Home |
Search |
Today's Posts |
#1
|
|||
|
|||
How to determine tranfer curve of a tetrode, and linearization
I need to figure out the plate current versus control grid voltage curve of
a 4X150A that I intend to use as a voltage controlled current sink (class A). I have not found any transfer curves in datasheets of the tube. What's the best way to determine this? Also, is it possible to linearize by varying, say, screen grid current or voltage as well? Thanks. |
#2
|
|||
|
|||
Prune wrote: I need to figure out the plate current versus control grid voltage curve of a 4X150A that I intend to use as a voltage controlled current sink (class A). I have not found any transfer curves in datasheets of the tube. What's the best way to determine this? To determine the transfer curves, you set up a sample of the tube and measure Vg1 input and Va output for a given load, and plot the graph. If it involves thousands of volts, beware, its dangerous. The transfer curve can be seen with a dual trace CRO in X-Y mode. Its possible to use a tube tracer to draw the Ra curves for values of grid voltage, but unless you have the piece of gear, its a lotta work to make one. Also, is it possible to linearize by varying, say, screen grid current or voltage as well? Linearizing a tetrode can only be done by applying NFB. But your aim is to use the tube for a current sink, ie, a high impedance sink. What for? If its for a DC plate supply so that the anode can then be cap coupled to an OPT, then the non linearity of the current source may have very little effect on the non linearity already present in the signal, since the current source load on the tube may be over ten times the value of the OPT primary load. So a 25% change in actual current source load may make almost no difference. We would have to know what your schematic is before offering more accurate advice. Patrick Turner. Thanks. |
#3
|
|||
|
|||
We would have to know what your schematic is before offering
more accurate advice. Right, well, I didn't describe it before as it's difficult to explain, and the application is nonstandard. I've posted about this here quite a while ago, so if it seems familiar, that's why. The tube is driving a glow discharge (Hill Plasmatronic speakers derivative, except he used TV sweep tubes, one per cathode; also, I'm using a complex cathode geometry, microhollow cathode discharges feeding the main discharge, to avoid the need for helium and provide discharge stability). Hill's patent is 4,219,705, with fig.8 being essentially what the actual Plasmatronics used (the fan just cools the heatsink's back and doesn't blow through). MHCD sustained discharges are described in various plasma physics papers; Hill didn't use them as they weren't known back then. Single channel description: I've got about 2500 V at the plasma anode (I plan to regulate it, just pi-filtered right now), and the tube is at the plasma cathodes as a current sink, to be fed the appropriately amplified audio signal (class A in the sense that maxium waveform will never cause current to go below a certain value, so it's always conducting). There are potentiometers for fine adjustment from the tube plate to the bottom halves of the five microhollow cathodes, and resistors between the top halves and plasma anode. I haven't actually run the tubes (I'm using salt-water ballast resistor for testing) as the cooling blowers haven't arrived yet. The microhollow electrodes I had made with copper-sandwitched mica, but it erododes too fast. I finally got free samples of sapphire wafer for the insulator, and I'm working on platinum plating some tungsten for the conductors. That should last very nicely. In the meantime I've been thinking of how to get good performance out of the output tetrode. One restriction is that the discharge should drop most of the voltage, otherwise power is just wasted in the tube. Sound generation -- instantaneous change in pressure at any time is proportional to the instantaneous change in power being dissipated in the discharge. A normal glow discharge is constant voltage within a range of current and is why Hill used current drive, but the I-V characteristics of an MHCD discharge are not so simple and vary based on the specific arrangement (they can also be resistive or negatively-resistive (+ an offset of course so there's always a minimal voltage drop across the discharge)). In either of the three cases, current drive works fine, and prevents a problem in the fault case of a glow-to-arc transition (if that happens, the arc extinguishes immediately due to insufficient current, the top half of the MHCDs are biased from the resistors from the plasma anode, and the discharge self-restarts). However, if the operating range encompasses more than one of these regions, then nonlinearity of the discharge itself becomes a problem, so I guess one way to deal with that specifically is to have feedback proportional to power rather than current or voltage. Current through the discharge and voltage drop across it can be sampled and with an analog multiplier gives such a feedback signal, but I hope it wouldn't be needed as it gets too complicated, and the best analog multipliers are only about 0.1% accurate. But before I worry about that, the 4X150 nonlinearity when used as a transconductance amplifier is what I'm trying to figure out now. |
