An Interesting Post concerning Cleaner Amplifier Techniques....

[K1JJ]

Taken off the Amps Reflector, here is an interesting post about techniques to building a cleaner linear amplifier. Notice the comments about bypassing certain components for audio. And shorting the tank coil with a bandswitch vs: tapping an open coil.   

Any comments? 

T

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

" Someone asked whether components such as capacitors can cause distortion
 in linear amplifiers.  Somehow the e-mail got side-tracked, but I have some
 comments in response.  The response is not quite on point, but raises some
 interesting points.

 The comments are not mine, but those of W2LPC and WA2PDI, both who worked
 for Amperex at the time they wrote the article.  The article was published
 in 73 Magazine in May of 1963.  The article is entitled "Low Distortion
 All Band Linear Kilowatt" and described about the design of an amplifier using
 a Amperex 8179, which may be similar to the 4-400.   They mentioned all of
 the normal things, such as regulated screen and bias - but let them finish the
 story:

 "When the amplifier was first constructed, the distortion was found to be
 quite bad (about -30 db).  This was caused largely by the fact that the
 current meters and the circuit breaker coils in the negative leads of the
 power supplies were not bypassed for audio frequencies, causing an audio
 voltage to be superimposed on the power suply voltages.  By bypassing the
 meters and circuit breakers coils with electrolyticcapacitors, the
 distortion was decreased considerably.

 The distortion was decreased still further by leaving the unused turns on
 the plate tank coil unshorted when switching from band to band, instead of
 shorting them as was done when the amplifier was first constructed.

 The 8179 data sheet states that the tube must be operated in a vertical
 position only.  During testing, the amplifier was operated on its side and
 it was noted that the distortion figures became slightly degraded,
 probably caused by misalignment of the control and screen grids due to sagging of
 the elements when the tube is operated on its side."

 Later in the article, they summarized design considerations for a low
 distortion amplifier.  In summary, they are:
 
1.  Any impedance in series with the power supplies must be bypassed for
 audio as well as rf.
 2.  In grounded grid operation, second harmonics of the driver must be
suppressed, normally with a tuned input.
 3.  The screen and control grid power supplies should be well regulated.
 4.  The excited distortion should be at least 20 db better than the
 amplifier in order not to increase the amplifier distortion by more than 1
 db.  If extra drive power is available reisistive swamping is advised for
 grounded grid operation.
 5.  The plate tank coil should be progressive opened instead of shorted
 when changing bands.

 I brought up these points a number of years ago and someone told me I was
 nuts for saying that shorting out the turns would affect distortion.
 Well, I am not saying that - two employees who worked for Amperex said it.
 The inference is that they had a spectrum analyzer and hard data to
 support  their statements - but I do not know.  Perhaps they were janitors and just
 made it up.

 I mentioned it, because if Tom is designing the perfect amplifier he at
 least ought to have the input, and it can be tested easily enough."

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

My Response:


Hi Colin,

That is some interesting info!

I was not aware of the need for audio bypassing in the areas you mentioned.

Also, I have always wondered about shorting the final's tank coil for band
changes. I figured it lowered Q and I usually just tapped it as suggested.
But on the higher bands with tapping, there is a Tesla coil effect and the
voltages can arc bandswitches, etc. So I've usually made the amp for 160-40M
and got away with tapping, not shorting.

In addition, I've found that when using a "shorted turns" jumper, it
sometimes got very warm, if not hot on the higher bands. Maybe I needed a
larger conductor.  There appears to be large circulating currents in the
shorted turns none the less.

The driver needs to be clean for sure, though I think someone here said that
the driver must be at least 6db cleaner than the final for a 1db final stage
degradation. They suggested 20db better, which is still easily obtainable in
class A drivers.

Swamping a GG amplifier is interesting too. I once measured the swing on my
GG amplifier input impedance and found it varying about 25% or so.

Anyway, I'll be watching to see comments on your points and maybe someone
will add more to it.

I'm in the metal and mounting construction stage on the new amplifier right
now and can make major changes if need be.

BTW, I was told today that in order to successfully apply 17db of RF
negative feedback around two stages, the phase has to be within ONE degree
of 180 degrees for reasonable stability. Pretty difficult! I'd like to do
20db NFB.  I'd imagine a lot depends upon layout, etc, but it appears that
it's a good thing I am building a 75M monobander rather than QSYing from band to band.... :-)

Thanks for the post, Colin!

