TL494 schematic by 0xff

Silvio

Well-known member
Silvio, is 10V enough for IRF740 or IRFZ44N to fully turn-on (0.25 ohm for IRF740 and 0.014 for IRFZ44N) ?

The gate-source threshold voltage for both mosfets is 4v so 10v should be enough to drive them. The maximum gate-source voltage for both of them is 20v.
 

Silvio

Well-known member
Regarding the drive circuit looks very promising and it should be tried in practice for real results.

As for the EI 33 core well it says you should widen the gap and this would lower the inductance of the core. An inductance meter will surly help you here as you can measure the primary inductance and adjust the gap accordingly until the required inductance is achieved.

I have seen videos on Youtube of Robert Balanos doing this method. This guy is a principal reserch engineer and designs SMPS for NASA. He is a very professional person and knows a lot about the subject but he also uses a lot of math and its a bit hard for me to understand at times. If you are patient enough you can see some of his videos. He is very calm and the videos are quite long.

Cheers Silvio
 

expressel

New member
Cheers!

Regarding the drive circuit looks very promising and it should be tried in practice for real results.

I'll try for sure.

As for the EI 33 core well it says you should widen the gap and this would lower the inductance of the core. An inductance meter will surly help you here as you can measure the primary inductance and adjust the gap accordingly until the required inductance is achieved.

I have seen videos on Youtube of Robert Balanos doing this method. This guy is a principal reserch engineer and designs SMPS for NASA. He is a very professional person and knows a lot about the subject but he also uses a lot of math and its a bit hard for me to understand at times. If you are patient enough you can see some of his videos. He is very calm and the videos are quite long.

Cheers Silvio

My EI-33 don't have a gap, I can make one myself or I can use another core I have with a ready-made gap, but I dont know what is it, is it ER core, or is it ETD ?
I dont know how to measure the dimensions too ?


From where I can buy ETD cores ? I can find on google but I dont know where is the best place :) Aliexpress good or there are fakes ?


However I succeeded to do a notebook adapter using boost converter topology but it works only if the supply voltage is greater than 13V
 
Last edited:

Silvio

Well-known member
As for the gap on the EI 33 you can make a spacer on the outer sides of the core, Be careful as the space on both sides will be equal to double. For example if you need a gap of 1mm in the center core then when you put spacer on the side it has to be 0.5mm to be equal.

You can buy cores from aliexpress but try to find cores with the core material specified, for example they say it is PC40 or 3c90 etc. This way you can determine the number of turns required for your transformer. I suggest you download Excellent IT software from diysmps there is a chap named Starichock51 The forum name is called Calculation programs for transformers and inductors

I bought an EE55 core from Aliexpress for a reasonable price.

I am posting a PDF file for you so that you can compare the size of the core you have and you will know if its ETD or not.

View attachment ferrite_mz_sw_e_en.pdf

Regards Silvio
 

expressel

New member
OK, Thank you again. Please keep guide me, I'll guide you too. :)

I readed somewhere that A4 paper is 0.1mm thick so for a 1mm gap I will use 5 pieces of paper for the left side as spacer and another 5 for the right side.
Or I can use my other core with the 2mm gap. If I remember correct its width or height is 42mm so its ER42 according to the pdf you shared.

I dont know the material but when I tried to switch between the materials in the program, I found out it gives me 5 turns primary and 5 turns secondary for the every material I selected.

--

My next project is 1kW and I have a ETD49 core but I dont know the core material.

I calculated for the material with the lowest count of primary turns and calculated again for the material with the highest count.
For example Kaschke K2004 gives me 23 turns primary and Ferroxcube 3C97 gives me 11

The worst case is the K2004 so I will wind 23 turns to prevent saturation.

How many watts I will lose if I wind 23 turns and turns out that I have Ferroxcube 3C97 instead that needs 11 turns ?

( 11 / 23 ) * 1000W = ~0.5 * 1000 = ~500W ? I will lose 500 watts in the worst case?

What is the most used material, Epcos N87 and PC40 ?

I suggest you download Excellent IT software from diysmps there is a chap named Starichock51 The forum name is called Calculation programs for transformers and inductors

You forgot that I already have ExcellentIT and use it for my calculations.

I bought an EE55 core from Aliexpress for a reasonable price.

Oh, this is 3kW core at 100khz ? I found one for 14$ with free shipping - click here

Regards
 
Last edited:

Silvio

Well-known member
Where did you get your etd49 core from? Did it had windings before? What I do myself is when I get hold of used cores from old equipment I simply try to look at the pcb where it came from and take note of the oscillator chip. Then I read the resistor capacitor value and make up the frequency it was operating with. This will give me a good guide of the number of turns per volt for a particular frequency. Form there I can make up also the core material by way of comparison.

