How do the diodes protect the OT?

[Alan0354]

I don't understand how the diodes D301 and D302 protect the OT. in this Prosonic schematic:

From my understanding, when the speaker is not connected, the primary of the OT show high impedance and is inductive. There should be no problem when the power tube pull current, the plate just goes low to the point the tube saturate at about +100V or so. The problem is when the tube turns off and stop sinking current, then the primary being inductive, will pull the plate to +HV.....say higher than +1KV. This is just like relay coil that need flyback diode to prevent the low side of the coil from shooting up when the current is switch off.

I understand the relay coil is not exactly representing the primary of the OT, the plate of the power tube do swing above the CT of the OT under normal operation because the opposite tube pull down and cause the plate of the first tube to fly above the CT. So the diode cannot be connected to the CT like the relay coil.

But how does the diode shown in the image works. The only way I can see it works is if the diode is a zener diode or a TVS that starting to conduct and clamp the plate from going beyond 1KV. But I understand those are not zeners.


This is my wild guess:

The only thing I can think of is say when V8 turns off, plate of V8 fly high, this will cause the plate of the V7 to swing low ( because of the two terminal of the primary of the OT always swing opposite due to the winding). When plate of V7 swing below ground, D302 turns on and prevent the plate of V7 from going any lower. This in turn prevent the plate of V8 from flying any higher.

But this is indirect clamping, is it effective?

Thanks

Alan

[jazbo8]

The diode from the output anode to ground is used to limit the peak voltage swing to the B+ supply, since the diode conducts only when either side of the primary windings tries to swing to a negative voltage caused by the back EMF (when the OPT is unloaded), this as you said would result in a very high "flyback" voltage, so the diode is there to prevents the OPT's insulation from potential demage casued by the HV, which often shows up as arcing between the OPT's windings, tube sockets/pins.

[Alan0354]

Thanks for the reply.

So my guessing is correct? That one side tries to swing high, the other side swing low until it reach below ground and turn on the diode and prevent from going any lower. THIS IN TURN prevent the original side from swinging any higher.

This is an indirect clamping. That the diode on one side is used to clamp the other side instead of direct clamping? It relies on the winding of the transformer to stop the original side from swinging any higher when the diode of the low side turns on. That really works?

So if I use two 1N4007 in series, that will be more than enough to protect the OT from arcing. 1N4007 should be enough as it's only going to clamp a few hundred mA max and power dissipation is very small.

[trobbins]

With an overdriven cranked squarish wave output, the point of switching from one valve conducting to the other causes large dI/dt in each half-winding of the OT primary. Each half-winding has leakage inductance, so when the conducting valve current ramps up from zero to max, it then abruptly stops increasing when saturation/grid-current conduction is reached. The high dI/dt and leakage inductance then pushes the plate voltage below zero. Loudthud presented some great oscilloscope images showing that.

The hassle with the diodes, is that they don't inherently protect the other half winding, which effectively remains unclamped, and hence not the best form of over-voltage protection for an OT.

[Alan0354]

With an overdriven cranked squarish wave output, the point of switching from one valve conducting to the other causes large dI/dt in each half-winding of the OT primary. Each half-winding has leakage inductance, so when the conducting valve current ramps up from zero to max, it then abruptly stops increasing when saturation/grid-current conduction is reached. The high dI/dt and leakage inductance then pushes the plate voltage below zero. Loudthud presented some great oscilloscope images showing that.

The hassle with the diodes, is that they don't inherently protect the other half winding, which effectively remains unclamped, and hence not the best form of over-voltage protection for an OT.

I think you are talking about a different situation. I am referring to an amp with no load. You are talking about driving an amp with speaker load to clipping or a square wave input.

Are you talking about normal situation with speaker load, then you OD and the tube behaving like a switch. When one side of the tube goes from off to saturation(0 grid), the dI/dt goes to zero abruptly? As long as you have a load on the secondary, the impedance is finite at the primary. I had seen spikes, but it's not bad. I don't think you need protection diode for an amp that is driven to square wave.

What I am talking about is if people forget to connect the speaker, the OT behave more like an inductor of a few henry, large amount of energy can be stored in the transformer. When you cut off the current, the inductor want to keep the current going by ramping up the voltage, that's when you create an arc.


