I think that Offline Half-Bridge converters are a complete waste of time....am i right?
The full-bridge is cheaper and better......
FET RMS current will be much higher with a half-bridge than a full bridge, and the half-bridge fets will need to be more pricey with bigger heatsinks.
-this is due to the unfortunate rail-splitting capacitors in a half-bridge, which divide the input voltage by two.
Also, the Half bridge needs these highly expensive , extremely low dissipation factor, rail-splitting film capacitors.....these are more expensive than a mosfet....so the extra mosfets in a full bridge, would be payed for by the fact that the full bridge doesn't need the expensive film capacitors.
FULL BRIDGE COSTS:
4 FETs, Two NCP5181 Bootstrap fet drivers.
HALF-BRIDGE COSTS SO FAR
2 FETs, one NCP5181 Bootstrap fet driver, 2 Pulse rated film capacitors
...Already the half-bridge is looking awry...........but there's more woes for it........
the half bridge cannot utilise low-side resistive current sensing, because when the high side fet conducts...most of the power current would not go through the low-side current sense resistor.......so the half-bridge must use an expensive current sense transformer in order to be able to sense any primary overcurrents.
..the Full-Bridge can just use a cheap current sense resistor.
Also, in the event of overload, the half-bridge's rail-splitting capacitors will end up with one discharged , and the other charged all the way up to the rail........which means that each rail-splitting capacitor must be rated for the full Vin of the DC Bus.....so thats two, expensive 400V rated pulse capacitors for the half bridge.
Also, the film capacitors for the half-bridge take up a large amount of space in the half bridge.
Is anyone seriously going to defend the half-bridge?
Even more woes for the half-bridge in that it cannot do peak current mode control without blowing itself up, or at the very least running in an unstable way...this is mitigated by adding a balancing winding, but thats more expense and complexity.
The full-bridge might need more heatsinks...but actually, no it won't...because if you use Insulated tab fets you can simply fix the four fets of a full bridge to the same heatsink, side-by-side.
Its already been said that primary current is much higher for a half-bridge than a full-bridge......so there's more woes for the half-bridge at the EMC testing stage.
So, Is anyone seriously going to defend the Half-Bridge vs the Full-Bridge?...Surely not?
The full-bridge is cheaper and better......
FET RMS current will be much higher with a half-bridge than a full bridge, and the half-bridge fets will need to be more pricey with bigger heatsinks.
-this is due to the unfortunate rail-splitting capacitors in a half-bridge, which divide the input voltage by two.
Also, the Half bridge needs these highly expensive , extremely low dissipation factor, rail-splitting film capacitors.....these are more expensive than a mosfet....so the extra mosfets in a full bridge, would be payed for by the fact that the full bridge doesn't need the expensive film capacitors.
FULL BRIDGE COSTS:
4 FETs, Two NCP5181 Bootstrap fet drivers.
HALF-BRIDGE COSTS SO FAR
2 FETs, one NCP5181 Bootstrap fet driver, 2 Pulse rated film capacitors
...Already the half-bridge is looking awry...........but there's more woes for it........
the half bridge cannot utilise low-side resistive current sensing, because when the high side fet conducts...most of the power current would not go through the low-side current sense resistor.......so the half-bridge must use an expensive current sense transformer in order to be able to sense any primary overcurrents.
..the Full-Bridge can just use a cheap current sense resistor.
Also, in the event of overload, the half-bridge's rail-splitting capacitors will end up with one discharged , and the other charged all the way up to the rail........which means that each rail-splitting capacitor must be rated for the full Vin of the DC Bus.....so thats two, expensive 400V rated pulse capacitors for the half bridge.
Also, the film capacitors for the half-bridge take up a large amount of space in the half bridge.
Is anyone seriously going to defend the half-bridge?
Even more woes for the half-bridge in that it cannot do peak current mode control without blowing itself up, or at the very least running in an unstable way...this is mitigated by adding a balancing winding, but thats more expense and complexity.
The full-bridge might need more heatsinks...but actually, no it won't...because if you use Insulated tab fets you can simply fix the four fets of a full bridge to the same heatsink, side-by-side.
Its already been said that primary current is much higher for a half-bridge than a full-bridge......so there's more woes for the half-bridge at the EMC testing stage.
So, Is anyone seriously going to defend the Half-Bridge vs the Full-Bridge?...Surely not?
eem2am,
I won't "defend" a half bridge circuit, but it can be made to work fine, even with current mode control. See: http://www.ti.com/lit/ml/slup083/slup083.pdf The energy storage bus caps don't need to be film, but paralleling the storage caps with a small film cap is a good idea.
I won't "defend" a half bridge circuit, but it can be made to work fine, even with current mode control. See: http://www.ti.com/lit/ml/slup083/slup083.pdf The energy storage bus caps don't need to be film, but paralleling the storage caps with a small film cap is a good idea.
