Gentlemen,
The real issue here in respect of this excellent discussion about diplexer
termination of harmonic energy is the profound impact on IMD and linearity.
These MOSFET and LDMOS RF power devices have non-linear voltage dependent
parametric capacitances that engender intermodulation distortion generation
and reflected harmonic energy control plays a key role; efficiency is just
one parameter consideration in the overall design objective that must be
judiciously balanced.
Leigh
VK5KLT
-----Original Message-----
From: Amps [mailto:amps-bounces@contesting.com] On Behalf Of Tom Georgens
Sent: Sunday, 11 December 2016 11:08 AM
To: 'Manfred Mornhinweg'; amps@contesting.com
Subject: Re: [Amps] SS amps and auto-tune. Is it even necessary?
Manfred
Thank you for the thoughtful responses. They triggered some ideas for my
next version and very valuable.
I will defer the questions on feedback and IMD improvement to someone with
more design experience. I am hardly disagreeing, and would surely love to
some of these techniques actually implemented and deployed.
The diplexer argument is an interesting one. The position that the third
harmonic energy is so high is aided by the fact that it is seeing a good
impedance match at this frequency is an interesting one. If it saw a big
mismatch, typical of a LPF, then maybe less power would be transferred at
those frequencies. That may be true, but it also might not. In fact it
could be worse. The output transformer, LPF and antenna would have to be
well modeled to understand this at all frequencies. That said, it is very
valid point to consider.
However, my design point is contesting so the tradeoff of reliability over
efficiency is one that I am willing to make. The diplexer makes the
performance much more predictable.
I agree with the need to monitor between the amp modules and the combiner
for a bunch of reasons. These are key measurements for the protection
circuitry. The Diplexer helps a lot here as well. Having the harmonics
reflected back from the LPF instead of sunk, creates a lot of variability in
the reflected power measurement by band. It is hard to tell the difference
between a mismatch on the primary or harmonic frequency. With the diplexer,
I see almost no reflected power from the combiner/filter so it is easy to
detect small excursions from the norm as a sign something is wrong (or
getting hot). As an aside, I spent hours with a commercial product dealing
with the fact that it was reporting a reflected power fault even though the
filters measured a perfect 1:1 SWR with my VNA. It surely was the
reflected power of the harmonics but it took me a long time to figure it
out. The manufacturer was no help. I eventually made a design change to
the monitoring circuit, which had already been changed from the published
schematic.
Overall, the points on the diplexer are well taken. However, it allows an
integrator like me to control a bunch of variables and I still recommend it.
A different design point could certainly warrant the opposite approach
The area were we will have to differ is on the relative reliability of tubes
versus these LDMOS parts. While nothing is indestructible under misuse, The
LDMOS are substantially more fragile than the tubes used in legal limit
amps. Risk can be reduced by wrapping a lot of protection circuitry around
the transistor pair, but I have yet to see this be sufficiently robust or
fast enough to be as reliable as a tube under fault conditions. It may be
theoretically possible to close the gap with better design, but none of the
products on the market today have demonstrably proved capable of reaching
this level. I remain unconvinced on this point, but would welcome being
proven wrong with actual user data.
My summary, based on observation alone, is that if the LDMOS part is kept
cool, sees a stable well matched load, and is not overdriven, it can work
reliably forever. However, excursions from these controlled conditions put
you on a much faster slope to failure than a tube amp.
One last anecdote. Every contest I have to put up some temporary antennas
for the low bands. I always power test them with the tube amps before using
the antennas with the SS amps.
As I said, I am not a designer like Manfred. I defer to him on design
possibilities and value his suggestions. My post was a summary of my
observations and possible tradeoffs for someone trying to integrate parts
that are generally available to hams.
Tom
-----Original Message-----
From: Amps [mailto:amps-bounces@contesting.com] On Behalf Of Manfred
Mornhinweg
Sent: Friday, December 09, 2016 8:26 AM
To: amps@contesting.com
Subject: Re: [Amps] SS amps and auto-tune. Is it even necessary?
