20 Volts In ... 2,000 Volts Out

[DigiTan]

I'm setting up an oscillator with one of those feedback transformers used in CCFLs. So far it's been successful in generating outputs as high as 1,300kV. The problem is the circuit draws an enormous amount of current, even when there's nothing on the high voltage secondary. The energy winds up mostly in the transformer where it heats up above the boiling point.

Everything seems in line with their decription, so I don't think this is a natural drawback of the circuit. It's just eating too much power for some reason. With the exception using L1 = 1mH and C1 = 330nF, mine is exactly the same as in the drawing.

Anyone have an idea of how to improve the efficiency of these Royer-type circuits?

Side question: Can the feedback coil be used to drive a second (identical) oscillator?

 

[Ubergeek63]

I'm setting up an oscillator with one of those feedback transformers used in CCFLs. So far it's been successful in generating outputs as high as 1,300kV. The problem is the circuit draws an enormous amount of current, even when there's nothing on the high voltage secondary. The energy winds up mostly in the transformer where it heats up above the boiling point.

Everything seems in line with their decription, so I don't think this is a natural drawback of the circuit. It's just eating too much power for some reason. With the exception using L1 = 1mH and C1 = 330nF, mine is exactly the same as in the drawing.

Anyone have an idea of how to improve the efficiency of these Royer-type circuits?

Side question: Can the feedback coil be used to drive a second (identical) oscillator?

what are you trying to do and why are you using a royer?

use a standard push pull regulator circuit. as to it's overheating, the gain is probably to high... add base emitter resistors and a resistor in the feedback coil and then try playing with those and the capacitor value. the other thing is the frequency, some transformers might not like where it is running

 

[DigiTan]

Well the thing with adding resistors is, while it lowered the current draw and heat--it also lowered the output voltage by the same amount. So you're forced to ramp up the input voltage and you're back where you started.

I did manage to solve the power problem. Since the output voltage depends on the primary's current and frequency, I decreased the capacitance so it would resonate faster. That let the circuit depend on running at high frequencies, rather than needing high flux values (with required too much current and heat). By doubling the frequency, I could cut the flux/current in the primary by about 1/2 (Vout = 2*pi*frequency*turns_ratio*flux). Since the joule heating energy is Q = R*I^2, doing this cut the thermal losses by 1/4.


Now the only problem is, at higher voltages around 800V and up, the output distorts from being a sine to triangle wave. The transistors are rated for 3MHz switching and I'm only at 60kHz. So I need to know if it's the transformer, transistors, or some other part causing it. Why would a transformer suddenly start outputting a triangle wave at high voltages?

 

[Ubergeek63]

Well the thing with adding resistors is, while it lowered the current draw and heat--it also lowered the output voltage by the same amount. So you're forced to ramp up the input voltage and you're back where you started.

I did manage to solve the power problem. Since the output voltage depends on the primary's current and frequency, I decreased the capacitance so it would resonate faster. That let the circuit depend on running at high frequencies, rather than needing high flux values (with required too much current and heat). By doubling the frequency, I could cut the flux/current in the primary by about 1/2 (Vout = 2*pi*frequency*turns_ratio*flux). Since the joule heating energy is Q = R*I^2, doing this cut the thermal losses by 1/4.


Now the only problem is, at higher voltages around 800V and up, the output distorts from being a sine to triangle wave. The transistors are rated for 3MHz switching and I'm only at 60kHz. So I need to know if it's the transformer, transistors, or some other part causing it. Why would a transformer suddenly start outputting a triangle wave at high voltages?

that could be your transistor and stuff ... those transformers are meant to run at 60KHz, that is the CCF peak efficiency frequency.

Dan

 

[DigiTan]

I was hoping to run at a lower frequency of switching reasons, but it looks like this was thermally designed to run between 55kHz and 70kHz. Otherwise, you have to put too much current in the primary to get the voltages you want.


The new problem is with rectifying the output. With the secondary windings totally unconnected, the output is close to a pure sine. When I connect the rectifier...even if the rectifier doesn't have load...the output shows repeating spikes instead of a rectified sine. The spikes have the same frequency as the normal output, except every even-numbered spike is bigger than every odd-numbered one. My understanding is this rectifier should resemble a large resistance except for the few ns or μs it takes to switch off. I don't know if these diodes broke down from previous testing, but they seem to short out for any output above 85Vpp.

 

[Ubergeek63]

I was hoping to run at a lower frequency of switching reasons, but it looks like this was thermally designed to run between 55kHz and 70kHz. Otherwise, you have to put too much current in the primary to get the voltages you want.


The new problem is with rectifying the output. With the secondary windings totally unconnected, the output is close to a pure sine. When I connect the rectifier...even if the rectifier doesn't have load...the output shows repeating spikes instead of a rectified sine. The spikes have the same frequency as the normal output, except every even-numbered spike is bigger than every odd-numbered one. My understanding is this rectifier should resemble a large resistance except for the few ns or μs it takes to switch off. I don't know if these diodes broke down from previous testing, but they seem to short out for any output above 85Vpp.

has nothing directly to do with thermals... it is directly related to transformer saturation. what are the diodes you are using and it will not work as a royer since that is a resonant circuit and the diodes will screw up the resonance.

dan

 

[MOSFET KILLER]

I believe I have misunderstood your first post, is it the transformer that is heating up? If it is what material are you using (ferrite, iron powder)?

 

[DigiTan]

I believe I have misunderstood your first post, is it the transformer that is heating up? If it is what material are you using (ferrite, iron powder)?

It's the transformer over-heating. It's an iron ferrite type with a very small-gauge primary that only has about 0.61 Ohms resistance. Previously I had to drive in over 5 watts to get 1000Vpp out of it, but now I can use way less.

has nothing directly to do with thermals... it is directly related to transformer saturation. what are the diodes you are using and it will not work as a royer since that is a resonant circuit and the diodes will screw up the resonance.

They're 1000V rated 1N4007's arranged as a bridge rectifier. I'm only using them until I get 1200V-rated STTH212S's, which also have faster switching. The resonance change is manageable, but the big killer is this rectifier thing.

Here's a rough plot of what it looks like with the bridge attached. It's almost like things are being half-rectified instead of fully-rectified and the pulses are event shorter than in my plot. You might say their duty cycle is 5% - 10%. It's like the voltage can't stay up for the rest of the half-sine.

 

[Ubergeek63]

which means you are trying to use diodes that you would be lucky if they worked through the audio range at triple that.

wait on the better diodes

 

[MOSFET KILLER]

Ok, so the heat in the core is due to the core not being able to handle the power and frequency, it represents wasted power, reduced efficency and can eventually lead to device failure. Some iron powder cores will melt if used for this application, such as the yellow-white filter choke cores in computer SMPS. Can you attach a picture of the core or a good description?

 

[schmitt trigger]

The 1N4007 can't rectify properly above 400 Hz. Faster than that, and the diodes cannot "close" fast enough and current flows back. In semiconductor parlance, that is called recovery time.

You have to use a ultra fast recovery diode. Google it.

 

[Ubergeek63]

Ok, so the heat in the core is due to the core not being able to handle the power and frequency, it represents wasted power, reduced efficency and can eventually lead to device failure. Some iron powder cores will melt if used for this application, such as the yellow-white filter choke cores in computer SMPS. Can you attach a picture of the core or a good description?

true ... they are meant for DC. however there is nothing that you can tell from from a picture or good description... and he is using a commercial transformer MEANT FOR WHAT HE IS DOING!!!

 

[DigiTan]

You have to use a ultra fast recovery diode. Google it.

Google the STTH212S I already ordered?


Here's that image. This type of transformer has other functions I need to investigate, so I prefer including it vs. something else like a cascade multiplier. I might add a cascade in addition to it, but the transformer stays.

Speaking of the cascade multiplier, what volt rating do the parts need to have? Is it the expected output voltage, twice that amount, or some other margin?

 

[DigiTan]

Okay, here's the other issue. It was recommended that the voltage rating on the Volt-doubler be rated at the output voltage of 2000V. With 2000V diodes being hard to come by, the obvious answer of putting two 1000V diodes in series also assumes that voltage will be evenly-distributed on every diode. Is that ever really the case?

 

[schmitt trigger]

To put two diodes in series, you have to equalize each with a shunt capacitor like the one shown in the attachment.

 

[DigiTan]

The diodes have 2.7pF to 13pF in the datasheet. I take it the capacitors should be large enough to make a difference, but not so large they make the diodes irrelevant when turned off? Maybe something around 13pF to 26pF and very close to the diode?