The great majority of chinese ATX computer power supplies are half bridge. No need for a film cap, they use two electrolytics in series for the input capacitance and connect the midpoint to one side of the primary.
Chinese fingers can wind gate drive transformers for less cost than gate drive ICs, so there's usually a single gate drive transformer for the main switches. Those are usually BJTs at lower power levels, MOSFET at higher power levels. Controller is a voltage mode TL494 or clone controller, sometimes SG3525. If it's a current mode supply (rare for a cheap supply, but they exist) they use a current transformer (again, chinese fingers are cheap) and a UC3846 or similar IC. TL431/opto feedback, with type 3 compensation split between the TL431 and the TL494.
Now if you're a boutique/custom power supply company that's making supplies in lower quantities or making them domestically, it might be cheaper to make a different topology.
Chinese fingers can wind gate drive transformers for less cost than gate drive ICs, so there's usually a single gate drive transformer for the main switches. Those are usually BJTs at lower power levels, MOSFET at higher power levels. Controller is a voltage mode TL494 or clone controller, sometimes SG3525. If it's a current mode supply (rare for a cheap supply, but they exist) they use a current transformer (again, chinese fingers are cheap) and a UC3846 or similar IC. TL431/opto feedback, with type 3 compensation split between the TL431 and the TL494.
Now if you're a boutique/custom power supply company that's making supplies in lower quantities or making them domestically, it might be cheaper to make a different topology.
Last edited:
Yeah
Full bridge is always better than half bridge, you have better regulation, better supply usage, simple current sensing, lesser primary current and more efficient use of ferrite core.
sawreyrw:
Thanks for your reply, but please see top right of page 6-2 of the document that you just referenced.........it admits that half bridge is unworkable unless an extra auxiliary winding on the transformer is added to balance the rail splitting caps.
-Another thing is that using electrolytic rail splitting caps will result in big overheating due to their esr unless you use really big ones.
Half-Bridge is a waste of time.
Admittedly having those "Chinese fingers" that you speak of makes the half bridge come better, but still not good enough.
Thanks for your reply, but please see top right of page 6-2 of the document that you just referenced.........it admits that half bridge is unworkable unless an extra auxiliary winding on the transformer is added to balance the rail splitting caps.
-Another thing is that using electrolytic rail splitting caps will result in big overheating due to their esr unless you use really big ones.
Half-Bridge is a waste of time.
Admittedly having those "Chinese fingers" that you speak of makes the half bridge come better, but still not good enough.
We've established already that you need a current transformer for current mode operation. Not an issue for voltage mode control.
Regarding ESR overheating of the primary... Keeping the same amount of primary side energy storage, you're splitting the input capacitor into two capacitors with double the uF and half the voltage rating. ESR is proportional to voltage rating (200V cap will have 1/2 the ESR of a 400V cap) and inversely proportional to capacitance (300uF will be 1/2 the ESR of 150uF), so the ESR of each cap will be 1/4 of a single 400V cap.
Primary side current in a half bridge is doubled, so you end up with the same amount of total I2R loss in the input capacitance. However, you're dissipating the same amount of power in two capacitors instead of one, which puts you at a thermal advantage over a single cap. Only downside is that two caps take up more room than one, but the physical room isn't any worse than adding a second pair of FETs and associated bits.
FYI, your "half bridge is a waste of time" opinion goes against literally hundreds of millions of power supplies that are out there in the field. Again, almost every ATX computer power supply ever made uses this architecture.
Regarding ESR overheating of the primary... Keeping the same amount of primary side energy storage, you're splitting the input capacitor into two capacitors with double the uF and half the voltage rating. ESR is proportional to voltage rating (200V cap will have 1/2 the ESR of a 400V cap) and inversely proportional to capacitance (300uF will be 1/2 the ESR of 150uF), so the ESR of each cap will be 1/4 of a single 400V cap.
Primary side current in a half bridge is doubled, so you end up with the same amount of total I2R loss in the input capacitance. However, you're dissipating the same amount of power in two capacitors instead of one, which puts you at a thermal advantage over a single cap. Only downside is that two caps take up more room than one, but the physical room isn't any worse than adding a second pair of FETs and associated bits.
FYI, your "half bridge is a waste of time" opinion goes against literally hundreds of millions of power supplies that are out there in the field. Again, almost every ATX computer power supply ever made uses this architecture.
sorry gmarsh, but any half-bridge, with either voltage or current mode needs a current transformer in primary side.
The Chinese do this type of power supply and they do it well....and as we have agreed, they CAN do it well, because they have lots and lots of willing, dextrous Chinese fingers to wind all the windings that the half-bridge needs.....
