On Topic: Distributed amplifier?

How much phase shift is necessary between stages before "a shitload of amps in parallel" is called "distributed amplifier"? And what is the role of that phase shift, what does it actually accomplish? Like, why not put a bunch of amps in a circle and feed their outputs radially towards the center and skip the phase shift BS?

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms
Reply to
Tim Williams
Loading thread data ...

Phase shift is not the objective, it's the consequence. Each amplifier has some input impedance. If you drive them all in parallel, you get LOW impedance. If you make them elements of a distributed transmission line, you make the whole thing easier to drive. Ditto on the output.

Reply to
mike

A distributed amp forms a transmission line from inductors (or t-coils) and a bunch of grid/gate capacitances, and another transmission line from inductors and the plate/drain capacitances. Such a transmission line can be very long and keep a fairly constant bandwidth.

This structure allows you to put the tranconductances of all the gain things in parallel without the capacitances being in parallel, so GBW goes up. Putting tubes or fets directly in parallel increases transconductance and capacitance together, so GBW doesn't get better.

A more fundamental concept is that a transmission line charges a lot of capacitance, but only one bit at a time, by spreading out the charging in time. A distributed amp trades time delay for GBW.

ftp://jjlarkin.lmi.net/DistAmp.JPG

The really fast e/o modulator drivers, DC to 20 or maybe 40 GHz, are GaAs integrated distributed amps.

formatting link

John

Reply to
John Larkin

As already pointed out, the input and output capacitances would be in parallel. The failure of a single component could take down the whole system.

A similar way to avoid this kind of problems is to use the Wilkinson power divider to split the input signal into several paths to be amplified and combining them back again with an other Wilkinson power divider. This has been often used to combine a large number of medium power transmitter modules to produce similar power levels than a big transmitter tube.

The failure of a single module only drops the total output power with that amount, but it does not affect the operation of the other modules.

Reply to
Paul Keinanen

ps

f
a

ter

It's the Percival distributed amplifer - he patented it back in 1936 when he was working on the EMI televison project

formatting link

Blumlein produced a lot more patents, but Percival managed to survive the war, and was stiil working at EMI Central Research in West London when I worked there from 1976 to 1979.

I know about his distributed amplifier before I went to work there, and pleased when I got to meet him, when he grilled me about some work that I was doing. Sadly the answers I gave him weren't the ones that he wanted to hear.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

On a sunny day (Sat, 09 Jan 2010 21:55:13 -0800) it happened John Larkin wrote in :

WOW! But 238 $ for RoHS... Good to know it exists though.

Reply to
Jan Panteltje

On a sunny day (Sun, 10 Jan 2010 00:01:51 -0800 (PST)) it happened Bill Sloman wrote in :

So he did not believe in global warming either, so what.

Reply to
Jan Panteltje

It could be difficult to match the operating conditions for devices connected in parallel; parallel is also prone to parasitic differential mode oscillations.

Long while ago, I worked with the distributed RF power amps using tubes. Somewhat 1.5kW of linear power from 30 to 100 MHz without tunning. The problem with it is different operating conditions for every active device: while going forward, the input wave gets weaker, and the output wave gets stronger. This results in lower power efficiency and gain. Each device can be used with different matching, and/or the line impeadances could be made variable; however it makes the amp complicated and narrows down the GBW.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

formatting link

Reply to
Vladimir Vassilevsky

loman

am=3D

e o=3D

ut =3D

cen=3D

Back in 1978, global warming wasn't a topic of any direct interest to EMI Central Research and we certainly didn't discuss it.

And I don't recall him being sceptical about what I told him (whateer it was) - he had hoped that the situation was different from the one I laid out, but he hadn't had any strong expectations either way.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

ps

f
a

ter

As you go from one amplifier to the next, there is a delay in the output side of things. You need to match this on the input side. A cute device that I have long wanted to make just for the fun is a distributed parametric amplifier like this:

-------+--------+-------A Repeat ! ! =3D=3D=3D V ^ =3D=3D=3D ! ! ------++---------------C ! ! --------------++-------D ! ! V =3D=3D=3D =3D=3D=3D ^ ! !

--------+------+--------B

The AB line is carrying the signal and the CD line is carrying the pump. At each point along the cable some of the pump energy is converted to the "sum" energy. Because the signal is a lot less than the pump or the sum, less of it goes towards making the sum. When you get to the far end the pump and the input signal are mostly gone and only the "sum" signal is used. A nice thing about such devices is that they are extremely low noise because there is no interaction with heat of the devices and no electrons are working alone.

The trick would be making the speed of the lines work out right so that the adding is always exactly in phase in the forward direction.

Reply to
MooseFET

The point of a transmission line is that you don't have to drive all the capacitance at once. The impedance of a lossless line is sqrt(L/C), where L and C are the inductance and capacitance per unit length. So how much distance you need between taps in a distributed amplifier depends on what characteristic impedance you want it to have.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
 Click to see the full signature
Reply to
Phil Hobbs

We tried building our own wideband (roughly 5 GHz) distributed amps using discrete GaAs fets. Results weren't very good, especially step response, which we needed to be very clean. We concluded that it would take a lot of work to do it right, and then we discovered the Hittite parts.

There's loss along both transmission lines, both from resistive losses in the inductors (or equivalent) and resistive components in series with the fet gates and drains. And a lumped LC transmission line doesn't behave like an ideal distributed line anyhow.

John

Reply to
John Larkin

When we were building these back in school, with our hand construction techniques, the law of diminishing returns hit around four FETs. Beyond that, the little inaccuracies in spacing starting digging into your bandwidth...

Charlie

Reply to
Charlie E.

center

It could probably be done, to a reasonable number of stages, on a pc board, but it would take a huge heap of engineering. The $200 Hittite parts are a relative bargain.

The IC distributed amps are usually cascodes, which improves isolation and increases drain resistances... gaasfets have pretty slopey drain curves. That would be even trickier to do with discretes. It's nice to not have package parasitics, too.

John

Reply to
John Larkin

Is there any loss in a lossless amp? There can't be, if both lines are lossless and there is no coupling (s12 = s21 = 0). That's never the case, of course -- the most common exception being a hell of a lot of transfer due to Miller C. But will that alone cause attenuation?

That's a pretty weird situation as it is, coupled transmission lines... well, not so weird, such a structure can be expressed as, for instance, differential lines over a ground. There's mutual and to-ground capacitances, pretty basic. But the quirky part is you've stashed amplifiers between them, so instead of coupling or loss, there's unidirectional* gain. That's not your average twinax.

*Hmm, it's not strictly one direction, since there's a nonzero reverse gain (s21)... 'anisotropic' would be better, but the word implies a spacial rather than graphical relationship. Oh well.

With lossy elements, there will obviously be loss. Say, how lossy are component capacitances, anyway? That's not something ever specified for semiconductors. That, and inductor loss, sum up the lossiness of a real distributed amplifier. It must've been pretty good back in the toob days, even the early Teks had like 13 6DJ8s in a row. Tubes are fairly ideal capacitors, which just so happen to have electrons flying around inside to do Work. Tubes always give you an honest RC time constant regardless of bias or amplitude. No voltage controlled capacitance, no storage time, no slew rate limiting, just RC. Do the same thing with transistors and you'll get all sorts of ugliness... and just imagine, someone tweaks the vertical offset knob and your total capacitance goes up by 100pF... argghh!

Yeah, I think Hittite *earns* their money...

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms
Reply to
Tim Williams

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.