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

[K1JJ]

** Update:  Another thread showing class E  AM rig IMD tests:
http://amfone.net/Amforum/index.php?topic=33555.msg259656#msg259656


It's been a long haul...  I've been slaving on a class A linear amplifer project for several months now.  It has been a difficult process, but I'm happy to say the system is even cleaner than I had hoped.

It starts with a low level  FT-1000D (4 mW)  output > into a (1 watt) ZHL-3A lab amplifier > into a homebrew 4CX-350FJ (5 watts) Mr. Clean >  into an 8877 (50 watts), > into a pair of 8877's (1500 watts).

Today, after a lot of tweaking, using a two-tone test, I am seeing -50dB to -55dB 3rd order IMD products.  Using voice LSB at 2.8 KHz bandwidth, I see side crud products down as much as -65 dB within 3khz of center frequency. This is outstanding and blows me away.  Expand the the pics and count the 5dB vertical grids to get an idea of signal to crud (splatter).

These numbers are hard to believe, but the secret is that the complete amplifier chain up until the final stage is running in class A. This is the way to break out above the -40dB 3rd ceiling.

Pictures 1 and 2 are 2-tone tests and pic #3 is my voice doing an atomic Yallo on 3775.  Each grid is 5 dB.  Notice the side frequencies and lack of side crud, especially in pic #3 using voice.  Check out the lower frequency side of pic #3 where the side crud falls into the noise floor at 3772 KHz, 3KHz away from center freq.    This demo is when running about 1.5 KW into a dummy load.  

Each pre-amplifier in the chain is capable of ten times the power it is putting out, thus the extreme cleanliness.

I have roller inductors in both the FJ and 8877 rigs to adjust tank Q.  The trick to getting the last -10Db of good IMD is being able to tune each stage perfectly to feed the next. And it is not necessarily 50 ohms J0 for best IMD.

The noise floor is limited to -75dB because of 1000D synthesizer noise. Never thought that wud ever be a concern to testing... :-)

I'd like to thank Frank/ WA1GFZ and Chuck / K1KW for their help fielding questions via email for this project. They both gave good advice.

As I told Frank, 25 years ago I would have given a left nut for an amplifer chain like this.  This, for me, is the attainment of a ham goal I've sought for a long time, but failed to achieve up to this point...  Yeah, I'm excited..


T

All pics at 1.5 KW out into a dummyload using a 2.8 KHz LSB signal.  This amplifer will be used on both SSB and AM.  BTW, I can flip a switch and drive this amplifer chain with the low level output of my HPSDR transceiver rather than the 1000D. This opens up a whole new world.

[KL7OF]

Tom....Way to go!  Your skirts are vertical.....So steep...Congrats...You are to be commended for your dedication toward making this work....
  TMC used 4 or 5 class "A" stages to end up cathode driving a 4CX 5000-A in the Navy FRT 39 (GPT10K) ...I don't think they achieved the results that you have...but I believe their goals  were  similar...
   I wait to see and hear your outfit in the near future...Pictures please...Perhaps we will get some East-West coast QSOs before the bands get too noisy..I'd love to see this TX on my SDR display......Thanks for sharing....Steve

[Mike/W8BAC]

Good Job Tom. The results are impressive.

Here is a way to capture your computer screen and save it or post it here. It's retentively simple and doesn't require any new software if your using a Windows machine. You can grab shots like those you posted and they will look great.

http://windows.microsoft.com/en-us/windows-vista/take-a-screen-capture-print-screen

Mike

[KA0HCP]

Wow, it's so slab sided it's like a block of marble.  Congratulations!

[ke7trp]

Great job man.  Very clean!

C

[W3RSW]

Very nice.

You are now the standard candle to which we will all aspire.  ..puns on top of puns for the King !

 

[K1JJ]

Heheheh.... thanks for the back slaps, guys.

I look at those IMD pics and still can't believe what I'm seeing. Like a dream come true.  For many, many years it was disappointment after disappointment. I tried every tube and config I cud think of.   Every time the IMD was worse than  -35dB... especially when QRO.

I think having an SDR with spec anayzer made all the difference to see what was really going on and how each  step affected the overall chain.

