Page 3 of 4 FirstFirst 1234 LastLast
Results 21 to 30 of 32

Thread: TL494 halbridge like ATX problem

  1. #21
    Thanks for very informative replies!

    As I wrote before, I have some quesions regarding this kind of psu's. The component names are from the schematic in attachment.

    1. What is purpose of using capacitor C2 in series with resistor R6 in the error opamp feedback? Maybe some "speed up" for feedback? How to choose this components - I have seen values from 10nF to 100nF and from 8.2k to 47k in different ATX schematics.

    2. How do we calculate (estimate) the value of output inductor? Currently I'm using 93uH measured, wound on Yellow core from ATX, 4 wires of 0.4mm diameter each. Should this be a litz wire or can be solid inductor? (I heard that litz is needed only when direction of current changes, so in transformers, not chokes). Also, when set to low voltage, and shorting output, there are some voltage spikes - are they coming from the inductor? Maybe it is too big? How to check if this inductor doesn't saturate at high currents?

    3. How do we calculate (estimate) output capacitance? Is that just to be in resonance with output inductor on working freuency? I think more caps=better, but when working in CC mode and shorting output, there is still huge charge in these capacitors - can kill connected device.

    4. Currently I'm using 2sc5027, and transistors are switching correctly (checked with oscilloscope), but they get warm (I know that's normal) But when using lower voltage drop transistors (eg. MJE13007 - only 1V at 2A) will the power dissipated in them lower?

    5. Is it worth to figth with synchronous rectification in such circuit?

    6. Is it worth to try PFC with such curcuit?

    Big thanks for Your time!

    atx_psp_lab_schema_01_2011.gif

  2. #22
    .... Silvio's Avatar
    Join Date
    Apr 2015
    Location
    Malta
    Posts
    887
    Blog Entries
    9
    Quote Originally Posted by fifi_22 View Post
    Thanks for very informative replies!

    As I wrote before, I have some quesions regarding this kind of psu's. The component names are from the schematic in attachment.

    1. What is purpose of using capacitor C2 in series with resistor R6 in the error opamp feedback? Maybe some "speed up" for feedback? How to choose this components - I have seen values from 10nF to 100nF and from 8.2k to 47k in different ATX schematics.

    2. How do we calculate (estimate) the value of output inductor? Currently I'm using 93uH measured, wound on Yellow core from ATX, 4 wires of 0.4mm diameter each. Should this be a litz wire or can be solid inductor? (I heard that litz is needed only when direction of current changes, so in transformers, not chokes). Also, when set to low voltage, and shorting output, there are some voltage spikes - are they coming from the inductor? Maybe it is too big? How to check if this inductor doesn't saturate at high currents?

    3. How do we calculate (estimate) output capacitance? Is that just to be in resonance with output inductor on working freuency? I think more caps=better, but when working in CC mode and shorting output, there is still huge charge in these capacitors - can kill connected device.

    4. Currently I'm using 2sc5027, and transistors are switching correctly (checked with oscilloscope), but they get warm (I know that's normal) But when using lower voltage drop transistors (eg. MJE13007 - only 1V at 2A) will the power dissipated in them lower?

    5. Is it worth to figth with synchronous rectification in such circuit?

    6. Is it worth to try PFC with such curcuit?

    Big thanks for Your time!

    atx_psp_lab_schema_01_2011.gif
    I will try to answer your questions to the best of my knowledge. I am also a hobbyist just like you.

    1) As I can really see it may be controlling hysteresis in the opamp (The reaction time between each change of state giving a little delay to prevent oscillation) This may vary between each setup according to the frequency of operation in each case.

    2) I do not really know how to calculate it but from my experience is that the lower the minimum current the higher the output choke would be. I have a software called Excellent IT which is also on this site to help out in calculation for smps transformers. IF in the input data of this software you will opt for a minimal current of say 0.2A the output inductor will get really high compared to say an a minimal current of 1A. You will see that this will offer the necessary continuous voltage at the output while the dead time gets bigger at minimum pulse width. I guess with higher inductance there is more stored voltage in the inductor to compensate for this. The impedance will also be larger but this at low current will not offer much voltage drop. The type of copper wire used being stranded or solid. I have always seen in ATX smps that this would be of the solid type in the DC line. If say an inductor is used where the magnetic field is changing (like the one used in LLC smps in the center of the half bridge) I have seen both solid and also stranded ecw in pro amplifiers.
    As you say the stored charge in the capacitors may overcome the CC regulation in the initial surge. This may not happen because often there should always be a minimum current shunt resistor for this purpose and also not to let the chip work in discontinuous mode. This resistor may be put before the ampere meter not to interfere with the actual output current going to the PSU terminals.
    [I]Saturation [/I]A blog exist on this site where a chap with the name Fortytwo made a saturation tester.
    here is the link. https://www.diysmps.com/forums/entry...ation-testerII

    3) I think the output capacitance will depend to what ripple voltage at load is estimated. This will also get calculated according to the output inductance (choke) used. Well to sum this up the higher the output choke the less capacitance needed to maintain the calculated ripple voltage at load. Keep in mind that its not only the capacitance used but also the ESR of these capacitors. Low ESR capacitors need be less in capacitance for the same performance when using normal capacitors at the output.

