Is it always necessary to use coupled output inductors with an offline current mode full-bridge smps supplying split class d rails.?
Power = 330W.
the Vout = +/-50V.
I hear that the method of increasing cross-regulation by using coupled output inductors is always needed when using current mode control.
Page 3, part 5 of this App Note states that current mode means coupled output inductors are needed.......
http://www.ti.com/lit/an/slua119/slua119.pdf
is this right?
Power = 330W.
the Vout = +/-50V.
I hear that the method of increasing cross-regulation by using coupled output inductors is always needed when using current mode control.
Page 3, part 5 of this App Note states that current mode means coupled output inductors are needed.......
http://www.ti.com/lit/an/slua119/slua119.pdf
is this right?
Strictly speaking, NO, coupled inductors are not required for current mode control. But for a +/-50V output, it would be advantageous to use a coupled inductor whether the power supply was current mode or voltage mode controlled. By adding slope compensation as Mammano suggests in note 2 of that same section, you add a fixed ramp to the current signal, which makes it more like a voltage mode controller.
Is there some objection to using a coupled inductor? They get messy when you have differing low voltage outputs because you have to control the leakage inductances to prevent circulating currents between the outputs, but with +/-50V it seems like a really good idea.
Is there some objection to using a coupled inductor? They get messy when you have differing low voltage outputs because you have to control the leakage inductances to prevent circulating currents between the outputs, but with +/-50V it seems like a really good idea.
The problem with using a coupled inductor is that when one half of the split output is fully loaded, the other half sees the coupled inductor as a very high inductance, and so its transient response, should its output suddenly become loaded, is very solw, and slow transient response is not good, as you know, for audio purpose.
You're right, an SMPS perfect for audio amplifiers, you can not develop with the chip manufacturers. pfc, phase shift and sync rectifiers, have nothing to do with the transients of an amplifier. I estimated at about 150ns response time to fully adapt to the demands of current.
Therefore, you can only increase the filtering capacity. or build something new specifically for amplifiers. (with your chips and your topology)
Regards
Therefore, you can only increase the filtering capacity. or build something new specifically for amplifiers. (with your chips and your topology)
Regards
If this is a forward converter with a +50V and a -50V output and a single control loop, then a coupled inductor will give you better load transient response on the unregulated output than having two separate inductors. The "coupled inductor" is also a transformer that keeps the voltages across the two windings the same during the ON and OFF switching periods. When the unregulated output has a load change, the currents are reflected to the regulated output through the transformer action of the "coupled inductor." This improves both DC regulation as well as transient response.
I have been assuming this converter is a single-ended forward converter like the ones shown in Bob's apnote, is that correct?
I have been assuming this converter is a single-ended forward converter like the ones shown in Bob's apnote, is that correct?
Hi,
the converter here is a full-bridge SMPS, supplying isolated split rails of +/-50V.
Would i be right to suggest that if i am using uncoupled inductors of 47uH......then if i want to switch to coupled inductors......i make them 47/4 uH?
....ie the inductance of one winding, with the other open, would be 47/4uH?
..Also, is there something about making the inductance of the coupled windings the same as the transformer secondary?.....i.e., it wont work if theyre not the same?
the converter here is a full-bridge SMPS, supplying isolated split rails of +/-50V.
Would i be right to suggest that if i am using uncoupled inductors of 47uH......then if i want to switch to coupled inductors......i make them 47/4 uH?
....ie the inductance of one winding, with the other open, would be 47/4uH?
..Also, is there something about making the inductance of the coupled windings the same as the transformer secondary?.....i.e., it wont work if theyre not the same?
Last edited:
Calculate the inductance as if they weren't coupled (so if 47 uH is the right number, the two windings on the core will each measure 47 uH).
Then size your output inductor core to handle the total expected DC (and AC Peak) output current.
If the outputs are the same voltage and you are using the same kinds of rectifiers on both outputs then the winding ratio is 1:1.