Roger, Tom, and all,
I can't help giving my 5 cents (or whatever) on this...
> With first generation SS amps, auto tune was almost a necessity, BUT
> with the new amps?
It totally depends. If you only operate in the band segments where your
antennas have a low SWR, say, under 1.5:1, then of course you don't need an
autotuner. If instead you want to operate with antennas having higher SWR,
_and_ you want your amp to actually work as it should, then you need some
sort of tuner. It can be automatic or manual, internal or external, located
in the shack or built into the antenna - but you do need one.
If the amp has good, effective protection, then a tuner is _not_ required
for safety. The protection handles that. At high SWR you would get lower
output power, or even the ampo might shut down, but there should be no
smoke. But the news for some people are that a tuner _does not_ improve
safety! Even with a tuner, the amp still needs the protective circuitry!
So a tuner is not a way to avoid damage to the amplifier, but a way to allow
operation at full power and normal IMD and efficiency into antennas having
high SWR.
> With the current flock of transistors that can handle the wrong
> antenna or even an open circuit, auto tune is just a convenient, but
> expensive accessory.
Not so. Please, everybody, _DO NOT BELIEVE THE FALSE CLAIMS ABOUT
INDESTRUCTIBLE LDMOSFETS!_ There has been much advertising lately about
LDMOSFETs that can survive 65:1 SWR or even more, and that cannot be
destroyed. This is false advertising! There are videos circulating that
fool people into believing things the video does not actually show. The
reality about these "extra rugged" LDMOSFETs is actually quite simple:
They can handle such a degree of avalanching that they are SWR-proof _as
long as not overheated_. Only if the input power to an amplifier is strictly
limited (in the power supply) to a level that the device/heatsink
combination can dissipate continuously, then the device is indeed pretty
much indestructible by load mismatch. But in a ham amplifier this is very
rarely the case! The amplifiers are about 50 to 65% efficient, which means
that the dissipation is about as much, or less than the ouput power; and
then the designers usually employ the smallest, lightest, least expensive
way to provide enough dissipation capability for this power level.
For example, a 60% efficient, 1500W output amp would have a maximum normal
dissipation of 1000W, and the designer would use two LDMOSFETs mounted on a
copper heat spreader and a heatsink with fans, that keeps the junctions at a
safe but high temperature under these conditions. The power supply needs to
deliver 2500W. If the amplifier is running like this, with the junctions
nearly borderline hot, then there is almost no additional dissipation
capability left. And this means that if now a mismatch happens, for example
because a bird sat on the antenna tip, the dissipation increases, possibly
all the way to 2500W, the junctions overheat, and the "indestructible"
LDMOSFETs blow up. They take a time to blow up - this time is given by the
amount of additional dissipation, and the difference between the normal
operating temperature and the absolute maximum one. The time might vary from
minutes to milliseconds.
So, even "indestructible" LDMOSFETs do need protective circuitry. The good
news is that a proper combination of their ruggedness with the protections
makes the whole amp pretty much indestructible. But it still cannot work at
full power into a high SWR.
> 50% Vs 60%. Who cares?
I do care! Better efficiency means lower power consumption, the ability to
run from a smaller genset in portable use, a smaller power supply, smaller
heatsink, so a much smaller and lighter amp, and less heat and noise in the
shack. It also makes thermal handling much easier. It's also better for the
planet and for the future of humanity, although I admit that the effect of
us hams on total worldwide energy use isn't such a big factor... :-)
Anyway, higher efficiency has such big advantages, that I made it the first
and foremost goal in my own amplifier development. If we can drive the
efficiency close to 100%, then we can basically do away with heatsinks and
noisy fans, and we can use power supplies of half the size and cost - all
while improving reliability!
> They are easily water cooled without the headaches of water cooling a
> tube amp..