    4) [I]The power dissipation in a given transistor will depend on the facts that I mention here.[/I] The saturation voltage (BJT) or RDS on (mosfet) The lower these get the less dissipation. The switching frequency has to be taken also in account as the more switching cycles per second the more switching losses. In each case one can check the datasheet and see the graph for a given current versus frequency in operation, with this you can calculate the total dissipation. Do not forget to check the temperature graph also.

    5) Synchronous rectification I would consider when I have higher currents over 15A depending also on the duty cycle of the smps in use. I would prefer to use scottkey diodes that can handle the maximum voltage used in your case.

    6) PFC up to 500w, I do not think it is really needed. You have the output regulated by PWM so if you have enough headroom in the output the output voltage at load will still be maintained. Active PFC will enhance efficiency but remember this is a lab PSU which will be used occasionally. [I]What I learnt and seen[/I]. Feed properly the input capacitor and voltage drop at the halfbridge will not be bad. Use a rule of thumb for 1uF to 3uF per watt of power across the the 320v rail. if you are using 200v capacitors than double the value of capacitance. This is because capacitors in series will halve the capacitance value.

    Regards Silvio
    Last edited by Silvio; 06-25-2020 at 08:48 PM.

  3. #23
    Thank for versetile reply!

    Currently I'm testing this supply, but the problem is, that the transistors are heating up very quickly - when drawing even only 10 or 20W the transistors (on heatsink from atx, without fan) reach 80 degrees celsius after a couple of minutes! What could be the problem then? Are some waveforms needed? I will try the same setup and check if working atx is also heating like that...
    About feedback - no more oscillations and I got regulation down to 40mA, while still being able to source 5-6A. But the full power is imposiible at this moment - transistors are going supper hot! Orginally author worte, that transistors will be cooler than output diode, but in my case, the transistors are blazzing hot (melating plastic insulators on screws) and diode is only slightly warm. Any Idea?

  4. #24
    .... Silvio's Avatar
    Join Date
    Apr 2015
    Location
    Malta
    Posts
    887
    Blog Entries
    9
    One of these things will make the transistors heat up.
    1) Transistors are not genuine and cannot withstand the load
    2) Dead time too short and one transistor is switching on before the other had completely shut off.
    3) When pulse width is very short the transistors are not fully on and working in the linear mode. See that the amplitude of the pulse going to the base has enough voltage to drive hard on the transistors.
    4) heat sink too small.
    5) If all is well from the above then try better transistors with more power dissipation
    Lastly try loading the smps at say 20v and see if it has the same result. It happend to me but when loading more than 3A at 12v. On higher voltage all was ok and could load to 5 or 6A

  5. #25
    Thank for a lot of options!

    I managed to get another atx psu - and checked waveforms (I will post them soon) And what I can see: My and commercial one looks quite the same - similar rise and fall times, similar shape etc. But my SMPS is working in the pwm range from very little to approx 50% (deadtime) when commercial one is never going to duty cycle less than about 10-15%. Thats probably due to the fact that it's output voltage is constant. And I have a thinking that this small pulses cause the heating up of the transistors.

    checklist:
    1. I tried with chinese mje13007 and also with 2sc5027 from atx supply - the same result (the atx was not running that hot, so not a problem with parameters of transistors)
    2. Heating is existant with small duty cycle, so not deadtime issue.
    3. Voltage is I think enough (I will post waveforms), and The too short time is, what might be te problem.
    4. Heatsink is from atx psu with similar power, currently without fan, but i want to add temperature controlled one.
    5. Transistors that I tried are I think even too large for only 200W... And commercial psu's work with them.

    I will make this test, when posting waveforms.

    So, as I observed, the psu is might work with too small duty cycle when set to too little voltage. I want to get the regulation from approx. 0V, And in low voltage I can live with such heating, but in higher powers this is impossible!

    What I have in mind:

    My power transformer is currently 58 turns primary and 2x18 turns secondary. Atx power supplies (of power <=350W ) have usually 40-50 turns primary and the 12V secondary is 2x4 turns (am I right?) so, if i wind for example 45 turns primary and only 2x12 turns secondary, I will force the smps to work with higher duty cycle (right?).