#4
|
|||
|
|||
Prune wrote: We would have to know what your schematic is before offering more accurate advice. Right, well, I didn't describe it before as it's difficult to explain, and the application is nonstandard. I've posted about this here quite a while ago, so if it seems familiar, that's why. The tube is driving a glow discharge (Hill Plasmatronic speakers derivative, except he used TV sweep tubes, one per cathode; also, I'm using a complex cathode geometry, microhollow cathode discharges feeding the main discharge, to avoid the need for helium and provide discharge stability). Hill's patent is 4,219,705, with fig.8 being essentially what the actual Plasmatronics used (the fan just cools the heatsink's back and doesn't blow through). MHCD sustained discharges are described in various plasma physics papers; Hill didn't use them as they weren't known back then. I remain entirely un-informed....... Single channel description: I've got about 2500 V at the plasma anode (I plan to regulate it, just pi-filtered right now), and the tube is at the plasma cathodes as a current sink, to be fed the appropriately amplified audio signal (class A in the sense that maxium waveform will never cause current to go below a certain value, so it's always conducting). There are potentiometers for fine adjustment from the tube plate to the bottom halves of the five microhollow cathodes, and resistors between the top halves and plasma anode. I haven't actually run the tubes (I'm using salt-water ballast resistor for testing) as the cooling blowers haven't arrived yet. The microhollow electrodes I had made with copper-sandwitched mica, but it erododes too fast. I finally got free samples of sapphire wafer for the insulator, and I'm working on platinum plating some tungsten for the conductors. That should last very nicely. In the meantime I've been thinking of how to get good performance out of the output tetrode. One restriction is that the discharge should drop most of the voltage, otherwise power is just wasted in the tube. Sound generation -- instantaneous change in pressure at any time is proportional to the instantaneous change in power being dissipated in the discharge. A normal glow discharge is constant voltage within a range of current and is why Hill used current drive, but the I-V characteristics of an MHCD discharge are not so simple and vary based on the specific arrangement (they can also be resistive or negatively-resistive (+ an offset of course so there's always a minimal voltage drop across the discharge)). In either of the three cases, current drive works fine, and prevents a problem in the fault case of a glow-to-arc transition (if that happens, the arc extinguishes immediately due to insufficient current, the top half of the MHCDs are biased from the resistors from the plasma anode, and the discharge self-restarts). However, if the operating range encompasses more than one of these regions, then nonlinearity of the discharge itself becomes a problem, so I guess one way to deal with that specifically is to have feedback proportional to power rather than current or voltage. Current through the discharge and voltage drop across it can be sampled and with an analog multiplier gives such a feedback signal, but I hope it wouldn't be needed as it gets too complicated, and the best analog multipliers are only about 0.1% accurate. But before I worry about that, the 4X150 nonlinearity when used as a transconductance amplifier is what I'm trying to figure out now. I do wish you all the very best of luck, but remain in a mood of dejected despair and am a complete and utter loss to be able to help you. I bet it all looks spectacular when its turned on. Patrick Turner. |
#5
|
|||
|
|||
Patrick Turner wrote in
: I remain entirely un-informed....... I know I'm not good at describing things. I'll sketch a diagram later today. I bet it all looks spectacular when its turned on. Doesn't matter very much what it looks like, it's about what it sounds like. There was a website (no longer functional) of a Plasmatronics owner who took measurements. The pulse response and waterfall plot looked unlike anything else, much cleaner than any tweeter I'd seen. And the frequency response was very flat. I wish I had saved the graphs on my PC. I completely forgot about my plans to build electrostatic speakers, I knew I had to go for a non-solid sound driver. |
#6
|
|||
|
|||
Hi,
I have an 8-page Eimac data sheet for the 4X150A and D if that would help. It has two full-page plots of transfer characteristics for screen voltages of 250V and 350V. I can xerox and mail this, for the cost of postage. Email me, adouglas at gis.net Alan |
#7
|
|||
|
|||
On Fri, 03 Jun 2005 13:17:24 GMT, Prune
wrote: all clipped for amazing coolness and complexity Are you working for a scientific understanding, or for a working solution? If the former, cool, good fortune, but nobody here will have a clue. If the latter, maybe you could do an impulse (or the mathematical equivalent) response test of a mockup and apply the response to your signal path. Big fun, Chris Hornbeck "He thought so little they rewarded he, By making him the ruler of the Queen's Navy". |
#8
|
|||
|
|||
Well, I was going to say both, but...
