So I am studying the ATX power supplies for few days.
Now I have few questions.
1. TL494 is switching the 340 DC volts at some frequency lets say 50khz. Now my question is what voltage goes to the EI ferrite core at the primary winding. At that frequency I think there wont be 340V or am I wrong? As far as I know by changing the switching frequency you change the primary voltage.
2. I would like to rewind ATX PSU EI ferrite transformer (EI-33) so I can get 2x36V out of PC ATX. Now my question is how to calculate the number of windings. Should I use 220V as input parameter for the primary winding or should I use 340V because AC-DC will be 340V or should I use other voltage because of the switching frequency?
3. I know how to calculate standard laminated EI transformer, but can I use those calculations for these EI ferrite?
4. Can someone tell me the way to calculate EI ferrite cores for windings wire tickness etc etc.
5. How can I know at what frequency the PSU is working and is that frequency variable? What will happen if it is changed.
Thanks.
Now I have few questions.
1. TL494 is switching the 340 DC volts at some frequency lets say 50khz. Now my question is what voltage goes to the EI ferrite core at the primary winding. At that frequency I think there wont be 340V or am I wrong? As far as I know by changing the switching frequency you change the primary voltage.
2. I would like to rewind ATX PSU EI ferrite transformer (EI-33) so I can get 2x36V out of PC ATX. Now my question is how to calculate the number of windings. Should I use 220V as input parameter for the primary winding or should I use 340V because AC-DC will be 340V or should I use other voltage because of the switching frequency?
3. I know how to calculate standard laminated EI transformer, but can I use those calculations for these EI ferrite?
4. Can someone tell me the way to calculate EI ferrite cores for windings wire tickness etc etc.
5. How can I know at what frequency the PSU is working and is that frequency variable? What will happen if it is changed.
Thanks.
Last edited:
1) The classic TL494 is normally operating at half bridge mode: the primary is connected between the two series transistor operating 180 deg out of phase, and the middle point of two eletro caps, via a bipolar 1uF to decouple DC in the winding and prevent saturation. The total primary voltage is the same as the peak rectified line voltage, in example for 220V supply, it is about 310VDC.
2) Unwind the xformer, you will find a half of primary, isolation, the secondaries, isolation, half primary. Simply unwound the secondary and scale the turns to match your desired voltage, in this case 3 times the 12V turns will give 36V.
3) It is equal, but the inductions are about 500 gauss.
4) There is a good book called "Switching Power supplies", from Abraham Pressman. I have one of them.
5) For large increase in frequency will give increased semiconductor looses, decreasing frequency will saturate the core and destroy power stage.
Have in mind that these topology of SMPS don´t support sudden changes in the output current, they are designed for steady loads. For Audio, it is better the Fly Back topology, I have one designed of my own since 12 years ago.
Good luck.
2) Unwind the xformer, you will find a half of primary, isolation, the secondaries, isolation, half primary. Simply unwound the secondary and scale the turns to match your desired voltage, in this case 3 times the 12V turns will give 36V.
3) It is equal, but the inductions are about 500 gauss.
4) There is a good book called "Switching Power supplies", from Abraham Pressman. I have one of them.
5) For large increase in frequency will give increased semiconductor looses, decreasing frequency will saturate the core and destroy power stage.
Have in mind that these topology of SMPS don´t support sudden changes in the output current, they are designed for steady loads. For Audio, it is better the Fly Back topology, I have one designed of my own since 12 years ago.
Good luck.
Last edited:
Wow man you helped me so much. All the questions are answered THANK YOU SO MUCH. I wanted to modify ATX for my amplifier, you say that is bad idea? What if I use LARGE Capacitors at the output, won't this stabilize the sudden current drops little?
The total primary voltage is the same as the peak rectified line voltage, in example for 220V supply, it is about 310VDC.
This is what I needed to know.
Simply unwound the secondary and scale the turns to match your desired voltage, in this case 3 times the 12V turns will give 36V.
I was thinking about this but I was not that sure now I know I am on right track. THANKS.
BTW I know how to modify the other parts of the PSU to stand higher voltages. I'm changing capacitors, I disabled over-voltage circuit, I make new divider resistors for the TL494 etc etc but my problem is the core.
The total primary voltage is the same as the peak rectified line voltage, in example for 220V supply, it is about 310VDC.
This is what I needed to know.
Simply unwound the secondary and scale the turns to match your desired voltage, in this case 3 times the 12V turns will give 36V.
I was thinking about this but I was not that sure now I know I am on right track. THANKS.
BTW I know how to modify the other parts of the PSU to stand higher voltages. I'm changing capacitors, I disabled over-voltage circuit, I make new divider resistors for the TL494 etc etc but my problem is the core.
Last edited:
No, augmenting output capacitors make things worse.
