At equal voltages, the 6AK5 has a current ratio of about 1:3. The ratio varies for plate-screen voltages above and below there, until you get fully into saturation where screen current takes over. The power rating is small, under a watt, so you do want to be mindful of the limit, even with small tubes like this.
Larger tubes, can simply melt the screen instantly. :-)
Not a problem within the linear range, of course -- if saturated operation is a regular thing, setting a screen power limit is a good idea, for which the resistor divider does a fine job. A triode cascode can do much the same thing but with less "screen" current and higher saturation voltage (although, don't underestimate the "on resistance" of a triode under forward grid bias, they can be quite good).
The simulation agrees with that to a degree at least showing that the circuit is very much a three legged stool, there's no way to get MHz wideband at large gains out of it. if 0dB gain is at 50MHz you can get a couple dB that's it, no matter how much you fiddle it.
To test its response IRL I'd have to lay out some tube sockets and BNC connectors on a slab of copperclad and bluh bluh bluh I can't say I've ever dead-bugged with tubes!
Sometimes one wants more voltage swing than an opamp can handle, like nanosecond 50V or 1200V pulses (both of which I'm doing now.) The current best way to do that is with SiC and GaN fets.
formatting link
formatting link
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
A Spice model should be dead-on for something simple like this.
The only gotcha might be the behavior of the V+ of the opamp as an output. That could be independently tested. Opamp Spice models are sometime wrong, sometimes absurdly wrong, as regards the power rails.
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
... I realize something peculiar now. I've come to understand this (I dare
*not* say "believe"), quite confidently over the last some years, but I haven't formed a concise and convincing argument or proof of its truth that I can share.
I suppose it's just as well, because a truly convincing proof would require knowledge of complex analysis and network theory and stuff, things that I'm not intimately familiar with (promiscuous with though?..). But that's a bit of a gap on my part; as they say, you haven't learned something until you've taught it to someone else.
I can safely say I have. :-)
formatting link
...Well, these are actually somewhat alive bugs... but I've done that too, SMT'd the pins against a copper clad board. Neither is exceptionally good for BW (pin-to-pin on that breadboard is about a pF) though.
????? Gain-bandwith _is_ a consequence of thermodynamics is that what you mean? It follows directly, you can't get more power out than you put power in.
In the case of electrical signals in the frequency domain it's a consequence of the Fourier transform being unitary. Plancherel's theorem:
Sure, you can get a hundred meg out of it -- see for example:
formatting link
What that (Cg1a) really limits though, is the maximum stable gain -- another topic I'm even less operative in, but suffice it to say, the more phase-shifted internal feedback a device has, the more unstable it is into certain questionable source/load conditions. (Resonant tanks are such a case, if not at mid-band, then at the band edges.)
As you raise the F (and Q) of the input and output networks, the amount of feedback that arises from that capacitance (Cg1a) increases. This works in much the same way as coupling does in the wireless power transfer thread -- higher Q means smaller coupling factors (in this case, capacitance ratios) have that much more of an exaggerated effect.
In any case, this is a further limit on the GBW product of a given device.
That's another equivalence worth making light of -- when you are in the lightly-coupled, high-Q regime, it doesn't much matter how the coupling is achieved, whether by series capacitors, series inductors, bits of common (shared) shunt inductors or capacitors, mutual inductance, resistance* -- those only tweak the resonant frequency**, which you can un-tweak by tuning the tanks appropriately. The important effect is coupling, which sets how the resonant tanks interact.
In the ideal case, you don't care at all how the tanks are coupled, there is simply some k factor coupling them together -- which can be mutual inductance k as such, or any other method. The pattern of k's in a bandpass filter (of some given prototype) is exactly the same as the ratios of impedances in a conventional (ladder prototype, alternating series-parallel or vice-versa) filter!
*Resistance is indeed different, as is the point to all of this about passive networks -- reactance conserves power, resistance loses it. For a bandpass network, I think resistive coupling limits its response in an identical way that an RCRC network can only have a maximum Q of 0.5 -- equivalent down to the bandpass transformation (any lowpass network can be transformed to a highpass or bandpass of equivalent response).