Yes. And indeed we should use more water cooling in high power amps, while
we don't yet have high efficiency ones.
I agree with a lot of Tom's comments, but not all:
> 1. The LDMOS parts are a huge step forward. The prior parts,
> typically needing matched pairs or quads were far more fragile and,
> failing one is failing two (or four). This was expensive, frustrating
> and frequent. As improved as the LDMOS part are, they are far from
> indestructible and no match for tube reliability. The
> 65:1 SWR spec does not apply to CW where far more modest mismatches
> will destroy the parts nearly instantly. I have had one of the 1.25KW
> parts fail on me.
Exactly. In CW you typically have the junctions running at a temperature
that doesn't allow much additional power dissipation.
> 2. A key to reliability is headroom. The closer you are to their
> rated limits the more delicate they are.
Exactly, again.
> I am running two 1.25KW parts at 1KW (total) out. I believe it has
> helped reliability, but efficiency goes way down when you run that far
> below max output. That means more heat
No, efficiency shouldn't be lower. You probably didn't run them at the
optimal load impedance, in that circuit. If you run a stage designed for 1kW
peak output at 500W, of course the efficiency suffers a lot. But if you
re-design the impedance matching for the optimum impedance resulting at 500W
load, then the efficiency will be back up - and even a tiny bit higher than
at 1kW, due to a lower drain saturation voltage!
> 3. The good news is that you do not need a matching network. The bad
> news is you do not have a matching network. Output varies quite a bit
> as the load impedance changes and you can't do anything about it.
> My solution is to flatten the SWR of all my antennas so I can switch
> between them without seeing big excursions in the output power. The
> worst my amps will see is 1.5:1 and most of the operating occurs below
> 1.3:1. Manufacturers will disagree, but I think if your SWR exceeds
> 1.5:1, you should use a tuner.
I fully agree.
> 4. The filter has been a bigger challenge than the RF module. I have
> had multiple cap failures early on. Some have been due to other
> design issues, RF in the selection logic etc.. I have not had a cap
> failure this year, but I really do not know how much headroom I have
> since the specs for RF current are sparse.
Yes, this is a problem. Many hams fail to understand it. Just a few days ago
I saw a web site of a ham who built an amp with low pass filters, and he
wrote that he "used 3kV rated capacitors because there is such a high
current there". He got it wrong! The point is that a capacitor rated for a
higher voltage _does not_ necessarily have a higher current rating too! Very
often higher voltage ratings go along with _lower_ current ratings!
For low pass filters, and many other RF power applications, designers need
to equally watch both the voltage and current ratings of the capacitors
used. A 1kV rating should give plenty of headroom in a legal limit low pass
filter used at reasonable SWR, but the capacitors will need to have current
ratings of 5 to 10 amperes, perhaps even a little more in some cases. And
there are very few capacitors available that meet this requirement. And they
tend to be expensive. Which makes homebrew metal clad mica or metal clad
teflon capacitors an attractive possibility. They are physically larger, but
are easy to make for those ratings.
Instead of a bank of low pass filters, possibly followed by an antenna
tuner, I see no reason why one shouldn't use a higher Q, band-switched tank
circuit, much like the one in a tube amp, proviidng both the filtering and
matching functions. Just like in a tube amp, this could be a manually or
automatically tuned one.
Due to impedance reasons, the tank needs to come after the broadband
transformers, so it cannot be directly connected to the LDMOSFETs. At least,
not while they run from only 50V. When we get practical, usable 300V
devices, we will be able to do away with the broadband transformers, as the
drain impedance will then be in a range where band switching is practical.
> Also, I concur with the comment that a diplexer is a must. The third
> harmonic is only about 10-11 dB down. It is good to know where that
> power is going.
Tom, I would like to dig deeper on this point. I do not believe that a
diplexer with a dummy load for harmonics is a good idea. So it might be
productive to discuss this.
