1kW smps project (based on MicrosiM design)

stewin

Member
hello all , i have an issue the smps i made heats up irfp460 or 21n50c fets but if i use irf740 the heat is not present at the fets. when i put the bulb in series with smps and load a class ab with the bulb still in series the irfp460/21n50c they do heat insanely with little load and must be mounted on the sink. but irf740 are ok when smps has little or no load.
i tried changing the gate resistors from 22ohms to 47ohms but the heating was worse so i settled on the 4.7ohms it reduced the heating a little bit but did not eliminate the heating.
but when i used irf740 the smps worked well also with a little load to the amplifier the irf740s did not overheat but warmed without a heatsink.

the smps was
- gate drive transformer is ee16 trifilliar 12turns
- gate resistors 4.7ohms
- i tried with two etd39 formers
former 1 etd39 - primary turns 23turns, secondary turns 7t - 0 - 7t aux 2turns
former 2 etd39 - primary turns 42turns, secondary turns 14t - 0 - 14t aux 4turns
both formers behaved the same regarding the heating of the irfp460/21n50c
- i used a1020 and c2655 as buffers for the gdt and 1ohm resistors .(i had used 10 ohms resistor but results were similar with 1ohm)
note the same circuit i tried using igbt an year ago but it over heated abnormally a change of gate resistor didn't help in anyway .
below are the pictures . kindly help out and any opinions are highly welcomed.
 

Attachments

  • gtG lp ps ver 1.2 to247 diodes schematic.png
    gtG lp ps ver 1.2 to247 diodes schematic.png
    50.7 KB · Views: 63
  • gtG lp ps ver 1.2 to247 diodes schematic.png
    gtG lp ps ver 1.2 to247 diodes schematic.png
    50.7 KB · Views: 34
  • IMG_20210517_135927_740-Optimized.jpg
    IMG_20210517_135927_740-Optimized.jpg
    57.8 KB · Views: 25
  • IMG_20210517_135935_786-Optimized.jpg
    IMG_20210517_135935_786-Optimized.jpg
    92.6 KB · Views: 16
  • IMG_20210517_174753_897-Optimized.jpg
    IMG_20210517_174753_897-Optimized.jpg
    63.4 KB · Views: 16
  • IMG_20210517_140146_464-Optimized.jpg
    IMG_20210517_140146_464-Optimized.jpg
    67.7 KB · Views: 15
  • IMG_20210517_135935_786-Optimized.jpg
    IMG_20210517_135935_786-Optimized.jpg
    92.6 KB · Views: 15
  • IMG_20210517_135959_108-Optimized.jpg
    IMG_20210517_135959_108-Optimized.jpg
    83.6 KB · Views: 15
  • IMG_20210517_135944_120-Optimized.jpg
    IMG_20210517_135944_120-Optimized.jpg
    88.1 KB · Views: 14
  • IMG_20210517_140146_464-Optimized.jpg
    IMG_20210517_140146_464-Optimized.jpg
    67.7 KB · Views: 27

Silvio

Active member
hello all , i have an issue the smps i made heats up irfp460 or 21n50c fets but if i use irf740 the heat is not present at the fets. when i put the bulb in series with smps and load a class ab with the bulb still in series the irfp460/21n50c they do heat insanely with little load and must be mounted on the sink. but irf740 are ok when smps has little or no load.
i tried changing the gate resistors from 22ohms to 47ohms but the heating was worse so i settled on the 4.7ohms it reduced the heating a little bit but did not eliminate the heating.
but when i used irf740 the smps worked well also with a little load to the amplifier the irf740s did not overheat but warmed without a heatsink.

the smps was
- gate drive transformer is ee16 trifilliar 12turns
- gate resistors 4.7ohms
- i tried with two etd39 formers
former 1 etd39 - primary turns 23turns, secondary turns 7t - 0 - 7t aux 2turns
former 2 etd39 - primary turns 42turns, secondary turns 14t - 0 - 14t aux 4turns
both formers behaved the same regarding the heating of the irfp460/21n50c
- i used a1020 and c2655 as buffers for the gdt and 1ohm resistors .(i had used 10 ohms resistor but results were similar with 1ohm)
note the same circuit i tried using igbt an year ago but it over heated abnormally a change of gate resistor didn't help in anyway .
below are the pictures . kindly help out and any opinions are highly welcomed.
Hello Stewin, well from what you are saying it seems that either you are not driving the big fets hard enough and due to that you may have a slow rise time. This will tend to switch the fets in a linear mode hence the heating. You are getting better results with smaller fets due to they do not need much to be switched hence they stay cool. Looking at the schematic I would rather put the snubber in the gating circuit across the GDT primary than at the output of the SG chip. The energy supplied from the chip I will expend it all to the totem pole bases. (47r-103 ceramic)

One other thing that is crossing my mind is that you may have a spike in the waveform from your gating transformer. This will also reflect itself on the main transformer winding. I am not sure if you have a scope to monitor the pulses but this will tell you for sure.
It also can be the case that the IRFP460 are fakes!!