In my 1000w smps I used N87 material. I used a new core bought from my electronics supplier. On this type of core my primary winding was with 18 turns for an input voltage of 240vac. My trafo is 1:1 for a max of 80-0-80 peak output voltage. I absorbed a maximum of 1200 watts from it which I could hold this peak output for 5 min and 900 watts continuous. Thanks to the efficient heatsink I put in. It is now in my amplifier. You can take a look at my youtube channel if you want. https://www.youtube.com/watch?v=sHH2GMITc7E&t=2s

As for the aliexpress buy I think the 3 piece mean 2 half cores and a bobbin. Do not be mistaken ok. I got my core and bobbin for 21 euro from my supplier, On ali it is 18.5 euro.

My EE55 is already wound up for the next project 2Kw smps for amplifier. I just finished the aux supply which is a small flyback type which is quite compact and gives an output of 17v and 470mA a total of 8 watts, This will have pins and will stand upright to save space. It has delay start up and short circuit protection. It consumes only 1w on standby.
 

expressel

New member
I get my ETD49 from a broken smps. Yes it has windings and I need to find a way to open it.

What I do myself is when I get hold of used cores from old equipment I simply try to look at the pcb where it came from and take note of the oscillator chip. Then I read the resistor capacitor value and make up the frequency it was operating with. This will give me a good guide of the number of turns per volt for a particular frequency.

I will not use the ready-made primary because there are many windings and I don't know which one is it.

These are bargains:

->click<-- 10 sets = 10$ -> you can get cheap 200W @ 350khz for 1$
->click<-- 5 sets = ~17$ -> you can get cheap 1kW @ 250khz for ~3$ (vertical)
->click<-- 5 sets = ~17$ -> you can get cheap 1kW @ 250khz for ~3$ (horizontal)

but its not said what exactly is:

is it:
- ER 32 / 21 / 11
or its:
- ER 32 / 20 / 11

 

Attachments

  • 1kW.jpg
    1kW.jpg
    20.7 KB · Views: 13
Last edited:

Silvio

Well-known member
I get my ETD49 from a broken smps. Yes it has windings and I need to find a way to open it.



I will not use the ready-made primary because there are many windings and I don't know which one is it.

These are bargains:

->click<-- 10 sets = 10$ -> you can get cheap 200W @ 350khz for 1$
->click<-- 5 sets = ~17$ -> you can get cheap 1kW @ 250khz for ~3$ (vertical)
->click<-- 5 sets = ~17$ -> you can get cheap 1kW @ 250khz for ~3$ (horizontal)

but its not said what exactly is:

is it:
- ER 32 / 21 / 11
or its:
- ER 32 / 20 / 11


Too many questions at one go lol

Opening ferrite cores

1) Try boiling the core for 5 min see if the resin looses a little bit. (Remove the outer binding tape on the core first). If this would not work try your mum's oven heat the core to around 130 deg celcius and with a pair of gloves or thick cloth grab the cores and try to separate them. Be careful do not go beyond 150 degrees as you ruin the core material. If it does not work either try immersing the core in Acetone or cellulose thinner for 24 hrs. All this is to try to dissolve the varnish which is gluing the core together. Sometimes they put super glue to hold the cores but this should give away with heat.

2) If you still have the old smps try to follow the traces and see which pins where on the half or full bridge mosfets / transistors etc

3) To decide which core it is 32/20/11 or 32/21/11 These are the dimensions of the core. Measure carefully the core and you will know. You can surly judge a 1mm difference.

As for the trafo turns please give more detail of the transistors you are going to use, output diode, your mains voltage at home (the usual you measure) The actual power needed for the amplifier you are going to use. The drive chip.

It is not so easy to squeeze all that output voltage in an ER35 with adequate current handling. I see your current density is a bit on the high side
 

Silvio

Well-known member
All the detail of this transformer is in the beginning of the video. I also posted the file in diysmps for the smps with the transformer in the video.

Link for load test of smps https://www.youtube.com/watch?v=re3cGUk2j60&t=1s

link for smps http://www.diysmps.com/forums/showthread.php?897-Audio-smps-700w-(IR2153)&highlight=smps+700w

Use genuine fets and chips if you are going to build an smps. I got lots of fakes from Aliexpress and the mosfets could not withstand half the load they suppose to handle.

Take a look here https://www.youtube.com/watch?v=gRRfb5YxvIA
 

expressel

New member
Too many questions at one go lol

Yes..

1) Try boiling the core for 5 min see if the resin looses a little bit. (Remove the outer binding tape on the core first). If this would not work try your mum's oven heat the core to around 130 deg celcius and with a pair of gloves or thick cloth grab the cores and try to separate them. Be careful do not go beyond 150 degrees as you ruin the core material. If it does not work either try immersing the core in Acetone or cellulose thinner for 24 hrs. All this is to try to dissolve the varnish which is gluing the core together. Sometimes they put super glue to hold the cores but this should give away with heat.