No, you do not need to match the inductance of the output choke to the transformer inductance. Some of the old unitrode apnotes are still around on the TI website that tell you to match the inductor ratios to the transformer ratios, but that approach is simplistic and can cause a lot of trouble.
Also, a word of caution about the Unitrode suggestions about slope compensation (the full bridge will operate at nearly full duty cycle, so you are over the 50% duty cycle that Mammano mentioned in the apnote article 2), so do your homework! the Unitrode apnotes were written before Ridley & Fred Lee published their work on current mode control, and some of the old cookbook stuff is not entirely the right way to go.
Good luck on this one!
Then size your output inductor core to handle the total expected DC (and AC Peak) output current.
If the outputs are the same voltage and you are using the same kinds of rectifiers on both outputs then the winding ratio is 1:1.
No, you do not need to match the inductance of the output choke to the transformer inductance. Some of the old unitrode apnotes are still around on the TI website that tell you to match the inductor ratios to the transformer ratios, but that approach is simplistic and can cause a lot of trouble.
Also, a word of caution about the Unitrode suggestions about slope compensation (the full bridge will operate at nearly full duty cycle, so you are over the 50% duty cycle that Mammano mentioned in the apnote article 2), so do your homework! the Unitrode apnotes were written before Ridley & Fred Lee published their work on current mode control, and some of the old cookbook stuff is not entirely the right way to go.
Good luck on this one!
OK thanks......i am using voltage mode......so i will not need slope comp..but thanks anyway.
My input DC Bus is 190V to 373V, so you are right, much slope compensation would have been needed.
You may agree with me that when a lot of slope compensation is needed, it is a bad idea, and one should simply change to voltage mode control.
...this is because when you have a lot of slope added, one has to reduce the Ohmic value of the current sense resistor, which degrades the short circuit and overload protection.
...not only that but Ridley says that one should never derive the slope ramp from the clock oscillator of the PWM ic, even if via a BJT buffer is used...since it makes the SMPS more noise prone.
Also, 47uH is the right value for the non-coupled inductors.....but if i simply make this the value for the coupled inductors.....then it will appear to be 94uH (due to mutual inductance).....and that surely is a bad thing because i need the inductor to be small Henry value , so as to allow a good, fast transient response.
....the increased value of inductance when coupled, is the same as in a SEPIC converter, where coupling the two inductors halves the input current ripple, since it makes the inductance twice as much.
So shouldnt i use 47/2 uH for the coupled inductor value (individually)?
["47/4", as in my previous , was an error, sorry]
My input DC Bus is 190V to 373V, so you are right, much slope compensation would have been needed.
You may agree with me that when a lot of slope compensation is needed, it is a bad idea, and one should simply change to voltage mode control.
...this is because when you have a lot of slope added, one has to reduce the Ohmic value of the current sense resistor, which degrades the short circuit and overload protection.
...not only that but Ridley says that one should never derive the slope ramp from the clock oscillator of the PWM ic, even if via a BJT buffer is used...since it makes the SMPS more noise prone.
Also, 47uH is the right value for the non-coupled inductors.....but if i simply make this the value for the coupled inductors.....then it will appear to be 94uH (due to mutual inductance).....and that surely is a bad thing because i need the inductor to be small Henry value , so as to allow a good, fast transient response.
....the increased value of inductance when coupled, is the same as in a SEPIC converter, where coupling the two inductors halves the input current ripple, since it makes the inductance twice as much.
So shouldnt i use 47/2 uH for the coupled inductor value (individually)?
["47/4", as in my previous , was an error, sorry]
Well, about the inductance value... You have most likely calculated your desired value for a single output based on a tradeoff between mainly minimum output current (for the "critical inductance" value) and ripple current (sizing the output capacitors for the desired ripple voltage).
In steady state balanced load operation, there isn't any "transformer action" going on in the coupled inductor core, so you can treat the winding inductances as if they were independent. You needed 47 uHy in a single choke, so you still need 47 uHy on each of the windings.
