The AM Forum

Here's a new homebrew 4CX-350FJ linear amplifer I just put together.   Only 3 milliWatts in -  100 watts out. (About 43 dB gain with a single stage)    It is grid driven with electronically regulated screen and grid power supplies and is neutralized. It appears to be pretty stable on 160 to 40M so far.

The 'J' version tube is one of the cleanest tubes made for linear service. I wanted something I could drive with the 3 mW  output of the FT-1000D driver stage and bring it to 25 to 50 watts to drive the 8877 amplifier to 1KW.

The IMD, according to Eimac, is about -45dB 3rd order.   1400V HV, 300V screen, -14V grid.  100 w pep output.

A cool feature: The screen circuit has a failsafe trip circuit. If screen current exceeds 15ma., then a relay opens the cathode return and grounds the screen. This will prevent many errors, like no HV, overdrive, bad antenna, mistuning, etc.  These J tubes are very fragile with  tiny screen and grid sturctures. During testing it triggered a few times and saved the tube.  I used a simple design: 339 comparator, 2n2222 and a sample off the negative lead of the screen.  I plan to build one for the grid circuit for each of my 8877 amps next.

Neutralization accessible from the front panel for touch up as well as the fan speed and trip reset..   I plan to add band preset markings on the front panel once zeroed in.

I used roller inductors in BOTH the input tuning and plate tank so I could experiment with the highest Q's possible to squeak out optimum IMD figures.  Higher Q, the better.

Notice the homebrew cabinets and color scheme. The baby blue cabinets and white panels with black letters are the theme throughout the shack.  

I once saw an ad for a precision target pistol that read, "Every man needs one highly accurate gun."  Well, I figgered every ham needs one highly clean amplifier.... ;D

Actually, this project is one cog in a long quest to build a 1500 watt amplifier chain that exhibits -45dB 3rd order IMD or better. It has been a difficult task so far. However, I just ran some tests and it appears that I have met this goal to the 100 watt level so far.

Pictured later, check out the new 8877 amps. There are three of them now, so I can park them on various bands.  "Mr. Clean"  will be the driver for Fabio, Dr. Love and Dr. Love Jr.

Thanks to Chuck, K1KW for his idea to use the 4CX-350FJ and his help with my many questions. Same for Frank, WA1GFJ.  Thanks again, guys!

T

The all-new Mr. Clean:

In the 3rd pic notice the power supply. HV, reg screen, reg grid and fils all on one platform.  It is on casters and has its own slide-off cabinet.  All rigs in the shack are easily accessible for servicing. All cabinets slide right off the top and rigs can be serviced in place.

#1 is the new 8877 Fabio using the same color scheme. It will be driven by Mr. Clean's impeccible 50 watts.  ;)
At first glance they all kinda look the same.

Pictured second is Dr. Jr., another 8877.

And of course, Dr. Love.   Another 8877.  (3CX-1500A7)

T

What??!?? No Plexiglass??? What happened to the Plexi-King??   :o

OMG, the Plexi-Police are here already!

It was painful, Todd.    Actually, Fabio, the 8877 pictured in the third post, had a Plexi panel and cabinet until last week.  Due to the high gain and chance of RF feedback in the shack, I've gone to all aluminum... [sigh]


The only remaining holdout is the front panel to the class E rig.  It is a stand-alone unit, so not involved with the other amp chain.

T

Rico Suave, the 24 pills class E -

* Update - I replaced it with a clearer pic:

What method of neutralization did you use that holds for all those bands?

Chris

Chris,

I use the same neutralization circuit on all rigs that need it. Take a look at this 813 X 813's rig. I used the same one for this linear.   Just a small stub sample put next to the anode that is fed into the bottom of the input tuning cap:

http://www.amwindow.org/tech/htm/813/813.htm

T

Hey, be thankful you're not a Collins collector. Johnny still twitches whenever someone mentions R-388 PTOs for the abuse he took - and he was never part of their group!  ;D


Pffft. That doesn't count. Nothing going on inside to see like the tube rigs. A vacuum cleaner is more exciting. Even the ones built into the wall. Being in there with the plexi-adorned transmitters all lit up and breathing fire was a sight to behold. I think Steve got some good shots last time we were there.

Honestly though, I was amazed there wasn't a fireworks show before now with the plexi panels due to the close proximities and strap factor.

R.I.P. Plexi-King. Seems fitting that he leaves us along with another stalwart of transparency: Dear Abby.

No wonder you have been so scarce, but obviously putting the time to good use!   

It is good to see matching meters on projects. :D

Meter hodge-podge drives me nuts.  Not that I'm not almost there anyway.

73DG

Very nice Tom.  It will be good to see the final numbers when you get to that point.  Mr. Clean is a great name for the beast!

Joe, GMS

Thanks, Joe -

The work could be better, but I have just so much patience for details and then move on to the next project.  I use recycled scrap yard aluminum and the parts I have lying around here. They are built strong and work well, but I have to stand in awe when looking at the professional work that some of the other guys do here... ;D

I just received a 2-tone kit board from ElectroCraft, so shud be able to run some accurate IMD tests soon. But it's easy to tell the splatter level just by talking into a dummyload and tuning with a receiver. This new amp is at least 15 db cleaner than my 200 watt FT-1000D, so it's doing the job.  After hearing how sharp the skirts are on a signal that is -45dB 3rd, I became hooked.  The difference is tremendous.   If a station is transmitting 5kc AM audio, after the 5kc point the signal just drops into the noise... no buckshot, no artifacts.  Same on ssb with those 3kc channelized QSO's. It's a worthy goal.

Hope to get on with the boys this weekend with the 'J'  driving the 8877 and get some on-air acid tests.


DG, yes meters are important to match. Those old round 3" relics are of beautiful quality. The plasdick junk of today cannot compare.  A hole punch or even a drill/ hole saw from ACE Hardware is all that's needed.

T

Well, I am sure when you get the two tone test run done you will be happy with the results.  Very impressive on the AM sideband performance!  I can think of some stations on 75M that could really use your gem!!!  

The constructions looks fine.  Sometimes its better to prove the concept and move on to other projects.  We are all cut from different cloths on that.    

You will have the best fidelity narrowest signal on the band!  

Joe, GMS

For the techies following the thread....