73,
Tom, K1JJ
--------------

[w3jn]

WHoever told you that is full of crap, Tom.  It all depends on how much NFB you have in the loop.  The more NFB, the less the margin for phase error.  I found this out firsthand messing with audio amps.

I still gotta copy that Collins book for ya, hopefully it'll put to rest some of these hammy hambone legends.

73 John

[K1JJ]

WHoever told you that is full of crap, Tom.  It all depends on how much NFB you have in the loop.  The more NFB, the less the margin for phase error.  I found this out firsthand messing with audio amps.

73 John

Yes, that's exactly what he meant.... maybe I stated it wrong.

The more NFB, the less phase error you need for the amp to take off.

For RF, in the real world layout loop, phase angle changes more rapidly than audio, and even loading the amp heavier with C2 can swing it 10 degrees. And phase angle can also change through the RF cycle - that sucks.  In contrast, getting 25db NFB in audio service is easier cuz of the smaller phase change in real world circuits..

He felt that 10-13db of RF NFB is more obtainable. I suppose using the 10 ohm unbypassed cathode resistor in the final will help add some NFB that will be 100% stable, but there is a limit to that technique.

I'm still shooting for a total 20db of NFB, and fully expect to break a lot of parts trying ..     Freq QSY is its enemy, thus the monoiband amp.

Bob/ZM also decided to build one up too. We both plan on doing some grid swamping for more stability. He even suggested to eliminate the grid tank circuit for a passive resistive network, thus elimnating one potential phase shift area in the loop. So it wud be just the final tank adding problems. The NFB is JUST from the final plate back to the driver unbypassed cathode.

The amp gets tuned up, THEN the NFB gets connected. That's how many commercial amps do it autotune.

T

[WA1GFZ]

Tom,
Tuning up an amp without nfb would produce a ton of extra gain. Then you connect it and have to do it again. ??

[w3jn]

The COllins book has an example where 10 degrees phase error at 15 dB NFB is acceptable.  The amp will take off at 19.1 dB NFB with no load.  1 degree is WAY too conservative, IMHO.

Remember an audio amp has to deal with several decades of BW while a 75M amp will only have to deal with a few percent - although with an audio amp you're not dealing with pi-networks etc.  Still there are plenty of opportunities for unwanted phase shift to creep into an audio amp, especially outside the 20-20KHz BW. 

I think to get the thing tuned up you'll need to install a couple of phase detectors (looking at input/output of each stage), monitor the phase det output with a 0-center meter, and tune for zero - much like the 30L-1 tuning meter arrangement.  That seems to be the easiest way to me.

73 John

[K1JJ]

Tom,
Tuning up an amp without nfb would produce a ton of extra gain. Then you connect it and have to do it again. ??

Hmmm... 20db extra gain in this case...   But once the two tanks are resonated, aren't they pretty well on the mark?   Would the loading requirements change with the gain change?

Hopefully we can get everything carefully set up once, then the amp will stay around the same freq area.

It was suggested that we make the combination plate /grid tank between the 4CX-250 and 4CX-3000, a pure resistive network. This would make only the final's tank a factor in the NFB phase change. More heat, but maybe it wud work being the 4CX-250 is a class A driver not needing a tank flywheel effect?

T

[WA1GFZ]

Tom,
I agree with John and I think that amp you posted a while ago had the phase detector.
I want to see anything you do with a 4cx3000 since I have a stash of them. fc

[K1JJ]

The COllins book has an example where 10 degrees phase error at 15 dB NFB is acceptable.  The amp will take off at 19.1 dB NFB with no load.  1 degree is WAY too conservative, IMHO.

Remember an audio amp has to deal with several decades of BW while a 75M amp will only have to deal with a few percent - although with an audio amp you're not dealing with pi-networks etc.  Still there are plenty of opportunities for unwanted phase shift to creep into an audio amp, especially outside the 20-20KHz BW. 

I think to get the thing tuned up you'll need to install a couple of phase detectors (looking at input/output of each stage), monitor the phase det output with a 0-center meter, and tune for zero - much like the 30L-1 tuning meter arrangement.  That seems to be the easiest way to me.
73 John

I'm glad to hear about the 10 degrees vs: 15db NFB, John!  Tnx , OM.

Yes, the Hughs amp circuit I have as a sample, uses a torroidal transformer at each grid return to sample phase. That may be a great idea, OM. In fact they auto tune it with the NFB disconnected and when the meters hit zero, they connect back the NFB.

