Can you improve these amplifiers?

I am currently trying to design an amplifier for the output of a generator. It needs to work at up to 1MHz, and have a negligible distortion, ideally in the ppm range whilst delivering 30 to 40Vpp.
This is my first tentative:

1687425104742.png

The THD is ~0.01% and there are cheats: spice sources used as ideal buffers, because the base currents of the transistors are too large and too distorted. Additional transistors would be required.
There is also a nasty, unexplained peak in the frequency response, in the region of 200MHz.
 

Attachments

  • OutAmp.asc
    8.4 KB · Views: 72
FETs and/or BJTs designed for RF use will be needed I think - you need individual component gains to hold up at 1MHz, suggesting GBW products of well over 100MHz would be wise. Layout will be tricky, good RF design experience essential, screening cans a requirement too I bet.

I hope the load impedance is a stable 50 ohms. I also presume it is DC-coupled?

You might want to consider using a design capable of much higher power so its running at low level, think of 3rd-order intercept behaviour for instance.

RF design is usually class A or class C - class B is seldom used without transformers, and I don't think you can get low enough distortion with transformers. I suspect you will not get good enough matching of complemetary devices at high frequencies to keep cross-over under control anyway. You need very low inherent distortion plus as much negative feedback as can be stabilized.
 
Member
Joined 2011
Paid Member
Link to thread: Winfield's 100W DC-10MHz 1000V/us amplifier, posted by Winfield Hill. He's a coauthor of the famous textbook "The Art Of Electronics" by Horowitz and Hill.

Output transistors are 5x 2SC4883 and 5x 2SA1859 , biased at approx 85mA per device. In other words, Class-AB. fT at that bias current is about 85 MHz, see datasheet graph below.

_
 

Attachments

  • ft_2sc4993a.png
    ft_2sc4993a.png
    25.2 KB · Views: 88
Check out the circuit of the older Tek FG504 or similar function generators.
Thanks, the circuit is interesting as always with Tek (or HP), but it is mostly speed-oriented (40Mhz) and the linearity, although good is suited to a function generator, not capable of ppm performance.

The output power is only 4W. Class A is very practical.
Yes, I may resort to class A, but with the current conveyor topology, some adaptations will be required
FETs and/or BJTs designed for RF use will be needed I think - you need individual component gains to hold up at 1MHz, suggesting GBW products of well over 100MHz would be wise. Layout will be tricky, good RF design experience essential, screening cans a requirement too I bet.

All the transistors have a Ft well in excess of 100MHz, thus speed per se is not an issue. The problem is linearity at high frequencies. Construction is not a problem for me, even on a breadboard, and after all the max frequency handled is 1MHz, which is not really challenging



I hope the load impedance is a stable 50 ohms. I also presume it is DC-coupled?

You might want to consider using a design capable of much higher power so its running at low level, think of 3rd-order intercept behaviour for instance.

RF design is usually class A or class C - class B is seldom used without transformers, and I don't think you can get low enough distortion with transformers. I suspect you will not get good enough matching of complemetary devices at high frequencies to keep cross-over under control anyway. You need very low inherent distortion plus as much negative feedback as can be stabilized.
The load will vary from 50 ohm to infinity; the sims are based on the worst case, a short at the output and 50 ohm for the amplifier.
The matching is indeed going to be a problem, and I may need to resort to a non-complementary output. And anyway, for the moment it is just a sim: the reality will probably somewhat harsher.

Link to thread: Winfield's 100W DC-10MHz 1000V/us amplifier, posted by Winfield Hill.
Yes, it is an inspiration, as are Stochinno designs, but the scope is not quite identical. Anyway, I try to gather as many inspiration sources as possible to arrive at a reasonable design: one that doesn't require exotic devices, large componets count or inordinate amounts of consumed power
 
Do you have a.asc for it? It will make things easier.
Yup , here. That cascoded circuit is my "go to" VAS. Bombproof , gets cooler and the red led's flash at overload.
I built a modified one with a cascode led's as infrared opto. Fed it to a MCU to count clip events.
Never tried to get it to amplify 1mhz ?? My EF3 limits both bandwidth and slew ... still fast @ 300V/uS.
 

Attachments

  • HELLRAISER V1.2.zip
    19.5 KB · Views: 101
Member
Joined 2009
Paid Member
Start with the distortion your diamond output stage generates at 1MHz by itself, without any NFB applied. Say it’s 0.1% (-60dB). If you want 1ppm (-120dB) with NFB, you‘d need 120-60=60dB of loop gain, at 1 MHz. With the usual dominant pole compensation this is equivalent to the crossover frequency of 1GHz. Commonly available transistors have Ft of up to 150-200 MHz, limiting the crossover frequency to 50-70MHz, so you have to use a more complicated compensation scheme. Even then, dropping from 60dB @ 1MHz to 0dB at 50MHz and retaining a reasonable phase margin is not an easy thing. And of course the actual hardware will need to operate at up to 50 MHz, so VHF construction techniques would be required.