**Except for mutual inductance, which doesn't shift the frequency. Very useful!
So, good IF tubes have f*ck-all Cg1a, so they can have massive -- if narrow -- gain. You only need two or three IF stages for a classic BCB FM receiver, including limiting (via excess voltage gain and subsequent saturation), and then add a discriminator and you're done.
When television was introduced (particularly color and UHF, in the 50s), things got a little annoying -- a practical IF frequency was fairly high (45MHz I think?), and the required BW is quite wide (6MHz -- compare to the entire FM BCB of 20MHz!). The first receivers used four or five IF stages cascaded (with IF tubes like 6CB6 -- much less Cg1a than 6AK5).
Later, audio and video IFs were split apart, so a conventional FM audio circuit could be used (more gain, less BW --> fewer tubes), and the video could be, hmm, I think, the gain was partially combined into the luma/chroma circuits. Hence, there was still a lot of pressure in terms of GBW of the luma and chroma amps, throughout the set -- but by distributing the gain among the mixing/matrixing circuits, they could save a few more tubes. And hence, /somewhat/ cheaper color TV was had by the early 60s (also, early hybrid designs).
And hence also why some tubes like 6GH8(A) (a very common triode/pentode) are so maligned. It's not that they're bad tubes -- quite the opposite, it's that they were so good (performance at price?), so popular, that they were beaten to shit by almost every set in existence. They needed every last millimho they could get from them, so they had to run them at or beyond ratings, with short lifetimes.
Well, we can put in arbitrarily much power, that's not the problem. The power out can be a minuscule signal, or full power.
In terms of maximum power, under band-limited conditions, I don't know that that follows from thermodynamics.
Offhand, the time dependency is the thing that doesn't make sense -- despite the name, thermo"dynamic"s concerns statics -- a system in equilbrium -- the infinite endpoint of a (necessarily and continuously) evolving system.
Or to put it another way; if there IS a frequency dependency driven by thermodynamics, then what is the maximum efficiency of:
- A switching power supply?
- A microwave oven?
- An LED?
- A gamma-ray laser?
I don't know of a fundamental prohibition on any of these having a maximum less than 100%, but that would be intriguing if so.
...Perhaps I shouldn't have used thermo as an example; GBW has a similar inevitability, but it /can/ be cheated, at least incrementally so, whereas so far no one has succeeded in repeating such a feat with energy balances (to a statistically significant degree*).
*Because quantum, of course. Damn you, Schroedinger. :^)
Yes, I believe that is about as concise as it can get -- thanks. :)
Or to phrase it differently a signal's _energy_ in the time domain must be equal to its energy in the frequency domain.
Then for a given signal power, if the signal's power spectral density is to remain constant in a given interval and the time-domain signal's amplitude increases, then the frequency-domain bandwidth must decrease such that the energy-equality always holds.
Tube circuits are sensitive to stray capacitances, due to the high impedance levels. Point-to-point wiring with components on terminal strips is here far better than printed circuit boards.
And not just signals, but functions of signals as well (transfer function), which is the important part we're after in regards to an amplifier. If we model a transistor or pentode as an ideal current source in parallel with some impedance load, including a fixed capacitance component, then all we need to know is: the current source creates some signal, which is filtered.
Or if it were a voltage source, it would be equally limited by ESL.
What if it's a Thevenin/Norton source (modest resistance)? Is it limited by anything? A resistive source can be matched into a transmission line at infinite bandwidth; what is its GBW? This doesn't magically flatten the time domain response.
Well, that's the thing, isn't it, time doesn't enter into it at all. We can do this (and usually do) entirely in the frequency domain. What causes the frequency to roll off at some slope (with a fixed offset)? It's not the unitarity of a transformation, it's the unitarity of something more basic in the system itself.
In my defense, I was drinking wine last night, hardly in a position to think deeply about this question... :^)
On Monday, 25 March 2019 11:09:37 UTC-7, John Larkin wrote: ...
...
Off topic comment - it looks like that schematic has errors - the anodes of the two sets of output stages are shorted together. They are are supposed to operate differentially but are wired in parallel.
For example V1104 anode connects to V1114 anode.
Also V1014 and V1024 have no DC path from cathode to negative supply.
Admittedly it does say 'simplified' but they introduced errors
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.