4.70Ω in the grid and 0.47Ω in the source and it will also work at 200kHz if you coil with litzwire. Don't forget the gap between the ferrite core.

Yes I know about the resistor. I am experimenting with various frequencies. Unfortunately I had to hand winding the transformer. Which may effect the Inductance. and so many other factors.

Thank you.

 

Have to look if I still have that one.

Thank you very much

 

The circuit looks like a flyback converter. if you intended this, you should change the polarity of the output winding. AND there must be an air gap in the transformers magnetic core. Wile the Mosfet is switched on, the transformer acts like an inductor, letting the primary current ramping up, while se output rectyfier is blocking the output. As soon an the current threshhold is reached (V over the source resistor), the FET switches off an lets the inductor current change over to the secondary side through the output rectifyer. One frequent problem is the missing low pass filter after the current sense resistor, to filter out the corrent spike an switching on of the mosfet.
If you really intended a foreward design, you will definitely need a additional inductor in the output path.
Unfortunately i just dropped my old Unitrode (te first manufacturer of that chip) data book with a lot of basics and sircuit examples... Good Luck. Hartmut

Thank you very much for your suggestions. I am using the flyback topology and changed the polarity as you mentioned.

Your suggestion to use a low pass filter at the current sense resistor is very much valuable. I used a 10K resister (period) to see the effects.

 

I've attempted several amps designs and never completed the designs.
For diy constructors, I found it difficult to get specs on the cores available to me, or alternatively the cost of cores from major suppliers was prohibitive ( and you may need to buy a few in order to experiment and get a final design)
I just resort these days to getting modules from aliexpress or similar. Whilst it's not the perfect design at least I get it at a cheap price and I know it works.

Exactly my problem too. It is almost impossible to know even the manufacturer of the core I get from retailers. If the required size is available , its proper bobbin may not and vice versa. However I think we can do it. May not be perfect (which is after all). atleast I can make a stable and long lasting even if the efficiency is little bit low. Except for those modules with transistors , most of them breakdown in a few months.

 

Well friends,

Thank a lot for your valuable contributions. One particularly from Ian0 was invaluable. His comment about starting with.... rang the bell. I was confused earlier with the equations used.

I just finished one. - basic design.

 

Thank you very much

Datasheet - here you have it.

Off for an espresso now. When I come back, AN will follow.

 

Datasheet - here you have it.

Off for an espresso now. When I come back, AN will follow.

AN U-100A - Unitrode
AN U-96A - Unitrode

 

You should ignore most of the advice telling you to use a gate driver. The UC384x series are all capable of a 1Apk gate drive current (it drops quite a bit around the gate threshold voltage of a typical mosfet but remains perfectly adequate up to useful frequencies with typical mosfets and your IRF840 is likely a more up to date version than the UC3845 was designed for. First thing I would suggest in this area is to start with a 10R resistor. Increase it if you observe excessive ringing in the gate signal. This is not a complete treatment of the gate resistor but I cannot squeeze 30+yrs experience into a single post. Be sure that the gate resistor is as close as possible to the mosfet gate lead. This is more important that the value of the resistor.
The reason you get high input current at lower fs is, if things are more or less working as the should is due to transformer core saturation. You should revisit that maths because no ferrite I know of can handle much beyond 500mT and most are happier with a design limit closer to 200mT or less.
If the topology is a flyback then you will need to factor in an air gap in the transformer. A flyback tx design is more than I can comfortably go into here. I suggest, if you need to, that you look at a few text books on the topic and then ask questions. In a forward converter the design focus is on handling the voltages and passing the current. In a flyback, the power literally goes through the transformer core (more accurately, the air gap in the transformer core, mostly).
Your regulation using a zener diode is a perfectly acceptable first pass method (although I have not looked at your circuit, the idea is not new and was common enough in some applications).
Set the converter to a fixed frequency and design for that. Varying the frequency on the fly is dodgy at best, and likely to make the smoke leak out somewhere.
The current signal into pin 3 usually wants an RCfilter to take out the leading edge gate current spike (look at the TI data sheets for these parts. TI bought out Unitrode, who designed the originals of these parts, when dinosaurs still roamed the earth. Unitrode and TI have published great data sheets and thousands of app notes for them. See if you can get a copy of the original parchment (data) scrolls . The filter needs to be effective at taking down the leading edge spike but not effect the waveform beyond the spike duration. Some balancing of gate drive and filter component values may be required here. If your tx has a large parasitic capacitance or the output rectifer diodes are a bit slow, they can make this leading edge current spike much larger.
I suggest starting your converter at low input voltages and low output currents.When you get that going then you should look at the waveforms and assess what is needed for a snubber, if a snubber is needed at all. There is a good and empirical method to determine snubber component values but to be frank, how you will manage any of that without a CRO is beyond me.
In fact, and I do not mean this to be mean or rude or to dissuade you from your ambitions, but to get an SMPS right, or even just off the ground, will be very very difficult at best, and IMHO just not possible.
Besides being a vital development tool for SMPS, a CRO will show you soooo much that will be informative, enlightening and very educational. Without a CRO, it would be virtually impossible, I think, for anybody to do this. I know I could not get an SMPS right, nor even working probably, without a CRO. Having said that, I do sincerely applaud your ambition and gumption and really hope you manage a miracle or two.

