EMRFD Message Archive 939
Message Date From Subject 939 2007-08-06 16:14:50 Allison Parent Understanding and basic design of linear power chains.
They come about as result of several questions asked
here and elsewhere.
One of the biggest sources of frustration in transmitters
is lack of power or instability and undesired signals
especially self oscillation.
Some ground work:
To get a 5W (carrier power) signal you need devices that can
deliver that and also enough gain. So the first step is how
much gain is enough?
For a SSB or hetrodyne CW radio the output of the typical DBM
mixer is in the range of -20 to -10DBM. For our cases we use
The -20 number as it's the worst case. We decided on 5W as
measured on a Bird wattmeter (carrier power). Another ways
to describe 5W power is +37DBM (37db greater than 1mW). That
notational change gives a clue on how much gain is needed.
Source power is -20DBM and output is +37DBM so we need 57DB
of gain. Also since we may have coupling losses along the way
plus a little excess gain (just in case) we add 3DB to that
for a nice round 60DB.
That number is a clue why many get oscillators rather than clean
smoothly operating amplifiers as 60DB of gain means the output
is 1,000,000 times greater than the input! Sound daunting, but
radios deal with gains and signal magnitudes like that all the
time. So whats the majik?
What the books rarely point out and is often only hinted at by
pictures of others design work is that with high gain circuits
and high power circuits layout is important. Specifically avoid
having the input near the output where possible and straight
line circuits usually accomplish this. This is rule one.
At RF impedences should be matched. We do this for antennas
to insure all of the available power is transfered. For amplifier
stages this is also important as any power not transfered is a
loss of gain and also the impedence mismatch will mean the stages
may actually produce more gain than desired or less depending on
the mismatch error. So designing for a 50 ohm interstage impedence
is handy as most measurement gear is standardized at this impedence.
It's is also handy as if all measurements are at the same working
impedence then gain calculations are simplified. So rule two is
50ohms in and out.
Let me add at this point that for sanity sake there is always a
factor added to insure that stages are not run at high gain or
expected to deliver their maximum power possible. The reasons
run from losses can be greater than anticipated or gains may
not be achieved. In the case of SSB signals linear amplification
is imprtant to signal quality and bandwidth epecially. In short
distortion is bad. So we never want a stage running near or at
the point where distortion from overload or lack of output
capability. Rule, never ask too much from a stage.
A forth consideration and part of the art is how much gain per
stage or how many stages? This is often a decision process that
involves what devices are available and the gain attainable from
them at the desired frequency of range of frequencies.
So look at out test case. We want 60DB of gain overall. We
could use 3 20DB amplifiers but that needs a sanity check. For
example if the final transistor is a 2SC1969 a robust 16 watt
HF device we note in the data sheet that device can only deliver
a guarenteed 12DB of gain which is typical for that class of
transistor. It may give more but we can not count on that. Well
if that's our final that means the preceeding stages must deliver
48DB of gain and this also requires a sanity check.
So we know the final can do 12DB gain and we want +37DBM
power(5W) that gives us out next needed information. What
power will the stage before it need to deliver. The
arithmetic of DBs says +25DBM (37-12=25) and for those scared
by that it's 320milliwatts. So the device before it has to
deliver with some comftable excess a half watt. A suitable
device that is often used at HF is the 2N3866. However they
do not tell us in the data sheet how much gain we can get
at HF. We can assume a number and design around that if
its reasonable. We will use 13DB as it's a good balance
between what's possible and trying to not have too many
stages.
Again working backward we believe the driver will do
what we ask but how about the one before that? Well
it has to deliver power too and if we assume the driver
can do 13db and deliver +25dbm (320mW) then we need to
deliver +12DBM (16mw). Again we design for more as losses
can happen so we use 15DBm (32mw) here (A 2N3904 can do
this. But one can't give over 30DB of gain from one
(reliabily!). So we say we'll use two and that gives us
two stages of 17DB and 18db gain which in experience
and practice is practical.
The proceeding has produces a gain plan and a list of devices
that can deliver the desired power, gain and potential bandwidth
if employed correctly.