The funny part is only a week ago I almost gave up on the whole project.  The drivers were putting out beautiful IMD, but the finals were stuck at -25 to -30 Db 3rd.  WTF??  I tried everything to solve it and finally figgered I had some bad pulls since they were used out of an MRI machine. (bad tube transfer curves)  

In final desperation I added a bigger cathode choke and more bypassing. The IMD got worse to -20 dB 3rd!  I was yallowfied.  Turns out I had a total of 20 ohms in the cathode in the form of big pie-wound BC type RF chokes. This was killing the bias regulation. I ripped out the chokes and used a simple wire thru the ferrite of the bifilar choke. Now it was a standard trifilar.  The IMD dropped like a rock into the -55 dB 3rd area as you see in the pics.  And I was one day before giving up and going with a different config...

Another thang I found is the input tuning to the final 8877's is VERY critical!  I can swing IMD up and down -10 to -15 dB by resonance or non-resonance. In the case of 75M, peak power output corresponds with best IMD, which is good. Without that input tuning the amp wud be back in the low 40's.  It's NOT a simple matter of matching 50 ohms J0.

There's so much more....  I added a 600 uf cap between the cathode choke cold side to the HV power supply capacitor negative. This hopefully brings the total dynamic cathode resistance down to less than an ohm. At 20 ohms it was swinging the bias all over the place.


I used as high a tank Q as I could get away with. Even now, at 1500w, the 3/8" copper tubing coil (only 7 turns) gets too hot to hold after a minute of RF carrier. (Circulating RF current is Q   X   plate current)  The tank Q is up around 35 to 40.  Same Q for the 8877 driver and 4CX-350FJ.  It seems to help bring down the IMD further.

Tank loading (C2) is also important. I find the sweet spot is when there is ample power output AND the grid current JUST starts to flow. No grid current is important to cleanliness for the drivers and especially for the finals. That's what class A1 means in addition to no fluctuation in plate current.

I am using a common 4KV HV supply for the 8877 driver and finals. On Chuck's suggestion, I added a 20H choke from the HV supply to the driver - and a 50 uf filter cap on the 8877 driver side. This reduces any HV ripple caused by the final into the driver HV. The voice syllabic rate ripple will be filtered out giving the 8877 driver half a chance to perform with a stiff HV supply.

There's more... maybe later.

Next, optimize the tuning / parameters on the other bands and see what happens. Then into the antennas. Each antenna is a different load  and requires pre-sets for optimization.  Being sloppy in the tuning can sacrifice -10dB or more of 3rd IMD.

Mike, thanks for the screen shot info. I've done them before but was too lazy this time to remember how to do it.  The next pics will be screen captures.

T

[K1JJ]

Some more notes for those curious engineers...

It may not be obvious, but much of this chain is devoted to replacing the "dirty" ricebox driver. Normally the average ricebox is lucky to be -30 to -34 dB 3rd order.  The IMD output of an amplifier is always worse than the driver itself. For example a -50db  1 KW amplifier driven by a -30dB ricebox will always be in the high -20's dB 3rd and no better.

The driver should always be at least -5 dB cleaner than the final - and preferably -10 dB cleaner for minimal degradation.

For years I was using the FT-1000D as a driver at 50  to 200 watts. The IMD sucked, so how could I ever expect the final amplifier to be any better?

About a year ago I discovered that the 1000D at 10 watts was about -45 dB 3rd and used a 3-500Z amp as a class A  IPA. This helped.  

Later I found the 1000D low level at 4mW was super clean at about -75dB 3rd.  I drove this into the ZHL-3A which is about -68dB 3rd at 100 mW.  The 4CX-350FJ amplifier is about -63dB 3rd when run at 10 w.  So here we are with a 10W driver at about -63dB 3rd.

Driving an 8877 in class A to 50 watts output is about -59 dB  3rd.  

So, we end up with a ~ 50 watt clean signal approaching  -60dB 3rd to drive whatever we wish.... whew....


I know that Frank has been diligently working on his solid state amplifers to push them above -40db 3rd and is making progress. I know he will get some good results in class A.   Though, I've yet to find a HF SS amplifier out there that does better than -40db 3rd when running 50 watts or more.  They exist in the feedforward, pre-distortion software and possibly military world, but I don't see any ham versions or inexpensive ones to buy.

So, I took the hard road using tubes....

More later -

T

[ke7trp]

I am curious to the power input figures for fun.  Have you ever calculated it?

C

[KF1Z]

.

[K1JJ]

I am curious to the power input figures for fun.  Have you ever calculated it?

C

Clark,

Interesting question. Well, we won't win any green awards.

For the drivers, I just calculated the efficiency at between 10% to 20% depending on how hard the drivers are pushed...  

This is realistic for class A considering the heavy stage loading and super Hi-Q I am using.  

