Rahul Bhai.....do send me the article...
mailto: floyds_ec@gawab.com
Me too interested in developing an 1500 w to 2500w SMPS with PFC,
My choice are Mosfets and IGBT's
Ease of driving,fficiency and cheapness my concern....
I have build one high power ballast 1000watt for driving number of flourescent lamps.i used a small gate drv torroid transfo Had to clamp the gate w a 15Volt Zener to reduce gate level..
think i still have it in my junk parts..
mailto: floyds_ec@gawab.com
Me too interested in developing an 1500 w to 2500w SMPS with PFC,
My choice are Mosfets and IGBT's
Ease of driving,fficiency and cheapness my concern....
I have build one high power ballast 1000watt for driving number of flourescent lamps.i used a small gate drv torroid transfo Had to clamp the gate w a 15Volt Zener to reduce gate level..
think i still have it in my junk parts..
This article shows a lot of interesting facts, but also shows some undesirable practices and mistakes
Remarkably, the most undesirable practice is to wind line-isolated secondaries directly over mains primaries without additional isolation. This is unreliable and potentially harmful because magnet-wire isolation is easily damaged by vibrations or by non-paralell crossing magnet wires, and it also results in very high primary-secondary capacitances producing high common-mode EMI
Three layers of electrical tape are advised as an additional mains isolation in transformers. Scotch 1350 or similar polyester tape is preferred since it withstands 150ºC and 4KV per layer with only 0.1mm thickness. As a bonus, common-mode EMI and primary-secondary leakage are dramatically reduced this way
But the most remarkable error is to leave the circuit in free-oscillation mode during start-up, this must be avoided as it's very unreliable. The reliable way to do it is to rectify the secondaries through a small value capacitor [~10uF] with no additional filtering, so the control circuit gets enough power during each start-up pulse to quickly shut down that pulse before any of the transformers gets saturated
This startup sequence is a bit tricky to obtain, but it's the one used in all AT PSUs. The following circuit achieves it and also keeps startup peak current smaller than maximum operating current
The start-up pulses are shut down as soon as Q1 or Q2 BE junctions become forward biased. In the first pulse C1 has no charge, so the pulse is terminated when the voltage across C2 reaches 4 diode drops, but this also forces some charge to be stored in C1, so the shutdown threshold is slightly increased in each pulse. It keeps increasing pulse by pulse until D1 and D2 start to conduct
At this point, we have reached three diode drops from Q1 and Q2 bases to ground [3x 650mV ~= 1.95V], so the voltage at C2 is 7.8V, just enough for the TL494 to start working and initiate its soft-start sequence. The circuit will prevent the voltage at C2 to increase above 7.8V until the TL494 takes control of the SMPS, so any failure is not likely to be catastrophic. Also, the duty cycle during startup will be very small, so nasty things like a heavy load or a dead short on the PSU output won't do any harm
These values of R1, R2, R3, R4 and C1 are only suitable for low power designs. At higher powers Q1 and Q2 are required to switch a higher Ic so its Ib has to be increased maintaining R1/R2 and R4/R3 proportions. The time constant introduced by C1 has to be also considered since it determines startup time and startup peak current. C2 may also have to be increased, finding optimum component values requires some experimentation
Remarkably, the most undesirable practice is to wind line-isolated secondaries directly over mains primaries without additional isolation. This is unreliable and potentially harmful because magnet-wire isolation is easily damaged by vibrations or by non-paralell crossing magnet wires, and it also results in very high primary-secondary capacitances producing high common-mode EMI
Three layers of electrical tape are advised as an additional mains isolation in transformers. Scotch 1350 or similar polyester tape is preferred since it withstands 150ºC and 4KV per layer with only 0.1mm thickness. As a bonus, common-mode EMI and primary-secondary leakage are dramatically reduced this way
But the most remarkable error is to leave the circuit in free-oscillation mode during start-up, this must be avoided as it's very unreliable. The reliable way to do it is to rectify the secondaries through a small value capacitor [~10uF] with no additional filtering, so the control circuit gets enough power during each start-up pulse to quickly shut down that pulse before any of the transformers gets saturated
This startup sequence is a bit tricky to obtain, but it's the one used in all AT PSUs. The following circuit achieves it and also keeps startup peak current smaller than maximum operating current
An externally hosted image should be here but it was not working when we last tested it.