My point is that a transistor does not "produce" a specific ratio of
fundamental and harmonic power. What it does, simply, is varying its own
conductivity according to the drive signal, following certain nonlinear
rules. How much power is developed in the process depends a huge lot on the
load impedance seen by the transistor. If it's loaded by a diplexer, the
load impedance will look mainly resistive at both the fundamental and the
harmonic frequencies, and a lot of power will be developed on the harmonics,
and dissipated in the diplexer's dummy load. That power is lost, and
represents a reduction in efficiency. If instead a plain simple low pass
filter is used, the load seen by the transistor will be mainly resistive on
the fundamental frequency only, but will be mostly reactive on the
harmonics, resulting in almost no power generated and dissipated on the
harmonic frequencies, and thus a better efficiency.
A low pass filter without a diplexer needs to be of a type that doesn't
cause the amplifier to develop excessive peak voltages or currents. An amp
with little or no negative feedback behaves largely as a current source, and
thus requires a capacitor-input low pass filter. Instead an amp with heavy
feedback works more like a voltage source, and thus requires a filter with
inductive input. And if the amp has a medium level of feedback, performing
like a "soft" current or voltage source, it can be hard to obtain good
operation with a plain lowpass filter of any kind! That's why some designers
have resorted to diplexers - but it's not a good reason!
> 5. The splitter/combiners have been simple and reliable.
Yes. And in some cases they can even be skipped. It is possible to design
the output transformers for 25 ohm, and place the secondaries in series. The
main disadvantage is that emergency operation with one half of the amp
burned out is not possible. And there is a larger chance for phasing issues,
unless all wires are very short.
> 6. IMD is not good. There may be clever feedback schemes to help
> with this, but it is over my head.
LDMOSFETs, just like other MOSFETs, bipolar transistors, and triodes, are
very non-linear. Tetrodes are better, but need a huge idling power for it,
resulting in ultra-poor average efficiency. So, indeed it's best to
linearize amplifiers by using negative feedback. And the very high gain of
LDMOSFETs makes them highly suitable for simple, effective linearization by
feedback.
That said, I find it highly desirable to operate amplifiers well into
saturation, for the sake of efficiency. This requires additional negative
feedback, to achieve decent linearity. Basically the feedback circuit has to
stabilize the gain at all drive levels to the exact amount the amplifier has
when saturated to the level decided by the designer.
Such a feedback circuit can no longer operate directly at RF, from drain to
gate. Instead it needs an external feedback loop: An envelope detector at
the output, another at the drive signal, an error amplifier comparing the
outputs of both detectors and amplifying the result, and applying it to some
point that controls the gain. This could be the bias of the LDMOSFET, or it
could be a PIN diode attenuator in the drive circuit, among other
possibilities.
The result should be very good IMD, along with a full-power efficiency
around 80%, and a corresponding improvement in average efficiency. The
circuit looks moderately complex on paper, but uses only cheap, small parts.
> My RF modules have been purchased as pallets. I do not think these
> parts are optimized for linearity and they are nowhere near as good as
> tubes.
Most pallets are indeed for broadcast or industrial uses, that don't require
linearity, and not for SSB transmission. So many of them have modest to poor
linearity. But surely not all. And of course, they cannot be driven into
saturation and still provide any acceptable linearity!
> I really am hoping that some clever "predistortion" algorithms will
> find their way into the SDR radios to offset some of the deficiencies
> of these parts
This is slowly but certainly finding its way into ham radio. It's an
alternative to the external feedback proposed above, and has the advantage
that it can also predistort and thus correct the phase distortion introduced
by amplifiers. At least in principle, it can...
But it's a technique mostly suited to amplifier stages integrated into the
radios, and not to add-on amplifiers.
> 7. Heat removal is manageable for CW and SSB. I have not tried RTTY,
> but there is a lot of heat in a very small area so it needs serious
> heat conduction and airflow. If I were to build one for home use, I
> would try water cooling. Some of the pallets now come mounted on
> heat spreaders milled for water flow.