The plate current stays steady from idle to full signal as well as no grid current flowing.

It's the only way to get into the -50 to -55dB  3rd area for a "brute force" amplifier.  ( No NFB or  predistortion software, etc)

The final amplifier is much more efficient, of course. It idles slightly higher than Eimac recommended specs and is run at only 1/3 of its rated power to achieve its cleanliness. Super high Q tank and input circuits

I'm using the sliding bias circuit on all amps that drops the idle power to about 1/4 during voice pauses, so that helps.

Right now I'm tuning it up for 40M.  So far I've found the proper settings all the way to the 8877 driver output. It shows about the same performance as 75M so far.  Next the final amp and full chain optimization.  I write this all down on a cheat sheet for fast pre-sets later on.

T

[ke7trp]

Thats not to bad really.. I guess the only real drawback is all the tuning.  But, It is purpose built.  I ran some tests on my rig and simple amp tonight.  I dont have the two tone. I just modulated the rig like you did in the last shot.  Your right, tuning really effects the IMD as does Power output of the rig.

C

[K1JJ]

Thats not to bad really.. I guess the only real drawback is all the tuning.  But, It is purpose built.  I ran some tests on my rig and simple amp tonight.  I dont have the two tone. I just modulated the rig like you did in the last shot.  Your right, tuning really effects the IMD as does Power output of the rig.

C

What kind of IMD numbers where you seeing when doing the voice tests?

There's a free tone generator to download that will do 2-tones.   Put it on your computer and run an audio cable to the rig.
http://www.dxzone.com/dx17728/sigjenny.html

You will find a good 2 tone test is very unforgiving and takes no prisoners. Kinda depressing at times. Whereas voice transmission looks much better and actually is in the real whirl. We are lucky in that respect.

"Purpose built"  - yep that's a good way to describe it. I will probably park it on certain frequencies for a week at a time. It will take a good three minutes to change bands until I get real familiar with the cheat sheets.


Well, I got the system playing on 40M into a dummy load, though I need to make some C2 mods in the 8877 driver. There is too much fixed capacitance used on 75M and I need to add a relay to toggle it in and out. Always something.

Right now it's not as clean as 75M by -10dB, but I'll get it.

Later -

T

[ke7trp]

Thanks for the link!

-47 DB for the radio and with the amp on 40.  But its not scientific.  Its just be watching an SDR while using voice.  The rig has a computer on it now. Just need to dig for the cable to go to the input. Then I will run the two tone with and without the amp.  Radio should do -45 easy but the ole SB220 might let me down   Eitherway. Its fun to test and to learn about this stuff.

[K1JJ]

Good.

Yep, get that 2 tone set up and you will have a rock solid standard to compare apples to apples in the shack and with th rest of the whirl.

What kind of radio do you have that will do -45dB 3rd at full power?  That wud be quite good indeed.

The 3-500Z's IMD will depend greatly on where it is idled, how hard you run it and how it's tuned and loaded. Also how good its HV regulation is under load as well as the 240 AC regulation.  I imagine the input tuning is fixed, so that will not help to tweak.  If you get real serious about it you could try a small tuner in between the rig and amp as an experiment to tweak the IMD.

Your radio may not like the varying amplifier input impedance. It pays to sample the driver for IMD when the whole system is running together and see if there is a bottleneck. Maybe custom interstage tuning can help the situation..  I'm finding that good IMD numbers is doing a lot of little things right... :-)

T

[KM1H]

Great work Tom but not exactly built for rapid QSY!  Can you even go from 3770 to 3885 without losing a lot of IMD?

[K1JJ]

Great work Tom but not exactly built for rapid QSY!  Can you even go from 3770 to 3885 without losing a lot of IMD?

Tnx, Carl.

Well, I thought it would be very sharp tuning too.  But the Q, being between 30 to 40, is only 3-4 times sharper than a standard amplifier with a Q of 12.  

I just tried moving 20 kHz on 40M and saw no noticable difference in IMD or in the power peak.  I'll try some bigger excusions later.

I think a Q up around 100 or more is when we would start to see extreme bandwidth QSY problems, akin to peaking a Hi-Q preselector in a receiver.

T

[K1JJ]

Update:

I just did a sweep test on 40M.   Into the dummyload (as I always use for tests)  I swung the VFO from 7150 up to 7200 and down to 7100.   This was at 1.5 KW output.   The power output changed about 10% as expected.  

