Electronics > Projects, Designs, and Technical Stuff
Analog challenge:High power/high speed amplifier for physics experiment
T3sl4co1l:
Folded cascode means there's no current gain between the diff and VAS, which gives shite for loop gain (perhaps a bit of a plus, to be fair), and means the diff needs to run stupid hot, so, needing so much base current as you note. You could tack on emitter followers to buffer the inputs (so it's, not quite a darlington diff), or put emitter followers in the diff collector circuits to boost them up (which would make a bit of a, I guess, cascaded normal-diff-pair into complementary-diff-pair for instance, depending on how exactly you choose to implement it), or use regular cascode VAS (thus driving the bases from the collectors, giving full current gain as usual).
Could maybe do something with JFETs or MOSFETs, but that's kind of ehh with respect to what kinds of parts are actually available -- JFETs are all 40V or less, and MOSFETs are almost all switching types with way more Ciss than Pd; that is to say, not good for wideband and small-signal amplification purposes. (There are power RF MOSFETs, in high voltages, but only N-ch of course.)
Could also maybe do something with low voltages, getting high gain and bandwidth with minimal phase shift, then level shifting (via cascodes most likely) to a (again cascoded?) VAS and output stage.
Incidentally, Apex makes amps very much like this, which as incredibly expensive as they are, are still looking like a very good investment compared to the effort to design and build this thing! If nothing else, it might be worth a look just to see how they do things -- they usually give equivalent circuits. :)
Tim
Mazo:
--- Quote from: T3sl4co1l on February 27, 2020, 09:30:00 pm ---Folded cascode means there's no current gain between the diff and VAS, which gives shite for loop gain (perhaps a bit of a plus, to be fair), and means the diff needs to run stupid hot, so, needing so much base current as you note. You could tack on emitter followers to buffer the inputs (so it's, not quite a darlington diff), or put emitter followers in the diff collector circuits to boost them up (which would make a bit of a, I guess, cascaded normal-diff-pair into complementary-diff-pair for instance, depending on how exactly you choose to implement it), or use regular cascode VAS (thus driving the bases from the collectors, giving full current gain as usual).
Could maybe do something with JFETs or MOSFETs, but that's kind of ehh with respect to what kinds of parts are actually available -- JFETs are all 40V or less, and MOSFETs are almost all switching types with way more Ciss than Pd; that is to say, not good for wideband and small-signal amplification purposes. (There are power RF MOSFETs, in high voltages, but only N-ch of course.)
Could also maybe do something with low voltages, getting high gain and bandwidth with minimal phase shift, then level shifting (via cascodes most likely) to a (again cascoded?) VAS and output stage.
Incidentally, Apex makes amps very much like this, which as incredibly expensive as they are, are still looking like a very good investment compared to the effort to design and build this thing! If nothing else, it might be worth a look just to see how they do things -- they usually give equivalent circuits. :)
Tim
--- End quote ---
The kinda lowish loop gain that will result from a cascode is to me as you said a bit of a plus->I don't need that stellar of a performance distortion-wise.Besides the high quiescent current makes for a high gm and the solid degeneration of the sinks+the fairly high input impedance of the output stage should make for a reasonable loop gain.
I thought about making the inputs from sziklai pairs,improving the input bias current,enhancing the transconductance of the diff pair,and improving the offset voltage that will be present from real world unmatched KSCs.I considered using a matched 2N5551 pair(I forgot the manufacturer,diodes inc probably??) as I don't want to add another cascode and that is the only high voltage matched pair,buut I am afraid such an arrangement will make up for a high frequency(few MHz to few tens of MHz) peaking of the response that I should then tame.
Thought about jfets and mosfets also but seems they are out of the question.
I know about Apex and searched their portfolio,if I am not mistaken they seem to not have a amp that is fast enough and powerful enough at the same time.
David Hess:
--- Quote from: Marco on February 26, 2020, 11:31:04 pm ---I wonder if a fast optically isolated floating driver to drive the output push-pull transistors wouldn't be easier to get high bandwidth with than having to ping-pong from the rails for voltage amplification.
--- End quote ---
I have tried that and direct level shifters are much faster if only because they can take advantage of capacitive bypassing to drive a high voltage node with low AC impedance. Consider the DC restorer circuit which can drive a signal thousands of volts away with transition times of nanoseconds.
thm_w:
Looks like the fastest apex on digikey is 65V/us, maybe 100kHz? https://www.digikey.ca/en/products/detail/apex-microtechnology/MP118FD/4916546
Their site has some better ones,
https://www.apexanalog.com/products/mp108.html 150V/us 200V supply 10A
https://www.apexanalog.com/products/pa107.html 3000V/us 180V supply 1.5A
but nothing matching the spec.
boblevy4321@sbcglobal.net:
hello
I am trying to do something similar.
I want a power amplifier for my function generator.
I have a design from LINEAR app note 87 that uses a lt1210 in a floating 100 volt arrangement .
I am retired and not that up on spice modeling.
I calculate the gain at -10.44
My question is what to drive this input with and what is the input current or voltage at this point.
I have an idea of using a lm7171 to drive this. it has low output impedance and 100mA of current.
I'm not use to dealing with current opamps.
All help is welcome and appreciated.
Thank you
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version