Chris Hornbeck wrote in : On Fri, 03 Jun 2005 13:17:24 GMT, Prune wrote: all clipped for amazing coolness and complexity Are you working for a scientific understanding, or for a working solution? If the former, cool, good fortune, but nobody here will have a clue. If the latter, maybe you could do an impulse (or the mathematical equivalent) response test of a mockup and apply the response to your signal path. Big fun, Chris Hornbeck "He thought so little they rewarded he, By making him the ruler of the Queen's Navy". |
#9
|
|||
|
|||
Actually I came across this:
http://www.webace.com.au/~electron/tubes/FIG23.jpg What sucks is that if I wanted to be in a linear region, it means a positive grid, and so grid current, which makes driving properly harder, not to mention that my tubes have gold-plated grid and grid current above 10 mA is a no no (I was thus told by Eimac tech, as I have special (read: obscure) audio versions of the 4X150A, but I can't complain given I paid $10 for a perfect condition NOS pair, and ironically the used air system sockets cost far more than the tubes themselves). I was thinking about ultralinear circuits, where the screen voltage is modulated (antiphase with control signal; doesn't necessarily have to be from an output transformer; e.g. OTL ultralinear circuits at tubecad journal). But that won't work for me, as it makes the tube more like a triode, away from my application as a current sink. If the modulation is instead in phase with the control signal, I'm not sure what it will do...I think it's kind of like positive feedback, just as the way the ultralinear connection is somewhat like a local NFB. In that graph, if g1 and g2 are interchanged, one would again obtain S- shaped curves. If the overall plate current function is (approximately at least) linearly separable, as in (for fixed Ea as this is a current sink) Ia = f(Eg1, Eg2) being expressable as Ia = f1(Eg1) + f2(Eg2) *, then modulating the screen as well should be able to be set up as to give a straight curve (if it's not separable, then the relationship between the needed signals is nonlinear and so impractical). Having considered this, it just doesn't look like any possible combination is possible anywhere near the operating range I'm looking for. I guess predistortion is the reasonable option I have, and then Hawkford EC and/or NFB. *of course E is potential, I current, a anode, g1/g2 the grids Alan Douglas adouglasatgis.net wrote in : Hi, I have an 8-page Eimac data sheet for the 4X150A and D if that would help. It has two full-page plots of transfer characteristics for screen voltages of 250V and 350V. I can xerox and mail this, for the cost of postage. Email me, adouglas at gis.net Alan |
#10
|
|||
|
|||
Hi,
I misread the Eimac plots; they're actually plate voltage vs. grid voltage, for plate currents from 10mA to 1.2A. You'd have to replot point by point to get transfer characteristics. A couple of operating points I checked, don't match this other plot that you referenced. I don't know where that one came from, or how they could run at up to 8A plate current. The Eimac sheet doesn't rate it beyond 0.25A. 73, Alan |
#11
|
|||
|
|||
"Prune" wrote
I was thinking about ultralinear circuits, where the screen voltage is modulated (antiphase with control signal; doesn't necessarily have to be from an output transformer; e.g. OTL ultralinear circuits at tubecad journal). But that won't work for me, as it makes the tube more like a triode, away from my application as a current sink. If the modulation is instead in phase with the control signal, I'm not sure what it will do...I think it's kind of like positive feedback, just as the way the ultralinear connection is somewhat like a local NFB. The UL connection is voltage derived, voltage applied feedback, which reduces Ra, which is what you don't want. If you use *current derived*, voltage applied negative feedback, the effect should tend to linearise the current, and hence increase effective Ra. However, assuming you are feeding a varying impedance which you cannot use as a sense resistance, and that you don't want the necessary additional sense resistance to be huge, you will need to amplify the feedback signal before applying it to the screen. Easier to use as pentode and arrange feedback to the grid. You need to be careful about screen current, which has a max rating, and which tends to rise sharply as anode current drops at low Vak. Under those circumstances, feedback trying to raise the screen voltage to counter the fall in cathode (or anode) current would put the screens in double jeopardy. cheers, Ian |
#12
|
|||
|
|||
"Ian Iveson" wrote in news:F61pe.106842
: If you use *current derived*, voltage applied negative feedback, the effect should tend to linearise the current, and hence increase effective Ra. Indeed, I can take care of the plasma load I-V nonlinearities by multiplying this current feedback with the voltage drop across the discharge. The problem with that is that I've not seen an analog multiplier more accurate than 0.1%. Can Hawksford error correction be adapted to this kind of output stage? The problem is that for a transconductance amplifier it would have to be a current error, and this stage has a big current gain. I was thinking about the parallel between this and typical class A, single ended amplifiers. They usually have a voltage output stage with a load parallel to the output device/ground column, supplied by a constant current source. Here I have a current output stage with a load in series with the output device, and with what should be a constant voltage source on the other end. Now, I've not built a regulator for the power supply yet (currently CLCRC filtered, handwound choke, godam heavy bulky capacitors..), but I have some appropriate high voltage solid state devices. I remember one discussion at tubecad.org where Broskie found that removing capacitance AFTER the regulator improved sound. I wonder about that. One advantage I can see in my case of not having a capacitance at the voltage side is that there won't be an out-of-phase voltage variation on the high side of the load, but if the regulator is very low output impedance, it shouldn't matter. I'm also thinking of doing a capacitance multiplier instead of fixed-reference regulator, so I can both have a small dropout and avoid problems from long-term mains voltage variation. Simulation of my current power supply shows ripple in the tens of millivolts, but say with a 500 Hz (my crossover frequency) load drawing DC offset 175 mA with 150 mA amplitude, times two channels, I'm getting about two volts of variation at the output. (BTW, what is the actual value of the primary inductance of a 1500 W transformer, is 50 mH a good estimate?) |
#13
|
|||
|
|||
Oh crap, forgot to mention it's a microwave oven transformer, rebuilt core
(interleaved laminations), 20% extra turns added to get rid of magnetic leakage (those cores are undersized). About 1 turn/volt initially, 1.2 now. About 16 lbs. 2100 VAC original output, I added also to the secondary winding to keep the voltage. I talked about it in this group some months ago. In the simulation I'm using 65 mH primary and 20 H secondary. Do these sound like reasonable guesses, or am I way off? |