Hi all,
nice to see so much activity in this matter! Here are my condensed
replies to some posts by several of you. I'm merging them all into one
post, instead of posting many separate replies.
Jim,
> ## why not use 4 x BLF188XR or 4 x MRFX1K80H ?
Cost is a powerful reason against that, in my opinion. It would be
expensive overkill. I do think that using 4 devices (or more) is a good
idea, but then these should be cheaper devices.
A few days ago I was looking at the IXZ210N50L2, which should be
suitable for a multi-device legal limit linear class AB amp. Each of
these individual (not Gemini) MOSFETs should be good for close to 200W
RF output in linear service, and they are 28 bucks apiece at Digikey, in
single quantity. An added benefit is that they run from 100 to 150V
instead of 50V, which should allow achieving better efficiency and
linearity.
> Is it even possible to use 2 x devices in parallel in each half of a
> push-pull amplifier ?
At 50V supply, the drain impedance for a single 1500W amp is so low that
it's really hard to handle properly. It's much better to make separate
lower power amp modules, and combine them.
> If it is possible, then the heat could be extracted over 4 x
> devices.
Yes. Or 8 or whatever. It makes for a much simpler, smaller, lighter
cooling solution.
> The TX imd, with 4 x devices run at 1.5 kw, on paper, should be good,
> since each device is running at 375 w pep output.
Don't let this fool you! If you put each of the 4 devices in the typical
circuit designed to saturate at 750W (1:9 transformer at the output),
then indeed you will get good IMD at 375W, but also you will get an
absolutely lousy efficiency. And if you instead use a lower ratio
transformer, to make each module saturate at 400W or so, then you are
back at square zero, with "normal" efficiency and IMD! The only thing
you gain that way is the easier heatsinking, and/or robustness against
overheating damage.
> Toss in pre-distortion, and IMD could be reduced further.
Yes. If it works. Yesterday I was experimenting with predistortion using
PowerSDR, the Red Pitaya board, and several homebrew amps, and I'm
getting erratic results. Typically the predistortion system implemented
in PowerSDR, in my setup will correct an already good amplifier to an
excellent quality (like IMD3 down at -80dB!), but will simply not work
with a dirty amplifier that starts with IMD3 at -20dB. I suppose that
this is due to some software-set limit, and surely can be improved.
> ## As is, I believe the SS amp manufacturers will be using
> pre-distortion techniques to try and get acceptable IMD out of fewer
> devices.
And for more efficient amps!
> Meanwhile, the pair of devices will still be running too hot.
My approach is that one should improve the efficiency of an amp, at the
cost of linearity, then use predistortion to fix the nonlinearity. But
so far my tests don't achieve this. I'm just starting to play with this,
though.
> ## A 4 x device SS amp, if feasible, IMO, would not increase the
> total cost of the amp by very much...but the benefits would be huge.
If you are talking about 7000 dollar amps, sure, the cost increase would
be modest, specially if less money gets spent on heatsinking, after
adding more MOSFETs. But in absolute terms, 1000 bucks in MOSFETs alone
seems too much to me.
Leigh,
I largely agree with your post, but I just can't like this approach:
> Solid-state QRO HF linear amps can be made to work very nicely
> particularly when conservatively driven and backed-off paralleled
> modules operating well clear of their 1 dB power compression points
> are combined and harmonic terminating diplexer LPF output filters are
> used to yield stable performance and good IMD specs. And this is
> achievable without resorting to esoteric pre-distortion and linearity
> correction techniques.
While that's all true, it should also be mentioned that this approach
produces an amplifier that is maybe 30% efficient at peak power, and
much less over the SSB cycle. I find this just too poor to be
acceptable. Along with the low efficiency comes large size, weight, cost
and power consumption.
Jim,
> ## How far apart can each device in a paralleled pair be....and
> still function in the RF environment ? IE: IF they have to be
> packed tight together, heat will still be an issue..if the pair of
> devices are just inches apart.
