In the transformer of the prototype I've shown I use 48 primary turns [24 turns + secondaries + 24 turns]. Secondaries are 4 + 4 turns
Whith this setup, the core requires about 360V p-p square wave at 36Khz to saturate at about 25ºC [empirically tested since I have no specs from any of the ferrites I use]. In normal working conditions this volts*time/tirms product is never reached since the regulation imposes a lower duty cycle for higher input voltages. Duty cycle reaches 100% at about 210V input and decreases proportionally with increasing input [about 65% for 320V]. Anyway, it has a more than reasonable security margin to accomodate core property changes with temperature [ferrites tend to saturate earlier when hot]
Whith this setup, the core requires about 360V p-p square wave at 36Khz to saturate at about 25ºC [empirically tested since I have no specs from any of the ferrites I use]. In normal working conditions this volts*time/tirms product is never reached since the regulation imposes a lower duty cycle for higher input voltages. Duty cycle reaches 100% at about 210V input and decreases proportionally with increasing input [about 65% for 320V]. Anyway, it has a more than reasonable security margin to accomodate core property changes with temperature [ferrites tend to saturate earlier when hot]
Hi Eva!
You are really showing impressive results with that BJTs!
I like BJTs in some applications as their conductive losses do
not increase at high temperatures. But I always was strugling to provide the base current without high losses in the base drive circuit as these beasts need the base current all the conductive time, not just during switching....
Only where I could use a transformer in a self resonant topology without much signal conditioning in the base drive... only in such
circuits I felt comfortable with the overall losses.
I can just envy you for your great results, also for your fast
switching!!!
Bye
Markus
You are really showing impressive results with that BJTs!
I like BJTs in some applications as their conductive losses do
not increase at high temperatures. But I always was strugling to provide the base current without high losses in the base drive circuit as these beasts need the base current all the conductive time, not just during switching....
Only where I could use a transformer in a self resonant topology without much signal conditioning in the base drive... only in such
circuits I felt comfortable with the overall losses.
I can just envy you for your great results, also for your fast
switching!!!
Bye
Markus
Bipolars are very tricky. They require a fast rising positive current pulse through the base at turn on to get quickly into saturation and reduce turn-on losses, and a big negative base current pulse at turn off to get fast crossover and fall times and low turn-off losses, and I'm using fast modern bipolars and both speed-up pulses
I will try to draw and post schematics for the base drive circuit I'm using. It's a compromise between complexity and losses. Adding more complexity would allow for lower losses but then I could use IGBTs instead of bipolars
I will try to draw and post schematics for the base drive circuit I'm using. It's a compromise between complexity and losses. Adding more complexity would allow for lower losses but then I could use IGBTs instead of bipolars
HI
I find the software to calculate the etd core and turn ratio.
software link :
http://henry.fbe.fh-darmstadt.de/smps_e/hgw_smps_in_e.html
I find the software to calculate the etd core and turn ratio.
software link :
http://henry.fbe.fh-darmstadt.de/smps_e/hgw_smps_in_e.html
This software suggests 54 turns primary for a 42/21/20 core while I'm succesfully using 48 in a full bridge and that would mean 24 in a half bridge [what the software actually simulates]
I think the best method to get minimum turn counts for a given core is to build actual transformer prototypes and test it measuring current vs time [looking at current vs time waveform, saturation appears very obvious] using a simple pulse generator with adjustable frequency and pulse width [allowing the core to reset between pulses]. I use a SG3525, a 60A MOSFET, a freewheel diode with suitable resistor in series, a 15V supply and 2x 10.000uF capacitors for that purpose [capacitors provide the peak currents required to mantain Vcc=15 V to some extent during saturarion]. I use the same setup to test inductors to know their inductance vs current curves [big gapped ferrite and iron powder inductors tend to show very progressive saturation characteristics so they are also usable in the saturation region]
Knowing how much time the core takes to saturate for a given turn count and Vcc=15V allows to calculate turn counts for any operating Vcc and frequency
Suppose a small core that with 11 turns and 15V applied takes 7.5uS to saturate at 25ºC [the time the current takes to skyrocket]. This means that with these 15V/11turns the flux takes 7.5us to travel from 0 to its maximum value before saturation. Giving some security margin for temperature variations let's assume only 5uS
In normal push-pull working conditions the flux travels from negative to positive so the core could withstand 10uS pulses without saturating. Two complementary pulses [full negative to positive plus full positive to negative flux swing] would last for 20uS so the minimum operating frequency at 15V/11turns should be 50Khz for that core [100khz PMW oscillator]
This is easily scalable, for example, 25Khz operation at 15V would require 22 turns and 30V operation at 25Khz would require 44 turns, etc...