However, the 3rd thru 7th IMD peaks simply tracked the main power tones and stayed rock solid in relative amplitude.  So it appears the IMD tweaking, once set up, is not going to be a bother.

One point:  At the present time the Q on the driver or finals is not as high as on 75M, maybe 20 - 30. Still trying to find the best optimization points.  Though I now have the 40M IMD performance within 3-4 dB of the 75M settings.  Getting there.

Update:  Just got there.  40M 3rd IMD now sitting at about -55dB !!

Tried an antenna on 40M for the first time - slightly different load.  The IMD held exactly the same as the dummy load. This is a good sign.

T

[KL7OF]

What are you using to couple your tx signal to the SDR RX/ analyzer?

[K1JJ]

What are you using to couple your tx signal to the SDR RX/ analyzer?

Hi Steve,

There's several ways to do it, but, I use a two-tier sampling method.  In an aluminum box I have a toroid wound with 40 turns. It is slipped over the inner conductor of the coax inside the box.  The 40 T winding connects to a coax which runs to a relay box that feeds the SDR reciever antenna input.   I use an additional  60 dB pad there consisting of   50 ohms to ground, 5K in series and 50 ohms to ground. (Or something like that - use an online pad calculator)    This connects to the SDR ant input.


For TX / RX transfer, in the PTT RX position, the relay connects the normal antenna to the SDR.  In the PTT TX position, the relay connects to the 60 dB pad.  The HPSDR is in DUPLEX mode, so I see the actual transmitted signal coming out of the sampled final amplifier.

Be careful with exposed leads and connections in both boxes. Just a small snub can pick up lots of signal and make the attenuator useless.

Use good quality coax too - that doesn't leak due to poor shield.   
T

[ke7trp]

Cant get Sig Geny to run on windows 7 64bit or windows 8 64bit. Looking for a new way to do the two tone.  I was using the Clean RF sample box with an attenuator that dennis gave me a few years ago.. I could adjust midrange to the SDR.

C

[K1JJ]

Cant get Sig Geny to run on windows 7 64bit or windows 8 64bit. Looking for a new way to do the two tone.  I was using the Clean RF sample box with an attenuator that dennis gave me a few years ago.. I could adjust midrange to the SDR.

C

Hmmm... maybe they need to update it.

Do a Google on 2-tone or audio signal generator software  and see if there are more out there. Shud be.

I ended up buying an Elecraft 2T, 2 tone kit board.  It works very FB.  I have it on a switch that can choose my voice audio, the pulse pecker or the 2 tone for fast use.

BTW, the pecker is a great way to test peak power with 10% load on the system. Very little heat.  But the 2-tone is a power hog, so be quick with the tests.

** Two-tone tests can be used on AM just as well as SSB.   The waveform will look somewhat different, but it will still show meaningful dB improvements and comparisons.  I plan to do some tests on my class E  AM rig for IMD, when I get caught up.  There may be some ways to improve on that through optimum loading, drive, better HV regulation, etc - who knows?



T

[WD5JKO]

Cant get Sig Geny to run on windows 7 64bit or windows 8 64bit. Looking for a new way to do the two tone.  I was using the Clean RF sample box with an attenuator that dennis gave me a few years ago.. I could adjust midrange to the SDR.

C

Try this one:

http://www.dr-jordan-design.de/Download/SigGen.zip

For two tone, run it twice...works here in W7...but 32 bit machine..YMMV

Jim
WD5JKO

[K1JJ]

Got 40M working.    I did a screen capture and expanded the screen to show farther out. This shows the noise floor a little better.   All pics at 1.5 KW output.

The first shot is a 2-tone test.   I believe I can tune the 5th to 7th order peaks down a little farther when I get back to it.


The second and third are VOICE yallos.  

** For some fun, start at the top of the peak in picture #3.  Count down in  5dB grid lines until you hit the crud level.   (Elevator going down!)  This is a perfect example of how friendly an amplifier can be to the average human voice vs: the harsh reality of a two tone test.  For this we are all fortunate.  Though, if we try, there ARE some voice noises that will produce wider crud.


I'm moving on to 20M now.

T
 

40 Meters:
 

[K1JJ]

The forth is the pulse pecker. This is a nasty pulse rich in harmonics that is about 10% duty cycle at 1.5KW peak.

The fifth is a two-tone using a narrower screen bandwidth.  I believe we are seeing the slope of the 1000D synthesizer noise, but not sure until I runs some different tests.  It may just be normal low level signal crud.

T