See: the rectangular high voltage is scaled inside the transformer, to the levels required at the output. Then, the bipolar squeare wave is full wave rectified, and integrated in a LC filter, designed to resonate at some frequency well below the switching frequency. (say 1KHz) Then, at the DC output, the IC senses the DC level and generate PWM accordingly to compensate it. In a sudden change in out voltage, the inductor momentarily "runs dry", the current in it goes zero during between pulses from transformer. The IC senses the drop out, and react to it, but the reaction can´t be instantaneously because of the time constant in the filter itself. Then, the voltage is reestablished. If you increase the cap´s, decrease simultaneously the cut off frequency of the filters, so the drop will be more pronounced and more large in time.
The inverse situation also is true when the supply is suddenly unloaded.
And, in case of you audio material appears at the same frequency of the cut off frequency of the filter, not only there will be a high output ripple at the output, the SMPS all itself can make unstable and there exist risk of short circuit of the output transistors.
See it like a spring loaded by a mass. If you apply the resonant frequency to it, oscillations can be so large with risk of break the wire of the spring.
Remember that if you increase the output voltage, the L of the inductor will be incremented proportionally, and the cap decreased in the same amount. In example, if you use 100uH and 1000uF for 12V, for a 36V you will need 300uHy and 330uF.
Best regards.
Last edited:
The output voltage will go very high and uncontrollable.
So you are saying that the best way for audio is Fly Back?
No doubt. It is the only that supports large current variation because of it intrinsic wide dynamic range when used in current mode (UC3842 or simmilar).
Btw can you send me on PM skype, msn? something? Thanks....
Sorry, I had Skype but there is large time I do not run such "social networks", nor skype, nor facebook, nor tweeter, nor... Only Email and chat.
Last edited:
Usually the seller don´t work about inside topology, but you quickly have an idea: search all type of SMPS´s that only have a unique transformer (and the line filters). IF near the diodes you see two or more cores of different styles, they aren´t Fly Back.
You can search an very old PC monitor from those that have power supply in a separate board and try to modify it. It surely is a Fly Back boundary mode type of SMPS.
You can search an very old PC monitor from those that have power supply in a separate board and try to modify it. It surely is a Fly Back boundary mode type of SMPS.
You are talking about the toroidal filters(the circle ferrites before the output capacitors LC circuit) or the EI cores (usually 3 of them 1 to separate high voltage from the driver circuit other to convert 220-12,5 and the last one is I think for current sensing)?
At least, a normal SMPS must have a input line filter located near the main rectifier, and the bulk cap. The power transformer, and CAN have a third ferrite core with output LC filter. (Usualy a toroid, but may be a RM, EI or other core shape.). All topologies, including half, full bridge, forward converter (1 or 2 transistors) DO NEED an LC filter at its output, except very bad designs. The unique that no need for LC filter is the Flyback, and for same power, the core for the mutually coupled inductors (IN FB it is wrong to call "transformer"), is bigger than for any other one topology.
The best way to know how it works, is to place working and see the oscillogram placing the probe near the transformer, no needly to connect it. The wave is very different, but too long to explain in words.
The best way to know how it works, is to place working and see the oscillogram placing the probe near the transformer, no needly to connect it. The wave is very different, but too long to explain in words.
> switching the 340 DC volts at some frequency lets say 50khz. Now my question is what voltage goes to the EI ferrite core at the primary winding.
Same voltage. >300V peak-to-peak pulses.
> At that frequency I think there wont be 340V or am I wrong?
Sure can be. Put your hand on top of a large AM broadcast tower, 1,000V at 1MHz.
> As far as I know by changing the switching frequency you change the primary voltage.
No, the chip changes the pulse duty cycle. If it hit 100%, the "5V" winding would go to 10V. The chip reduces the duty cycle to get 5V, increasing or decreasing to compensate for load and line variations.
> I can get 2x36V out of PC ATX. Now my question is how to calculate the number of windings.
What Osvaldo said. Do NOT change the 300V or 5V windings. The 5V (or 3.3V) winding is used for the feed-back to the controller chip. The 12V winding is not monitored for duty-cycle. Count its turns and compute 36V/12V.
BUT any good ATX supply has a second chip which monitors all the important outputs and will signal "Power Bad" and shut-down if ANY of them are out of bounds. So 36V into the "12V monitor" is a shut-down, so is zero V if you find and disconnect this monitor line.
You could count the turns on the 12V winding, replace it, and add an additional winding scaled for 36V. However a good design already uses large wire for low resistance and there may be no room in the window for an additional high-power low-sag winding.
And you have to have a pretty good load on the 5V/3.3V output to keep the main switcher happy. (The Duty Cycle can't go very-very low.) I often use an 8 ohm >5W on the 5V line.
And there's a trick to starting ATX but I'm sure you know that.