And that is WHY the Chinese do the half-bridge....because they know that no Western compeny will ever be able to copy their design because in the west, we simply dont have the staff to wind all those current transformers and transformers with balancing windings.
By the way, do you understand that without the balancing winding, all rail splitting caps need to be rated to the full supply DC bus voltage.
The Chinese do this type of power supply and they do it well....and as we have agreed, they CAN do it well, because they have lots and lots of willing, dextrous Chinese fingers to wind all the windings that the half-bridge needs.....
And that is WHY the Chinese do the half-bridge....because they know that no Western compeny will ever be able to copy their design because in the west, we simply dont have the staff to wind all those current transformers and transformers with balancing windings.
By the way, do you understand that without the balancing winding, all rail splitting caps need to be rated to the full supply DC bus voltage.
You don't need a current transformer for a voltage mode supply. What would you use it for, OCP? In that case it's adequate to just measure the current through the low side switch with a sense resistor - even with a shorted output, it'll take a number of cycles for current in the output inductance to charge up so you don't need cycle by cycle limiting.
You also don't need balancing windings - in voltage mode operation, the converter will naturally balance the voltage across the two caps. If there's a small difference in duty cycle between the upper and lower primary switches, the balance will shift slightly, but it's not hard to keep duty cycle variation to a reasonable amount.
Don't believe me? take the primary schematic of a half bridge supply, replace the transformer primary with an inductor, draw the switches on the left and the capacitors on the right. Call the center node of the two capacitors "Vout". You've just drawn a synchronous buck regulator, but with a capacitor going between Vin and Vout which makes no difference in circuit operation other than precharging Vout to Vin/2 when you ramp up Vin. Provided the on time of upper and lower switches are the same, the circuit will force Vout to be Vin/2. If the on times are different, it's pretty easy to calculate what "Vout" will be - I'll leave derivation of that equation in your capable hands.
in 2006 i scrapped a cubic meter of used/failed ATX computer power supplies and none of them directly connected the transformer to the 200v electrolytic caps.
All of them had a .47-2.0 uf polyester cap in series.
most were two transistor driven half bridges, current transformer and and base drive transformer using the primary current to provide the base drive power.
about a third were forward or flyback converters.
honestly i have never seen an ATX supply directly connect the primary to the rail splitting caps.
but maybe i live near an expensive city (seattle) and that's why.. idk
anyhow, if OP has to plan around a failure mode and buy 400V caps to split a 400V rail.. then why not throw an extra 5$ at it and go with a full bridge?
some designers just put a 230v MOV across the 200 v caps..
thanks gmarsh
-Sorry i was unclear here, i am talking about voltage mode half-bridges which have gone into cycle-bycycle current limiting due to overload or heavy load transient
-when a voltaage mode controller is overloaded, or suffers a large load transient, it goes into cycle-by-cycle current limiting
Please view this
http://www.ti.com/lit/ds/symlink/lm5039.pdf
Its the voltage mode LM5039, which controls half bridges.
It has a special feature whereby it goes into average current mode control when it is overloaded......and this stops the imbalance of the rail-splitting capacitors.
-but this is the only controller in the world that has this feature.
So if you are not using the LM5039 then you need to make your rail-splitting caps rated to the full rail voltage.
the first page of the datasheet tells of the "balancing act"
-Sorry i was unclear here, i am talking about voltage mode half-bridges which have gone into cycle-bycycle current limiting due to overload or heavy load transient
-when a voltaage mode controller is overloaded, or suffers a large load transient, it goes into cycle-by-cycle current limiting
Please view this
http://www.ti.com/lit/ds/symlink/lm5039.pdf
Its the voltage mode LM5039, which controls half bridges.
It has a special feature whereby it goes into average current mode control when it is overloaded......and this stops the imbalance of the rail-splitting capacitors.
-but this is the only controller in the world that has this feature.
So if you are not using the LM5039 then you need to make your rail-splitting caps rated to the full rail voltage.
the first page of the datasheet tells of the "balancing act"
Only matters if you're using cycle by cycle current limiting - which is effectively the same as current mode control when it's active, just with a fixed threshold instead of a threshold controlled by the output voltage.
If you use latched/hiccup overcurrent protection, or an output current regulation loop that overrides the voltage feedback loop like you'd find on a bench supply, you're fine.
Certainly if cycle-by-cycle limiting is part of your design criteria, you're definitely better off with a different topology than a stacked-cap half-bridge.
If you use latched/hiccup overcurrent protection, or an output current regulation loop that overrides the voltage feedback loop like you'd find on a bench supply, you're fine.
Certainly if cycle-by-cycle limiting is part of your design criteria, you're definitely better off with a different topology than a stacked-cap half-bridge.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.