    And last thing: Is EI 33/34 core from chinese atx psu's sufficent for this 200W converter? (of course continous load). The core has about 1,2cm(2) center leg cross section area. It will be in front of fun (like in atx supplies). If this is sufficient, what current density should I choose for such transformer 5A/mm(2) is ok?

    And that's the moment when I see, that regulated SMPS is harder to make than unregulated one...
    Last edited by fifi_22; 07-20-2020 at 06:33 AM.

  6. #26
    .... Silvio's Avatar
    Join Date
    Apr 2015
    Location
    Malta
    Posts
    887
    Blog Entries
    9
    Hi FIFI, so now you know that its not that simple to make a variable regulated power supply as these issues will crop up at times. Please note that due to different operating frequency the number of turns varies accordingly. Note also that usually the switching frequency of such ATX supplies are around 20Khz that is why they have around 50 turns in the primary. If your operating frequency is 50Khz if I am not mistaken than your transformer needs less turns in the primary.

    You can work out again your number of turns with this formula and see what you get

    155 x 10^8 / 4 x freq in Hz x flux density in guass (2000g for 20Khz operation) x core area in cm^2

    For 50Khz use 1700 guass.

    Regarding the short pulse could be the problem as there is not enough time to switch the transistor fully. I suggest you try a higher voltage at the base so that at short pulse width it switches faster. Be careful not to exceed the maximum base voltage on the transistors. You can rise the voltage at the primary of the BDT so that it will produce a higher voltage at the base or add more turns to the BDT secondaries.

    One other task you can adopt is to pre load the smps with a couple of high wattage resistors so that the pulse width will widen a little more. You can always place the ampere meter after the resistors and read what is actually adding to the pre loading on the output.

    One other option is to divide the secondary voltages in half by adding another tapping. which will have say a maximum voltage of 12v. This will allow the pulse width to go wider on minimum voltages. The tappings then can be switched externally according to the voltage needed at the output.

    Regarding current density depend on the current needed at the output. You will have to make a compromise here as at lower voltages will need more current for a given power.
    What I can tell you here is this For continuous rating without forced cooling you will need a current density of 3.2A per mm^2 while at 50% duty cycle you may use 6A per mm^2. Some cooling of the heatsink and on the transformer is recommended. The intensity of the cooling force depend on the current drawn and also on the current density of the copper wire used. The heatsink also plays an important part here the larger the better. Why not try a cpu heatsink from an old computer? These are very efficient and have a lot of dissipation in them. You can solder your transistors underneath the pcb on the copper side and fit the heatsink below the pcb. This will allow more room to fit the large heatsink .

    Regards, Silvio
    Last edited by Silvio; 07-20-2020 at 07:45 AM.

  7. #27
    Thanks for superfast help.

    spec: 0-35V 20mA-6A sf = 50kHz
    Using this formula I got ~32 turns for primary, so about 2x5 or 2x6 turns for secondary.
    And primary current = 200W / 325V = 0.8A with 3.5A/mm^2 it gives 0.22mm^2, so 0.5mm wire diameter. I have the closest 0.4mm, so I will use 2 strands of it. In the end: 2x0.4mm
    Secondary current = 6A with 3.5A/mm^2 it gives 1.7mm^2, so about 1.5mm wire diameter. I will use 6 strands of 0.6mm wire. In the end: 2x(6x0.6mm)

    I want to start by making the transformer correct, and then try other things. Are those numbers ok?

    Loading SMPS would work, but then it doeasn't make sense to use switching technology - linear one would give tha same heat...
    I've got some CPU coolers, but I want to save them for some more powerful projects. Milioons of these chinese atx psu's work everyday in milions of computers with those small heatsinks. And the fan will be temperature controlled.

    Also: Should I split the primary winding in half?

  8. #28
    .... Silvio's Avatar
    Join Date
    Apr 2015
    Location
    Malta
    Posts
    887
    Blog Entries
    9
    Hi FIFI please note that the voltage at the half bridge is 155v and not 320. The current has to be calculated at 155v so for 250w out (taking into account that the smps is 80% efficient) so 250 / 155 =1.6 Amps.

    Ampere turns. Copper wire is 0.4mm so 0.2 x 0.2 x 3.142 = 0.12568 mm^2 for a single wire. so 0.12568 x 3.5A= 0.44A

    Primary number of 0.4mm wires = 1.6A / 0.44 = 3.6 wires. You have to either choose 3 or 4 wires.

    Here If you choose 3 wires this becomes a current density of around 5A per mm^2.