The start-up pulses are shut down as soon as Q1 or Q2 BE junctions become forward biased. In the first pulse C1 has no charge, so the pulse is terminated when the voltage across C2 reaches 4 diode drops, but this also forces some charge to be stored in C1, so the shutdown threshold is slightly increased in each pulse. It keeps increasing pulse by pulse until D1 and D2 start to conduct
At this point, we have reached three diode drops from Q1 and Q2 bases to ground [3x 650mV ~= 1.95V], so the voltage at C2 is 7.8V, just enough for the TL494 to start working and initiate its soft-start sequence. The circuit will prevent the voltage at C2 to increase above 7.8V until the TL494 takes control of the SMPS, so any failure is not likely to be catastrophic. Also, the duty cycle during startup will be very small, so nasty things like a heavy load or a dead short on the PSU output won't do any harm
These values of R1, R2, R3, R4 and C1 are only suitable for low power designs. At higher powers Q1 and Q2 are required to switch a higher Ic so its Ib has to be increased maintaining R1/R2 and R4/R3 proportions. The time constant introduced by C1 has to be also considered since it determines startup time and startup peak current. C2 may also have to be increased, finding optimum component values requires some experimentation
Hi!
Eva learnt many new things abut SMPS design by going thru your posting. This new configration I will first try on old ATX supply and then make a proper board for the final ckt. I am sure we will get improved performance.
You are right about the primary sec winding isolation this page here gives good idea on transformer winding technique:
http://w5jgv.com/hv-ps1/index.htm
Just one question while going thru some designs I found a common mode choke on AC mains and then one after the rectifiers on DC. Now should these chokes have to be of different value or same can be used?
Sivan this is one of my first SMPS psu based on local ferrites. I would look forward to your comments on performance of different stuff available locally.
Regards
Rahul
Eva learnt many new things abut SMPS design by going thru your posting. This new configration I will first try on old ATX supply and then make a proper board for the final ckt. I am sure we will get improved performance.
You are right about the primary sec winding isolation this page here gives good idea on transformer winding technique:
http://w5jgv.com/hv-ps1/index.htm
Just one question while going thru some designs I found a common mode choke on AC mains and then one after the rectifiers on DC. Now should these chokes have to be of different value or same can be used?
Sivan this is one of my first SMPS psu based on local ferrites. I would look forward to your comments on performance of different stuff available locally.
Regards
Rahul
Transformers show primary to secondary capacitances and switching allways causes high frequency AC currents to flow between primary ground and secondary ground
If these currents are left flowing freely through input and output wiring then a lot of EMI is radiated, so this HF AC current has to be forced to flow through a known path and has to be prevented to flow through the wires
A capacitor of ~4.7nF rated at 2KV and placed between primary side ground and secondary side ground is routinely used to create a short path for these 'common mode' currents
That capacitor improves things a lot, but common mode inductors are still required to reduce HF AC currents flowing through wires to reasonable values
Common mode filtering is much easier to perform at low currents, so the primary side is the preferred place for the common mode filter. This dramatically reduces EMI radiation in the wires
But in applications where the load is not in the same case as the SMPS and the output side has long wires, high frequencies above 10Mhz get still radiated when they appear as a voltage between the casing and the wiring, so additional common mode filtering on the secondary side is advised. This filtering is only required to supress >10Mhz frequencies so single ferrite beads do the job quite well
In contrast, primary side common mode filtering requires inductors with several mH of common mode inductance and dozens of turns to get high common mode impedances. I tend to use the biggest core and the biggest turn counts I can fit
As a bonus, using big turn counts and making two separate windings [using double section formers or doing one in each half of a toroid core] produces high leakage inductance and with the help of a pair of capacitors this provides also differential mode filtering
Be careful, there must be no common mode DC current flowing through the common mode filter, otherise the filter will be saturated and it won't filter