Yes. And the thermal bottleneck is usually the parts closest to the silicon.
That's the metal flange of the device, its interface to the heat spreader or
watercooled block, and its metal immediately under and around the device. A
large block of copper with water flowing through channels on its other side
is often far from optimal. It's better to find one milled such that the
water flows through narrow channels very close to the device. Like directly
underneath the device mounting area, with the block milled such that it has
a star-like structure of metal fins protruding from the mounting surface
into the water, combining the most metal-water interface area with the
shortest and widest possible thermal path through metal. And the water
should flow fast through that part, causing high turbulence, which reduces
the thermal resistance of the metal-water interface. The fast flow is
achieved by having narrow water passages, and a sufficiently powerful
circulation pump.
> 8. While energy is still energy, it is a lot easier to work with the
> cover off when there is no 3500V exposed. That said, it is still
> dangerous if you are not careful
Working under the hood of a running engine is far more dangerous, and many
rather modestly educated mechanics have survived that for many years. No big
deal. One just has to use reasonable caution.
> This is a must. IF the amp does not have automatic band selection,
> you will blow it up some day, protection circuitry notwithstanding.
If the protection circuit is good enough, there should be no way to blow it
up. But that means some sensing _before_ the low pass filter! Many amps
don't have it.
> 10. No doubt solid state is the future and, in a very controlled
> situation, can be made to work well under demanding circumstances.
> However, will they be as reliable and as linear as tubes for a wide
> range of load mismatches, high duty cycles and operator errors? Not
> today
I think they can well be MORE reliable than tubes. I have seen more than
enough tubes destroyed because of operator error. Of course that's mostly
because many tube amps don't have much protection circuitry. The basic point
is: With correctly implemented protection, an amp should be impossible to
blow up. But no amplifying device - neither solid state nor hollow state -
is indestructible on its own. Tubes tend to be far more forgiving than
transistors, in terms of mismatch and the resulting overheating, but
transistors are enormously more forgiving in terms of rough handling, like
impact, shock, vibration, etc. Shipping a tube amp can be a problem.
Shipping a solid state amp is not.
> Like I said, I am speaking as an integrator, not as a designer. Many
> of the shortcomings can be improved with better designs, but the
> underlying technology still has its limits.
I see it from the designer's point of view. It is possible to embed an
LDMOSFET in an amp in such a way that it is protected from all sorts of
mishandling the designer thought about. And it isn't even expensive to do
that. But designers are fallible humans, and often overlook some combination
of conditions that can result in blowing up an amp despite the protection
circuitry. So it comes down to quality in design and construction, and lots
of testing under all imaginable conditions.
Engineering is a combination of brains and material. The more you use of
one, the less you need of the other. In low-volume products like ham
amplifiers, companies tend to invest rather modest amounts of brains, so
there are many components for a modest result. If ham consumers want bullet
proof, highly linear amplifiers, they have to pay either for huge amounts of
expensive parts in an amplifier that follows the brute force principle, or
they have to pay for lots of engineering hours that get diluted among only
relatively few amplifiers sold. That explains the high price of today's
solid state amps.
If there were a hundred million demanding hams in the world wanting to buy a
good amplifier, soon there would be many companies offering excellent
products at ever lower, competing prices. But given our real numbers, and
the lack of interest in high IMD performance and efficiency
that most hams show, there are only a few companies serving this market,
and they can't invest enormous engineering resources in it. It simply
doesn't pay.
A technically minded ham who is willing and able to spend lots of hours
designing and optimizing a good amp, can end up with an efficient, highly
linear amp that uses relatively few parts, and avoids high cost ones. But
there are very few such hams remaining. And that's the core problem.
Phew, this post got longer than intended. Sorry for boring you all!
Manfred
========================
Visit my hobby homepage!
http://ludens.cl
========================
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|