Even though the FJ amplifier works well at 43dB of stage gain, I decided to reduce the gain and add a one watt lab amplifier (ZHL-3A, 24 dB gain) between the 3 mW FT-1000D driver and the 4CX-350FJ.   43dB is axing a lot of one stage and can be prone to instabilities.

I added a 3db pad before -  and a 6dB pad after this one watt amplifier.  This adds isolation and a level of protection.  Then I added a 4.7K swamping resistor from grid to ground of the 4CX-350FJ stage reducing its gain down to about 30 dB.  So the amplifier should become even more stable at a lower gain - good when changing bands and tuning up on odd frequencies.  I saw no effect on IMD, being that the lab amp is super clean.  Just a way to make the overall system more bulletproof.

The FT-1000D's  3 mW driver works a little easier now too.   The neutralization seemed to tweak in sightly better too, which is the sign of a more stable stage.  I now see a perfect screen current peak, plate current dip and power output peak coincidence. It was close before, but right on now.  

So, we're basically axing a tube rated for 350 watts  to put out 25 to 50 watts PEP in service. What an EZ life!

T

Love the very Navy like blue cabinetry.

Tom, can you post a picture of the grid drive input circuitry?

Tnx, Bob!

I really like that blue color. Kinda like the old IBM computer cabinets I think.

The amp is all put together and in the operating position, so next time I have it apart I'll snap a shot of the bottom.

I just finished up the little lab amp and the two pads and put them in service too.  This is the pre-IPA between the 3 mW  1000D driver and the 4CX-350FJ input.

Notice the 24 vdc 2.4A Sola supply. This regulated supply is only about $10 on eBay, brand new. What a deal.  I also picked up a + - 12V and 5V Sola to power my HPSDR modules.  The wimpy supplies that were stock are now replaced with some man-size Solas - cheap.

This lab amp runs about 12 watts in class A to put out 1 CLEAN watt.  I wasn't happy with the heat sink temp so added a small 12 vdc muffin fan running on 6V.

Shown below:

T

Mr. Clean is real pretty Tom. He can walk the walk, but can he talk the talk?

Close..

You bery vely funny man, Jeff!

Yes, there are striking similarities, I must admit.  Unfortunately, National went out of biz about the time they adopted that slick color scheme...  ;)


Well, I put together the Electrocraft 2-tone tester board tonight. It appears to produce a very clean two-tone, good enuf for reasonable IMD testing.  The lab method would be to inject two separate RF carriers, mixed and spaced 1200 hz apart for the purest test.  But this 2-tone method will use the same station transmitter with audio tones, so will simulate the total enviroment as is.  

Due to station artifact pickup, I was only able to see down to about -45db 3rd with the HPSDR, and that's what the FJ is producing.  I then drove the 8877 linear with it and I still see about -45db 3rd at 500 watts output.  When I go above about 750 watts output, the 8877 starts to draw normal GG grid current and the IMD starts to go south. Still it's better than most ricebox IMD figures even then.

It was interesting to see how drive, loading, screen current and many other factors caused the 3rd to drop but the 5th and 7th orders to increase. And vice-versa.  They go up and down and change positions, just like squeezing a balloon. There are definately sweet spots to tuning an amplifer and if we do our homework and premark these settings, they may help us to tweak in an overall better IMD result.

All in all, so far this is the cleanest lashup I have had to date and will continue working to see if there are any more combinations of current, voltage and other parameters that improve things.   The days of being stuck at -31 db 3rd because of my marginal FT-1000D final amplifer are over... ;D

Hope to get on tmw with the 8877 and get some daytime-quiet-band bandwidth reports. Caw mawn.


BTW, Frank/ GFZ says he's making some good progress cleaning up his SS amplifier. His IMD is getting good at low power levels so far.  It's amazing how much an amplifier cleans up when run 1/2 to 1/4 power. Running full strap and drawing grid current is like letting the wolves out.

T

Tom,
No need to pull it apart for a picture on my account.  I meant schematic of grid input circuitry.  Brain fart on my end for not using the correct wording.   Is the input circuitry also similar to the 813 setup?  I have enough stuff to build a linear but wasn't sure what I wanted to do on the input side of things.

Nice metal work, BTW!

Go Tom !
Your a true experimenter.
You've got me thinking. I think I'll give a low IMD/IP3 intercept amplifier chain a try.
Right now the best, low IMD exciter I have is the QS1E (mating to the top of the QS1R.)  The QS1E has only 2mw output though.
http://qs1r.wikispaces.com/QS1E+Exciter (http://qs1r.wikispaces.com/QS1E+Exciter)

I've been experimentally running it into an old TR7 brick (preamp and 150pep amp on same board w/sink) with manually switched TR7 output filters I got off epay and now see the error of my ways.  ;D

So to maximize the "1E's" very clean signal and get up to a super clean 1/4 watt or so I think I'll build Clifton Labs quad GALi 74+ board,  http://www.cliftonlaboratories.com/z10046a_high_gain_amplifier.htm (http://www.cliftonlaboratories.com/z10046a_high_gain_amplifier.htm)
You'll note that this board's specs. are pretty nice. It is intended as a super clean receiver pre-amp but can also be used to get the QS1E's output up to real world levels.

BTW, Phil, N8VB, designed the QS1E at that low output so it could be used as a lab generator for, say, designing filters by simply feeding the E's output through the filters and right back into the QS1R's input to see the passband, attenuation, etc.

So back to the Clifton board:
Well that gets me to 1/4 watt for feeding into decent Class A tube amp circuits, or maybe another clean solid state circuit before going directly to a clean high power final (8874, 3Cx800 or 8877); the 8873/4/5 series alone at low feed levels should easily duplicate your 350.  I have a couple of 8875's and a spare Johnson socket.

You said ElectroCraft for the two tone gen.; did you mean Elecraft?  I need to look up the spurious specs on the two tone generator.

Hi Rick,

Glad to see you are excited enuff to build one up too!

Sounds like you have a nice clean starting point at a few mW.  I talked to Jack Smith from Clifton at length about using several of his preamps for this job. The bottom line is he recommended going with the Mini-Labs ZHL-3A. He ran IMD tests on his own at higher power levels (50 - 100 mW) and found they were good, (-45 area) but the MiniLab was about 10-12 dB better.   It takes a real honest guy to rec another competitor's product.  So I bought the ZHL-3A. We both measured it and it's about 58 to 63 dB 3rd IMD. That is stellar performance and complements the low level you or I use.

