Designing a Clean Linear Amplifier with RF Negative Feedback - Opinions Wanted

[K1JJ]

Hola,

I posted this message below on the AMPs Reflector today. I'm looking for all the ideas I can get before making a decision on a design. Please feel free to add. [If I can find a good design, Mr Ugly may be used for parts... ]

T

---------------------------------------------

Hi All,

I'm planning to build up a 75M monoband amplifier that has at least -45db
to -50db 3rd order IMD using heavy RF negative feedback. 10MW in, 1500W out.

I wanted to get opinions here what others thought was the best general
configuration and perhaps suggest some real world tubes.

It seems that the best amps that use RF feedback start with a much larger
tube than generally used in open loop linears and run it about half power .

I plan to use the 10 milliwatt output from my FT-1000D as a source. This is
the low level stage just before it goes into the power amp board. I'm hoping
this is class A stuff and a  much cleaner starting point than its -32db IMD
200W solid state power amplifier board output. Any opinions on that?

For this amplifier I am thinking of either a chain of grounded grid stages
so that they have individual built in feedback and an additional loop of
feedback around the whole chain.

OR, use a tetrode final with regulated screen voltage with stages going back
to amplify the 10MW level. Hopefully, use up to 15-20db of  loop feedback
assuming the gain is available.

I would appreciate any pointers to URLs showing working schematics of the
big commerical amps that Collins, Harris, etc use, and ideas you may have
for a design.

For example... I don't know what good low level tubes to use, but lets just
say it went something like,  6AU6, 12BY7, 6146, 811A, 3-500Z,  GS-35B  - all
in grounded grid with an additional feedback loop.

Or, for grid driven:  The same low level stages, then 4CX-150, 4-1000A,
4CX-3000.  Looking for conservative ratings at 1500W and high IMD figures.

Another thing I wonder about is if a tube like the 6146 or 4-1000A which has
modest IMD figures open looped,  can successfully participate in a -45
to -50db IMD chain when big feedback is used?

I'll be building this amp over the next few months and will post the
results once completed. Thanks in advance for any help!

73,
Tom, K1JJ

[w3jn]

Get the Collins book I mentioned in response to your post in the Vanguard thread.

[n2bc]

Hi Tom,   just want to pass along another potential topology.  The Harris RF-110 amplifier IMD is spec'd at 40dB below full power output. It uses   a pair of 4CX1500Bs in the PA at 1KW output. These are driven by a pair of 8122s running Class A (how's that for overkill!).

There is no RF feedback, no neutralization. The 8122s are driven with 100mW across a 56 ohm resistor. The interstage and output circuits are broadband transformers.  2 to 30mHz in 19 bands. Electrically the RF side of things is quite straightforward.

When I initially got the Harris, I drove it from the transvertor output of a TS-850 through a single 2N5109 broadband amp stage. I now drive it with an IC-756 through a 30dB attenuator. The ICOM's transvertor in/out scheme was not workable without mods.

I had a discussion with the designer, W2DMR, some years back. Harrris worked with EIMAC to spec the 4CX1500Bs specifically regarding IMD performance.

Both stages in the amp are run at relatively low voltage. The 8122s @ 500V and the 4CX1500Bs @ 2200V.  I don't have any objective measurements. But when I first put the amp on the air I did a bunch of tests with locals and everybody felt it was clean.

It's interesting that the datasheet for the 4CX1500B shows IMD improving as plate voltage increases... -38 @ 2500V, -40 @ 2750, -43 @ 2900, yet Harris runs them at 2200. 

I think there's a manual for the RF-110 on BAMA - or just hollar & I can email to you (3MB or so).

Have fun with your project!

73, Bill   N2BC

[K1JJ]

Hi Bill,

You're the 2nd one who mentuioned that Harris amp.   Wow, 500V on 8122's and 2200V on 4CX-1500's. Conservative or what?  Though a good point about IMD improving with higher voltage.... strange choice.  I think it had something to do with 100% duty on RTTY.

I've already received a few private responses to the AMPS Reflector post. One guy suggested to add a resistor to the cathode of ANY linear amp. Like 10 ohms or so. This, of course, is negative feedback like on any unbypassed audio stage.  I just don't know what exact value to use or how much it adds in db to IMD figures. I may try it with my current amps as a test anyway.

I'll pass on the schematic, Bill, since it doesn't have neg feedback.   And, John, I'll have to think where to get an old Collins ssb manual. If there is just a few pages on neg feedback in it, maybe I can get it from someone online...

Tnx, OM!