In another way to put it..............If you have a matching load at the secondary, the transformer almost behave like a transformer that transform impedance. Leakage inductance is low, very little energy is stored. BUT if you have an open circuit secondary, the transformer is operating way out of it's normal parameter, the primary behave more like an inductor. The inductance is much higher than the normal leakage inductance. Large amount of energy can be stored in the transformer, that will be transform into high voltage spike and self destruct.

[J M Fahey]

The hassle with the diodes, is that they don't inherently protect the other half winding, which effectively remains unclamped, and hence not the best form of over-voltage protection for an OT.

Sorry but they do protect the other half.
That's the point.

Imagine the transformer as a kid's seesaw.



Primary is the seesaw,+B is the centerpoint and kids are plates.

At idle, all 3 points (CT and both plates) are at the same level=+B .

At saturation, one kid is at ground level (0V) and the other is twice as high as before (2X +B)

Suppose below each kid there is a hole, so he can swing *below* ground if "something" pushes him strong enough ... which automatically means the other kid will swing above 2X +B

Usually, and specially with a load absorbing energy on the secondary (think the kids are not in free air but submerged in molasses or similar, something which absorbes their energy), even if they have inertia which would make them move beyond Earth and 2X , there is no problem.

But remove load (molasses) and inertia may and will make them overtravel, one below (no big deal in a tube, deadly in a transistor) and the other way too high (arcing or voltage breakdown in a transistor).

Inductance by definition IS inertia, because it "tries to keep current moving the same way"

The diodes are similar to filling those holes below each kid with molasses.

Under normal (loaded) operation , kids/plates do not reach them.

Under lack of load, they move freely and inertia (inductance) raises its head until the one swinging down "hits the molasses", the diode becomes forward biased just with 0.7V below ground and it becomes a short to ground.

Downwards swinging gets brutally absorbed/stopped and by the seesaw action, the other kid/plate gets also brutally stopped not reaching more than 2X +B (plus 1 or 2 diode drops, a couple volts).

Only Real World problem in this idyllic history is that both primary halves are not perfectly coupled (100%) but somewhat less (still in the high 90%) so in fact there is some overshoot (in fact you see it as a leading edge overshoot even when driving resistive loads) but it's quite contained, nothing deadly.

Another considereation is that the unclamped side (the one flying high) will apply that voltage to the clamping diodes.

Although in theory a single 1N4007 should be just fine, that's like playing with matches and gasoline, so in practice 2 in series is fine and 3 in series does not hurt.

By the way, vertical MosFets (IRFxxxx) and some power transistors (TIP14x) already come with built in protective diodes, to cope with inductive loads, the foirmer because of their primary use in switching and the latter because speaker loads are inductive.

Big heavy 15"/18"woofers or subwoofers are specially dangerous.

[Dave H]

As Juan said above it’s a (auto) transformer. In a perfect world the two halves of the primary would be perfectly coupled. The centre tap is at B+ so if one end of the OT tries to fly up to B+ plus 1kV (say) the other end will try to fly down to B+ minus 1kV but it can’t as it is clamped to -0.7V by the diode so the positive side can only fly up to 2 x B+. Of course in the real world it won't work as well as that.

In the real world the diodes could fail short circuit which is also bad for the OT. I think I would prefer to always have a load of a few hundred ohms on the OT secondary.

[Leo_Gnardo]

Juan mentioned big woofers as a factor. Let's expand on that. I've stared at the scope screen for decades and never seen overshoot "spikes" when driving a load resistor. Because the inductance is negligible, that's one reason. Here's another: speakers don't move instantly when "pushed" by the amp. The cone/former/voicecoil assembly has mass and takes a little bit of time to get going. Also to stop. When the spring forces in the speaker spider and surround pull the coil back to rest position, there's a conductive coil moving through a magnetic field - that's a generator. The current developed by that generator goes through the OT secondary, multiplied by turns ratio*, it shows up on the primary, and can develop "spikes" of some 3000 volts. Big woofers as Juan says are especially troublesome because there's so much mass in the cone/former/vc. Not many popular output tubes are rated for 3000V, and OT insulation usually tests out to 2500V. It's a miracle of "tube toughness" that they put up with this abuse.