The Clifton will work FB, but we never know what we will be doing in the future, so might as well get the best.
I was able to add some pads (3dB and 6db) to make the ZHL happy.

IMD figures are generally hard to come by. They depend on a lot of factors but the biggest is how hard are we driving the stage.  For example, for extreme cleanliness, I never want to drive the 8877 harder than about 750 max watts output.  This is when it begins to draw grid current. Sure, we can get 2KW out of it and it will probably be about -30db 3rd, but maybe there's a way to do 1.5 KW with a bigger tube running 1/2 its max power.

Or do like some guys do... scream at full power in the pile up, and then back it way off when the contact is made... ;D

Yes, I mean the Elecraft kit two-tone board.  It went together in 45 minutes and worked the first time. They say it's good to -55db harmonic distrotion. It apears as clean as I need for measurments.  It was cool trying different idle bias settings with the 8877. Running the two-tone, I could clearly see the best idle. The IMD dropped like clockwork whenever the idle was near 150 ma at 3500V. That's about 525 watts of diss, which is roughly 1/3 of plate diss. That's a good rule of thumb and works in this case.   Funny that if I idle it harder, the IMD creeps up slightly, opposite of what we normally think.

Keep us posted as your plans evolve for the amplifier chain. As a word of encouragement... last night as I tweaked the settings of the overall chain and saw the interactions, I felt like I was looking at an SDR pan adapter (actually I was) and finally seeing the real picture for the first time.  I tuned my other receiver that uses a 6" antenna stub. The two-tones were at 60 over while the 3rd and 5th sounded like weak whistles that disappeared with the higher orders up the band. This was at 500 w out.  When I increased output to 1KW, these tones popped up out of the noise as usual when seeing grid current.  The AM signal using voice disappears after 5kc with gentle wisps. No tearing garbage.  It's well worth the extra effort, and it IS a lot of effort.  


Bob: I do have a working schematic with iron burns - and will take a pic for you once redrawn..  The input circuit is the same as the 813 rig, minus the grid leak resistor.   I also used a tranzorb on the screen for protection. See G3SEK's tetrode board and suggested circuit precautions for more info.  I designed up my own screen trip circuit and regulated supply, so avoided the cost fo that board.


T

BTW, Rick, speaking of choosing the right tube....  Check out the typical operating data on page two for the 4CX-350FJ.   1400V, etc =  -45dB 3rd.  I've proven to myself that the tube will do about 43dB of stage gain for a single stage.  Use Hi-Q circuits for both in and out.  

In contrast, I think the power grid triode triode family will do 15dB of stage gain?  So, even though a tetrode is more of a hassle, we might need two or three triodes to do the same job for total gain.  

http://www.g8wrb.org/data/Eimac/4CX350FJ.pdf

The alternative is a solid state amp for 10 watts IPA, but I've had trouble getting the usual MRF-150's to do any better than -32db 3rd or so. The CCI kit boards were no better.   The Yaesu FT-5000 DXXX thing advertises class A at 75 watts, super clean, but I can't seem to duplicate this with regular ss amps.  Frank/GFZ is working on the problem as we speak and feels he is making some progress. All about cores and windings.

Let me know what tube chain you settle on for a design there.

T

Tom,
Ok on the ZHL-3a. At 24 db gain that would get me to 500mw clean output.
Well I was hoping for 20 db gain from a xxxA7 triode giving me 47dbm or 50 watts at such low drive levels but your 15 db gain from a 8874 etc is probably more realistic.  Perhaps one clean class A stage in between, say an 8874 feeding a pair of 8877's (very expensive, and redundant, given excellent tetrode specs as you say.)

I like the idea of very clean amps but there is a point of no return in having huge power supplies to achieve it in the last two or three stages.  Vis, starting with the Mini circuits amp, for 1 or 2 watts out you have 24 volts at .6 amps or 14 watt supply and a comparatively huge heat sink.  Full high power voltages, cabinets, etc. for a 50 to 100 watt output amp also is very eye-opening.  There is a middle ground particularly at hf where even atmospheric noise will swamp IM products, even at 35 db down, etc. Yes distortion products are cumulative with the addition of stages, so I will be seeking something a lot better than the usual fried transistor and positive grid voltage class of amp but maybe not as perfect as the art allows.

Also of note to you, audio is injected to the QS1e via USB, either from a USB mike or normal mike through USB digitization, another area of distortion concern to me. There are software two tone generators right from sound cards so I need to research the distortion products from these sources too. 

Concerning the Yaesu 5k, wonder if a lot of their success comes from super regulated bias too, i.e. keeping it in tightly same point class A.  That seems to be a recently, better explored design hot topic.

Really cool amp and my hats off to you for trying to have a clean transmit signal.  Most simply do not care Tom.  That is a real problem.

I am on 20 meter SSB today searching for some DX and the band is crowded due to a contest.

Its filled with people overdriving amps, audio cranked. and splattering the entire band.  We are forced to build/buy better and better receivers to cope with idiots with cheap amplifiers and bad practices.

w9RE is on 14.224.4 right now taking out nearly 60 KC of band calling CQ contest on a voice keyer. His lookup shows a 781 with an alpha 87A.  Its so bad, I had to stop using the radio.

C

More like 25 years Tom!

Here is mine plus I also have 3 NCL-2000's on line. Note where the Mode switch is ;D

Rick,

You made some valid comments. Yes, the infrastructure to support the high level stages is huge if scaled up.

As far as QRO amplifiers go, I think -35db 3rd would be quite acceptable on the bands IF that was an average number. But I think it's probably in the high 20's these days. It's a bell curve with some very clean signals and some not so clean. The wild card is the human operating factor that can easily take a once clean amp and run it overdriven, etc.

There are times when we need a clean signal and times when it's not so important. In the daytime on 75M when no one's around, we can run about anything and no one will mind. But when the band opens up and DX starts rolling in, locals can be 40db over and the DX PW at  S5.  This is the time when it's easy to take out adjacent channels and get complaints.   The same holds true when QSOs get close, like the normal 3kc spacing for ssb or 5kc spacing for AM.  I hear it all the time... guys complaining about the next QSO up the band.  I'm just tired of worrying about it.