T

[WA1GFZ]

bama has the amp.

[n2bc]

Tom, I missed the cathode resistors!

The Harris has 10 ohms to ground on each of the 8122s and 5 ohms on each 4CX1500B.

[K1JJ]

Tom, I missed the cathode resistors!

The Harris has 10 ohms to ground on each of the 8122s and 5 ohms on each 4CX1500B.

Bill,

Tnx for the values.

For some reason, I've never used that as a regular practice. I added a 10 ohm today to the GG triode amp and it halved my idle, so I switched out a few diodes and all is OK. BUT, would not the DC bias swing around under ssb modulation... a bad thing?  One guy suggests using a choke across the resistor for DC pass, so RF is not affected for neg feedback.

Though, I found that the amp drives with about the same power as w/o the resistor. ie, if it added -3db neg feedback, it shud take twice the power to drive it. I saw little change, telling me it added little neg feedback.

T

[w3jn]

The Harris amps were used in high speed (300-600 baud) full-duplex HF data circuits.  They were also used with the Harris tunable pre/post selector.  When you're running full duplex, IMD, harmonics, and other crap MUST be minimized so as to be able to hear with the receiver with the xmitter shooting out data.  I don't recall what the frequency separation was, but it was on the order of a MHz or less.

73 John

[K1JJ]

And the plot thickens.....

I just received this from a guy who axed an RF engineer about putting a cathode resistor in a GG amp for neg feedback.   He is commenting about a technique recommended by a well known amplifier guru....

"
> His logic is WRONG! Adding a CATHODE RESISTOR to a GROUNDED GRID (Cathode
> Driven ) amp does NOT increase negative feedback!  The drive and the
output  are
> IN PHASE. His logic applies to GRID-DRIVEN amps, where the CATHODE and
> SCREEN are at/very near R.F. ground. (Usually TETRODE or PENTODE
amps)-With an  amp
> such as he described,(Grounded-Grid) he will need to add resistance IN
> SERIES with the R.F. Drive, NOT the cathode ground return.- "


Guess that blows THAT idea.    [oh I feel so scrotless sometimes]

He  said put it  "in series with the RF drive"  ??

T

[WA1GFZ]

Tom the 30L1 had a third winding on the fil choke that fed feedback into the cathode from a the tank circuit tube area if I remember. this would be a way to introduce feedback. Heavy class A bias is a must.

[K1JJ]

Tom the 30L1 had a third winding on the fil choke that fed feedback into the cathode from a the tank circuit tube area if I remember. this would be a way to introduce feedback. Heavy class A bias is a must.

Hmmmm.... interesting thought, Franz...  I like it and will have to think more about that.  Better yet, I will axe the guys on the AMPS reflector.   

----  edit:
Whoops, looks like you already did!  HA!

T

[WA1GFZ]

seems simple enough, just wrap some wire over the fil choke and put a probe near the tube. Adjust the turns and length of probe to set fb level. This way there are no series elements to introduce additional phase shift. the other end of the third winding would just go to ground, a series cap or inductor.... boss needed me....to continue the 3 options would allow you to play with the phase shift of the feedback. maybe even a variable cap/inductor to get real crazy. 

[K1JJ]

seems simple enough, just wrap some wire over the fil choke and put a probe near the tube. Adjust the turns and length of probe to set fb level. This way there are no series elements to introduce additional phase shift. the other end of the third winding would just go to ground, a series cap or inductor.... boss needed me....to continue the 3 options would allow you to play with the phase shift of the feedback. maybe even a variable cap/inductor to get real crazy. 

Sounds interesting.

The filament/cathode is a real low impedance spot. The bias from idle to cut-off is only a matter of about 20 volts or so.  Not sure how that affects the amount of turns and coupling needed.

Do you think a little 1" X 1" plate standing 2" away in front of the tube making maybe 10 pf and connected to a coax is enough pickup? How many turns around the choke would you try? 

The problem is that the fil choke is under a sub chassis and lot's of work to get at. It wud be nice to start with the right amount of turns around the choke and bring everything to the outside to JS with.

Guess what... I just checked Mr Uglie's fil choke. It's just two wires run straight with 12" of big ferrites over them. No turns.  Think that will still couple OK with X turns?