    It is up to you to see what is the duty cycle of the smps and the current drawn. If you are going to use forced cooling or not.
    You also have to see how the wire will fit in the transformer. You can twist the wires together to make a single bundle or you can put them side by side. Try to avoid partial layers. If possible use full layers on the bobbin to get maximum coupling. Do not forget to leave some margin space at the ends for safety. Use mylar or capton tape and nothing else. Windings tend to get hot and inferior tapes will go soft, shrink or deteriorate with heat.

    It is good practice to split the primary winding and sandwich the secondary in between. This will enhance better coupling in the transformer and minimizes leakage inductance.

    Regarding the ATX supplies do not forget that every voltage has its own winding and non of them is variable. The 5v usually carries most of the power and is regulated. The other voltages have to get there somehow. The 3.3v usually has its own regulator. Yours may be a different story.

    I am posting a small schematic to give you an idea how to split the voltages in half when using low voltage to help with widening the pulse width. You will also get double the current at the low voltage setting because you be using two windings instead of one

    I suggest watching my video on youtube how to wind transformers for smps. Take time to see it all, there is a lot to learn from it. Link https://www.youtube.com/watch?v=_K3ixhcTYFg

    I cannot upload any pictures there is still some problems with this site.

    Here you can find the diagram https://imgur.com/a/5p3dpEx

    Regards Silvio

  9. #29
    Once more, thanks for BIG help.

    There is a progres!
    I rewound transformer in such a way:
    Primary: 2x0.45 32 turns
    Secondary: 5x0.6 8+8 turns

    1/2 primary - 1/2 secondary - 2/2 secondary - 2/2 primary

    So output voltage for maximum duty cycle (50%) should be not less than: 8*155 / 32 = 38.5V Unfortunetly, under load (4A) the voltage falls to 30V... the smps is giving maximum duty cycle, but it cannot give enough voltage. I knew that there will be some drop, but I wanted to have at least 35V output. Is increasing secondary turns a good way to do it?

    Note, that te voltage on 325V caps never falls to less than 310V - measured

    Transistors doesn't heat up anymore! The snubber on the primary winding had 2n2 and 50ohm, It was taken from atx, but one that worked with 20Khz, changed capacitor to 680pF and now the temperature after 5minutes without load rizes not more than few degrees!

    When loaded to 100W, the transistors heat up to 60 degrees (heatsink is placed badly (horizontally) - no convecion ), but adding only a small fan brings the temp down to 36 degrees, which I LIKE!

    Output diode at 4Amps and massive heatsink (only for test) is cold!

    So for now there is only the voltage drop, and it looks better and better.

  10. #30
    .... Silvio's Avatar
    Join Date
    Apr 2015
    Location
    Malta
    Posts
    887
    Blog Entries
    9
    Hi FIFI, good to hear that you are making progress. Regarding voltage drop it is inevitable especially this is due to some more factors that have to be considered.

    1 Voltage drop at the input capacitor
    2 switching losses in transistors
    3 copper losses in the transformer
    4 core losses
    5 loses in the output diodes
    6 voltage drop in the output inductor
    7 voltage drop in the input bridge rectifier.

    Summing all this up as a rule of thumb your calculation for a regulated power supply this should be calculated by multiplying the maximum output voltage needed by 1.414.

    This is in your case is 35v x 1.414 = 49.5v. So your secondary voltage must be around 49v unregulated. This will provide adequate headroom to compensate for the losses mentioned above and also there will be room for the chip to regulate the output voltage when under load. A difference in a volt here and there will not make much difference. In your case if the secondary winding had to be with 10 turns it will bring an output unregulated voltage of around 48v which will also suffice.

    Lastly 4 wires of 0.6mm dia will give you enough current (4.5A @ 4A pre mm^2) If you are making a double winding in the output then at lower voltage range this will give you 9A-10A which I think is more than enough.

    Regards Silvio
    Last edited by Silvio; 07-23-2020 at 08:34 PM.

Similar Threads

  1. TL494 Lab PSU
    By bogdan2011 in forum diysmps
    Replies: 27
    Last Post: 07-07-2019, 02:29 PM
  2. Tl494 smps
    By noorworkshop in forum Switching Power Supplies Informations
    Replies: 26
    Last Post: 06-13-2018, 06:59 PM
  3. TL494 Class D Amp
    By noorworkshop in forum Class D
    Replies: 1
    Last Post: 05-15-2018, 02:49 PM
  4. Replies: 7
    Last Post: 07-24-2014, 06:20 PM
  5. 1kW halbridge smps with PWM control
    By mych3 in forum diysmps
    Replies: 4
    Last Post: 07-22-2011, 09:37 PM

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •