Adding RF Negative feedback to a complete exciter / linear system?

[K1JJ]

I wonder if we could get some comments on this idea before I try it...

Background:

I did some accurate measurements today. Using a computer two-tone generator and the SoftRock spectrum analyzer, I found the 3rd order IMD of my FT-1000D was about -32db. That's normal. When I switched on the GG linear amp, I saw almost no change in IMD, even at full power. This tells me the amplifier is probably better than -32db 3rd, which is what I expected.  Now I wish to increase the purity of the complete system.

Does the following make sense or is it a Rube Goldberg?

What if I sampled some RF at the final linear tube and fed it back via coax to the input of the FT-1000D's low level RF amplifier board? There is a good point where the 100mw pre-driver feeds the 200 watt  SS board  where I could tap into. The signal at that point is very clean, like -70db 3rd.

For example, on 3.8 mhz, the coax running out to and back from the amplifier will be about 40' long which is about 90 degree phase lag of the FT-1000D signal.  If I add an adjustable inductance for another 90 degree lag, this will equal about 180 degrees out-of-phase signal back to the input. This should give negative feedback.  Of course this assumes the whole chain ahead is 0 degrees, which is unlikey.  An RF transformer could flip it for an additional 180 degrees if needed, etc.

Most likely I would have to play around to find 180 degrees of feedback on the fly.

To optimize, I could drive the rig with a carrier and tune the inductor until the power dropped down, indicating maximum 180 degree NFB.  The whole amplifier chain has about 6db of excess gain that could be used when the NFB reduces gain. I would use a pot to adjust the amount of NFB.

This would only be used on 75 and 40M so I would need two preset phasing settings and the ability to tweak it on the fly when QSYing...  

Does this idea make sense?   I thought it would be an interesting way to increase the FT-1000D’s IMD figure, since there’s no way to do it any other way. The linear would improve too.

The other alternative is to build a new SDR exciter and SS amplifier with better IMD and also figger a way to add NFB into the GG  linear amp. Maybe that is easier and more stable all around. Just thinking out loud here…

T

[W7TFO]

Hi Tom,

Western Electric, Continental Electronics, and RCA all used global NFB in their 1940's QRO AM broadcast transmitters to get distortion and noise down.  It worked well, but was in use to improve final audio product.  Most of those rigs used low-level grid mod followed by a string of linears.

Keeping RF harmonics down usually was the job of either single ended, Ampliphase, or Doherty linear finals, PI-network output coupling, and finely matched antennas.  Of course, not having to move about the band made the job a lot easier.

Your idea has merit, and may already be the subject of previous design.  If I were to tackle it, trying to filter/limit the FB loop to the harmonic in question might provide a viable route.

At -70dB it might be just an excercise.  In the -30's it would make a difference.

One hams' opinion.

73DG

ps...I never realized the Russkis had a one-man version of Lawrence Welk....too funny.

[W2VW]

Have you looked into adjustment of the final and driver bias?

[K1JJ]

Dave,

Yes, I already optimized the linear final for maximum idle. I tried various settings and left it where the IMD doesn't improve any more with higher idle.  The driver (FT-1000D) is set to idle above specs, but going any higher would cause heat damage according to a guy who works on them and has seen the effects of high idles.. :-)

Dennis,

OK on the past use of NFB.  It has been done with stages in close proximity, but a long cable with phase corrections is the challenge.

It would be cool if someday the SDR rigs had an input that would accept final amplifier RF and then process it for phase and amplitude that gets added into the low level driver. This is akin to the null-steering routine that is now used in the dual diversity receiver software routines.

Imagine average hambone ham with his -28db 3rd IMD  amplifier being able to run a cable back to his SDR exciter and push the mouse around for optimization. He's then at  -45db 3rd. Pretty cool.

T

[W2XR]

Tom,

Collins employed RF NFB in their KWS-1 transmitter. TMC also used it their model PAL-1K and RFC-1 1000 watt linear amplifiers, and their continously-tunable 2 to 32 Mhz SSB transmitters always had exceptional 3rd order IM distortion product performance, even those rigs operating at the 10 KW, 40 KW, and 100 KW PEP output levels. You may wish to take a look at the schematics for these transmitters and get an idea as to how the NFB was implemented. For what it is worth, the IPA and HPA tubes in all of these TMC products were configured in common cathode, and not grounded-grid. Same thing, of course, with the KWS-1.

The implementation of RF NFB in a frequency agile RF amplifier is not easy.

Good luck!

73,

Bruce

[WD5JKO]

The implementation of RF NFB in a frequency agile RF amplifier is not easy.