They need to be very close together, if you want to feed them all into a
single output transformer. So this approach is not as good as separate
amp modules and combiners, from a thermal point of view.
I have been toying with the idea of a distributed amplifier that has a
long string of MOSFET push pull pairs, each pair or small group of pairs
feeding a local output transformer, and being driven from a local input
transformer. That is, just the cores and MOSFET-facing windings would be
local. The input primary and output secondary windings would be one
single wire each, threading once through each core. That might allow
enough physical device separation to achieve thermal advantages, while
still not running into excessive phasing problems. I need to give this
idea a more detailed look.
> ## I forget what the typ push pull SS amp does for harmonics. I
> think it was high for the even harmonics, like 24- 40 db..and very
> poor for the odd harmonics..like 12 db.
A perfect linear push-pull amplifier theoretically generates no
harmonics at all, because the two currents from the two devices are
combined just right to produce a clean sine wave. If linearity is poor
(and it always is, to some extent), the even harmonics will largely
cancel, while the odd ones will not. How much harmonic distortion you
get, depends on how bad the amplifier is, or how much you overdrive it.
Getting the 3rd harmonic at -12dB is really gross. You would have to
overdrive an amp so much that you get almost a square wave, to get that
much 3rd harmonic!
> Without the use of a diplexer, that is one helluva lot of power being
> reflected back by the LP filter... reflected right back into the
> final pa section...further cooking the final pa.. degrading imd, etc,
> etc.
No, that's not the right way to see it. The amp is not "producing" a
certain amount of harmonic power. What it does is try to pull a certain
current waveform, or if it has strong negative feedback, it will try to
produce a certain voltage waveform. Those waveforms might have a large
harmonic content, if the amp is very nonlinear (overdriven). But power
appears only if current and voltage appear together. If you slap a plain
simple capacitor-input low pass filter on the drain of a MOSFET, you
will get the distorted current, but a pretty clean drain voltage. That
means that there is almost no harmonic voltage, and thus almost no
harmonic power, even if there is a strong harmonic current. So, harmonic
power is not produced and reflected back - it's simply not produced!
If it wasn't like that, it would be impossible to make single-ended
class C amplifiers having 80% efficiency. Yet the world is full of them.
And might I mention class D amplifiers, which can be 97% efficient,
despite having an almost perfect square wave at the drain?
But if you use a diplexer filter, then the amplifier will develop the
same amount of harmonic power as it would develop into a dummy load
without any filter. And that does cause power loss, and thus an
efficiency reduction.
The problem happens when using broadband transformers between the
MOSFETs and the filters, as is usually the case. These transformers have
a significant time delay, limited frequency response, and so they
prevent the low pass filter from properly acting upon the MOSFET drains.
The typical result is highly variable behaviour depending on the band,
with some good and some bad bands, depending on just what phase
relationship exists between fundamental and harmonic components of the
voltage and current waveforms. These problems are reduced by using a
diplexer filter, at the cost of reduced efficiency and added complexity
and cost.
So it's important to understand that a diplexer filter is a workaround
to permit the use of a poor broadband transformer. The sad fact is that
we can't build really good broadband transformers for this application!
And that's why I would prefer to use transformerless broadband
amplifiers, that connect the lowpass filters directly to the MOSFET
drains. To do this, we need drain impedances high enough to allow
bandswitching by relays or whatever. And that means either running low
power, or using a relatively high supply voltage, like 300V and above.
And that's exactly what we should do: Use 300V supplies. A few MOSFET
types are already available that can use this voltage and work up to
30MHz. Goodbye, lousy broadband transformers!
Roger,
> On the parallel, PP, the convention has been PP, parallel. Each unit
> runs two devices, PP with the outputs into a combiner.. Those new 65
> Volt devices,
>
http://www.richardsonrfpd.com/Pages/Product-Details.aspx?productId=1241241
> rated at 1800 watts ea (one page lists Max as 2KW Carrier) at $250
> ea. 4 would be $1,000 and only 375W per device. Less than a quarter
> of their ratings which should require far fewer efforts at the most
> efficient cooling per device. Actually with 4 of these, 3 KW out is
> still less than 50% leaving them running well away from the "1%
> knee"
But the efficiency would be very poor.
