Electrostatic deflection CRTs in television sets.

[Neutrino]

The Sylvania Reference Data for the 6SN7GT lists ratings for operation as a Deflection Oscillator:
http://www.r-type.org/pdfs/6sn7gt.pdf
I do not know whether it is of any interest but my estimate of the average current through each transformer winding for a ΔV of 450 V across each 0.001 µF capacitor during the sweep time of 54 µS, when the triode is not conducting, is about 8.3 mA, given by I=CΔV/Δt. To have a reasonably linear sweep the inductance of the transformer would need to be sufficient to maintain a reasonably constant current during the sweep. This would also be the average of the current through the anode during t1 (scan - zero current) and t2 (flyback - high current) .
Similarly the average current through each 0.001 µF capacitor during flyback is estimated at more than 37.5 mA to get a ΔV of 450 V in a Δt of less than 12 µS. The peak current at the start of flyback would be higher.
The anode current during flyback would be the sum of the current through the 0.001 µF capacitor and transformer.
They do seem to push the envelope of this little triode.

[Argus25]

It would be a very difficult proposition with a circuit like this to make any close estimates of tube load/stress etc without this particular circuit in front of them for analysis with the scope. The DC axis of the waveforms are surprising and interesting and the blocking oscillator primary waveform complicates the plate to cathode voltage. I have attached some scope images which will help.

Neutrino's calculation of the 37.5 mA current required to alter the capacitor voltages by 450V during the flyback time is quite correct, but well within spec, used as an H deflection oscillator the 6SN7GTB is rated for a peak cathode current of 300mA.

It also appears that the 6SN7GT has a 3.5 watt plate dissipation, but the GTB 5 watts. The circuit was specified for a GT type though.
The tube in this circuit has a low power dissipation because the only time plate current occurs is during flyback, the 6SN7 is cut off during scan time. To calculate the tube's dissipation then, very roughly, it is the plate current x the average of the plate to cathode voltage during flyback x the flyback duty cycle (about 18%). However the dissipation is even less because for about half of the flyback time, it turns out, the plate to cathode voltage is very very low, so the actual duty where there is any significant voltage across the conducting tube only about 9% to 10% of the time.

The images attached are of the cathode voltage and lower 0.001uF tuning capacitor, the voltage across the upper 0.001uF tuning capacitor, and a 2 trace recording of the anode voltage and the cathode voltage.
The width control was set to give 400v amplitude of the sawtooth to keep it on the scope screen with 50V/cm. Firstly looking at the cathode voltage the waveform straddles the ground axis exactly as expected.
Looking at the waveform on the upper 0.001uF tuning capacitor the DC axis is interesting. The voltage here falls 62 volts lower than ground. On the other hand the plate voltage, due to the voltage on the primary side of the blocking oscillator transformer, swings positive halfway through flyback, enabling plate current even though the cathode voltage is climbing toward +200V.

The synergy of this feature is more of the brilliance and elegance of the design.

The other photo is a dual trace recording of the plate and cathode voltage, I had to decrease the amplitude to fit it on the screen. However correcting the scale for this, a rough estimate of the plate-cathode voltage over about the first half of the flyback time (because it is negligible over that second half of the flyback time) is about 570V (but it would be less on the average due to the wave shapes).

So absolute worst case average plate power dissipation = 570V x 37.5mA x 10% = 2.1 watts, well in spec for the 6SN7. Due to the actual wave shapes it is probably more like 1.5 watts.

Since the circuit does not deliver any significant output power to a load ( less than 100mW to the def plate load resistors) I will measure the power consumption of it from the HT supply, and see how that compares, will post later.

[Argus25]

It turns out the tube plate dissipation estimate above was extremely close.

With a 223V supply the HT current is 10mA average with 400V sawtooths being produced.

The voltage drop across the 25k pot (when at this setting is 4.7k to attain 400V sawtooths with a 223V supply) was 47volts, so the power dissipation in the pot is around 0.47w. Therefore the 6SN7 tube dissipation (ignoring transformer losses so as to over-estimate it) is close to:

223 x 0.01 - 0.48 = 1.75 watts Well within the ratings for a 6SN7.

(Its not necessary to subtract the energy delivered to the load which is 4.7 meg deflection plate tie resistors in the set, as the rms value of a 400v pp or 200V peak sawtooth wave is 200/ root3 = 115v rms, so the power delivered to the deflection plate resistors is only about 3mW each, as pointed out before the power required for electrostatic deflection is very low unlike magnetic deflection)

So in summary:

1) The 6sn7 doesn't have to be a "power triode" for this application, its not a power application.

2) The peak cathode voltage is 200V peak when the circuit set for 400v pp sawtooths which give the correct width in the TV set for the image, this is within ratings, albeit just.

3) The peak cathode current, correctly estimated at around 37.5mA, is well under the max rating of 300mA.

4) The power(plate dissipation) is only about 1.75 watts or less, well under the 3.5 watts max rating, so a lot is not being asked of this little triode.

