a receiver only?

Is this a hard sample from the ricebox output back into the SDR receiver?


The voice test is hard to tell until you get a real 2 tone generator going.


The acid test is to use a regular second receiver and use a 6" stub wire for an antenna off the back.  Then adjust the stub length until your 100 watt signal into a dummy load is S9 +40.  Then tune the receiver and see how far down the side crud level is 3, 4, 5 khz away from center. You will hear the raspy noise of IMD once tuned away from the real signal. The s meter may not be perfect, but you will get a rough idea how much drop off there is. Then you can look again at the test set up and figgger out what's going on.

T

 

[ke7trp]

The SDR has no TX.   I will try to use a piece of wire for an antenna next.  The SDR is hooked to a sample box I use for an output to my O-scope. Its possible its just to much signal period.

Thanks

C

[K1JJ]

Use a second RX, like a tube RX for the ant stub test so you can eliminate the SDR set up and software possibility problem.

Actually too much signal will usually give you poorer IMD results from RX overload, etc.  

Don't feel bad - it took me a long time to get MOST of the bugs out of my own high power IMD testing set up. It is very sensitive to stray and overload. Getting down to the noise floor with reasoanably accurate results is not easy. One step at a time.

T

[K1JJ]

20 Meter AM


Here's some IMD shots of the amplifer chain running AM on 20M using my voice.   A combination of audio roll off and the lack of side splatter makes for sharp skirts.  Using 1500 watts output for all pics.

I show both +- 4 KC  and +- 5KC  audio examples, which is where I intend to run it most of the time.  

The first two shots are using the FT-1000D.


The third shot uses the HPSDR xcvr. Notice how sharp the skirts roll off with digital filtering. Of course, the amplifier must be clean to support that kind of roll off without side crud.

The third pic shows an "SSSS" which fills out the AM bandwidth.


I meant to run thru some  AM 2-tone tests, but forgot.  Next time.


T

[K1JJ]

And, some 1500 watt SSB  USB IMD pics from 20M.  

#1 is 2-tone and #2 is voice.
  

Now on to 160M and then printed cheat sheets to get fast repeatability when changing bands.


T

[Tom WA3KLR]

Congraturations Mr Tommy Vu, you have vely cleen USB siganel.  Yur low frequencie energy is vely high.  Yu must have big BAs.

[W2VW]

Open S.P.S.T. switch my IMD drops below grass.

[K1JJ]

Tom, it's tough uttering highs and lows at the same time...

I'm gonna build up a white noise generator. That oughta be a good way to fill out the audio bandwidth for tests.
(or as Tron calls it, a "Brown noise generator")


Dave, I figgered your comment would be, "Now that you got this thing working, are you finally gonna get on the air?"  Or, "It doesn't matter how clean it is if ya never hook up the antenna and push the PTT switch... "  

T

[WD5JKO]

I'm gonna build up a white noise generator. That oughta be a good way to fill out the audio bandwidth for tests.
(or as Tron calls it, a "Brown noise generator")

Tom, The Dr. Jordon PC based Audio Generator has the "White Noise" feature enabled in the free version. I linked to this earlier..Just in case you want to try a PC based audio source..

Jim
WD5JKO

[K1JJ]

Thanks for the info, Jim.

I think I will stay with a board version so I can put in on the same switch as my pulse pecker and 2 tone generators. This way there's a fast choice of any three.

Here's a simple two transistor version I will build.

T


White Noise Generator:

[ke7trp]

I cant get good results with the SDR.  Its showing me a false positive.  Not sure why. Its not possible for my rig and 2x3-500 amp to show -65 db.  Should be -35 to -45.  I have yet to figure out why. 

One thing I would like you to do Tom is to run the test but go to the second harmonic Frequency on the SDR.  I did and the PIP was right there showing -43 db. That is believeable for my little amp and radio.

C

[W2VW]

Dave, I figgered your comment would be, "Now that you got this thing working, are you finally gonna get on the air?"  Or, "It doesn't matter how clean it is if ya never hook up the antenna and push the PTT switch... "  

T

I usually save that for internet warriors 

When the wx warms up you might not want all that dummy load heat around....

[K1JJ]

I cant get good results with the SDR.  Its showing me a false positive.  amp and radio.
C

Clark,

First step is to tune yourself in on a second tube receiver and determine how far down the crud is. Look at my post a few back.  This will ball park it.

Then get a real 2-tone generator going to get a reference that is standardized.