There's also the matter of setting up for arcing at output tube bases due to spiking. I just solved a case 2 days ago where a customer dimed his Champ and got a good arc going between plate and filament pins. On repair I added "anti-flyback diodes" 3 series 1N4007 and he shouldnt have this problem again. Even in a lowly Champ, it's a problem. Just bought a pack of 1000 '4007's so I'm set to go for a good long time.

Some amps like MusicMan had spike suppressor diodes in place early 70's. Ken Fisher brought this "fix" to our attention in the Trainwreck Pages so those paying attention late 70's - 1980 and on were twigged on to the technique, saving countless tubes and OT's. Thank you Saint Ken!

--- --- ---

Or is that turns ratio squared? Yikes, even worse. Experts, what's the answer?

[dai h.]

I think I would prefer to always have a load of a few hundred ohms on the OT secondary.

would I be correct in thinking the Vox (protection) scheme (of added 250R0 5W across 8ohm output? IIRC) is because the combination of no feedback (in the output section) plus use of pentode outputs makes things especially bad (high voltages) when open load?

[trobbins]

Yes I was talking about a loaded secondary. An unloaded secondary will cause the plate voltage to hit the end-stops with very little grid drive. If the plate-cathode was a switch then the primary current would certainly ramp up, eg. 500V B+ and 50H half winding inductance would ramp plate current by 100mA for a 50Hz signal. But the plate-cathode is not a switch, and would follow the plate curves.

Leakage inductance and high dI/dt is one form of plate voltage overshoot. If leakage was down at 100mH, and 100mA tube current was turned off in 10us,then you have 1kV. This is separate from standard turns-ratio transformer voltage 'see-sawing' action around the B+ CT point, as Juan describes.

As Leo pointed out, a speaker load pushes the loadline all over the place, and that interaction appears to have the capability of pushing plate voltages significantly outside of the 0V to 2xB+ range.

The diode clamping of one valve plate helps, but its control of the other half-winding voltage is 'loose' as it relies on a perfect transformer model. To my mind, a MOV across each primary half winding is simpler and more appropriate to the application as it directly suppresses the peak voltage level at the winding terminals.

[Dave H]

Or is that turns ratio squared? Yikes, even worse. Experts, what's the answer?

I could be wrong (I'm no expert) but I think the coil moving through the magnetic field generates an EMF of -dB/dt and it's that which is multiplied by the turns ratio (not ^2). If you square wave it dB/dt x the turns ratio could be very high.

[Alan0354]

Sorry but they do protect the other half.
That's the point.

Imagine the transformer as a kid's seesaw.



Primary is the seesaw,+B is the centerpoint and kids are plates.

At idle, all 3 points (CT and both plates) are at the same level=+B .

At saturation, one kid is at ground level (0V) and the other is twice as high as before (2X +B)

Suppose below each kid there is a hole, so he can swing *below* ground if "something" pushes him strong enough ... which automatically means the other kid will swing above 2X +B

Usually, and specially with a load absorbing energy on the secondary (think the kids are not in free air but submerged in molasses or similar, something which absorbes their energy), even if they have inertia which would make them move beyond Earth and 2X , there is no problem.

But remove load (molasses) and inertia may and will make them overtravel, one below (no big deal in a tube, deadly in a transistor) and the other way too high (arcing or voltage breakdown in a transistor).

Inductance by definition IS inertia, because it "tries to keep current moving the same way"

The diodes are similar to filling those holes below each kid with molasses.

Under normal (loaded) operation , kids/plates do not reach them.

Under lack of load, they move freely and inertia (inductance) raises its head until the one swinging down "hits the molasses", the diode becomes forward biased just with 0.7V below ground and it becomes a short to ground.

Downwards swinging gets brutally absorbed/stopped and by the seesaw action, the other kid/plate gets also brutally stopped not reaching more than 2X +B (plus 1 or 2 diode drops, a couple volts).

Only Real World problem in this idyllic history is that both primary halves are not perfectly coupled (100%) but somewhat less (still in the high 90%) so in fact there is some overshoot (in fact you see it as a leading edge overshoot even when driving resistive loads) but it's quite contained, nothing deadly.

Another considereation is that the unclamped side (the one flying high) will apply that voltage to the clamping diodes.

Although in theory a single 1N4007 should be just fine, that's like playing with matches and gasoline, so in practice 2 in series is fine and 3 in series does not hurt.

By the way, vertical MosFets (IRFxxxx) and some power transistors (TIP14x) already come with built in protective diodes, to cope with inductive loads, the foirmer because of their primary use in switching and the latter because speaker loads are inductive.

Big heavy 15"/18"woofers or subwoofers are specially dangerous.

You just have a much better and simpler way to explain this, particularly the hole on the ground and fill the hole. The result is the same, but I have a slightly different take.

I was thinking the mechanism is the high going side is the one that want to swing the voltage up when the tube turn off. This is because the OT at open secondary being an inductor, wants to keep the current going ( initia as you put it) by raising the voltage. That cause the other side to swing low. The diode clamp it to -0.7V. In turn this clamp the high going side to 2 times the +B plus 0.7V.

But this is potato and potarto. Close enough.

[Alan0354]

Yes I was talking about a loaded secondary. An unloaded secondary will cause the plate voltage to hit the end-stops with very little grid drive. If the plate-cathode was a switch then the primary current would certainly ramp up, eg. 500V B+ and 50H half winding inductance would ramp plate current by 100mA for a 50Hz signal. But the plate-cathode is not a switch, and would follow the plate curves.

Leakage inductance and high dI/dt is one form of plate voltage overshoot. If leakage was down at 100mH, and 100mA tube current was turned off in 10us,then you have 1kV. This is separate from standard turns-ratio transformer voltage 'see-sawing' action around the B+ CT point, as Juan describes.

As Leo pointed out, a speaker load pushes the loadline all over the place, and that interaction appears to have the capability of pushing plate voltages significantly outside of the 0V to 2xB+ range.

The diode clamping of one valve plate helps, but its control of the other half-winding voltage is 'loose' as it relies on a perfect transformer model. To my mind, a MOV across each primary half winding is simpler and more appropriate to the application as it directly suppresses the peak voltage level at the winding terminals.

That's exactly my question, in the existing circuits, you are relying on the transformer behave as transformer, that clamping the low side to -0.7V will stop the high going swing of the high side.

I was thinking of using TVS ( MOV) P6KExx or 1.5KExx trans sob. But problem is they are only up to about 450V, you have to stack like 3 of them to clamp. Also power dissipation is very high using high side clamping. Say if you have 100mA, swinging to 1KV, that will be W=0.1X1000=100W. Trans sob is good for arcing situation when it arc once a while. They are designed to take a lot of "instantaneous" surge......that happens once a while. In this case, it is repeating like 50Hz!!! I'm afraid this will not work. Even the 1.5KExx are the size of a 5W diode, 3 of them is only 15W. You fry the diodes very soon.

[Alan0354]

I could be wrong (I'm no expert) but I think the coil moving through the magnetic field generates an EMF of -dB/dt and it's that which is multiplied by the turns ratio (not ^2). If you square wave it dB/dt x the turns ratio could be very high.

I am no expert on inductors, I think back EMF control the impedance of the speaker, it's the impedance of the speaker that reflects back to the primary.

[trobbins]

Imho, that outcome is not a valid view. A MOV would be selected that just starts to conduct (ie. 1mA spec) when voltage on half-winding exceeds B+ at least. MOV has spec tolerance, and B+ level would be sized for mains high and idle conditions, and normal resistive-loaded plate voltage doesn't swing +/- B+, so there is at least a significant margin between resistive load type plate voltage swing and MOV min tolerance 1mA conduction level. Each 'MOV' is acting on only one side of winding. A MOV may well be a series string of MOVs - I have a large number of 330VDC min 1mA MOVs, so commonly end up using 2 or even 3 in series per half-winding. The MOV soaks up the energy in the voltage transient that is in excess of the MOV 1mA level voltage rating, not the whole voltage transient energy level [ie. not Vpk2, but rather (Vpk-V)2]. A voltage transient may well reach 500-1kV in excess of 2.B+ on plate in cut-off, but that is with little or no 'loading' (ie. stray capacitance type loading, plus coupling efficiency to other windings). With the voltage transient starting to be loaded above 2.B+, its actual peak voltage will end up being less. A MOV also has a soft clip characteristic, so will progressively take up more energy from the voltage transient if the energy in the transient is sufficient to keep pushing the peak voltage up.

Plots from Loudthud were in these threads.
http://music-electronics-forum.com/t28096/
http://music-electronics-forum.com/t28786/

I have some in:
http://dalmura.com.au/projects/Outpu...protection.pdf