-----Original Message-----
From: Jon Ogden [SMTP:jono@enteract.com]
Sent: Tuesday, September 01, 1998 8:36 AM
To: craig; amps@contesting.com
Subject: Re: [AMPS] High IMD Drivers
Craig,
>Has anybody built or played with high IMD solid state amplifiers? There is a
>system called " Feedforward" which is well explained in "Single Sideband
>Systems & Circuits" But i would like to know if anybody has implemented a
>working model. Third order imd figures of greater than 60dbc is achieved. It
>certainly would make a great driver.
I have spent many hours playing with these. I used to design them! Most
of the current "linear" amplifiers used in digital cellular service use
feed-forward amplifiers. It's an amazing technique and really works
quite well. It's amazing to see 16 or 20 tones going through an amp but
with -70 dBc IM!
But all this benefit is not without cost. Efficiency of a big
feed-forward amplifier is very low. Our amplifiers at 900 MHz would draw
about 80 amps of current and fed with 27 volts. The maximum average
power out of the amp was about 150 watts composite power. So, as you can
see there is quite a bit of overhead. And yes, transistor devices are
much less efficient at higher frequencies as well.
The system aspects are also VERY complicated. Although, manual
controlled error correction loops are possible, we did all of ours with
software control that was pretty complex.
Here is roughly what happens. If you want more info, I can probably dig
up tons of literature/notes on FF designs and mail it to you. OK....
The signal comes into the amplfier and is split equally in two
directions. One we call the "error loop" path and the other is the
"main" path.
The signal in the main path proceeds to get amplified by a couple of
different stages of amplifiers. The initial stages are class A and the
final stages are class AB.
The error loop signal is amplified and some "initial" IM is generated.
The first stage of amplification in the error loop also has some gain and
phase adjustment controls. These are used to control the amount of IM
cancellation. Initially the job of this "pre-error" as we'll call it is
to cancel out the main carriers. This way, we are left only with
amplifying IM in the error path (hence the name). So we sample a bit of
power from the main path after the first driver stage (still a clean
linear signal at this point). We then apply phase and amplitidue shits ...
snip ... HAVING PLAYED JUST A TINY BIT WITH THIS TECHNIQUE I CAN APPRECIATE
YOUR FREUDIAN SLIP, JON. 73, Dick ... snip
... to the error signal and combine it with the sampled signal and this way
we are able to cancel out the carriers. Now we are left with just IM.
The trick now is to amplify the IM products by the same amount that the
main path has. So we amplify the signals in both loops by the same
amount. Now in the main path we have carriers and IM products. In the
error loop path we have just the IM products that have been amplified
(since we cancelled out the carriers). What is now done is to add a
phase shift of 180 degrees between the main and error loop paths. Once
done, we combine the two paths back together.
What happens? You guessed it. The vector sums on the IM is a net of
zero while the carriers remain! IM cancellation!
Obviously, it is a bit more complex than this, but I believe I accurately
described this technique (it's been about 3 years since I worked with
these beasts).
There are additional difficulties too. In order to do the final
combining you must have a high power delay line. We used a large piece
of heliax wound in a bundle or high power cavity delay filters. Neither
are cheap. And at HF frequencies, the amount of cable needed to make a
180 degree phase shift would be VERY large.
Phase cancellation requirements also have another inherent limitation:
bandwidth. Cellular telephone base stations transmit in the range of 869
to 894 MHz (in the US). That is only about a 3% bandwidth. A multi-band
FF HF linear would be difficult indeed.
Heat is a problem too. You effectively waste half of your power
requirements in the error loop. That is because you are using all of
that power to amplify the IM signal only. All that power does nothing
for your carrier.
Now, you were wondering about if you could implement such as system as a
driver amp. Well, it certainly would be easier to do it at low power,
but I don't know about the benefits. How much power do you consider to
be a "driver"? If it's 100 watts, then we still have the power
consumption issues, etc. If you are talking about 10 watts, it's a bit
easier. However, for low power amplification at HF where you want good
linearity, you probably could make a straight class A amplifier a lot
easier and with less headache.
The main question is how low a level of IM do you need? Celllular
systems use such low IM levels because the modern digital modulations
require it, particularly TDMA cellular (I have heard news that CDMA
modulation does NOT require 60 dBc or better of IM but can still pass
mask specs with a level of 45 dBc). For SSB, one certainly does not need
60 dBc IM products. It would be nice, but we must look at the
cost/benefit level here.
>
>Alternatively could your achieve high level of imd performance of greater
>than 60db by using say a 4cx600J, with low plate voltage and low drive power
>to achieve good third order imd performance? Maybe there is a high
>performance imd tube that i am not aware of. Certainly the vast majority of
>ham transceivers the imd performance is certainly very poor. Anybody looked
>at the FT847? its a SHOCKER, maybe i just have bad examp
How bad is the 847? The FCC has regs that specify how much IM a
transceiver can generate. My guess is the rig is probably about 25 dBc
or so. While this isn't good, let's remember that for every 3 dB drop,
we cut our power in half. So if an IM product is 24 dB down, it's power
is 1/256 of the carrier. At 100 watts carrier, that puts the IM at about
0.4 watts. Significant? Sure, but not deadly.
Don't get me wrong. I am all for signals as clean as possible. But
there are just tradeoffs. The FT-847 is a great rig, but it isn't
Yaesu's top of the line by any means. You get what you pay for in rigs.
In order to be able to put other features in and keep the cost down, the
manufacturers gotta cut something somewhere. I would bet the PA is one
place where the cut costs considerably. A smaller PA means less
dissipation, less power requiremets, etc. So if you compared the FT-847
to an FT-1000D, you'd probably see a huge difference in IM performance.
As to your question on the 4cx600j, I have no idea. Never worked with
the tube.
If you'd like more info on feed-forward amps and how they work, mail me
your address. I'd be happy to dig stuff up.
One last note, if you want a FF amp in the VHF/UHF range, Mini-Circuits
is closing out some little half watt FF amps for around $300. But I
don't know if it's worth it for just 1/2 watt, which is probably why they
are being closed out!
I hope my long diatribe helped.
73,
Jon
KE9NA
--------------------------------------------------------------------------
Jon Ogden
jono@enteract.com
www.qsl.net/ke9na
"A life lived in fear is a life half lived."
--
FAQ on WWW: http://www.contesting.com/ampfaq.html
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
Search: http://www.contesting.com/km9p/search.htm
--
FAQ on WWW: http://www.contesting.com/ampfaq.html
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
Search: http://www.contesting.com/km9p/search.htm
|