I have no EMI measurement equipment, but a loop of wire of 10cm diameter connected to the oscilloscope probe helps a lot. I usually measure with the wire loop 1 meter away from the prototype [raw boards without metal case], at 1mV/div, and I allways try to reduce EMI until I see nothing but the the ambient and oscilloscope noise floors on the display. Sometimes the noise floor dominates even at distances below 50cm. I use a Hameg HM407 hybrid analog-digital oscilloscope, in digital mode since the EMI pulses get blurred in analog mode
If these currents are left flowing freely through input and output wiring then a lot of EMI is radiated, so this HF AC current has to be forced to flow through a known path and has to be prevented to flow through the wires
A capacitor of ~4.7nF rated at 2KV and placed between primary side ground and secondary side ground is routinely used to create a short path for these 'common mode' currents
That capacitor improves things a lot, but common mode inductors are still required to reduce HF AC currents flowing through wires to reasonable values
Common mode filtering is much easier to perform at low currents, so the primary side is the preferred place for the common mode filter. This dramatically reduces EMI radiation in the wires
But in applications where the load is not in the same case as the SMPS and the output side has long wires, high frequencies above 10Mhz get still radiated when they appear as a voltage between the casing and the wiring, so additional common mode filtering on the secondary side is advised. This filtering is only required to supress >10Mhz frequencies so single ferrite beads do the job quite well
In contrast, primary side common mode filtering requires inductors with several mH of common mode inductance and dozens of turns to get high common mode impedances. I tend to use the biggest core and the biggest turn counts I can fit
As a bonus, using big turn counts and making two separate windings [using double section formers or doing one in each half of a toroid core] produces high leakage inductance and with the help of a pair of capacitors this provides also differential mode filtering
Be careful, there must be no common mode DC current flowing through the common mode filter, otherise the filter will be saturated and it won't filter
I have no EMI measurement equipment, but a loop of wire of 10cm diameter connected to the oscilloscope probe helps a lot. I usually measure with the wire loop 1 meter away from the prototype [raw boards without metal case], at 1mV/div, and I allways try to reduce EMI until I see nothing but the the ambient and oscilloscope noise floors on the display. Sometimes the noise floor dominates even at distances below 50cm. I use a Hameg HM407 hybrid analog-digital oscilloscope, in digital mode since the EMI pulses get blurred in analog mode
The old prototype, that proved to work quite well at 1KW continuous can be seen here :
http://www.diyaudio.com/forums/showthread.php?postid=421162#post421162
And this is a preliminary version of the layout of the new 1.5KW [or more] prototype, it will be all mounted in a single PCB and will feature a single forced cooling tunnel for the heatsinks and the transformers. Currently it measures only 22.5 by 17.5cm
Note that there are still a lot of things missing on that PCB : Mains rectification, common mode filtering, TL494 control circuit, pulse transformer drive, synchronous rectifier logic and gate drive
http://www.diyaudio.com/forums/showthread.php?postid=421162#post421162
And this is a preliminary version of the layout of the new 1.5KW [or more] prototype, it will be all mounted in a single PCB and will feature a single forced cooling tunnel for the heatsinks and the transformers. Currently it measures only 22.5 by 17.5cm
Note that there are still a lot of things missing on that PCB : Mains rectification, common mode filtering, TL494 control circuit, pulse transformer drive, synchronous rectifier logic and gate drive
An externally hosted image should be here but it was not working when we last tested it.
Hi!
Eva thanks for all the valuable information. I went thru most of the postings on related topic. I am realy impressed by the work done by you.
Based on your suggestion I would go in for MJE13009 and try out the supply with sandwiched secondary winding and anther link to supply for faster control loop action.
Presently my requirements are very modest just 20 Amps intermittent at about 50% on off cycle so I assume will not be facing much problem, but then never to take Murphy too lightly.
I would appreciate a look at your your 1Kw schematics as and when they are done.
Thanks & Regards
Rahul
Eva thanks for all the valuable information. I went thru most of the postings on related topic. I am realy impressed by the work done by you.