My goal is to get on and not be concerned about how close I am to the next QSO. If I am far enuff away to run a certain bandwidth of audio, then FB. I don't want to have to add in extra margin due to my own splatter, assuming it is excessive. Anything less than -30db 3rd, to me, will create problems when running QRO.

 I was talking with Frank again today and he's closing in on -40dB 3rd on his 300 watt SS amp. That's running it at 40 watts out, but it's getting better.

Carl:  When did National fold?  I figgered it was after the NCX-5, NCL-2000 and 500 stuff, but guess they contined on.

Clark:  OK on 20M.  I've noticed 20M has always been the roughest band aof all. I think it's the national competition, the ease of getting up a Yagi and especially the chance for RF to get into the audio more than the lower bands.  I think many of the severe splatter problems are related to RF in the audio, followed by being overdriven.

Well, doing some fine tuning mods and touches to the FJ amp today. I'm finding that C2 needs to be very big for best IMD. Q is important in the tank, but there appears to be a limit where higher Q degrades it.   The input Q appears to have little effect, though I am adding another cap to better match 50 ohms and allow the Q higher.

At this point I can run a pecker pulse thru the FJ amp chain and tune a receiver located about 200' away. The skirts of the signal appear very sharp with crud disappearing quickly off the sides.  In contrast, my FT-1000D at 200 watts cud be heard 5-6 kc up the band doing the same test.  The extra -15 dB makes quite a difference.

The downside is having an extra set of knobs to turn when QSYing...  ;)

T

T.,
 I've been thinking about how to best right-size clean, two stage tetrode amps. 
Your 4x350 uses 300v screen at a measly 9ma and 1400v plate at say, 80ma. Those are minor compared to a second stage using the same screen voltage and, say, twice the plate voltage and much more current.

So a choke input HV supply with the choke in neg. rail could supply the final amp tube's HV (off the Positive post choke and filter cap. point) and the driver amp off a pick off point before the neg. rail choke.   Since your running considerable constant quiescent current at the no signal point, current drawn would almost eliminate any sort of high wattage bleeder.  Alternatively, the plate xfor could simply be full wave tapped at two or even three points for plate and screen voltages.
 
The screen supplies would both use the same transformer or same taps on HV xfor. with the driver pulling off a simple resistive divider if even needed for slightly lower voltage, say 400 down to 300.  Decent solid state control could be implemented for the driver screen's small current also if needed. 

Hi Q input and interstage bandswitched networks, not to mention a Pi-L output network (really big) would be somewhat bulky, but not a lot more so that all the relay switched networks needed for SS amp filtering and antenna tuning.

Bias supplies:, first stage -14v could be from a SS regulated supply taken from a small 24 volt @ 2 amp xfor. used for that and 4x350's fil. 24 volt relays, other stuff.  Second stage a common 120 vac. pri./sec. transformer.

All this should easily fit in a 32V-2 sized cabinet or a little smaller. After all, your shooting for 1000 to 1500 clean watts instead of 2500 "dirtys."   ...Aren't you?  ;D

Rick,

What you've described is pretty much what I have now.  The screen definately needs electronic regulation to +- 100 mV. No more than 1/2 volt is possible. Must be a SHUNT regulator, not a pass.  A precision elec reg is what is needed to gets us into the -45dB 3rd area. Even a zener will drop 3-4 volts with only 8 mA draw.   The grid can be a simple zener shunt reg since there is no current drawn.  

I use a 47K in series with the grid for protection and isolation and stability swamping, since the bottom end is bypassed to ground.

The 4CX-350FJ plate idles at 80 ma but peaks are about 160ma.

The screen protection is important and covers other mistakes too.

I use the exact Eimac specs at 1400V as shown in that last link I gave you.  25 watts out is about 800 watts from the 8877, so it's still in the very clean area.  It will drive the 8877 easily to  2KW, but then the IMD specs of the 8877 are the bottleneck, as you know.  An 8877 will do -45dB 3rd when run below 600 watts.

BTW, from testing I found that the 8877, when running in the low power clean mode under 600 watts out, requires LOWER  plate loading for best IMD. IE, use MORE C2. This is opposite of what we usually do. For higher power closer to saturation, then running the tube more heavily loaded is cleaner. (less C2)  There is a cross over point probably when the stage starts drawing grid current at 700 watts or so.  These results are from experimental IMD tests with the FJ driving the 8877.    I was amazed yestersday how clean I could tune the whole system and how EASY it was to mess it up with bad IMD by improper tuning.

As you know, an IMD testing setup is mandatory when setting up these high performance systems.

Hope you build something up!

T

BTW, we probably think this IMD stuff only applies to linears, but there's no reason why a standard plate modulated rig, class E rig, DX-100 or ANY AM transmitter cannot be tested for IMD. After all, we are mixing audio with a carrier and if not perfect will generate artifacts in the sidebands. Any AM rig could probably be tuned cleaner with adjustments in voltage, current, loading, ratios of modulation, running less power, different biasing in the final or modulators, etc.

I plan to test my 24 pill KW class E rig in the next day or so and will post the results. I've gotten good reports with it from the SDR guys, so shud be a reasonable example.

Someone correct me if I'm wrong, but on AM, I wud think a single tone for audio and viewed on the SDR spec analyzer is all that's needed.  But how do we compare it to ssb to get a similar standard? Is the main, single, audio  peak compared to the next peak up the band (3rd order) and subtracted in dB , just as on ssb?

T

I'd start with just the carrier and look at (compare) the harmonics uo to say, the seventh.
Then inject single tone at slightly less than 100 p.c. Modulation.   Look at that.  Then go ahead and use your two tone gen.  It just represents constant amplitude speech so should be useful for spectrum analysis.  In each injection case you're comparing apples to apples, so decibel addition should apply.  I'd write more, but this darn iPad is not typing friendly.

Ya know, I've never checked the harmonics on any of my rigs.  With the HPSDR, I can easily dial up there and take a look.  Maybe I've been whirlwide on all bands all the time... :-)

I'll get a feel for AM IMD when I fire up the 1000D tmw and go from there on the class E rig IMD tests.