> NOTE the base, rather than being insulated is the source,
Yes.
> so the copper spreader would be at 65 VDC
No. The source is usually at ground potential.
But you have a point there: It's that using such transistors requires a
very good, very short, very wide connection path between the spreader
and the circuit board. The bypass caps at the bifiliar choke need to
connect to the MOSFET's flange through a very direct, short, wide path,
to get any chance at correct class AB linear operation. And the gate
drive circuit also needs this direct access to the flange. Otherwise
there is a big chance of inducing enough RF voltage from the drain
circuit to the gate circuit, to burn out the gates, which tolerate only
a very small voltage range. In addition to stability considerations, of
course.
For those reasons I feel more comfortable using transistors that have an
insulated thermal surface, and several source straps that solder
directly to the board. Like the MRF150, or the IXZ210N50L2.
> Running at those levels would require less protective circuitry and
> an ability to handle higher SWR. Of course, with that much overhead
> there would be those who would want every watt they could get out of
> it even though the circuits were optimized for the legal limit, or
> relatively close to it.
The power supply should be dimensioned so as to disallow running
excessive power.
> With 4 devices at $1,000, we are very close to the cost of tubes
> capable of running any mode at the legal limit.
Yes. I don't find it attractive to spend that much money, neither on
MOSFETs nor on tubes.
> Even at the 1800 W limit we're looking at 7200 Max which 4 devices
> should do on SSB.
No, they can't handle the dissipation associated to 1800W each, in
linear operation.
> How ever you look at it these new LDMOS are capable of working the
> legal limit from 160 through 440 although the LP filters could get
> kinda messy, but ALL bands with one amp! Now there's something to
> think about. OTOH the layout for HF and low VHF wold probably be a
> problem at high VHF and UHF.
An unsurmountable problem. For VHF and UHF operation, such transistors
are used in narrow-band circuits, with the impedance transformation
network starting right at the very body of the device. Actually the
impedance you get at the device's terminals is very different from the
one that's present just a few millimeters inside, at the die! That's
part of the fun of working with very low impedances at UHF and higher.
There are no conductors at all, everything is highly inductive! Even a
flat, wide, short piece of copper strap is an inductor. Let alone a wire.
In fact this extends down to HF, at these low impedances, although to a
lesser degree. Still that's the reason why two devices used in push pull
need to be close together. You are connecting them by de-facto coils,
not by conductors, and you have to keep these coils small enough.
Roger, Bill,
>> OTOH we should not forget that a tiny (short duration) voltage
>> spike can take out a SS device, while tubes are relatively
>> forgiving. Very forgiving when compared to SS devices
>>
>> 73, Roger (K8RI)
>
> REPLY:
>
> Good point. Induced voltage from a nearby lightning strike could be a
> serious problem.
Forget it. MOSFETs can clamp an overvoltage pulse, and absorb a pretty
large amount of energy in avalanche mode. They are overvoltage-sensitive
on the gates, but not on the drains, where a lightning-induced transient
would arrive.
I started repairing radios in 1980. I have seen my fair share of radios
damaged by nearby lightning. What fails is always the receiver section,
not the transmitter. Often it's just a protection diode or other
protective device, in some cases somehing more fails - but always in the
receiver section, or right at the antenna connector, where some
protective devices are often placed. The transmitters are robust, even
in 100W radios and often even in QRP radios!
A direct lightning hit will fry a radio beyond repair, and possibly the
operator too. Regardless of whether it was in TX mode, RX, or off. But a
nearby hit might damage the receiver frontend, not the transmitter. If
nearby lightning happens during TX, most likely nothing bad will happen
to the radio.