I didn't meaned that. I know how to open a core with 220V bulb in series with the ferrite.
I meaned that the transformer is closed with iron that I dont know how to remove. Maybe with angle grinder but there is a chance to damage the ferrite.

2) If you still have the old smps try to follow the traces and see which pins where on the half or full bridge mosfets / transistors etc

I dont have it more.

3) To decide which core it is 32/20/11 or 32/21/11 These are the dimensions of the core. Measure carefully the core and you will know. You can surly judge a 1mm difference.

The calculations for both are the same so I dont need it to measure it. The only difference is the max power but its a small difference.

As for the trafo turns please give more detail of the transistors you are going to use, output diode, your mains voltage at home (the usual you measure) The actual power needed for the amplifier you are going to use. The drive chip.

Im going to use 2x IRF740 in parallel. I dont know what output diodes I will use. Maybe MBR10100 in parallel. Mine mains voltage is 220Vrms. The actual power is 4 class D channels * 250W = 1kW. Im going to use TL494 or UC3525 with discrete mosfet drive.
 
Last edited:

Silvio

Well-known member
The ferrite core with iron strap if I understood correctly

Usually large ETD cores have metal clips which can easily be removed. If this time is a strap wrapped around the core its a different story.
It is very odd to see ferrite transformers strapped this way. Well the factory did it somehow so now I think it is best to grind a small slot at the iron strap carefully until you nearly reach the ferrite then tear slowly away with a pair of long nose pliers. Be careful if you intend to tie the core in a vice it is very brittle and you can break it. What I suggest for you is to tie the grinder in the vice instead and hold the core in your hand. Use a pair of safety gloves and gogles to protect your hands and eyes.

Well for the ETD49 I would go for 20 to 22 turns for the primary for an unknown material. If the nominal voltage at home is usually 220v then I guess even 20 turns will do.

Some considerations you should take.
Smps is not 100% efficient neighter is your amplifier so take into consideration these factors. Your smps will usually be 85% efficient and the class D amp will be in the 90% region. However extra power from the smps is needed in the input to compensate for these losses. I am not saying that it has to withstand the full power all the time but it needs to peak to the power needed.

Making a quick assumption
So smps at 85% The fets in the input need to handle 1150W The amps at 90% need 275w each so multiply this by 4 = 1100w
All this extra power must be added to the input of the smps and a total of 250w of extra power is needed in the input. Give a little headroom to this your smps must be capable of at least 1300w peak.
In my smps I used IRFP460 fets and I got a peak of 1200w before I experienced severe voltage drop in the output. My switching frequency was 65Khz and the current density of the copper wire is 6A/mm²

Try not to go beyond 100Khz with the switching frequency as you will experience a lot of switching losses from your fets. There is also a compromise here as due to skin effect the higher the frequency the thinner the copper strands will get. The more strands that you have the larger the diameter for a given cross sectional area will be when the strands are twisted together.
The flux density will also gets lower with higher frequencies.
My Favorite frequency is 60 to 70Khz here I can use a 0.5mm copper wire which is suitable for these frequencies and the number of strands for a given cross section is not so large. The flux I use here is 1600 guass or 0.16 Tesla

For audio we usually divide the current handling of fets by 2 so for a fet that can handle 10A then we do not load it more than say 5 amps.
Keep in mind that as the junction temperature goes up the power of the fet goes down. A good heatsink will be necessary for stable operation.
Lastly use genuine transistors from known suppliers as otherwise you will be in for surprises when it comes to power handling.

Input capacitance is important for a good punch and low ripple in the smps. You will need around 1500uF across the 320v rail. this means you either have 3 X 500uF X 400v or 6 X 500uf X 200v for your half bridge. Output capacitance is 4000uF per rail.
You will need a good soft start not to blow your input bridge rectifier. I guess 30 to 50 ohms is a good choice and a 3 second delay for the bridging relay. These are observations that I make from the professional amplifiers that I repair.

As for the PWM chip you can use either one of them as long as you have a good circuit around them. Use the overload protection with your circuitry this will save you a lot trouble in the event of something going wrong in your amplifiers or the smps itself.
 
Last edited:

expressel

New member
Try not to go beyond 100Khz with the switching frequency as you will experience a lot of switching losses from your fets.

There aren't loses if you drive them correctly with more current than usual. I mean 2-3 totem pole stages.
And also 10 ohm gate resistor is very high for high frequencies. It needs to be 1 ohm.
There is no ringing problem if you have 2-3 totem pole stages. :)
 

Silvio

Well-known member
Well if its that what you say. So your transistors do not have any switching losses neither does the transformer and also no dead time, RDS on???? I just wait to see when you load the smps and calculate the input and output power. There is nothing 100% efficient in this world.

Good luck :)
 
Last edited:

expressel

New member
I found out that the zener don't provide the neccessary current for the discrete mosfet driver.
Also 470n is not enough for the bootstrap capacitor. I used 10uF

schematic update:

 
Top