T

[WA1GFZ]

Tom,
Just string a third wire through a few of the beads near the tube. Or multiple turns on the first bead to keep the current loop small The coupling C vs coupling T will set the amount of power fed back. Yes this will be a low Z but that should be more stable to control. I would just mount a ceramic feed through on the chassis. The type with #6 threaded rod so you can attach a plate above the chassis and a wire to the fil choke below. Find a place near the tube on top and also near the choke below so the leads will be short. You will have to play with it to get the desired effect.
You could monitor the feedthrough under the chassis to see how much voltage is dumped into the fil choke third winding. At some point more drive will be required so you know you are getting negative feedback. Short direct connections will make it easier to control.  Multiple bands may require an additional control like a variable cap in series to find the sweet spot for each band. Who knows you might get lucky like a neutralized amplifier.

[W2VW]

How about an additional L on the output with an adjustable link nearby? A few pF back to low impedance doesn't seem like it will amount to a hill of beans in this crazy world.

[K1JJ]

Yes, Dave, it IS a frickin clazy whirl. We all shud really run spark gaps and  go on the offensive..

Frank,

Since the plate is in phase with the cathode, we need the 180 degree phase reversal. So, is it just a matter of having the 3rd winding going in the proper direction around the fil choke for neg feedback? 

ie, Lets say the feedback comes in at zero degrees. So since the plate sampler is a fixed zero degrees, the only way to 180 reverse it is by changing the direction of the turns, correct?

T

[WA1GFZ]

yup just reverse the transformer leads to get 180 degrees. you see how short direct connections are important. Or you could build a cool free running sbe rig.

[Steve - WB3HUZ]

Put a switch on it, and you can also run it as a power oscillator.

[K1JJ]

Now here's a VERY interesting and informative reply from a ham in CA from the AMPS reflector. This has helped a lot.  Check out the schematic he posted on his website for a 4X10 using negative feedback...

Blow it up by saving to your own drive and then magnifying.

http://wc6w.50webs.com/wc6wamps/index.html?fr432.html

Read his comments below.  I like his class A, solid state chip for the low level stuff. Also the feed forward configuration comments.

73,
T

-------------------

Hi Tom,
  Well, you have a certainly started a string here.

  A couple of suggestions for you.

  Use the 4CX350FJ.  May as well start with the lowest distortion.

  Don't try to employ RF feedback across more than two stages.

  Use one of the old TRW - Motorola - M/ACom class A bricks for the first
stage.

  Since you are operating narrowband, consider the feed forward
configuration.  If you start with a -50dB distortion amp, you should be able
to slice another 30dB off that figure for a -80dB system.

  I doubt if your exciter, at any level, is clean enough to drive the above
FF system.

  Fortunately, a clean 80M SSB generator is easy to construct compared to
the balance of this undertaking.

73 & Good afternoon,
  Marv WC6W


------------------------------------
My reply post:

Hi Marv,

Now THAT'S  the kind of info I've been looking for, OM.

OK on using the 'J' tube versions and the TRW  class A chips. I imagine the
J's are harder to find used.

I went to a site that has a CA5106  TRW RF Broadband stick... is this the
right one?

Here's the link - they have lots of interesting low level RF chips:
http://www.sphere.bc.ca/test/rf-semis.html

Your website schematic of the amplifier with negative feedback is quite
helpful. I can see where they neutralize each stage, grid driven, and see
the simple neg feedback from the final to the driver using a 20pf cap from
plate to the driver cathode.  Though, I didn't see any feedback for the pre
driver, the 6146. I guess this amplifier was is just for an example since
the 6146 w/o feedback is not that clean, I understand.  I don't think this
will be too hard to pull off using other tubes and these basic techniques.

I just did some reading about adaptive feed forward systems that you
mentioned can approach -80db 3rd order IMD.  They talk about cancellors. I'm
not sure if I should persue that - might be too complex for the job. Then,
as you said, the FT-1000D even at the low level 10mW tap, would not be a
good driver for this super clean system..  The homebrew ssb generator is
intriguing. I do have experience constructing SS stuff, but wonder how much
of the FT-1000D I would retain as a unit with this outboard stuff.  I think
the 10mW FT-1000D output with simple negative feedback in the linear might
be all I need to grab another -15 to-20db IMD over the 3-500Z amps.  That's
what I'm looking for.

Thanks again, Marv.

73,
Tom, K1JJ

[W3SLK]

Tom,
I like Marv, WC6W. He is a pretty smart cookie and dispenses his advice with out any horseshit. I also have bought a few items from him too. Very easy to deal with.

[Bacon, WA3WDR]

Ulrich Rohde wrote something back in the 70s about feedback applied to receive preamplifiers.  He used transformer coupling to apply a combination of voltage and current feedback to get the desired amount of negative feedback, and the desired input impedance, at the same time.  I think that this could be applied to power amplifiers, even the grounded grid variety.