For my 42/21/20 core, minimum operating frequency at 15V with 2 turns appears to be 41.6Khz [with some security margin] so 48 turns would allow for 315V at 36Khz reliably. At 25ºC the core would actually require more than 360V to saturate, may be even 420V [not likely to happen]. Due to the nature of a transformer coupled buck converter, the duty cycle is reduced for higher input voltages so a 315V input would only be applied to the core for 66% time and the core would be shorted for the remaining 34% time. This is the same as applying 208V with 100% duty cycle so there is plenty of margin for the dynamic behavior of the converter [duty cycle is slightly bigger than expected during a brief period of time following any increase in load and shorter than expected for a moment following any load reduction]
I could reduce those 48 turns to maybe even 36 turns without saturation but then, if for any reason the converter goes into discontinuous mode and some pulses are missed or asymmetric [no proper flux resetting], the core would saturate easier and the transistors of the full bridge would blow easier
[I've seen that behavior when playing with the feedback loop, when testing current limiting behavior, etc...]
I think the best method to get minimum turn counts for a given core is to build actual transformer prototypes and test it measuring current vs time [looking at current vs time waveform, saturation appears very obvious] using a simple pulse generator with adjustable frequency and pulse width [allowing the core to reset between pulses]. I use a SG3525, a 60A MOSFET, a freewheel diode with suitable resistor in series, a 15V supply and 2x 10.000uF capacitors for that purpose [capacitors provide the peak currents required to mantain Vcc=15 V to some extent during saturarion]. I use the same setup to test inductors to know their inductance vs current curves [big gapped ferrite and iron powder inductors tend to show very progressive saturation characteristics so they are also usable in the saturation region]
Knowing how much time the core takes to saturate for a given turn count and Vcc=15V allows to calculate turn counts for any operating Vcc and frequency
Suppose a small core that with 11 turns and 15V applied takes 7.5uS to saturate at 25ºC [the time the current takes to skyrocket]. This means that with these 15V/11turns the flux takes 7.5us to travel from 0 to its maximum value before saturation. Giving some security margin for temperature variations let's assume only 5uS
In normal push-pull working conditions the flux travels from negative to positive so the core could withstand 10uS pulses without saturating. Two complementary pulses [full negative to positive plus full positive to negative flux swing] would last for 20uS so the minimum operating frequency at 15V/11turns should be 50Khz for that core [100khz PMW oscillator]
This is easily scalable, for example, 25Khz operation at 15V would require 22 turns and 30V operation at 25Khz would require 44 turns, etc...
For my 42/21/20 core, minimum operating frequency at 15V with 2 turns appears to be 41.6Khz [with some security margin] so 48 turns would allow for 315V at 36Khz reliably. At 25ºC the core would actually require more than 360V to saturate, may be even 420V [not likely to happen]. Due to the nature of a transformer coupled buck converter, the duty cycle is reduced for higher input voltages so a 315V input would only be applied to the core for 66% time and the core would be shorted for the remaining 34% time. This is the same as applying 208V with 100% duty cycle so there is plenty of margin for the dynamic behavior of the converter [duty cycle is slightly bigger than expected during a brief period of time following any increase in load and shorter than expected for a moment following any load reduction]
I could reduce those 48 turns to maybe even 36 turns without saturation but then, if for any reason the converter goes into discontinuous mode and some pulses are missed or asymmetric [no proper flux resetting], the core would saturate easier and the transistors of the full bridge would blow easier
[I've seen that behavior when playing with the feedback loop, when testing current limiting behavior, etc...]Hi Eva!
...yes bipolars are tricky...
Another approach to get them fast is to cascode them with a cheap low voltage MosFet... In fact this also achieves the required high positive base current peak for turn on and the
strong negative current peak for turn off.