I'll take a look at my 2-tone results on the second harmonic today. I would expect it to be a duplicate of the fundamental, though way down in amplitude and possibly distorted from low level test set-up artifacts.

* BTW, looking again at your picture... is there a chance that the grid divisions are 2 dB per rather than 5 Db per?  If so, I see about -42 dB to the crud level, which is what you are expecting.  You don't have the scale numbers showing in the pic, so I can't tell.  Though you did say the lower shelf is -105 dB, so maybe it is 5 dB per.


Dave, if I can just get out of this building phase, you might hear more of me than you can take over the next few months...  

T

[ke7trp]

The divisions are normal.   105 for the crud and -40 for the peak.  I cant get good results with the SDR this way.


C

[WD5JKO]

The divisions are normal.   105 for the crud and -40 for the peak.  I cant get good results with the SDR this way.
C

   Clark,   Maybe you have too much attenuation at the SDR receiver input. If you were to get the peak signal up 50-60 db above the noise, then you'd be good so long as the SDR system has the dynamic range. If the dynamic range isn't there, then the SDR will create its own IM crud. I believe you own a spectrum analyzer also;if it has a resolution bandwidth down to 300 hz or lower, then that would be an option as well. What Tom is doing is tough, and that makes the test setup even tougher to get useful and believable data from....

Jim
WD5JKO

[K1JJ]

   What Tom is doing is tough, and that makes the test setup even tougher to get useful and believable data from....

Jim
WD5JKO

It's akin to tuning a Yagi antenna for -45 dB front to back.  Any wind or weather will detune it back to -25dB.

Holding -50dB 3rd or better in a "brute-force" amplifier without NFB or predistortion is difficult.  All the tuning has to be right on.  In fact when I switch bands and then go back to preset band settings, I'm finding the physical markings are not good enuff and my IMD is off by -10 dB. I have to put thru a 2-tone and re-optimize for the full -50 to -55 dB 3rd.  I'm getting better at it though.


The strange thing about Clark's issue is that usually the test setup gives poorer IMD results due to overload, artifacts, etc. But he is seeing results that are -60 dB better than what is expected, so I can understand  why he doesn't trust what he is seeing.


Well, I finished with the 75 - 20M  IMD band preset settings and now moving on to other things.  Built a new 8877 final input tuning system yesterday that switches between pi-network and T.  I'm finding I get the best IMD on 75M with a pi input - but best on 20, 40M using a T.  It's quite dramatic.  At least it gives me more versatility to handle all situations..   This is like tuning a dragster...

T

[K1JJ]

After all this work, I wondered to myself... "Why you go thru so much trouble Mr. Vu?    Did you waste plecious time? Are you clazy?"


I decided to see if there were any real improvements from the old driver system to the new...

So I hooked up the old FT-1000D driver and ran it at 100 watts as a stand-alone for an IMD test. This is a 200 w rig. Pic #1. At 100 w out, the IMD 3rd is barely -28 dB 3rd.

Then I used the 1000D at 15 watts to directly drive the 8877 amplifer chain.   This is pic #2.      When used as a driver at about 15 watts, it is slightly better at about -31 dB 3rd.  IMD generally sucks, but this is about average for a ricebox.  This is how I used to run my rig...


The new setup:

The 3rd pic is using the new driver setup: The FT-1000D low level 4 mW  > Mini Circuits 1 watt amp >  4CX-350FJ driver into the 8877 amp system.  Loaded up to 1500 watts. The difference is stunning.     The IMD is about -22dB better.... to -55 dB 3rd.  That is like night and day when it comes to splatter.

It goes to show if we have a clean final amplifier, it pays to be careful in our driver (ricebox) selection - or consider alternatives.


 Tom Vu feel better now...

T

[ke7trp]

Yes, My tests are showing better results than what I can believe and as a result, I just dont trust SDR in this kind of test.  Its showing what I believe is 10 to 15DB false positive. Unless this latest board is really that clean.  But I need to use the Spec AN to know for sure.

C

[KA3EKH]

The display on the Flex SDR is amazing! The dynamic range exceeds what many of the analyzers that are out there in the industry today. Having used various spectrum and sideband analyzers over the years find it almost too good to be true the quality of the display that the SDR produces. Maybe it’s just like when most people migrated to DVM instead of analog meters, where people use to round numbers off with analog with digital they can go out one or two decimal points? And the SDR radio and software now makes this possible. Have to say however you do this by SDR or analyzer that’s one clean amp you got there for that power level.