Today finished up a QRP SWR bridge KIT that is good for SWR down to 100mW and also reads power from 100mW to 8 watts.  Nice digital readout. I plan to use it with the 500 mW output to quickly PRETUNE the various ampifier input networks as I change bands.  

For only $50 it's a nice deal. Went together in about 2.5 hours and worked well.
http://www.4sqrp.com/QRPometer.php

I added grid current monitoring to all my GG amps today. For some reason I didn't fully understand how the shunt resistor in the negative lead worked. Everyone said to just stick a 25 ohm resistor there and hook up the grid meter. It didn't make any sense to me. Well, finally figured out that the shunt really is a shunt resistor and must be closely calculated and matched with the grid meter. I ended up with a 1 ohm shunt and a 1 mA meter movement for a 100ma full scale.. No wonder my grid current readings were always screwy.  


Here is an excellent paper about plate and grid meter monitoring for GG amps. Everything needed to do the whole job and do it safely:

http://www.w8ji.com/metering_amplifier.htm#Shunts_and_Multipliers_calculating_or_selecting_metering_resistance_values

T

Update of testing:

I have come across an interesting combination of parameters that have brought the IMD performance of the 4CX-350FJ driver/ IPA up a notch.

By adding at least two times the C2 loading capacitance above the normal tank calculation for a given Q, I am able to get an even cleaner IMD number.  This also requires some more C1 and less L which increases Q, but the C2 seems a bit out of proportion.   This is the opposite of what we normally do to a GG amplifier that runs at full power. Both the tetrode and triode amps are showing this IMD improvement.

The downside is that the power output is reduced before flat topping occurs. Normally, by adding a lot of C2 (lightly loaded) and running at full power would create a nasty signal.  But we are looking for a super clean, low power signal from a driver stage. By running X2 C2 and pulling less power uncovers another higher level sweet spot.  Under these conditions, I can pull only about 50 watts from the 4CX-350FJ before flat topping occurs, but I estimate the IMD is approaching -50 dB 3rd.  This is the best I've seen so far.   The 8877 actually needs less than 25 watts drive for my purposes, so FB.

The same holds true for the 8877 when running about 500-750 watts out. It can tolerate much more C2, flat tops earlier, but gives better IMD at this lower power level. Never wuda thunk it.  

I am finding it best to tune each amplifier in the chain independently of one another. I make sure it is matched 50 ohms in and 50 ohms out, tune for the best IMD and then run them together.  Most of the time the amplifers need no further adjustments once cascaded in a chain for operation. Simpler and more standardized this way.

It appears we can get away with some things when demanding low power and running higher Q vs: running near sturation.


Current tank values:

C1 = 276 pf     L = 7 uh      C2 = 3100 pf.     Running at 25w output gives a plate load of about 7K.  The Q works out to be about 50!   Quite a bit of flywheel effect there.


Two-tone efficiency:   126 watts in /  25 watts out = 20%.   That's about normal for a quasi-class A1 stage.  It draws no grid current and when running 25 watts, the plate current barely moves. 

Take home:  IMD approaching -50 dB 3rd order.

T

Officially they folded in 1991 Tom when the IRS had their auction. The NCL-2000 was in limited production till the end and lots of completed and partially completed NCL's were there along with the NCX-1000.

If I wanted -45 IMD Id run a 3-500 at 1500V and be done with it with less complications. A pair in a SB-220 with 160 added, run in the CW position and zero bias, would make a YC-156 very happy.

Single band amps rule (or are an absolute necessity in this case).  ;D

Carl: What would you use to drive the 3-500?

Yep, been there.  The 3-500Z's worked FB at 1500V, but as HUZ axed, what to use for a 50 watt driver? The bottleneck became the driver in the system. The FT-1000D is clean until 10 watts and then falls apart to normal ricebox standards.   No ham rigs rigs around will do -50dB third at 50 watts or so.  As I understand it even that FT-5000 DXX rig with the class A SS amp was measured by the ARRL at about -40dB 3rd.   Subtract their 6 dB IMD marketing factor and we are down to -34dB.

The 4CX-350FJ will be the liason disco duck diplomat  between the HPSDR 100 mW or the FT-1000D 3mW driver.   The 1000D driver and and HPSDR are both about -60dB 3rd.   So a clean 50w driver is important to complement the chain.  


Update:

Still playing around with the 8877, drivers, bells and whistles.  I'm trying hard to push the 8877 beyond -45dB 3rd. Presently I'm playing with a Q= 40 to 50 to see what effect it has on IMD.   This requires about 4,000 pf in C2 and only 5 uH for L1.  This is exciting stuff when good results are seen. I'm trying to apply some of the techniques I learned from the FJ project to the 8877.

Built up a universal grid current trip circuit for the GG amps in the shack. Since they feed off the same HV plate supply, I can use the same grid shunt resistor to trigger the same trip circuit and open the PTT circuit. No more drunken stuper mistakes with 200 watts into the grid of the 8877 or forgetting to turn on HV, etc.....  :o

I mounted a large grid current meter (8" X 6") to monitor grid current in all the GG amps. My goal is to run all of the GG amps without any or little grid current. It seems IMD starts to fall apart when it occurs. There are ways to set up the stage so that grid current is minimum for a given power output, but the tradeoff is less power output.  Stilll experimenting.


Getting there.


T

You bring up a good point, Big Country.  Probably the major reason for single band is the tank coil optimization.  I have overcome that by using a roller inductor in both the grid and plate circuits of the FJ amp.  With a link input, the greatest gain can be had. Plus the roller lets me zero in the 50 ohm input.

The plate tank is also sensitive to optimum L, as shown how the IMD has a sweet spot when found.

I don't have a roller in the 8877 tank, though there is one in the cathode input.  If I'm gonna play the IMD game with the 8877, think I'll take a look around for a one KW roller inductor for the 8877.  Thanks for the thought, OM.

T

When your running GG there has to be grid current.  Your feeding power through a power amplifier, not a voltage amplifier, and that will be added to the output...   Well, you know what I mean, some current has to be passed to make power.  So adding some driver power at low IMD will cumulatively help the output depending on how much of a fraction it is.  The distortion curves cross at some proportion point.