ST microelectronics call this "emitter switched"....
http://www.st.com/stonline/books/pdf/docs/9464.pdf
http://www.spoerle.com/binaries/1071219559997-STC03DE150_Nov03.pdf
And further more:
http://www.circuitprotection.com/techpapers/BJT6.pdf
But of course all this also requires some increased efforts in the driving circuits.
...looking curiously forward to your drive circuits.
Bye
Markus
P.S.
...always great to see your support for George Ho...
...yes bipolars are tricky...
Another approach to get them fast is to cascode them with a cheap low voltage MosFet... In fact this also achieves the required high positive base current peak for turn on and the
strong negative current peak for turn off.
ST microelectronics call this "emitter switched"....
http://www.st.com/stonline/books/pdf/docs/9464.pdf
http://www.spoerle.com/binaries/1071219559997-STC03DE150_Nov03.pdf
And further more:
http://www.circuitprotection.com/techpapers/BJT6.pdf
But of course all this also requires some increased efforts in the driving circuits.
...looking curiously forward to your drive circuits.
Bye
Markus
P.S.
...always great to see your support for George Ho...
Very good discussion about transformer saturation calculations, EVA!
Are you using a half-bridge topology for your offline SMPS?
You talked about 40turns in the primary, 4+4 in the secondary, more or less. That's about what you get for a +/-40V smps from 12V in a car, although connected backwards (not using the primary center-tap).
Just for curiosity: have you tried to build an offline SMPS with the same ferrite toroid of a CAR SMPS?
Almost all the designs I have seen in the web are built on E / ETD cores, is a ferrite toroid also suitable?
Another question: given a theorethical 50% duty cycle in each mosfet of a half-bridge design, what is the required turns ratio for a given rail voltage at the output and a given input voltage? For example, for +/-40V output and 320V input, assuming 40 turns in the primary, what should I wind in the secondary side, 10+10 turns or so?
Thanks EVA, your help is greatly appreciated
If you don't mind to answer, do you live in Madrid?
Are you using a half-bridge topology for your offline SMPS?
You talked about 40turns in the primary, 4+4 in the secondary, more or less. That's about what you get for a +/-40V smps from 12V in a car, although connected backwards (not using the primary center-tap).
Just for curiosity: have you tried to build an offline SMPS with the same ferrite toroid of a CAR SMPS?
Almost all the designs I have seen in the web are built on E / ETD cores, is a ferrite toroid also suitable?
Another question: given a theorethical 50% duty cycle in each mosfet of a half-bridge design, what is the required turns ratio for a given rail voltage at the output and a given input voltage? For example, for +/-40V output and 320V input, assuming 40 turns in the primary, what should I wind in the secondary side, 10+10 turns or so?
Thanks EVA, your help is greatly appreciated
If you don't mind to answer, do you live in Madrid?
Chocoholic :
Sorry, I've been very busy and I still have no decent schematics drawn
ssanmor :
I'm using a full bridge to keep current requirements low and be able to use cheap bipolar transistors
I think I've already discussed the use of toroid cores for off-line SMPS. These cores are advantageous for car-audio where turn counts and ratios are low and decent line isolation [3,7 KV] is not required. I find toroid cores not suitable for off-line SMPS because it's hard to add proper isolation, even harder with sandwitched secondaries, high turn counts are hard to achieve and leakage inductance is not better than in EE cores of similar size
Turn ratios are not hard to calculate. All you need to know is the minimum rectified voltage including ripple [usually 200V for 160V to 250V AC at full load], the maximum duty cycle the circuit can achieve [usually 90% to 95%] and the voltage drops on the primary and the secondary sides due to resistive losses
(Input_DC - primary_V_drop) * (turns_sec / turns_pri) * max_duty_cycle = output_DC + secondary_V_drop
This gives the sec/pri turn ratio. Then the actual turn counts have to be adjusted in order to avoid saturation [too few turns] and excessive leakage inductance or not enough winding space [too much turns]
For my design : (215 - 5) * 4/48 * .90 = 14.4 + 1
Pretty simple I think
Sorry, I've been very busy and I still have no decent schematics drawn
ssanmor :
I'm using a full bridge to keep current requirements low and be able to use cheap bipolar transistors
I think I've already discussed the use of toroid cores for off-line SMPS. These cores are advantageous for car-audio where turn counts and ratios are low and decent line isolation [3,7 KV] is not required. I find toroid cores not suitable for off-line SMPS because it's hard to add proper isolation, even harder with sandwitched secondaries, high turn counts are hard to achieve and leakage inductance is not better than in EE cores of similar size
Turn ratios are not hard to calculate. All you need to know is the minimum rectified voltage including ripple [usually 200V for 160V to 250V AC at full load], the maximum duty cycle the circuit can achieve [usually 90% to 95%] and the voltage drops on the primary and the secondary sides due to resistive losses
(Input_DC - primary_V_drop) * (turns_sec / turns_pri) * max_duty_cycle = output_DC + secondary_V_drop
This gives the sec/pri turn ratio. Then the actual turn counts have to be adjusted in order to avoid saturation [too few turns] and excessive leakage inductance or not enough winding space [too much turns]
For my design : (215 - 5) * 4/48 * .90 = 14.4 + 1
Pretty simple I think
*OFFTOPIC*
Hi Eva!