[ of the RD016HHF1 were use that device is also a 16W and robust
but unlike the 2SC1969 it can produce 16DB of gain. Consider how
that 4db increase (2.5x) in gain would impact the gain and power
calculations!]
list of devices and gain for each stage:
2SC1969 12DB 5W po (it can do more but, we want 5)
2n3866 13DB .32w po (plan for 500mW in case..)
2n3904 17db .016w po (capable of 50mw)
2n3904 18db .0008w po (not even stressed)
We have defined 4 stages with attainable gains, some excess
gain if needed and reasonable power outputs. As the desire
for multiband is there we'll assume a wideband design using
torid cores for coupling and resistive feedback wideband
amplifiers as describes in section 2.7 of EMDRF. This will
insure was meet the first three criteria as well as others.
It's possible to have other gain place and power distributions
and device choices can affect this greatly.
The circuits that embody this are in EMDRF and worth study.
Every designer will make tradeoffs and have favorites that
end up as part of the end result. The reasons are often more
infomative than the end result.
However the "howto" for construction of those is not in any
one place but part of the descriptive text of the entire
handbook. I can point to far to many places so please read
it, it's there. There is a lot of basics up front and them
later there are examples scattered as whole or partial projects
that embody that.
I gave a few bits early on here but there are curcuit tricks.
Well not really. The handbook(EMDRF) discusses this and the
"tricks" are just good construction methods. One favorite
is "deadbug", "Manhatten" or just plain "Ugly" construction
on a ground plane of scrap PC board. At first glance this
appears to be a sloppy cheap assembly but has hidden
qualities that are hard to realize with other techniques.
One being ground is right under you and everywhere. That
feature alone is hard to duplicate with a etched board
without careful hand work and often rework on the part of
the layout designer. Ground, the curcuit common has to be
of high integrity as remember we have a gain of 1,000,000
to deal with. It also allows us to put up shields between
stages and use parts with very short leads. It is also
fast to build, inexpensive and easy if care is taken to
modify.
Speaking of lead lengths, Short leads are desireable.
Not only can long leads act as undesired antennas but at
higher frequencies or in very low impedence circuits the
inductance of a wire can have an impact to the operation
of the overall circuit.
One other consideration is modular construction as it allows
us to group stages or widely seperate them. For a power
amplifier chain as described we already assume 50ohm input
and outputs and each stage can be tested in isolation
to see it does do as we expect or to determine why not.
In this case modules could be anything from scraps of
PC board or even Altoids (mint) tins (they are solderable
with lids) if shielded modules are needed. Of course there
commercial packages but most are aluminmum and that means a
PC board either etched or ugly construction inside. In the
end scraps of PC board material are both a good plane to
build on and can even make RF tight boxes as needed in
any size or shape.
A moment on cooling. The last two stages are running at
a power level where cooling is usually in the form of
heat sinks and again modular construction helps as your
not forced to have a heatsink in a awkward location or
worse.
Last item, power. I mean power as in DC power. Often
forgotten is that a linear amp chain like this sucks up
Amps. Typical for a no signal condition for the described
amp is around 150 or more milliamps as standing bias and
to feed the bias resistors and at full power over 1.0A.
Why, class A, AB,and B amplifiers have efficiency of
around 50% or worse. So a 5W out means around 12W of
power or more for DC in! This points out another detail,
your power distribution has to be capable of providing
amps of current and also be adaquately bypassed so the
DC bus is not the transport for undesired feedback.
There is more but paying attention to these basic details
and understanding their impact will help greatly in
producing successful rf power chains and may help those
diagnose possible reasons why a project doesn't behave
as it should.
Just thinking out loud.. hope it's useful and stimulates
ideas and questions.
Allison942 2007-08-06 21:29:08 K6WV Re: Understanding and basic design of linear power chains.
helpful.
Jon, K6WV
Allison Parent wrote:
> Some thoughts on RF transmit power train construction.
> They come about as result of several questions asked
> here and elsewhere.
>943 2007-08-07 06:59:56 Bruce Beford Re: Understanding and basic design of linear power chains.
am primarily a lurker here, but thoroughly enjoy your posts. They
always are very informative, and you have a clear and easy to
understand writing style.
Obviously, your many years of experience and experimentation have
provided you with a wealth of knowledge from the 'School of Hard
Knocks' 8-). Your willingness to help others through their musings is a
great asset to this group. Thank you so much for taking the time to
share your experience and knowledge here.
