Multiple gain stages - how much where?

Evenly divided is the best way to use gain-bandwidth.

Use low-noise amps in the first and maybe second stage.

If DC coupled, use low-drift amps in the first and maybe second stage.

But other considerations may apply.

I'm doing the layout on a test board right now. It's the prototype of a photodiode amplifier that will have 120 MHz overall bandwidth. There are 5 opamp gain stages, three with switchable gains. Maxed out, each stage has a gain of 10, for a net gain-bandwidth of 12 terahertz.

I like to lay out a board now and then.

John

"John Larkin" wrote in message news: snipped-for-privacy@4ax.com...

I'd say Tektronix would be jealous, but then, I remembered they were already doing that while you were still in diapers. :-)

Tim

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

he

ing

Ohh, 10^5 gain at 120 MHz all on one board? I had issues when I tried this with only a ~2MHz bandwidth. A wee bit of capacitive coupling back to the input and you've got gain peaking. A little more and it's an oscillator. I cut it back so I only had 10^4 on one board. But you know all this.

What opamps are you using and what's the input impedance of the first stage?

George H.

What a setup! However, being the kind and gentle person that I am, I will refrain from the obvious old-person diaper humor.

Bob

--
== All google group posts are automatically deleted due to spam ==

No, they weren't.

John

It depends.

The first stage defines small signal parameters, the next stages define large signal parameters. The first stage is the interface to the source, the last stage is the interface to the load. There are other considerations such as stability, power consumption, cost. etc. In the other words, put all data into equation and find the optimum.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Enough to be a little scared.

AD8009s, sot23, most likely, maybe a couple of THS3201s if we need the slew rate. Input is 50 ohms, with my favorite trick matching network:

pd |\ in-----+-------50r-----+--------| + AD8009 | | | \ 50r | | ----------- | 18pf | / | | | - 50nh | |/ | | | | gnd gnd

The pd pulse is so fast we need to slow it down a little before it hits the first opamp. This does it, and also presents a wideband 50r load to the pd input, to prevent reflections.

John

Oh, I thought the 140dB @ 100kHz for 1THz GBW I once did for a precision tuned IF amplifier was already pretty good :-)

The required gain and phase stability (down to the m° and mdB level) called for a really careful power supply scheme, thanks to opamps poor PSU rejection.

Aren't some designs real fun?

--
Thanks,
Fred.

Interestingly, given some number of identical tubes, and a choice of plate resistor value (which trades off lf gain vs 3dB rolloff frequency), the best gain-bandwidth results from a per-stage voltage gain of sqrt(e).

e pops up everywhere.

Of course, a distributed amplifier can do better.

John

Sometimes.

Often.

Yes.

If all you're doing is applying some straight DC gain to a well-behaved but small-amplitude signal, then it's hard to go wrong with a pair of similar amplifiers in two equal-gain stages.

If you're doing filtering, either by capacitively coupling the signal, or by low-passing it, or something yet again, then you need to design the overall amplifier so that you get the gain you want without hitting any voltage or slew limits anywhere -- this may require a different gain distribution than simply setting the two DC gains equal.

I would say, though, that any time you need so much gain that you have to cascade op-amp circuits to do it you should make sure to analyze any proposed circuit for gain, bandwidth, overflow (both hitting the rails and slew rate), noise, offset, and maybe distortion as well. Just fumbling around with a concatenation of cookbook solutions will get you something that works if you're lucky, but endless hours of frustrated puzzlement in the lab if you're not -- and there's no reason you have to depend on luck.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" was written for you.
See details at http://www.wescottdesign.com/actfes/actfes.html

Yeah, but (out of necessity) Tektronix probably spent several order of magnitude more to make it happen back then...

They would've killed for PHEMTs ;-)

Tim

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

Since each amp has its own rolloff, that product is gain* (BW**5) not gain*BW. ;)

Cheers

Phil 'Confederate Correct-orr wannabee' Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

The old one-pole diplexer trick, eh, 99? It's a goodie, I agree.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

For N stages, each with bandwidth BW, the effective bandwidth becomes:

BWeff = BW*sqrt(2^(1/N)-1)

For 2 stages, the "shrinkage" factor is 0.643 ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

       The Ground Zero Mosque IS Appropriate When Renamed...
            The Obama Monument to American Impotence

On a sunny day (Mon, 20 Sep 2010 12:02:43 -0500) it happened "Tim Williams" wrote in :

They would kill for a buck. When the first gulf war I was at Tek, and they were so happy with the extra sales that created. Made me sick.

Maybe that is an aspect of capitalist system, but not a good aspect. In the end we all come from cave-men.

What does it matter, the Universe keeps going.

the

ling

Well this is the most on-topic response. I'll just add that the first stage will determine the noise figure of the cascaded amplifiers. Less gain and more feedback in the 2nd stage provides a lower output impedance.

sales that created.

That was LONG after the good days at Tek, which basically preceeded 1982 -- before they offered early retirement to over 800 employees, trying to get their revenue per employee number improved before going on to fire people wholesale. About 800 took the offer, gutting the company. Executive staff expected to see only 200 accept it (I spoke personally with Jim Castles about this, who was on the board.)

By late 1980's, Tek had been completely transformed into something no longer anything like the old Tek.

Tek had, circa 1982, $600 million dollars at US National, no short term or long term debt and were terribly worried about leveraged buyouts (big thing, then and I suppose now.) They were also at less then $65k/employee gross revenue, which was

1/2 of what IBM was, that year. But rather than borrowing say $2B money to add to their $600M, going into debt, and walking into any marketplace they wanted to with excess employees already in their seats and hired up and tons of cash to plow into it -- completely removing the threat of leveraged buyout and putting themselves in a situation that anyone would drool to be in -- they had no self-confidence in being able to succeed and instead chose the safe bean-counting approach of reducing employees and buying their own stock back (you _NEVER_ buy your own stock, as a rule, because if you can't do better things with the money then you shouldn't be in business.)

It is our ideals and dreams which move us forward. And we do not have to succumb to the lowest common denominator, if we choose.

Depressed? Or just unable to maintain the facade of denying reality that keeps most people 'sane?'

Jon

On a sunny day (Mon, 20 Sep 2010 11:26:24 -0700) it happened Jon Kirwan wrote in :

No, just seeing some really stupid remarks on usenet, by the usual people. I wrote a little about free will in sci.physics a little while ago: Personally I do not think we have any free will at all. We just think we do [some do, sometimes]. The 'decisions' we make are prepared in the subconscious, and that is just a machine with has as input the genetics of evolution, the experience of our life, the senses (eyes ears, feeling, etc), Running like clockwork in this universe, just like suns get born and die, planets orbit - for a while - etc.

It is funny for such a peace of complicated matter as we are, to think we are 'in control' of anything, or even are at the centre. OTOH is everything in the universe and beyond, because what do we really know, is coupled in a FTL way - say immediate way - by some quantum coupling or something beyond that, then even we are just a wheel in a big machine, like a wheel in a clock that causes it to indicate time. Such a wheel, with 'I am in control' engraved on it.

Evolution, cave-men -> us -> whats next will be as vicious. Professor Hawkins was on C4 TV in the UK Saturday, talking about aliens, if those existed. IF those exist and ever come here, we would be in big trouble. See what we do to apes. (Because if they managed to come here their technology would be more advanced then ours).