Nice to see you again.
No need to appologize..., no need to hurry...
Somehow my subwoofer is also keeping me busy.
...just wanted to design and built a better sub...
....and now I am learning about class D amps,
piezo accelerometers and SMPS...
The only difference between men and boys
is the price of their toys.....
*ONTOPIC*
I think for safety of off line SMPS usuall not only a high isolation voltage is mandatory, but also minimum creepages, clearances, number of isolation layers and thickness of isolation...
I assume for audio applications the standards for SELV and SELV equivalent requirements would be applicable.
Anybody out there, who is in charge of safety standards
for audio electronics?
My current focus would be IEC/Europe, less ANSI/NAFTA .
Bye
Markus
Hi Eva!
Nice to see you again.
No need to appologize..., no need to hurry...
Somehow my subwoofer is also keeping me busy.
...just wanted to design and built a better sub...
....and now I am learning about class D amps,
piezo accelerometers and SMPS...
The only difference between men and boys
is the price of their toys.....
*ONTOPIC*
I think for safety of off line SMPS usuall not only a high isolation voltage is mandatory, but also minimum creepages, clearances, number of isolation layers and thickness of isolation...
I assume for audio applications the standards for SELV and SELV equivalent requirements would be applicable.
Anybody out there, who is in charge of safety standards
for audio electronics?
My current focus would be IEC/Europe, less ANSI/NAFTA .
Bye
Markus
Thanks, EVA.
Are the equations for half-bridge the same as for full-bridge?
About toroids, I can understand that isolation can't be as good as with ETDs. But what if a coated ferrite is chosen, so no problems with the primary, and then add a layer of isolation tape or something like that, then the secondary? Is there some other problem I am missing?
And about the number of turns, well, 48 : 4+4 is not very different of what you need in a car supply. At most, you would need 48 : 7+7 for +/-40V output or so at 320VDC, right?
Thanks!
Sergio from Madrid
Are the equations for half-bridge the same as for full-bridge?
About toroids, I can understand that isolation can't be as good as with ETDs. But what if a coated ferrite is chosen, so no problems with the primary, and then add a layer of isolation tape or something like that, then the secondary? Is there some other problem I am missing?
And about the number of turns, well, 48 : 4+4 is not very different of what you need in a car supply. At most, you would need 48 : 7+7 for +/-40V output or so at 320VDC, right?
Thanks!
Sergio from Madrid
I was trying to point out that applying a layer of isolation tape to a wound toroid is a real pain, not to say about applying two or more isolation layers due to sandwitching
About your calculations : You can't assume 320V DC rectified mains since mains voltage fluctuates and sags badly at full load, even falling to 200V DC. My calculations were for a full bridge transformer-coupled buck converter [regulated]. I strongly recommend regulated output. For a half bridge take half the DC rectified input voltage
About your calculations : You can't assume 320V DC rectified mains since mains voltage fluctuates and sags badly at full load, even falling to 200V DC. My calculations were for a full bridge transformer-coupled buck converter [regulated]. I strongly recommend regulated output. For a half bridge take half the DC rectified input voltage
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.