Some things you may try-
1) Rise the auxiliary voltage to 15-18v dc
2) Remove the emitter resistors from the totem pole transistors to enhance current in the GDT primary
3) Put a snubber across the main transformer primary or a snubber (across each switching Fet 1nF x 1Kv- 5.6R x 3W)
4) Ask a friend to take a look with the oscilloscope and monitor the low side wave form (Gate to source) The rise and fall times need to be sharp as possible, any slopes here will heat up your fets for sure.
5) Check the dead time and see it is adequate so that while one fet is switching off the other will not start switching on.
6) As for the gate resistors use between 4.7R to 10R. Do not go for a higher value than this on the large fets.
7) Put a parallel ultra fast diode with the body diode (drain source) MUR 440 or MUR460. across each fet.

Good luck regards Silvio
 
Last edited:

stewin

Member
thanks silvio , thanks for the quick reply and your help, but i have one question which is puzzling me . all this problems i am facing because i am using a class ab amp and have bridged it. now even the irf740 at test volume and a load of 4ohm speaker . the bridged ab amp has 4 transistors per channel and the power i feed to it from the smps is +/-52volts.

today when i used the irf740fets, the smps at idle does not heat , but at powering the amp with 4ohms woofer for five seconds . the irf 740fets(although they are mounted on a heatsink) heated up and de natured and failed. the same case happened at irfp460(but they lasted longer)
why is it that i have ever used class d amp circuits +/-49volts with a 4 channel class d and 2channels are bridged driving hard a 4ohm woofer, and other two channels driving 6ohms mid each individually. but the smps similar to this one ,but never heated to this extremes like when using this ab amp. also the smps never failed when driving the classd amps . unlike now it is over heating and fets failing driving only 2channels bridged class ab amp ??
 

Silvio

Active member
Well class AB amplifiers draw more juice than class D. Class ab may also have some idle current. The amp for sure will draw more power than class d for the same power output. Class AB amplifiers are only 60% efficient. In my 1200 watt smps I drive a class AB amp and with 8 ohms 15" full range speakers I can drive it up before clipping with an output of 335W rms per channel. I tried the amp with 4 ohms load on the dummy load but the smps tripped on overload. Peaks where in excess of 1300w hence the tripping. At this power the fets are seeing around 14 amps. Through experience and of what I have read driving IRFP460 more than this they are likely to blow.

The input voltage to the amp is 80v- 0 - 80v at an input voltage of 240v. I am also using a couple of IRFP460. The trafo is ETD49. My windings are 18 turns primary and 9-0-9 turns secondary. The frequency is 62Khz. The only difference I guess is that I mounted the fets and output diodes on a cpu heatsink form an old computer. I tested it with the fan on with 900w for 1 hour and the heatsink temperature only came up to 55 degrees If I remember correctly. Why not try a larger heatsink.
 

stewin

Member
hi Silvio , thanks for the reply , i think it is because i use 7812 regulator even before the gdt , and maybe the home wound gdt requires more current than the 7812 can provide . also not using bd139/140 as buffer transistors and using a1020 and c2655 as buffers for the gdt
 

Silvio

Active member
hi Silvio , thanks for the reply , i think it is because i use 7812 regulator even before the gdt , and maybe the home wound gdt requires more current than the 7812 can provide . also not using bd139/140 as buffer transistors and using a1020 and c2655 as buffers for the gdt
I don't think its that. 1A current is more than enough. My 1Kw smps aux supply can only source 600mA and I don't have any problems. I think your GDT may be consuming a lot of current due to low inductance. I made GDTs and the inductance is to be around 1.4 to 1.5mH. This was made on a green round core of around 20mm dia. This works very well and can drive the mosfets very well. The waveform was very faitfull to the output from the chip.
Only an oscilloscope can give you these answers.
All input filters in smps have high permeability cores so that there is no need to wind a lot of turns to get to the right inductance.
Good luck
 

stewin

Member
I don't think its that. 1A current is more than enough. My 1Kw smps aux supply can only source 600mA and I don't have any problems. I think your GDT may be consuming a lot of current due to low inductance. I made GDTs and the inductance is to be around 1.4 to 1.5mH. This was made on a green round core of around 20mm dia. This works very well and can drive the mosfets very well. The waveform was very faitfull to the output from the chip.
Only an oscilloscope can give you these answers.
All input filters in smps have high permeability cores so that there is no need to wind a lot of turns to get to the right inductance.
Good luck
how many turns did you use for your gdt? also can the heating and the failing of the fets be because i am not using full wave rectification at the dc +/-60 . i am just using one diode per rail! ??
 