Based on your suggestion I would go in for MJE13009 and try out the supply with sandwiched secondary winding and anther link to supply for faster control loop action.
Presently my requirements are very modest just 20 Amps intermittent at about 50% on off cycle so I assume will not be facing much problem, but then never to take Murphy too lightly.
I would appreciate a look at your your 1Kw schematics as and when they are done.
Thanks & Regards
Rahul
I can't post complete and detailed schematics nor construction details because there are some people on that forum that will start inmediately to build an sell my SMPS as if they were their own designs, altough they understand little or nothing about SMPS
But I can help you with your project. For 20A you don't need to expend money on MJE13009, the transistors used in any >200W AT or ATX PSU will be fine for that power level
15V * 20A = 300W. At that power level you can reuse almost everything from an AT or ATX SMPS. For 176 to 250V input I recommend a half bridge topology with approx. 6:1 turn ratio on the transformer. This way 20A output translates into 3.3A in the primary side and this current level is pretty easy to switch with the 6 to 8A transistors commonly found on these PSUs
If you blow the original transistors and have to get new ones, get MJE13007 instead. This is because bipolars with higher current capability switch slower at low currents than bipolars with low current capability. MJE13007 will switch faster than MJE13009 in the 0 to 3A range
I'm, using MJE13009 because I planing to switch up to 8A or more if it proves to be reliable, but remember that these transistors are actually quite slow below 3A
Having crossover times below 100nS and negligible current tail when they are operated at medium to high currents, modern bipolars outperform any MOSFET with similar price/size and allow to use smaller heatsinks and fans. Computer PSU manufacturers know this fact quite well
Also note that MOSFETs driven directly from transformers provide poor crossover times ~500ns. Getting below 100nS requires some kind of buffer between the transformer and the gate, so isolated proportional bipolar drive isn't actually much more complex than isolated buffered MOSFET drive
But I can help you with your project. For 20A you don't need to expend money on MJE13009, the transistors used in any >200W AT or ATX PSU will be fine for that power level
15V * 20A = 300W. At that power level you can reuse almost everything from an AT or ATX SMPS. For 176 to 250V input I recommend a half bridge topology with approx. 6:1 turn ratio on the transformer. This way 20A output translates into 3.3A in the primary side and this current level is pretty easy to switch with the 6 to 8A transistors commonly found on these PSUs
If you blow the original transistors and have to get new ones, get MJE13007 instead. This is because bipolars with higher current capability switch slower at low currents than bipolars with low current capability. MJE13007 will switch faster than MJE13009 in the 0 to 3A range
I'm, using MJE13009 because I planing to switch up to 8A or more if it proves to be reliable, but remember that these transistors are actually quite slow below 3A
Having crossover times below 100nS and negligible current tail when they are operated at medium to high currents, modern bipolars outperform any MOSFET with similar price/size and allow to use smaller heatsinks and fans. Computer PSU manufacturers know this fact quite well
Also note that MOSFETs driven directly from transformers provide poor crossover times ~500ns. Getting below 100nS requires some kind of buffer between the transformer and the gate, so isolated proportional bipolar drive isn't actually much more complex than isolated buffered MOSFET drive
Thanks Eva for your guidance. I very well understand your reasons for not posting finer details. There are lot in the category who think they have copyright ie right to copy.
I checked the local mkt I was told that 2SC2625 is available at 2 different price level cheaper one is probably of far eastern origin. Surprisingly MJE13009 was available at only one shop but MJE13007 is commonly available for 0.30$ ir Rs 15/ Indian, make ST from Morrocco.
I will keep the turns ratio you have mentioned and go ahead with the first prototype.
Thanks & Regards
Rahul
I checked the local mkt I was told that 2SC2625 is available at 2 different price level cheaper one is probably of far eastern origin. Surprisingly MJE13009 was available at only one shop but MJE13007 is commonly available for 0.30$ ir Rs 15/ Indian, make ST from Morrocco.
I will keep the turns ratio you have mentioned and go ahead with the first prototype.
Thanks & Regards
Rahul
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.