As Osvaldo says, sudden change of load may not stay in regulation. I'm not sure that audio is "sudden" compared to PC CPU loads, modern CPU "speed stepping", and unanticipated CPU-sleep.
I used ATX to power 12V car-audio. It worked but I did not like the sound. (And I am not very fussy.) Maybe Osvaldo is right, the power wobbles with the audio.
People have been trying switchers for 30 years, and there are still people who prefer Heavy Iron. The interaction between audio and a switcher control loop, and the supersonic garbage, are all difficult issues.
Same voltage. >300V peak-to-peak pulses.
> At that frequency I think there wont be 340V or am I wrong?
Sure can be. Put your hand on top of a large AM broadcast tower, 1,000V at 1MHz.
> As far as I know by changing the switching frequency you change the primary voltage.
No, the chip changes the pulse duty cycle. If it hit 100%, the "5V" winding would go to 10V. The chip reduces the duty cycle to get 5V, increasing or decreasing to compensate for load and line variations.
> I can get 2x36V out of PC ATX. Now my question is how to calculate the number of windings.
What Osvaldo said. Do NOT change the 300V or 5V windings. The 5V (or 3.3V) winding is used for the feed-back to the controller chip. The 12V winding is not monitored for duty-cycle. Count its turns and compute 36V/12V.
BUT any good ATX supply has a second chip which monitors all the important outputs and will signal "Power Bad" and shut-down if ANY of them are out of bounds. So 36V into the "12V monitor" is a shut-down, so is zero V if you find and disconnect this monitor line.
You could count the turns on the 12V winding, replace it, and add an additional winding scaled for 36V. However a good design already uses large wire for low resistance and there may be no room in the window for an additional high-power low-sag winding.
And you have to have a pretty good load on the 5V/3.3V output to keep the main switcher happy. (The Duty Cycle can't go very-very low.) I often use an 8 ohm >5W on the 5V line.
And there's a trick to starting ATX but I'm sure you know that.
As Osvaldo says, sudden change of load may not stay in regulation. I'm not sure that audio is "sudden" compared to PC CPU loads, modern CPU "speed stepping", and unanticipated CPU-sleep.
I used ATX to power 12V car-audio. It worked but I did not like the sound. (And I am not very fussy.) Maybe Osvaldo is right, the power wobbles with the audio.
People have been trying switchers for 30 years, and there are still people who prefer Heavy Iron. The interaction between audio and a switcher control loop, and the supersonic garbage, are all difficult issues.
PRR thank you for the nice post man.
As Osvaldo says, sudden change of load may not stay in regulation. I'm not sure that audio is "sudden" compared to PC CPU loads, modern CPU "speed stepping", and unanticipated CPU-sleep.
This was my tough too. :S
It worked but I did not like the sound. (And I am not very fussy.)
Really? Can you make that big difference by hearing it compared to standard linear PSU?
As Osvaldo says, sudden change of load may not stay in regulation. I'm not sure that audio is "sudden" compared to PC CPU loads, modern CPU "speed stepping", and unanticipated CPU-sleep.
This was my tough too. :S
It worked but I did not like the sound. (And I am not very fussy.)
Really? Can you make that big difference by hearing it compared to standard linear PSU?
To PRR:
Dear friend, I did the SMPS 12+ years ago, using PC monitor scrap, Nichicon Caps, two line filters, UC3842 and STP7N60 Flyback topology. From secondary, I get 3 voltages: +14V, -14V for TDA2006 woofer driver * 3 units, and +12V for TDA2003 for tweeters outputs, and I post-regulate 12V using L4974 synchronized with primary to get +5V for some logic IC´s, and two very old CD player (I don´t know if 1X or 2X) for PC, that has all buttons for FF, rew, play, stop, etc. This set also has AM and FM receiver made with TDA1220B and TDA7211 for FM front end. The SMPS is inside a box made of aluminium, and well earthed. Neither AM nor FM get noise from SMPS, but the TV yes does some interference noise in the screen (less than 50cm from supply). So the design is not too bad, given home brewed. So, I couldn´t understand why still using iron in commercial sets. The set drives 3 Karlson Speaker projects, 2 of 15" + 4" and a third 12" + 4" for difference channel (both L & R substracted and amplified indepently with it own volume control. I normally play electronic music with half full power and the supply doesn´t regulate power lines too bad. For the time I did it, I didn´t know about regulate both negative and positive rails to TL431, and only +14V is sensed. So, negative line has relative poor regulation. And in the primary (transistor side of PC817), I can see the audio frequency like if it was rectified, both peaks to positive. Although you possibly cannot believe me, the last week I had a supply trouble at my house: a lightning bolt strike in the town high voltage lines, so one of three power phase was broken, and I have between 20 to 170V (normally there must be 220V). The audio set above described, continues working properly from 220V down to only 36V on line, in where a high LF noise sounds in the core of the power mutually coupled inductor (transformer), and then it switch off by a too low voltage. I didn´t do such a test on my bench !!!