    Tom,

I think Bruce pretty much summed it up in one statement. NFB is easier with AM since the RF output envelope should match the audio input waveshape. Here we detect the RF out, and use audio NFB instead of RF feedback. With SSB this is not applicable unless we demodulate the RF output with a product detector. To do this you would need to tap the RF carrier frequency from the exciter as an input to the product detector. Once done, use audio NFB back to the balanced modulator stage at the audio input to it.

Just thinking out loud, I'm sure the sharks could pick my idea to pieces. 

Jim
WD5JKO

[K1JJ]

Bruce,

Yes, I have studied some of those grid driven amplifiers. The 4CX-10,000 driven by 4CX-250's? stands out. They fed the NFB back to the lower one watt level.  But again, they had the tubes near each other. 40' of coax is another animal when dealing with RF... :-)  I think a simple L/C network with a 180 degree reversable transformer placed that the input to the 100mw driver might work to tame the RF coming back from the final, no matter what phase it is.

I'm modeling it after the NFB used in conventional AM audio modulator service. I have a 5' cable coming back to the driver for NFB.  

The good thing is this will be somewhat a one frequency  situation, so maybe will work.  I'll give it some more thunking and give it a try.   The worst that can happen is the NFB is at 0 degrees and the amp takes off and smokes right away...

Jim... I think RF NFB is RF NFB no matter what kind of signal is going thru the amp. It wud be pure RF coming off the final plate and pure RF injected (180 degres out) into the low level RF driver. No need to demodulate it.

T

[WA1GFZ]

Tom,
Why not find the stage in the FT1000D producing the IMD first. It could be all the way back in the mixer where feedback will not help. It could be a low level stage that just needs more bias. You might just need a fly swatter rather than a shot gun.

[K1JJ]

Tom,
Why not find the stage in the FT1000D producing the IMD first. It could be all the way back in the mixer where feedback will not help. It could be a low level stage that just needs more bias. You might just need a fly swatter rather than a shot gun.

I'm pretty sure it's the driver or finals in the FT-1000D. That is their design and it gets no better for transistors in these type rigs of the 80's.  The predriver RF into the final board is very clean, at -70 3rd.

I had it in at Yaesu five years ago and axed him to check the finals for IMD. He replaced the drivers and said it was FB.  The specs say about 31 - 34db (depending on power) and I measured -32db.


I did already idle the SS driver and finals higher. To idle the finals much higher makes the white paste dry up and crack according to the tech and wasted the finals quickly.  So, bottom line, is any improvements made on the 1000D wud be tapping off the -70db 100mw port and amplifying up from there with a new system. I tried that with those Erb amps, but got stopped with the quad MRF-150 amps which were dirty - remember?

T

[KA2QFX]

Tom,
I have to agree with Frank. Wrapping some NFB around the exciter's final/driver system would likely get you closer to that -72dB f3. Kenwood employed NFB in the TS-830S with great success. Assuming the amplifier system is as clean as you indicate...

In considering NFB from the amp, 40' away, I'd be very concerned about even minor phase errors being compounded by distance. Even the tuning/loading procedure will exhibit significant phase shift until you get "on".  Either way, I would only consider broadband methods to perform RF sampling and summing since the phase of all the harmonics present also need to be preserved.

But I like the way you think!

[WA1GFZ]

Well you can mount Erb on a big heat sink that can be used later. I never checked erb on the big heat sink but it was pretty easy to get it close to 40 dB. The small heat sink gets real hot fast when you crank the bias.
MRF150s will do -50 dB at 6 amps bias per pair but only 100 watts out. That is 300 watts of heat. I'm using a heatsink with 750 square inches of area on my HPSDR Erb. The whole strip needs to be biased at 9 amps between 3 stages. So hard class A you are talking over 400 watts of heat.

[W2XR]

Bruce,

Yes, I have studied some of those grid driven amplifiers. The 4CX-10,000 driven by 4CX-250's? stands out. They fed the NFB back to the lower one watt level. Jim... I think RF NFB is RF NFB no matter what kind of signal is going thru the amp. It wud be pure RF coming off the final plate and pure RF injected (180 degres out) into the low level RF driver. No need to demodulate it.

Tom,

Actually, the HPA in the TMC GPT-10K, GPT-40K, GPT-100K, and their 200 KW GPT-200K was driven by the extremely linear Penta PL-172 external anode tube. TMC always selected tubes that would  provide the most linear service for SSB operation, with perhaps the exception of the 2x 4-250As as used in the GPT-750D-2. I think they used Machlett tubes in the >1 kw PEP output transmitters, and these too were highly linear amplifying devices. And they biased the tubes so they pulled a significant amount of quiescent current to maximize the linearity, as you are similarly doing in your rig; the PL-172 in my TMC PAL-1K 1000 watt PEP output linear amplifier pulls 200 ma at 2500 VDC with zero signal. That's 500 watts of plate dissipation at idle.

TMC used extremely conservative design principles in their SSB equipment; for example, their model SBE SSB exciter used a single 6146 biased for class A operation, and it delivered a maximum of 3 watts (no, not 30 watts!) PEP output; however, it was typically operated at the 100 mw output level to further maximize linearity. At this level of output, the 3rd order IMD products were a minimum of -45 dBc.

I would tend to agree with Frank; make sure every stage in the exciter is operating as linearly and cleanly as possible; perhaps the source of any perceived distortion is emanating from a very early low-level stage within the exciter.

At it's worst, NFB in any amplifier, RF or AF, is a band-aid for other perhaps less obvious problems. It is best suited to to those amplifiers where linearity is at it's optimum level. And it can sometimes actually increase the presence of higher-order odd-order distortion products.

Just my thoughts........

73,

Bruce

73,

Bruce

[K1JJ]

Tom,
I have to agree with Frank. Wrapping some NFB around the exciter's final/driver system would likely get you closer to that -72dB f3. Kenwood employed NFB in the TS-830S with great success. Assuming the amplifier system is as clean as you indicate...

The problem would be that for every db of NFB, the overall gain is reduced by 1db. So, the FT-1000D is already optimized for gain and would probably need an additional 40db of IPA inserted...  yikes!

I have my currrent exciter/linear system backed off about 6db, so could afford 6db of NFB. But I think youse guys are right. Make the driver as clean as possible and then work on the linear independently.  I still wud like to find an easy way to add some more NFB to a single stage GG linear amp. I must revist that. we talked here once about cathode resistors, grid L/C thingies, etc.  I was never happy with what I read, so never tried it. There is SOME NFB already in a GG linear, but I'd like more.

In the meantime, I'm talking with my SDR mentor, Frank, about getting a pair of Mercury SDR RX's and their Penelope exciter that is very clean.  I might give it another go with the class A SS linear amps over time. The problem is the HPSDR boards are not in stock and have big lead times when they are made. Wait -wait, I guess.

T

[Steve - K4HX]

Just use your FT-102. It beats the FT-1000 IMD spec by 6-8 dB.

[K1JJ]

Just use your FT-102. It beats the FT-1000 IMD spec by 6-8 dB.

Hmmm... think I'll put one on the bench and test it for IMD. Nothing like a 6146/12BY7 with NFB. 

That slow attack ALC will certainly trash up the band, but running it w/o ALC might work.


T

[Steve - K4HX]

No synthesizer noise too. 

[w3jn]

Collins' commericial SW xmitters used NFB also.  Tom, as I recall I sent you some scans from the Collins SSB handbook where they described their NFB scheme.   I have TMC's SSB handbook as well, which details their strategies for reducing distortion.  One is using a 6146 in class A (about a watt output) in their SSB exciter.

I think it's a losing proposition to try and feed back into the pre-driver.  Too much phase shift uncertainty thru the various bandpass filters, the pi-net, etc.  You'll end up making a FB oscillator.

Why not build up a low level exciter with NFB and class A stages, with no phase shifting L/C stages, that'll take NFB input from your existing amp?

L/C oscillator (lo phase noise) --> 6JH8 balanced mudulator --> pair a 6550s in class AB1 push-pull --> yer amp
-                                                                                          ^----------------------------------v

[AB2EZ]

Tom

This sounds to me like it would work, even with the proposed 40 foot cable, provided (as others have pointed out) you don't have a huge percentage of r.f. harmonic content in the feedback signal, and you are aiming for a moderate improvement in the 3rd order IMD.

In doing this, you will need to keep in mind the center frequency and the bandwidth of the modulated rf signal that is entering the amplifier chain.

For example: consider a clean, AM signal, produced by a low power transmitter (or a low level stage); with a 5kHz brick wall filter in the audio chain, and centered at 3885 kHz.

The output of the amplifier chain is going to have 3rd order IMD components over the frequency range 3885 kHz +/- 15 kHz. I.e. the 3rd order products extend around the carrier by +/- 3x the bandwidth of the modulating audio,

As you know, to achieve the full, theoretical benefits of negative feedback, in terms of improved linearity of the amplifier chain, the phase shift around the feedback loop will have to be close to 180 degrees for all frequencies within the above band of frequencies.

If you have a cable whose electrical length is (for example) 0.25 wavelengths, at the 3885 kHz center frequency of operation, and whose corresponding phase shift is, therefore, 90 degrees, then the phase shift at 3900 kHz will be (approximately) 90 degrees x [1 + 15/3885] = 90.35 degrees. This doesn't sound (to me, without simulating it) like it would be much of a problem for obtaining 10 dB of 3rd order IMD improvement. I don't know how much more IMD improvement you would be able to obtain. The phase v. frequency characteristics of the (nominally) 90 degree inductive phase shift network would add to this.

Separately, if you change frequency from 3885 kHz to 3835 kHz, the phase shift, through just the cable, would change from 90 degrees to approximately 88.84 degrees.

Stu


 

[W2XR]

Just use your FT-102. It beats the FT-1000 IMD spec by 6-8 dB.

Hmmm... think I'll put one on the bench and test it for IMD. Nothing like a 6146/12BY7 with NFB.  

That slow attack ALC will certainly trash up the band, but running it w/o ALC might work.


T

Tom,

If 100 mw PEP of drive is not a problem, you might want to try and source a TMC Model SBE SSB exciter. As I mentioned in a previous post, the 3rd order IMD from this unit is on the order of -45dBc at this output power level. It is an exceptionally clean exciter from an IMD standpoint.

I have a spare sitting in the W2XR parts depot. If you would like to borrow it for testing purposes on a long-term basis, you are welcome to it. It does require an outboard source of 2 to 4 mhz of RF at about 2 volts RMS (across 50 ohms) for output over the range of 2 to 32 Mhz; any RF synthesizer or generator can, of course, provide this drive voltage for you.

Let me know.

73,

Bruce

[WA1GFZ]

Or just get an HPSDR penny with -50 dB IMD3

[K1JJ]

Thanks for the offer gear, Bruce.  That wud definately work. Though, I'm commited (yes I should be)  to building up a fully solid state SDR design eventually. The FT-1000D stuff is just a temporary thang until the SDR is on line. I don't want to get off the track too far.

Yep, Frank is right, the HPSDR Penny, or soon to be relased "PennyLane"  will be a single board exciter with 100mw at -50db 3rd. Wow.  Frank has one now and says it works FB. That and a pair of Mercury receiver boards in dual diversity with null steering software is the final goal.  I have decided on that above the Flex 5K. Might as well do it myself and learn something about SDR, with Frank's help.

Stu, the coax phase shift doesn't sound too bad for QSYing and I do have about 6db of gain to sacrifice for NFB. It wud be an easy thang to try, but there is risk, of course.  I'll have to think about it some more. The cool thing is it would improve the FT-1000D AND the linear at the same time.

BTW, anyone want to do some reading on the cool stuff the HPSDR guys are doing:
http://openhpsdr.org/

And here's a nice free two-tone test generator I came across for IMD testing: (try it free)
http://www.nch.com.au/tonegen/index.html?gclid=CIHOraW7sKYCFQ915QodlEJgpg


T

[K1JJ]

BTW, this guy has already built where I'd like to be eventually - the complete HPSDR disco duck transceiver and amplifier system.  Nice.

At 0:41 take a close look at the Mercury/Penny/Ozy stack where the exciter, receiver and interface cards reside.

These modules and their coming updates are at the leading edge of ham design for a decent price. They do group buying so thangs are cheap. Join the forum, read the manuals and check it out.

http://www.youtube.com/watch?v=2WHw4g_if0E&feature=related

[WA1GFZ]

That is my rig but need another Mercury after I recover from a roofing job and new truck transmission.

This stuff is not for the faint of heart and if you don't think you need to worry about ESD be ready to pay for repair work or buy a good microscope.

[K1JJ]

To me, the best part is this 800 member HPSDR group (high performance SDR) is using and developing the already FB Flex PowerSDR software for their boards. Everything is open to development and the sharpest guys take on their own specialties. I plan to join (only $25) though I'll just be a lurker for now. They give member discounts too.  Start small with the basic Mercury RX and infrastructure and go from there. 

The resale value on these boards is usually high. There are rarely any offered used. A few boards are presently out of stock due to high demand, but other than that, if you're into learning and using SDR, what's not to like? 

T

[WA1GFZ]

HPSDR is free but TAPR is $25. You will make your money back when you buy the first module. Members usually get $30 discount per module. TAPR isn't making any money on this deal.
Tom, I just sent you some pictures of my set up. I use amplified computer stereo speakers. I'm not sure how audio is routed in diversity mode