Best regards,
Bruce Beford, N1RX944 2007-08-07 11:37:50 Hari G Re: Understanding and basic design of linear power chains. 946 2007-08-07 21:12:19 Allison Parent Re: Understanding and basic design of linear power chains. 948 2007-08-08 11:39:38 Hari G Re: Understanding and basic design of linear power chains.
On 8/8/07, Allis1092 2007-10-15 21:44:43 Bill Meara Understanding and basic design of linear power chains.
mind if I post it on my web site.
http://www.gadgeteer.us and mention it on our
podcast http://www.soldersmoke.com
Thanks, 73 Bill M0HBR
1095 2007-10-16 05:50:15 Allison Parent Re: Understanding and basic design of linear power chains. 1101 2007-10-17 00:49:08 neomag_magneo Re: Understanding and basic design of linear power chains.
This is something I have been tackling maybe even harder than the
receiver issue expained in another thread. I have been able to reach a
satisfactory amp chain up to some 1 W (30 dBm) level but from that on
the power levels start to reduce. Most probably, this is due to the
active devices I am relying, i.e. power mosfet IRF510 which works
quite nicely up to 7 MHz but the gain goes down above 10 MHz. For
nonlinear modes like CW and RTTY I have been able to achieve some 10 -
20 W but for SSB and PSK31 the output is more like 5 W in 14 - 28 MHz
bands.
The plan right now is to carry on with IRF510 since it is robust and
inexpensive, aim at 5 W level in all bands up tp 28 MHz (also in
linear modes) and consider something else beyond that. One opti1102 2007-10-17 09:32:19 Allison Parent Re: Understanding and basic design of linear power chains. 1103 2007-10-18 01:40:18 Heikki Ahola (OH... Re: Understanding and basic design of linear power chains.
Thanks for advice and info!
>To clarify my situation, I started with 24 - 28 V for IRF510´s from
> The real problem with IRF510 is it's a poor device at 12V. At 24V
> it does perform far better. A pair of them at 24V in the WA2EBY
> PP amplifier [also known as K5OOR HFpacker] does an easy 45W at
> 12M and still does 40 at 10M with 1W drive. That amp at 28V and
> pushed a bit harder can do as much as 75W but the devices are being
> abused! The magic is 24V and the drive cicruit (input side) has to
> be designed with the high input capacitance of the MOSFET in mind.
> Most amp designs around IRF510 forget that and thats why the power
> drops badly above 7-10mhz.
>
> The articles for HFpacker and Wa2EBY are available on line.
>
the very beginning, so this should not be any problem or limitation.
Furthermore, I was the schematic by Mike Kossor in QST I followed.
After reading the messages in this group ( in particular the original
by Allison) I start to realize that my problems may be related both
the input drive and DC power available for my amplifier. The drive is
from a transverter with class-A output delivering hardly 0.5 W, also
the DC power is 50 W max (about 2 amps), these combined cannot
produce more than 10-15 W at best. Need to increase the drive somehow
and change the power supply. I did try without the attenuator at the
amp input but this is obviously needed for stability, at leat in my
case the trial ended up with wild oscillations !
One further question regarding the biasing of the IRF510 amps.
According to the articles the HF Packer Amp is class AB linear, still
they specify 10 mA bias for each device. According to my experience
and other sources this should be close to 0.5 A for good linear
operation, the discrepancy is quite evident !
BR
Heikki (OH2LZI)1104 2007-10-18 04:16:16 w4zcb77 Re: Understanding and basic design of linear power chains.
website). Inexpensive Mitsubishi 12 Volt TO220 devices, (MUCH easier on
the backpacking), very conservative design, 60 dB of unconditionally
stable gain and IMD in the basement. Adaptable, I've built it in just
the first 2 stage format as well as the complete 3 stages. You even get
a circuit board pattern layout for your etching enjoyment.
W4ZCB1105 2007-10-18 10:40:58 Hari G Re: Understanding and basic design of linear power chains.
On 10/18/07, w4zcb77 <W4ZCB@mchsi.com> wrote:You might be interested in taking a look at the G6ALU design. (His
website). Inexpensive Mitsubishi 12 Volt TO220 devices, (MUCH easier on
the backpacking), very conservative design, 60 dB of unconditionally
stable gain and IMD in the basement. Adaptable, I've built it in just
the first 2 stage format as well as the complete 3 stages. You even get
a circuit board pattern layout for your etching enjoyment.