 

AN U-100A - Unitrode
AN U-96A - Unitrode

Unless TI have updated them, the Unitrode app notes are going to be a bit out of date. There are subtle differences between the modern parts and their early predecessors even with the same part numbers.

 

//Off-the-shelf transformers simply are not available that work at 10kHz.//

Very true. Off the shelf ? they are not available. One has to get them wound. They are asking for oders in 500's minimum qty.
//Start at 50kHz.// I am working on 50Khz to 132 Khz range.

//Buy a scope. You can probably get a usable scope for the cost of a few dozen dead MOSFETs //

Not possible here. I can get two mosfets for 1 dollar in India. But a used good scope will cost almost 150 to 200 dollars.

If buying, I think would need a 200Mhz scope , since I work with microcontrollers too. And they don't come cheap.

For the SMPS stuff, a 100MHz CRO is usually good enough and there are some newer CROs around that have mutlichannel digital inputs you can use for all manner of digital circuit tests and measurements as well as logic and protocol analyser type functions and at far higher frequencies than 100MHz. Is there something like that around where you are that suits your purposes and your budget? I only ask because I know what a difference it would make for you in developing your SMPS. It is not just a difference of night and day, it s more like complete black out and the light from a fission bomb up close and personal.

 

A CRO? I use one of those fancy new digital ones with a colour LCD display.

 

You should ignore most of the advice telling you to use a gate driver. The UC384x series are all capable of a 1Apk gate drive current (it drops quite a bit around the gate threshold voltage of a typical mosfet but remains perfectly adequate up to useful frequencies with typical mosfets and your IRF840 is likely a more up to date version than the UC3845 was designed for. First thing I would suggest in this area is to start with a 10R resistor. Increase it if you observe excessive ringing in the gate signal. This is not a complete treatment of the gate resistor but I cannot squeeze 30+yrs experience into a single post. Be sure that the gate resistor is as close as possible to the mosfet gate lead. This is more important that the value of the resistor.
The reason you get high input current at lower fs is, if things are more or less working as the should is due to transformer core saturation. You should revisit that maths because no ferrite I know of can handle much beyond 500mT and most are happier with a design limit closer to 200mT or less.
If the topology is a flyback then you will need to factor in an air gap in the transformer. A flyback tx design is more than I can comfortably go into here. I suggest, if you need to, that you look at a few text books on the topic and then ask questions. In a forward converter the design focus is on handling the voltages and passing the current. In a flyback, the power literally goes through the transformer core (more accurately, the air gap in the transformer core, mostly).
Your regulation using a zener diode is a perfectly acceptable first pass method (although I have not looked at your circuit, the idea is not new and was common enough in some applications).
Set the converter to a fixed frequency and design for that. Varying the frequency on the fly is dodgy at best, and likely to make the smoke leak out somewhere.
The current signal into pin 3 usually wants an RCfilter to take out the leading edge gate current spike (look at the TI data sheets for these parts. TI bought out Unitrode, who designed the originals of these parts, when dinosaurs still roamed the earth. Unitrode and TI have published great data sheets and thousands of app notes for them. See if you can get a copy of the original parchment (data) scrolls . The filter needs to be effective at taking down the leading edge spike but not effect the waveform beyond the spike duration. Some balancing of gate drive and filter component values may be required here. If your tx has a large parasitic capacitance or the output rectifer diodes are a bit slow, they can make this leading edge current spike much larger.
I suggest starting your converter at low input voltages and low output currents.When you get that going then you should look at the waveforms and assess what is needed for a snubber, if a snubber is needed at all. There is a good and empirical method to determine snubber component values but to be frank, how you will manage any of that without a CRO is beyond me.
In fact, and I do not mean this to be mean or rude or to dissuade you from your ambitions, but to get an SMPS right, or even just off the ground, will be very very difficult at best, and IMHO just not possible.
Besides being a vital development tool for SMPS, a CRO will show you soooo much that will be informative, enlightening and very educational. Without a CRO, it would be virtually impossible, I think, for anybody to do this. I know I could not get an SMPS right, nor even working probably, without a CRO. Having said that, I do sincerely applaud your ambition and gumption and really hope you manage a miracle or two.