Are these findings really surprising ? I don't think so. It would be highly unlikely that the genius with the imagination and technical acumen who designed this circuit would would have put the 6SN7 in peril.The designer was well ahead of the normal curve. But it is the case that when we are confronted by such an astonishingly creative and interesting circuit (that appears to do the impossible such as creating high voltage linear sawtooth waves from a low voltage HT and few parts) that it is tempting to try to find something wrong with it.

(nobody asked what the other circuit was that was cleverer than this one)

[dominicbeesley]

I've been looking at deflection oscillators for a mini-crt type application and was thinking I would have to go with a 3 or 4 valve solution. I need roughly 150-200V p-p on each plate.

Do you have any idea of the rough inductances of the windings. I'm tempted to have a brief skive this afternoon and see what transformers I have lying around. If I could make this small then a single sub-miniature valve and a couple of small psu driver transformers might well do the trick!

(I've tried doing some rough calcs 1nf @ 2khz gives me 6H per winding for the output transformer which seems a bit steep to me!)

[Argus25]

Hi, dominicbeesley, a good question:

When I was given a stripped Admiral 19A11S chasis back in the 1980's I was confronted with the fact I would have to re-manufacture the transformers. This is how I found out about this unique circuit, I was forced to figure it out to re-build it from scratch. I think people in the TV service sector must have fixed these sets many times back in the '50's, but not paid much attention to the underlying design and the genius of it.

I had no idea of the core sizes in the original. I decided to use two ferrite cores from some very small transistor TV line output transformers of the sort where two U shaped halves are clamped together, with small threaded rods that were about 35mm apart, The core cross section is probably in the order of 8 to 10mm. Two bobbins were fitted on each side, about 15mm high.

For the blocking oscillator its normally a 1:4 ratio, the DC resistance of the primary on my hand made one is about 18 Ohms, the secondary around 47 ohms. I think for that you could just use an "off the shelf" H blocking osc transformer.

For the tuned circuit, its an interesting analysis what frequency it resonates at with two identical tuned circuits tightly coupled on the same core, electrically, from the resonant frequency perspective, is behaves as though it is one winding tuned by twice the capacitance. Another method to get to the same answer is to imagine the two resonant circuits connected in series on the same core(one link) otherwise isolated.Then disconnect the link between the junction of the capacitors and the junction of the inductors. Then you have a coil of twice the length (4 times the inductance) tuned by half the capacitance value, so the resonant frequency calculation is the same.

In other words to calculate the inductance you must double the capacitor value for your calculation and you will find each winding should measure about 3H.

Winding each bobbin on my substitute transformers to achieve this they ended up with a DC resistance on each bobbin of about 140 Ohms. I didn't have an inductance meter back then and had to wing it by iteration. Since the TV is on my desk right now I disconnected the transformer and measured it, it came out at 4.29H per winding, which wasn't too bad. Obviously if the inductance is too low, the waveform will flatten down toward the end of scan, as too much of the sine wave cycle will be used.

Later I found out that the original transformers from Admiral & Motorola for this application were smaller in size than the ones I had made.

[Argus25]

I checked the Transformers in the Motorola VT71 set. Not the same as my Admiral 19A11S , but the transformers in my set as noted were made from scratch.

In the VT71 the H blocking oscillator is a very small transformer. It looks like a coil wound on a 1/4" diameter 1 inch long ferrite slug, or it could be iron. The outer diameter of the winding is about 1 inch and its only about 3/4 inch tall and coated in some sort of white resin looking like a fat choke but with 4 wires exiting its ends rather than two.. The secondary has a DC resistance of about 160 ohms and the primary about 60 Ohms, I'm pretty sure it would be a 1:4 or similar ratio, not too critical as long as the primary winding resistance is not too high so as to slow the charging/discharging of the tuning caps and slow the flyback.

In the VT71 the resonant transformer is a small iron cored(standard E-I looking lamination stack) looking just like a typical 2 or 3W sized audio transformer with a 1/2 square core cross sectional area (as noted I wound mine on a small ferrite core) One winding has a resistance of close to 250 ohms, the other 200 ohms (The VT71 manual says 260 and 210).

The inductance was a surprise at 1.62H for each winding. Also the VT71 manual says the tuning caps are 900pF and 680pF not both 1000pF as in the Admiral circuit. Perhaps they balanced the circuit tuning for the additional load capacitance on the 680pF connection to the other circuit components of the cathode circuit of the blocking osc.

So my home made transformer may have had a higher inductance than it required, but it works perfectly.

One other trick worth mentioning in this generation of TV set where necessity seems to be the mother of invention; To get the high HT voltage to run the vertical scan output stage (a 6SL7 with 4.7meg anode load resistors) some bright spark realized that since the alternating currents in the two anode load resistors were equal and opposite the total vertical output stage current would be a constant. So they dropped the vertical output stage circuit into the ground leg of the EHT bleeder chain and acquired the high voltage supply that way. It creates about a 1000V HT rail there.