When referring to sudden changes, I mean deep bass like tom´s. Such a fast transient in a SMPS not designed for support it can loss stability and if the audio material is just on the resonance of the output LC filter, it may be hardly amplified by the Q of such filter, causing the SMPS entirely to oscillate and possibly destroyed. Fly back topology DOES NOT HAVE SUCH OUTPUT FILTER, in place I put a very small inductor before the first cap (3900uF 25V) and a small second cap (470uF) about 15KHz where normally very few amplitude transients of those frequency may appear at the amplifiers. (The supply is running about 70KHz). The sense resistors are located before such a filter, so it is not included in the feedback loop.
I still believe that Fly Back topology (of the classical ones) is the one can support audio loads, I don´t try yet some more new topologies such as resonant phase shift, etc.
Dear friend, I did the SMPS 12+ years ago, using PC monitor scrap, Nichicon Caps, two line filters, UC3842 and STP7N60 Flyback topology. From secondary, I get 3 voltages: +14V, -14V for TDA2006 woofer driver * 3 units, and +12V for TDA2003 for tweeters outputs, and I post-regulate 12V using L4974 synchronized with primary to get +5V for some logic IC´s, and two very old CD player (I don´t know if 1X or 2X) for PC, that has all buttons for FF, rew, play, stop, etc. This set also has AM and FM receiver made with TDA1220B and TDA7211 for FM front end. The SMPS is inside a box made of aluminium, and well earthed. Neither AM nor FM get noise from SMPS, but the TV yes does some interference noise in the screen (less than 50cm from supply). So the design is not too bad, given home brewed. So, I couldn´t understand why still using iron in commercial sets. The set drives 3 Karlson Speaker projects, 2 of 15" + 4" and a third 12" + 4" for difference channel (both L & R substracted and amplified indepently with it own volume control. I normally play electronic music with half full power and the supply doesn´t regulate power lines too bad. For the time I did it, I didn´t know about regulate both negative and positive rails to TL431, and only +14V is sensed. So, negative line has relative poor regulation. And in the primary (transistor side of PC817), I can see the audio frequency like if it was rectified, both peaks to positive. Although you possibly cannot believe me, the last week I had a supply trouble at my house: a lightning bolt strike in the town high voltage lines, so one of three power phase was broken, and I have between 20 to 170V (normally there must be 220V). The audio set above described, continues working properly from 220V down to only 36V on line, in where a high LF noise sounds in the core of the power mutually coupled inductor (transformer), and then it switch off by a too low voltage. I didn´t do such a test on my bench !!!
When referring to sudden changes, I mean deep bass like tom´s. Such a fast transient in a SMPS not designed for support it can loss stability and if the audio material is just on the resonance of the output LC filter, it may be hardly amplified by the Q of such filter, causing the SMPS entirely to oscillate and possibly destroyed. Fly back topology DOES NOT HAVE SUCH OUTPUT FILTER, in place I put a very small inductor before the first cap (3900uF 25V) and a small second cap (470uF) about 15KHz where normally very few amplitude transients of those frequency may appear at the amplifiers. (The supply is running about 70KHz). The sense resistors are located before such a filter, so it is not included in the feedback loop.
I still believe that Fly Back topology (of the classical ones) is the one can support audio loads, I don´t try yet some more new topologies such as resonant phase shift, etc.
When referring to sudden changes, I mean deep bass like tom´s. Such a fast transient in a SMPS not designed for support it can loss stability and if the audio material is just on the resonance of the output LC filter,
yes I understand what you mean but there must be some kind of isolation circuit or filter that can amortize the sudden current spikes. :S
yes I understand what you mean but there must be some kind of isolation circuit or filter that can amortize the sudden current spikes. :S
Unfortunately, don´t exist, or I don´t know them. And if would exist, the this device can´t agree whit some mathematical laws as Kirchoff´s ones. Current in a node must summate 0 (zero). If such a device exist, summation is not zero, and the accumulation in some place must be.
Only solution may be incorporate more filters, but it gets new troubles. And, if you limit this spikes, then the amplifier can´t put the right amount of current to the output, distorting and clipping the it.
Only solution may be incorporate more filters, but it gets new troubles. And, if you limit this spikes, then the amplifier can´t put the right amount of current to the output, distorting and clipping the it.
Last edited:
Google "start ATX power supply"
start atx power supply - Google Search
(The "power supply" bit avoids a LOT of cruft about Austin TeXas, tax software, yadayada.)
start atx power supply - Google Search
(The "power supply" bit avoids a LOT of cruft about Austin TeXas, tax software, yadayada.)
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.