No paragraphs? Not for me.

 

No paragraphs? Not for me.

Sorry. Did I reply to you instead of the OP? I meant to reply to the OP.

 

For the SMPS stuff, a 100MHz CRO is usually good enough and there are some newer CROs around that have mutlichannel digital inputs you can use for all manner of digital circuit tests and measurements as well as logic and protocol analyser type functions and at far higher frequencies than 100MHz. Is there something like that around where you are that suits your purposes and your budget? I only ask because I know what a difference it would make for you in developing your SMPS. It is not just a difference of night and day, it s more like complete black out and the light from a fission bomb up close and personal.

Ooops. My apologies. I meant to reply to the OP. Not sure how I got that wrong (twice!!!). Like I said, oooops. Sorry.

 

Exactly my problem too. It is almost impossible to know even the manufacturer of the core I get from retailers. If the required size is available , its proper bobbin may not and vice versa. However I think we can do it. May not be perfect (which is after all). atleast I can make a stable and long lasting even if the efficiency is little bit low. Except for those modules with transistors , most of them breakdown in a few months.

If they are an option for you, Digikey and Mouser stock a number of core geometries and matching bobbins and yokes; for power transformers and chokes the popular ETD, E, EFD and RM for example. The usual power focused materials as well, Epcos/TDK N27 and N87 and the Ferroxcube 3C92.... 3C97 ferrite materials. They are not cheap but also not robbery.
Some manufacturers are a bit slack about identifying themselves or the material but TDK and Ferroxcube do print the material specification and the gap on the side but not their name. So if you have a core with no printing on it you can at least know it is neither of them, if that is any help.

 

If they are an option for you, Digikey and Mouser stock a number of core geometries and matching bobbins and yokes; for power transformers and chokes the popular ETD, E, EFD and RM for example. The usual power focused materials as well, Epcos/TDK N27 and N87 and the Ferroxcube 3C92.... 3C97 ferrite materials. They are not cheap but also not robbery.
Some manufacturers are a bit slack about identifying themselves or the material but TDK and Ferroxcube do print the material specification and the gap on the side but not their name. So if you have a core with no printing on it you can at least know it is neither of them, if that is any help.

Thank you.

// Some manufacturers are a bit slack about identifying themselves or the material but TDK and Ferroxcube do print the material specification and the gap on the side but not their name. So if you have a core with no printing on it you can at least know it is neither of them, if that is any help.//

That is great information. Thanks a lot.
No nothing is stamped on the core that we get here.

The delivery charges of digikey and Mouser are exorbitantly high here. So that is not an option.

 

Ooops. My apologies. I meant to reply to the OP. Not sure how I got that wrong (twice!!!). Like I said, oooops. Sorry.

Thank you for your suggestion. I am searching for a good used scope.

 

Just look at the 3C90 datasheet
https://www.ferroxcube.com/upload/media/product/file/MDS/3c90.pdf
The tolerance on the permeability is ±25% and it varies from 1800 to almost 4000 over the normal operating temperature range. Why are you worrying about accuracy?
The permeance of the assembled core is dominated by the permeability of air which is always 1.

 

No paragraphs? Not for me.

Appreciate a lot for your insights. I fully understand the need for a scope. I am trying to get a good used one.

nevertheless I have made it, and is studying it more carefully. as you mentioned, I have collected many articles on the subject..

I have to read your msg a couple more times , so as not to miss anything. Thank you so much.

 

Just look at the 3C90 datasheet
https://www.ferroxcube.com/upload/media/product/file/MDS/3c90.pdf
The tolerance on the permeability is ±25% and it varies from 1800 to almost 4000 over the normal operating temperature range. Why are you worrying about accuracy?
The permeance of the assembled core is dominated by the permeability of air which is always 1.

Thank you. I have downloaded it. hmmm. That is good info.