I was planning on eliminating those detectors, but now that you mention it, I will give it some more thought and talk it over with ZM...  It wud be better than flying blind and blowing up parts. My last attempt with an 813 class A driving a tetrode ruined lots of parts. I had to PEEL the melted layers of bunched wire off the plate choke form several times until I finally gave up...

T

[w3jn]

Tom, the phase dictectors are simple.  Capacitive voltage dividers, a couple of 1N34s, and a couple of pots to adjust the balance, and finally -50-0-50 uA meter.

There is a lot in the Collins book on how grid current and screen current come into play in distortion reduction and phase shift.  I agree with Frank, I think you're gonna hafta tune up with the NFB network in place or you're just gonna hafta tune it up twice.

73 John

[K1JJ]

Tom, the phase dictectors are simple.  Capacitive voltage dividers, a couple of 1N34s, and a couple of pots to adjust the balance, and finally -50-0-50 uA meter.

There is a lot in the Collins book on how grid current and screen current come into play in distortion reduction and phase shift.  I agree with Frank, I think you're gonna hafta tune up with the NFB network in place or you're just gonna hafta tune it up twice.

73 John

John,

Here's the model we are using. It's the auto tuned Hughs 10KW amplifier, but instead we are substituting a 4CX3000.

Notice the diode detectors on each stage. I see they sample detected RF, but do you see the phase sample derived too in those circuits?  I believe the "Tune Signal" is the correct output we are looking for to provide phase info.

Copy and paste into the browser - the link does not work as a hot link:

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

[WA1GFZ]

Tom there are a number of phase detectors and power detectors. Study each one and look at the signal going into each side. The phase detectors are set up to null when things are properly tuned . Look at the output stage where there are 2 15 pf caps You can see how the circuit output is 0 when the tuning is right. I suppose the control circuit tries to null phase and tune to max power at the same time. the null helps the control to know which way to turn the inductors. one side generates a positive voltage and the other a negative so at perfect phase the output voltage is 0.

[w3jn]

Those phase detectors are a bit more complicated than the Collins ones, IIRC.  If you look at the pic I posted in the other thread I think you can see how Collins did the phase detectors.

In any event, the RF voltages sampled on the input and the output will be radically different and the phase detector most likely won't give you the result you're looking for.  So you hafta hook it up, check the phase, and have a pot to adjust exactly where the meter will zero AT THE EXACT POINT the phase relationships are ideal.  Sometimes for ideal linearity you don't want a zero phase relationship between input and output = the 30L-1 is one example of this.  You just adjust the loading for zero on the meter, because Collins figured out what the ideal loading will be considering the relationship between input and output.

Y' know, T, I'd maybe try this all out with a bit lower power setup with cheap toobs before you set your mind to building something with pricey toobs.  Maybe try something like a 6CL6-->6146-->813 and mess around with the FB loop as well as some phase detectors to get a grasp of what's going on here before you put the pedal to the metal with the big stuff.

73 John

[Bacon, WA3WDR]

Interesting that the resistance and inductance of metering could cause distortion, and likewise interesting about shorted turns in the tank coil... I would think that the current in the shorted turns would cause voltage drop across slightly corroded contacts and cause distortion.  Of course, I believe that a certain amount of impedance in series with a screen grid will reduce distortion in many tetrodes, but the wrong impedance in the wrong place will have the wrong results.

I think that there would be a real danger of damage if large RF negative feedback was used, and then the antenna shorted, opened, etc.  All phase and gain bets would be off.  There needs to be some protection system that looks at phase angles, load impedance, etc, and shuts the system down fast if there is a fault.

This, and the low efficiency of heavier amplifier bias and class A driver stages, makes me think that a subtle reverse pre-distortion system is a better way to go.  I imagine a more or less normal tune-up, and maybe a computer monitoring the low level signal and the output, and dynamically tweaking the anti-distortion system as the operating session progresses.

[K1JJ]

Here's some comments on the subject from the Amps Reflector.  Check out the stuff about IMD and sweeping the two tone test... Good stuff.