I admire trying to keep grid current very low but  this looks a very inefficient process with far more than one comparative stage to achieve low IMD.   Check the microwave, HDTV people to see how to amplify with very low IMD.

I'm awaiting the Clifton board (its size meets my needs better than the mini circuits) and will install it inside the qs1r cab with shielding and schrack 24012 relays to use it as both a rcv preamp and post qs1e amp.  This will get me to 1/4 very clean watts w/o any post filtering.
Then the 4cx350 or similar class A, little B,will yield some decent SSB (gasp).  Then  I'll feed my pair of 8875s GG and experiment with basing, load and drive.  The GM3SEK board will allow any degree of variable bias  desired.  I will be happy with real IMD's of 40 or so. My amp also has rotary coil on 80 and 40. Pix before 3SEK mod on QRZ.com.

Sounds like a good plan with the Clifton amp and plans, Rick.  You should have everything you need to play around and get some outstanding results.

Let me know (or anyone else) if you have a one KW roller inductor for sale.  I need one for my 8877 amp.

The HDTV people - are you talking about feed forward systems and the like?  After reading about it, I've yet to find one real life working system that can be built by an average guy like me. There's a lot of ivory tower talk on the web. Way too complex and digital for me. What other way did you mean?

This extreme IMD game is like drag racing. The last improvements are more difficult and often seem like, "why bother?"   But that's where the fun is to me - trying to see how far I can push things. In the end, after a lot of work, we sometimes come up with a combination that woks surprisingly well.

As far as grid current... there are definately parameters that cause more or less grid current. They seem related to loading, Q and drive. The trend is the lower the grid current, the lower the IMD. For example, with the 8877, I started out with about 30ma of grid current for 600 watts out. By changing parameters, I was able to get the same output power, 10ma of grid current and better IMD.  But remember that I am axing for only about 1/3 of the 8877's output power.  All of this IMD stuff will certainly mean an extra stage of amplification in the end. It is not meant to be an efficient game at all, at least in the lower level stages.

BTW, I was listening to 75 AM the other night and heard a 30 minute squabble between some southern and northeastern AM stations concerning splatter.   This bandwidth issue is still with us, AM or SSB.  It's  a worthy goal to run our own rigs through IMD tests to be sure they're reasonably clean.  This is something that cannot be seen on an  o'scope alone.   An SDR spec scope or even tuning a good receiver with a 6" antenna stub can tell the story of our own signals.


T

Nice work tom.  Keep the updates comming.

SH*t in and SH*t out.   Have you seen the Elecraft and Flex radio IMD numbers?  27 and 22DB?  Garbage. So tired of hearing the Flex's trash the band. Run that into a two or three holer and its buckshot city. The 2000D has a real world 41DB 3rd order test by the way. Not sure if the 5000 has the exact same output section or not.  At any rate. The Class A option is nice for low drive amps.

Keep up the good work

That -41 number is in Class A and is the ARRL measurement, so subtract 6 dB to compare to Tom's numbers. In other words, it's -35. Not exceptional but much better than most of the new crap out these days. The FT5000 supposedly beat the 2000 by 2 dB. I wonder how many people run either of these in the Class A setting?

The Flex 5000 was measured by QST at -35 (29 for comparison with Tom's numbers), better than many of the newer radios but not as good at the class A output radios. I've heard some claim at reduced power output the IMD number improves on the Flex 5000.

To add to what Steve said, here's a few postings from the HPSDR refector about how IMD is measured, the ARRL way and the Mil spec way. These guys know much more than I do on the subject so I will defer to them:



"Hi All,
ARRL method of measuring IMD covers up the poor quality of rice boxes on the market.
I'm here to get away from that over priced junk. The mil method accounts for PEP by running at
a power level that puts the combined waveform peak at the amplifier power rating. The spectrum analyzer
displays the carriers and IMD with the same demodulation so how is it valid to add 6 dB to the carriers and
not the distortion.  It is all about high performance as in HPSDR.
Frank WA1GFZ"


"When you look at a signal on Hermes and Mercury and see IMD distortion it is X dB down. What you see is the real number.
I don't agree with the commercial method claiming another 6dB for PEP. The distortion should also get another 6dB in my view so you end up with the same number. The Mil Method checks IMD at 6dB below max power (3dB on an average power meter) Then  the IMD is the difference between IMD and carrier peak on the display.
Funky specs are for rice boxes.....Frank WA1GFZ"


"Hi Frank,
 
completely agreed.  What you mention is not only a Mil Method, it’s  a commercial standard in IMD evaluation e.g. for base stations in mobile networks where linearity is even more important than in our ssb application. The 6 dB PEP bonus is not more than an advertising trap.
73, Helmut"      [A Talented German engineer]

Here's two pics of the new grid current monitor and trip circuit for the 8877 GG amplifier. It also works for the other amps in the shack since they all use the same HV supply and grid shunt resistor.

Whenever I make a mistake to overdrive an amplifier, run without HV, use the wrong antenna, etc., the grid current will hit a predetermined point and open the PTT.  Works pretty well.

The BIG meters -  I bought them back in the 80's at Deerfield thinking maybe I'd use them someday. They measure  6" X 8".   Used for HV and grid current.  The top scale of each meter is accurate - 0 to 12KV  HV,  and 0 to 120 mA grid current.  [That's in SonoRays, of course]

Also notice the temperature gauge that can be preset and trigger an alarm if the tube exit temp gets too high.  I've done some dumb things in the past like forget to turn on a blower or have a ducting fall out - ruined tube.  It presently says that the exit temp is 63F and the alarm goes off at 110F.

Notice Mr. Clean in the third pic sports a new C2 padder. This is for the higher Q experiments.


Thanks for the comments, Clark.  I think we're all learning some new things here.

T

Nice.  I respond, The post is up for few minutes and then is removed.  Had lots of good data in that post too.  I will email you a copy Tom. You would have enjoyed it before I got sensored.

c

Clark,

Yep, losing a post can be frustrating...

Sometimes I type up a reply and when sent, the little icon goes round and round and never updates.  This is why I always copy the message just in case. It's easy then to paste it and try again.

Post it again and it will probably go thru OK.