[dominicbeesley]

Thanks Argus, do you have any rough measurements of the blocking oscillator inductance? I have to admit that I have never managed to get my head around how to calculate the required number of turns, the only one I ever got to work was a frame oscillator and I just kept trying bigger and bigger transformers...the one that worked was about 10lb!

For the output trans - it looks like my rough and ready approximation wasn't far out then - I wasn't sure whether to calculate for each leg or add them together but it looks like something >1.5H should do the trick - I'll go for 3H. I'll have a look in my junk box for some formers, I'd like to keep them small. I'm using Soviet pencil valves and large transformers look a bit incongruous! The limiting factor on size will probably be what is the finest wire I have that wont' snap on my meccano and hope coil winder.

I doubt I'll get into the workshop today but I may spend an hour knocking up a simulation this afternoon...

Thanks

D

that wont' snap on my meccano and hope coil winder

Handy hint, take the wire off axially not unwinding it off the drum. Most good drums are smooth enough for this.

[Argus25]

D:

"knocking up a simulation this afternoon"

I sincerely hope not, if you are talking about something like a Spice simulation ? you might not be but..... The blocking oscillator is one of the poorest circuit propositions to run in a simulator. There is the self winding capacitance and the transformer core properties to consider, but moreover it is a regenerative circuit.

When plate current flows the grid is driven positive and grid current flows too so that has to be simulated too. This develops a grid leak bias. So the oscillations of the regenerative effect are only seen for about 1 cycle because on the second half of the cycle the tube is cut off because of the RC time constant of the grid's input network and grid current charging of the coupling capacitor.The actual grid current is dependent on the exact type of tube.

During this regenerative phase, lasting just one cycle, all of the complex properties of the transformer come into play. The circuit initialization also relies on an amount of initial plate current.

At the start of flyback the plate voltage initially swings negative and then positive on the next half cycle, and that frequency is determined not only by the transformer's inductance but by the self capacitance of the transformer windings, especially the secondary winding. The self capacitance is not insignificant in layer wound transformers. The damping on the primary by the tube also affects the resonant frequency.

The winding L and self C values though would relate to roughly the resonant frequency of the secondary inductance combined with its winding capacitance and that period needs to be less than or equal to about twice the flyback time. So an inductance value alone won't help you at all with a simulation......if that's what you were intending.

In this case it is better to get an actual blocking osc transformer, or attempt to wind a 1:4 ratio one on a small core and start to experiment.

The simulation for this sort of thing will lead you "up a garden path" in my experience. This is not the only type of circuit that gives problems in simulators, another is the classic is the multi-vibrator that won't start in a simulator, but does in reality.

Perhaps if someone has a blocking H osc transformer on hand they could measure the inductance of the secondary, then you could approximate the self capacitance of the secondary as a value that would oscillate at a period of twice the flyback time, then estimate the primary self capacity to be a 1/4 of the secondary's self capacitance. The primary inductance for a 1:4 ratio transformer would be 1/16 of the secondary. That would go some way toward a reasonable simulation if the tube properties relating to grid current were close enough.

another is the classic is the multi-vibrator that won't start in a simulator, but does in reality

I found this (of course demonstrating ltSpice!) and fluffed around the problem by saying simulators are perfect, no noise (etc.). So I put a microvolt of noise (a sine wave for simplicity, I couldn't find a noise source in ltSpice) in one on the base connexions, then it started. So did changing one resistor a weeny bit. I have just tried the latest version of ltSpice and it started after 570ms, they must have added noise somewhere.

all of the complex properties of the transformer come into play

To approximate saturation I find the voltage controlled switch device quite useful.

[dominicbeesley]

Thanks both,

I have had trouble in LTSpice getting oscillators to start, like MM I add a noise source somewhere in the circuit to give it a tickle, or sometimes just step the power supply.

Yes, I know there are gotchas with simulations, any simulation, but I am not prepared to spend a week and waste miles of copper wire for the sake of having a go in Spice and come up with a rough number for the inductance of the blocking transformer. I got something that looked to work in LTSpice after a bit of fiddling - see attached pictures. This does show the usual spice tendency to have ridiculously large spikes and other "features" but has at least given me a starting point...whether reality concurrs is the fun part!

D

Another advantage of simulation programmes is that if you save it it's documented unlike scraps of paper!

[Argus25]

D:

Most likely the conduction time of the tube is too short and the energy delivery to the resonant circuit too low by the look of the waveform voltages. Look at the plate waveform on spice and add capacitance across the blocking transformer secondary until the period of the first cycle of plate oscillation is about twice the flyback period, about 20uS would do. It might be easier to run the blocking osc on its own first with a current limiting resistor in series instead of the resonant circuit.

With these blocking osc transformers, they are made so that the self capacity does the job. As you know there could be a whole range of 1:4 ratio transformers with low inductances and low self winding capacity and few turns up to a larger physical size, with much higher inductance and self winding capacity. There is a correct proportion for the task.