T

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



>"When the amplifier was first constructed, the distortion was found to be
>quite bad (about -30 db).  This was caused largely by the fact that the
>current meters and the circuit breaker coils in the negative leads of the
>power supplies were not bypassed for audio frequencies, causing an audio
>voltage to be superimposed on the power suply voltages.  By bypassing the
>meters and circuit breakers coils with electrolyticcapacitors, the
>distortion was decreased considerably.
> 
>


Right. All the voltage need to have low impedance across the audio
spectrum. The classic test for uses uses the two tone intermodulation test method and
sweeps the spacing of the two tones from about 300 Hz to 3 KHz.
If the level of IMD varies with tone spacing, then one or more of the
voltages needs better filtering.

In solid state amplifiers, thermal effects can through off the tuning
because the surface of the die can have a very fast thermal time constant. The
transistors actually change temperature at the beat frequency between the tones. In the
literature, this is call "memory effect".

>The distortion was decreased still further by leaving the unused turns on
>the plate tank coil unshorted when switching from band to band, instead of
>shorting them as was done when the amplifier was first constructed.
> 
>
Interesting... It's not discussed very much, but IMD products come from
multiple sources. People describe 3rd order IMD usually as 2*F1-F2. But
the amp will generate the same spur frequency from 3*F1-2*F2, 4*F1-3*F2,
etc. Each of these components have their own phase and can either add or cancel
with each other. It's possible for a two tone test to show very
different IMD results from slight variations (0.1 dB) in the power to one tone. The
FCC has changed the test method for PA's to require modulated signals
specifically to prevent manufacturers from hunting around for a sweet spot in IMD.

It's possible that the change in results from a short vs open turns
comes from the different impedance the Pi-Net presents to the tube has the higher
harmonics.

>The 8179 data sheet states that the tube must be operated in a vertical
>position only.  During testing, the amplifier was operated on its side and
>it was noted that the distortion figures became slightly degraded, probably
>caused by misalignment of the control and screen grids due to sagging of the
>elements when the tube is operated on its side."
>
>Later in the article, they summarized design considerations for a low
>distortion amplifier.  In summary, they are:
>1.  Any impedance in series with the power supplies must be bypassed for
>audio as well as rf.
> 
>

yup

>2.  In grounded grid operation, second harmonics of the driver must be
>suppressed, normally with a tuned input.
> 
>

yup

>3.  The screen and control grid power supplies should be well regulated.
> 
>

yup

>4.  The excited distortion should be at least 20 db better than the
>amplifier in order not to increase the amplifier distortion by more than 1
>db. 
>

I'd say at least 10 dB better for less than 1 dB effect


>If extra drive power is available reisistive swamping is advised for
>grounded grid operation.
>5.  The plate tank coil should be progressive opened instead of shorted when
>changing bands.
> 
>
Probably depends on the details of the frequency response of the network at
harmonics...and probably not a general rule to follow.

>I brought up these points a number of years ago and someone told me I was
>nuts for saying that shorting out the turns would affect distortion.  Well,
>I am not saying that - two employees who worked for Amperex said it.
> 
>
Possible...but not easily controlled at design time.

>The inference is that they had a spectrum analyzer and hard data to support
>their statements - but I do not know.  Perhaps they were janitors and just
>made it up.
>
>I mentioned it, because if Tom is designing the perfect amplifier he at
>least ought to have the input, and it can be tested easily enough.
>
>73,  Colin  K7FM

[WA1GFZ]

Tom,
I use a little swamping in my 4cx3000A rig. The reason for this is simple. The fil choke
can also have wild swings of reactance just like the plate choke. I've seen that make a stable amp take off. fc

[K1JJ]

Frank,

Yep. Put an swr meter in line from the exciter to the linear's input. Go from min to full drive.  I have seen the swr swing from 1:1 to 1.5:1 on the big amp.

Yes, swamping is good if ya have the spare exciter drive to use.  But most guys are maxed out at 100w driving the standard hambone GG linear.

T

[WA1GFZ]

I think I did 100 ohms in series with 100 pf dork knob

[K1JJ]

Frank:

Remember the Dead Poets Society movie?

Can you hear the whispers now ?

"Sayyaas  th   Deyyy......   Seezzz  the duyyy.....   SEIZE THE DAY......"


There's another one coming thru...:   "  Qaro  peedm  Madltor - far See X theethosand...
QRO  PDM Modulator  - 4CX-3000 "...


Jacob Marley is telling you to repent. Fegit the SDR receivers for a few days and get the big rigs finished so we can hear ya!   


T

[WA1GFZ]

Tom,
I got the solid state rig back on 160 and went through a RX jones for a couple weeks. I know a guy who bought a 6830 from me who is going to do the drafting for the SDR interface so can't pass up a good deal.