T

It was up on the site and removed.  It listed out private party testing of rigs not ARRL and it was showing Rob sherwoods tests of the Flex and elecraft and I think that got a Flex owner upset to see his rig is dirty and the post was deleted. The 6DB thing came from Rob sherwood. He used that in a power point presentation. You cannot just take 6DB off ARRL listings. It was a blanket statement used to prove a point.  Oh well The power point is on the web for transmitting signals if anyone wants to read it.  Most radios where tested private party in format once his gripe with the ARRL came out. People wanted to know what the real figures are with the Sherwoods suggested meathod.

In short, The Class A radios Mk5, 2kd, 5kd, 9k ect are the cleanest by far and the BS figure listed here is just that, BS.  The real world figures are in the low to mid 40's.  It has merit when using a Low Drive amplifier like a two holer or three holer that needs just 15 to 25 watts drive for full legal limit.

It thought it was pertinent info for this thread but I guess not.. Sorry.
C

Nonsense Clark. I just checked the logs. No post of yours or anyone else was deleted in the past three days. Your Flex conspiracy is silly and completely out of line. It tells me you really don't know what you are talking about relative to IMD measurements.

The 6 dB thing did not come from Rob Sherwood. It comes from specifying IMD on a per tone basis versus a PEP basis. Using PEP gives a 6 dB higher number when compared to the per tone basis. It's simple math. No conspiracies, no bashing and no scarey Flex boogie men involved. It doesn't mean one method is right and one is wrong. It just means you have to pay attention to how the measurements were made when doing comparisons. So, you can't compare measurements made by the ARRL using the PEP approach to measurements made with the per tone approach without applying a 6 dB correction factor to the ARRL measurements.

All this has been covered in professional journals, magazines, T&E app notes. It has nothing to do with amateur radio per se. IMD measurements are hardly unique to the ham radio realm. Let me quote from an Analog devices app note:


In other words, your "real figure" and "real world figures" statements are meaningless.

Then there is error. To measure -41 dB IMD, the sources must have IMD much less than -41. If the sources were 10 db cleaner (-51dB) there could still be as much as a +2.4 dB and a -3.3 dB error. Once again, exact info on how the measurements were done is important.

Further, no one in their right mind is going to claim two-tone IMD measurements (regardless of how they are made) represent real world dynamic conditions. This is why three-tone (and multi-tone) and pulsed tests are done (although apparently not in the amateur radio world).

All the rest of your rant makes even less sense than the Flex conspiracy part. No one ever claimed that the Class A radios weren't the cleanest on the ham market at the moment. Please stop with the nonsense so we can discuss things here like adults.

Tom...Can you post some pictures of the HPSDR display?   I am curious about how you feel on the accuracy of the measurements you are taking from the HPSDR...Is it easy to read?  Does the display compare to a spec analyzer presentation...?..I'm trying to get a softrock Ensemble ll going  and I'm wondering if I will be able to measure my rigs like you are....I have NO experience with SDR to speak of...I tried to get something going last winter and gave up after I couldn't master the software and wound up messing u p the hardware as well....I have a new "tony built" Ensemble ll now and am ready to start over....I'm on travel in AZ right now so I'm away from the shack..This is all very interesting to me... Keep up the good work..It is inspiring...Steve

Hi Steve,

The HPSDR or any SDR system will open up a new whirl for you. I understand that VE7KHZ just got his HPSDR system working in software the last few days. The HPSDR guys are there to help and won't let you fail.

Well, first let me say that I don't do this for a living like some of the engineer guys here. I was trained as a technician and don't have a BSEE so I am a TECHNICIAN... ;D   I axe a lot of questions of my mentors when I get stuck  - K1KW, WA1GFZ and a few others.  I do what I must to get by to complete a new project and no more. So you will not find a lot of theoretical stuff from me - mostly practical stuff I have "discovered" which many already know....:-)

I like to work in relative comparisons, not absolutes. For example, I have trained myself to recognize what a clean signal looks and sounds like on an S-meter and on the HPSDR. I have learned what a -31dB 3rd IMD FT-1000D looks and sounds like in a relative way. When I hear a signal that is -45dB 3rd, it jumps out at me. And a dirty -22dB 3rd signal is obvious too.

Once these relative standards are in place, I can then work on what I have and make incremental improvements by comparison.  Notice I try not to become absolute. Absolute precision requires an excellent test setup using a solid sampler, pads, and repeatability. The HPSDR screen is accurate enough, but the stray crap and pickups, poor sources and different environmental conditions can change readings. I eventually want to get a stable setup, and realize the limitations of what I presently have.

So, if you want to be able to give out absolute numbers compared to the real whirl, then focus on good and accurate accessories to support the HPSDR.  Look at what Clifton Labs uses for his IMD tests. He uses two separate transmitters (or sig gens) sources that are mixed into a combiner. This can give a source -60dB 3rd or better.  In contrast, I use my own trasmitter with a two-tone audio signal that is only as good as the 2-tone and transmitter itself.  Go to his website and see how much $ he has invested in his.

BUT, if you go with a modest setup like I did, you can still make meaningful measurements and improve on your own equipment. Just do it in a relative manner and don't expect to pecker match with the real whirl.

You can also get a good idea of whats going on on the air.  In fact, Frank sent out some 2-tone audio on the air the other day and I was able to tell him (good guess) within about 3dB what his 3rd order IMD was by looking at my HPSDR.  He has a very nice and accurate lab setup there.

I will take some pics later, once I get everything finalized, but it's just a matter of using the curser to take a reading of your peak of the 2-tone and comparing it against the next significant (and lower) peak up the band. (3rd order)     The following lower peak will be your 5th, etc.  The difference of these peaks in dB is a rough idea of your 3rd, 5th, etc IMD figures.

Keep reading, test some of your own stuff and you'll catch on quickly.   And keep posting the results you are finding with your rigs.

Hope this helps.

T

This is the per tone method of IMD measurement. All other things being equal, it will yield numbers 6 dB less (worse) than the PEP method.


Below is a simplified spectrum (showing only 3rd-order products) of what you are likely to see on your SDR. The relative measurement shown is what is done for the per tone type measurement.

(http://www.amwindow.org/misc/png/imdspectrum.png)

Very cool Tom...Thanks for the explanation....You are getting results that you can compare and use to tell if things are better or worse....and you have been able to put some values/ (numbers)on them...and that is what I am looking to do..
   My first attempt didn't work out so good for me......Frank helped me some and encouraged me when I was having trouble as did a few others here on the Forum....When I get back to the shack, I'll be trying again ....It is all fun and interesting to me ...