Leigh,
> For adventurous homebrewers here's an interesting article on an
> alternative novel approach to SSB generation and its efficient
> switched-mode amplification by K1LI and K1KP in the March/April 2017
> QEX magazine, pages 3 to 9. It uses a modern digital implementation
> of novel modulation concepts long ago proposed by Leonard Kahn.
I would be interested in a copy of that article.
> Whilst I like the polar IQ modulation and efficient switching RF
> amplifier architecture and its novelty I must admit I'm a bit old
> school when it comes to SSB generation and its associated low
> efficiency linear amplification...I also don't mind having a lot of
> "big iron" to crank out legal limit PEP and above for copious
> headroom.
That position seems to be shared by most hams, which places me in a
rather lonely position! :-(
> Unlike radio broadcasters, for the modest amount of time an amateur
> station is on-air, efficiency and the electricity
> consumption/conservation is not such a serious driver...
That's right. But it doesn't apply to me: I don't have a connection to
the power grid, because I literally live in the woods. In exchange I
have a great, manmade-noise-free radio environment, complete with
hilltop propagation. I generate my power with a turbine fed by a nearby
creek. In winter I have plenty of power to run my old NCL-2000 amp. But
during the other three seasons of the year, I can't use it! Instead I
do have enough power available year-round to run a legal limit high
efficiency amp in SSB (not in RTTY). This is what makes me so keen on
improving amplifier efficiency!
But while my case is special and rare, all those hams interested in
mobile or portable QRO, including DXpeditioneers, are much better served
by a small, lightweight, low power consumption amp, than by a big, heavy
and inefficient behemot. These hams are the ones that might join me in
my quest for high efficiency.
Kevin,
The guys running those LDMOS devices full tilt, meaning past 1dB
compression, will have them fail,
Not necessarily. When driving an amplifier into compression, its
efficiency improves so much that in a certain range the dissipation gets
LOWER rather than higher. And the current ratings of these MOSFETs are
plenty high enough. So, as long as the low pass filter caps, etc, keep
going, the MOSFETs will, too. It's wishful thinking that a MOSFET
failure will take those splatterers out of the air!
A friend of mine has an Ameritron 1306. 8 MRF-150's, amplifier rated for
1200W out SSB and CW. He runs it at 1KW. He doesn't vary. When doing
RTTY he knocks it back to 700W.
The dissipation at 700W might not be much lower than at 1200W. The best
way to implement a power reduction for RTTY is to reduce the supply
voltage, instead of only throttling back the drive. That will reduce the
power while maintaining efficiency, thus reducing the dissipation.
That's old news, it was done in the tube era too. Many tube amps, like
my NCL-2000, have "SSB" and "CW" settings, the main change being lower
voltage in CW mode. The NCL-2000 is rated for 2kW PEP input in SSB, and
1kW DC input in CW. I understand that this was in line with FCC rules
and regulations in those years.
RF design of an SSPA is pretty straight forward,
The devil lies in the details...
The most expensive part of
any SSPA is the LPF...by a long shot. Ceramic chip caps capable of
taking the high current involved are more expensive than silver if sold
by the ounce.
We should roll our own. There is no reason why these HAVE to be chip
caps. Metal-clad mica caps work fine, and these are easy to homebrew.
I'm not willing to pay 20 dollars for each capacitor, if I can build a
perfectly usable although physically larger equivalent for three cents
in materials plus 5 minutes of my time. I enjoy doing it!
Why do we need four LDMOS devices for 1500W out?
It's useful for relaxing the otherwise too stringent cooling
requirements, as I explained above and before.
What's the 1dB compression point on the devices?
It's undefined. The 1dB compression point can only be specified for a
complete circuit, not for a bare transistor.
The LDMOS devices I know about have a 1KW compression point
That must be a specification given for a specific test circuit. It will
change depending on what circuit you use.
Well, enough for today! Let's leave something for tomorrow! :-)
Manfred
========================
Visit my hobby homepage!
http://ludens.cl
========================
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|