W4ZCB
--
De Hari
VU2GHB1110 2007-10-18 19:19:09 Allison Parent Re: Understanding and basic design of linear power chains. 1111 2007-10-18 19:23:17 Allison Parent Re: Understanding and basic design of linear power chains. 1112 2007-10-18 21:11:26 jabauzit Re: Understanding and basic design of linear power chains. 1113 2007-10-19 15:57:34 Allison Parent Re: Understanding and basic design of linear power chains. 1114 2007-10-19 16:04:57 Allison Parent Re: Understanding and basic design of linear power chains. 1115 2007-10-20 05:30:37 w4zcb77 Re: Understanding and basic design of linear power chains.
> satisfy a lot of needs. How is the IMD?
>
all bands and for several power levels. At the 5-6 Watts required to
drive a 4CX1500B, the IMD is comparable!
Feedback is nice, particularly for containing the I/O return loss
and gain flattening. Biggest problem with Steves design is the
marginal output transformer for 160 Meters.
If you want more power, Bill Sabins old 1999 article shows what can
be done with more Voltage. I have built that with some mods, (Change
of first stage to 48 Volt MRF148's as well to exempt the 12/24 Volt
requirement). 42 dB of gain (WITH a 3 dB pad on the input to flatten
the line) and flat gain within 0.5 dB from 1.8-30 MHz. Matter of
fact, it only droops about 7-8 dB at something like 36 MHz and comes
back up to 37 dB of gain at 50 MHz.
Bill has a thousand years of commercial design under his belt and it
shows! Running it class A, (2 amps for the pair of 148's and 3 Amps
each for the MRF150's, it's clean enough by itself to get past the
FCC specs without an LPF. Does take a lot of heatsink for that
though!
Regards
W4ZCB1116 2007-10-20 07:00:50 Allison Parent Re: Understanding and basic design of linear power chains. 1117 2007-10-20 10:45:55 w4zcb77 Re: Understanding and basic design of linear power chains.
> a huge heatsink or as I did a far smaller one and use a very
> quiet fan. The difference between forced air cooling and natural
> convection is great enough to warrent choosing a small fan turning
> at reduced rpm that runs quietly. Most of the small DC fans
> require trivial current and run at voltages as low as 5V. A good
> example of this is a 2M MRF247 amp from junkbox parts. The problem
> was the heatsink on hand was about 40% the area I'd have chosen
> for sustained full power (100W). Operating 1 minute at full power
> would raise the heat sink temperature to 120 degrees F. This did
> not cause problems but the device temperature was reaching 155F at
> the case. One could argue this is well within the ratings and leave
> it at that. however for sake of testing I grabbed a 25 cfm 80mm
> 12V DC fan and ran it at 9V placed on the heat sink and tried the
> same test and the heatsink never exceeded 95F even after 5 minutes
> and the the transistor ran proportionately cooler. Also the
> recovery time from operating temperature to within 10 degrees of
> room temperature was far faster. The fan was hard to hear and the
> cooling was well worth the inclusion. Fan power requirement was
> a very small 70mA at 9V. Something to consider when cooling seems
> marginal.
Indeed. My 100 Watt class A amplifier has a 5/8 inch Aluminum web,
and a 3/8 copper heat spreader. (One inch thick overall!)With the
fins spreading over a volume of 7 x 9 x 12, it takes 35 minutes of
key down to heat things up to uncomfortable.
I well remember the value of moving air despite my rather intense
hatred of fan noise. My old KWM-2, during the heat of a contest,
would fry eggs on top. Or possibly boil them if you liked them
poached. A tiny "Whisper fan", 16 cubic feet per minute as I recall,
and nearly silent, seated on 4 rubber grommets and just laying on
the lid over the top of the 6146 compartment, sucked air in from all
over the radio. After a contest, you could hardly tell that it had
been turned on.
My little G6ALU amplifier has the TO-220 devices mounted on a CPU
heatsink ($4 at some hamfest, brand new and in the original factory
box!) with a fan that only comes on when the bias is supplied. Never
gets much warmer than ambient.
W4ZCB