Below is an excerpt from "QST Product Reviews: A Look Behind the Scenes", QST, October 1994, pp 35-38. The article gives details on how the various performance measurements found in the product reviews are performed. Below is the section on IMD.

Two-Tone IMD

  Just like your receiver, your transmitter produces IMD distortion products. To measure IMD we use an audio generator. The generator produces two tones, one at 700 Hz and the other at 1900 Hz. We put the transmitter in the desired SSB mode and modulate it with the two-tone generator. The output is measured by the special PEP-reading watt-meter and observed on the spectrum analyzer.

  Figure 5 shows a typical IMD plot. Each vertical division is 10 dB; each horizontal division is 2 kHz. You can see the two desired signals in the center of the display. All other pips spreading out to the left and right are the odd-order distortion products. In the example shown, the worst-case third-order product is 35 dB below PEP, the worst-case fifth-order product is 39 dB down and the worst-case seventh-order product is down 45 dB.

(http://www.amwindow.org/misc/png/imdspectrumarrl.png)

  We test the transmitter in both upper and lower sideband; QST Product Reviews always shows the worst case. Poor IMD performance can result in unnecessary interference to others on the band and illegal out-of-band products. The greater this number, the better. Typical modern transmitters should have IMD performance better than 25 dB below PEP for the worst case low-order products. High-order products must be much less.


As you can see in Figure 5, the two test tones are 6 dB below the top of the spectrum display, or -6 dB since the top is zero. Yet the article says the worst case third-order product is 35 dB down (and it is 35 dB down from the top of the display). If you compare the level of the worst-case third-order product (the first pip to the right of the right hand test tone in Figure 5), to either of the two test tones (per-tone based measurement), you can see the difference is much less than 35 dB. It is 29 dB. This is because the PEP of the two tones together is 6 dB greater than either one alone. Since the ARRL references their IMD measurements to PEP, there will be a 6 dB difference between their numbers and any measurements done on a per tone basis.

Very illustrative Steve.
The closest ARRL comes to mentioning the center line or "phantom algebraically added PEP" between the two 6db down tones on the spectrum analyzer is:

Yes, too bad they didn't mention that "observed" was from the top of the grid at 0db.

Class AB mode graph. Not QST.

The output is measured by a "special  PEP wattmeter  ???

I am told Special means a NIST calibrated meter Steve. 

Tom,  How about his idea of testing the amp:

http://www.w0qe.com/Technical_Topics/imd_testing_of_amplifiers.html

I wonder what you use to test in your lab.

Attached is the Yaesu vs K3.  Another real world test thats interesting to compare to other published figures.

C

Hi Clark,

The method he is suggesting is the best way to generate a two-tone signal. We can use two RF signal generators or two CW carriers from two riceboxes and combine them using a combiner of some sort.  This will give a signal worthy of any test.

The Sherwood example is using white noise. It's difficult to compare this numerically to two-tone tests, but it does show that the MKV class A is cleaner than the K3.  I sometimes use my pulse pecker signal to get a rough idea of how an amplider bandwidth is doing too. It becomes more of a relative comparison thing, but works when making improvements within one test setup that doesn't change.

In my shack I use a standard two-tone audio signal generated from a Elecraft two-tone board.(The 2T)   When using my 3mW signal source and 1 watt ZHL-3A MiniCircuits lab amp, the HPSDR spec analyzer shows the source signal to be at least -60dB 3rd order or better. So this signal source is plenty good for my purposes, since I do not expect my amplifier chain to be any better than -45 to -50dB 3rd.   We want the signal source to be at least -10 dB cleaner than the amplifer we are testing, so this fits the bill.

If I were to do testing on high end receiver preamps, etc., I would need to look into dual signal generators, a combiner and better, isolated  sampling, which raises the price.

T

THanks for the information Tom.  The combiner looks easy enough to build.  The issue is that if one rice box does not get keyed up, then you could damage the receiver.  So you need to make sure you have mode, level and single PTT all working. 

I will look into the elecraft generator.

C

Damage to the unkeyed ricebox - that's a good point and I never thought of that.  And even if they are both keying properly, if one keys up 100 mS before the other, there could be a problem - damaged RX front end.  

The best way might be to get into each ricebox and disconnect the receiver at the internal T/R relay.  I did this on my FT-1000D to generate a second ant to my HPSDR receiver, but that is another story.

Remember, when using the two tone audio gen board, you still need ONE transmitter that is very clean to drive the amplifier.  Using just a two-tone with a standard ricebox may put you back into the -31dB category, which is unacceptable for testing.

You do not need the Elecraft 2-T gen when using two riceboxes and a combiner. They themselves create the two tone.

This is why they use two riceboxes to combine into one clean, high level signal.

In my case, I use clean drivers to generate the signal using the 2-T board.  The 2-T generator drives the 3mW into the ZHL-3A and into the 4CX-350FJ. This gives a 50 watt signal with about -50dB 3rd. This is barely clean enuff to test a big amplifier.

T

Yes. I understood the two rice boxes would generate the tones. I was thinking of the two tone gen for receiver testing. I am going to modify one of my Yaesu rigs here with a different roof filter and I would like to do the testing before and after and adjust/tweak the filter transformers for optimum performance.

I agree its dangerous to use two rice boxes like that. 
C

Not worst case.



You took that image from this link.

http://www.webalice.it/romano.cartoceti/yaesu_ft-2000.htm#PICTURES GALLERY


You failed to look at the image right next to it - the same rig tested on 20 meters. The IMD is almost 10 dB WORSE! This is why QST reports worst case, as stated in the article except I previously posted.

(http://www.webalice.it/romano.cartoceti/images/FT-2000%20TX%20TT_20M.jpg)

More from that page.

Nobody knows what you are talking about..   All you do is follow me around the net and post things nobody can decipher.

C

You are delusional. Name one other site I've ever posted where you were involved.

Those who are technically competent understand. Come back when you have something intelligent to post.

I have no problem following what Steve has presented. He's spot on. I do this almost everyday at 2.4 and 5 GHz.