Silvio

Active member
The GDT has around 18 to 20 turns. It is trifilar wound with 0.5 enamel copper wire.
In the secondary using one diode per rail does not really heat the fets but I only seen this method when used in low power such as a preamp supply or to power a fan motor. If you want proper juice to come of your secondary it is best to use a fast bridge with 4 diodes for dual supply or a double winding with 2 diodes for a single supply rail.
 

Attachments

  • GDT.PNG
    GDT.PNG
    1.9 MB · Views: 10
Good evening, excellent work from Ludo + MicrosiM!
I finished making a smps of a PCBshare Facebook page that had errors and with the help of colleagues I was able to make it 100% functional. Looking for info I found this forum and noticed that the creator of the smps I made is Silvio from this DIYSMPS forum ..... I made it with another core, turns, frequency and different wire but it is still Silvio's. I have been a hobbyist for 7 or 8 years in electronics (I love smps and audio amplifiers), I have several cores EI33, EE42 and especially EE55 (with the latter I want to make some smps with protections). I have been reading many pages of this thread, from what I see they have used SG3525 + IR2110 to control the mosfets, 555 as a delay timer, LM358 for voltage comparator protection and they sense the current with a toroid or something like that ..... also the auxiliary transformer that feeds the ICs has a soft start. I am an electromechanical (graduated 2009), electrical engineering student and passionate about electronics. My intention is to know if you can recommend any other smps topic focused on EE55 or if I should start a new topic.
Greetings from Argentina!
IvanElectric
 

Silvio

Active member
You can open a new tread and we can help you out during your new build. It is best to make some reserch on the internet to take a few ideas then you can create your own smps and pcb.
 

P.S.RAJU

Member
Hi sir silvio
how to find the( Bmax) of smps transformer of ( Etd 29,34,39,44,49,54,59) iam trying to learn transformer winding formula.
do different core size have different Bmax ,i dont know ,i found the (Ae) in the data sheet of (Tdk etd tx) i could not find
the (Bmax )in the datasheet, so please help me to find the (Bmax) of different etd core.


Thanking you
p.s.rajuturns calculation smps transformator.png
 

Silvio

Active member
B is the flux density. This varies according to the core material
Not all core materials take the same flux density. This is usually applied according to the core material. There exist a lot of types of materials some are siuted for high frequency around 100khz or more and some are siuted to work at lower frequency around 10-20khz.
I don't know how to find a correct flux density but I usually use a rule of thumb.
In general for a switching frequency of 20khz I use a flux density of 2000 guass (0.2 tesla) for a frequency between 40 and 50 Khz 1700-1800 and for 60 to 80khz 1500 to 1600, from 80 to 100khz 1200 to 1400. guass. I found this range to be siutable for most cores.
For a given number of turns the flux density varies with the rail voltage applied and we must take it in account as if the main input supply voltage is say 220vac we see a lower flux density than when a smps is working at 245v input. That is why I am a bit low on flux density in my assumption.
If however the smps has a boost PFC then the input DC voltage will be stabilized and the flux does not vary.
Lastly there is a calculator on this site that work things out for you. It is called Excellent IT. You can download it.
Regards Silvio
 

P.S.RAJU

Member
Thanks for your reply silvio
i use cf139cosmoferrites, core material ,for my transformer so what will be the Bmax if use smps
circuit half bridge topology,and my supply is 230vac. can just give an example of Excellent it
and how to use it ,my supply is 85volt 10amp.
thanking you
p.s.raju
 

Silvio

Active member
You can download the file on the link below. This program does not run on windows 10 but works well on windows 7 and possibly 8 also. You will find the zip files on post #85 download the 7300 file

See if it works after you unzip the file by pressing on the colored icon to open the program.
 

Silvio

Active member
This is a modified file that works on windows 10
 

Attachments

  • ExcellentIT(7300)-20200320T081334Z-001.zip
    259.7 KB · Views: 20
Top