Here is someting I found that discusses similar ideas, as applied to common-base amplifier designs.
http://www.kongsfjord.no/dl/Amplifiers/Ultralinear%202N5109%20And%202N3053%20Amplifiers.pdf

Also, the characteristic of the slight non-linearity of the amplifier device is known.  If possible, I would try to apply slight additional distortion in the feedback path, possibly by passing it through a similar, but smaller, amplifier tube.  This would introduce more distortion in the feedback, resulting in more correction with less reduction in gain.  This way, it is possible to not only reduce, but actually to cancel the amplifier distortion, and it need not require massive reduction in gain.  Distortion in preceeding stages is an issue, and it too can be cancelled by reverse distortion at some point, possibly at the output stage.

This concept can be applied to a level-comparing feedback scheme as well.  You can take two matched level detectors, and feed one with signal directly from a low-level stage, and feed the other with signal from the output, with additional distortion from one or more small amplifier stages in the feedback path.  Then compare the detector outputs, take the difference between them, and use that to modulate the driver stage somehow.  If the additional distortion in the feedback path matches the distortion characteristic of the amplifier, then the distortion reduction effect of the feedback can actually become distortion cancellation.

In fact, the additional distortion can be applied in the baseband control voltage itself, by using multiple variable diode-break loading of the baseband feedback.  That could be adjusted rather precisely.

[w3jn]

Collins managed to do it over 3 stages IIRC.  Now, if I could find that damn book....

[WA1GFZ]

Bacon,
I've built that norton push pull amplifier with 2N5109s. I run 60 ma of bias with a gain of about 10 dB. My HP8640B does not saturate it

[Bacon, WA3WDR]

That's a lot of headroom for a receiver preamp!  But when there are a lot of signals present, the composite waveform gets really peaky.  A preamp has to handle those peaks, or it's IM and desense city.  Fortunately, there isn't much RMS power unless some of the signals are really strong.

The problem with classic negative feedback is that it can only reduce distortion, it can not eliminate it - and for every dB of distortion reduction, you lose a dB of gain.  This limitation can be overcome by feed-forward pre-distortion (reverse-distorting the source such that the amplifier distortion makes the signal clean again), or by duplicating or even exceeding the amplifier distortion in a stage that is inserted in the feedback loop.  I think the second approach is simpler, because you only need to insert one or more small amplifiers with distortion characteristics similar to the main amplifier distortion characteristic into the feedback path.  It should be possible to pretty much null the main amplifier distortion with minor reduction in gain, or possibly even with no reduction of gain at all if the clean signal is nulled out of the resulting feedback.  But I think the balances will drift over voltage and temperature, etc, so some feedback gain reduction will keep things more stable.

[K1JJ]

Another interesting response below from Marv, WC6W off the Amps Reflector.  He certainly gets right to the meat of the questions.

Bacon: OK on the error correction and feed forward. Marv seems to feel it is do-able too.  If done, not sure I want to go to the trouble of a new ssb generator too, as required...  I think four tubes in a chain with heavy feedback and the 10MW  output from the FT-1000D shud do the job.  I just want another 15-20db better IMD than the usual 3-500Z -35db type amps out there.

T

------------------------------------------

Marv's Response:


Hi Tom,
   That '5106 is probably for Cable TV use and likely cuts off around 40 MHz (on the low end!).

   If I recall correctly, some of the general purpose units that go down to a MHz use numbers from CA28XX group.  They come in both single ended and push-pull versions.
   
   I've seen a few of the 350J's for sale on the web over the years.  I don't believe the demand is very high for those as they have a 26V heater and as such can't be dropped into a 4CX250B socket.  There is also the 4CX600J/JA/JB.

  The 6146 shown on the Hughes schematic is run deep into Class A.  Note the low screen & plate voltages.  I bet the numbers were just fine running all of a watt output.

  For feed forward, an error amp with perhaps another 4CX350FJ would be necessary.  It is not a difficult scheme to implement but, it would double the parts count for the project.  If you read about the technique on the web, keep in mind that "they" are typically working with transistor amps that start with distortion numbers 20dB worse than tubes.  Therefore, in this case, less correction power will be required, the output combiner ratio will optimally be a bit higher, and power lost from the main amp will be lower.

  The subject line said you wanted "the Cleanest Linear".  To avoid
degrading the system, the driver should exhibit distortion specs at least 10dB better than the amplifier.   Your FT-1K will still be useful for the receiver.

73 & Good morning,
  Marv WC6W