Above the DC critical load current, the AC ripple current in the +50V winding goes through the +50V output capactitor, then through the -50V output capacitor and into the -50V leg of the coupled inductor. AC current in the return leg cancels. The voltages across the two windings are the same during the ON and OFF switching intervals.
Each winding has to have the value of inductance that you would have chosen for separate cores if you are going to have the desired ripple current in your output capacitors. Each winding will have the same number of turns on it that you would have had if you used two of the same type of core and wound two inductors. The H in the single core will be twice the H of the separate cores, so it doesn't save any core material to use a coupled inductor.
The effective "load" that the output with coupled inductors presents to the secondary of the transformer is exactly the same as if you had separate cores (i.e. it "looks" as if you have two 47 uHy inductors in parallel and the +/-50V output caps in parallel driven by a single secondary).
There's a lot more to be said about the inductor if you think more needs to be said.
In steady state balanced load operation, there isn't any "transformer action" going on in the coupled inductor core, so you can treat the winding inductances as if they were independent. You needed 47 uHy in a single choke, so you still need 47 uHy on each of the windings.
Above the DC critical load current, the AC ripple current in the +50V winding goes through the +50V output capactitor, then through the -50V output capacitor and into the -50V leg of the coupled inductor. AC current in the return leg cancels. The voltages across the two windings are the same during the ON and OFF switching intervals.
Each winding has to have the value of inductance that you would have chosen for separate cores if you are going to have the desired ripple current in your output capacitors. Each winding will have the same number of turns on it that you would have had if you used two of the same type of core and wound two inductors. The H in the single core will be twice the H of the separate cores, so it doesn't save any core material to use a coupled inductor.
The effective "load" that the output with coupled inductors presents to the secondary of the transformer is exactly the same as if you had separate cores (i.e. it "looks" as if you have two 47 uHy inductors in parallel and the +/-50V output caps in parallel driven by a single secondary).
There's a lot more to be said about the inductor if you think more needs to be said.
Hi,
Thanks neophyte, that is interesting.....
I chose the inductor so that it has a lot of ripple.....to give a fast transient response.
-but i didnt want too high peak current, so i chose the inductor value such that the inductor currnt was in boundary mode at Maximum load, Maximum input line voltage.
....So basically, in half load, or light load, my output inductors should be in discontinuous mode...............and i am wondering if the coupled inductor principle still works when there is discontinuous mode?
So i take it from the fact that a bigger core is needed for the single coupled part that each inductor of the coupled inductor is connected in like flyback transformer inductors.?...i.e. they both make the flux lines go the same way when they flow current.?......
Thanks neophyte, that is interesting.....
I chose the inductor so that it has a lot of ripple.....to give a fast transient response.
-but i didnt want too high peak current, so i chose the inductor value such that the inductor currnt was in boundary mode at Maximum load, Maximum input line voltage.
....So basically, in half load, or light load, my output inductors should be in discontinuous mode...............and i am wondering if the coupled inductor principle still works when there is discontinuous mode?
So i take it from the fact that a bigger core is needed for the single coupled part that each inductor of the coupled inductor is connected in like flyback transformer inductors.?...i.e. they both make the flux lines go the same way when they flow current.?......
Hm. OK....
Discontinuous mode inductors are going to have a high frequency, high voltage resonant associated with them, which may be annoying. For your converter, it's probably in the 250 kHz-350 kHz range.
Fast transient response is a nice goal, but there are some fundamental limitations to worry about. First, with a linear type voltage mode controller, is the loop bandwidth. You don't want to set it too high or else you'll have too much gain at the switching frequency. To have adequate margin, you will want the unity gain bandwidth to be no more than 25% of the switching frequency, and may find that the practical upper limit for your circuit is 20%. For voltage mode control, the power stage transfer function changes at the continuous/discontinuous mode boundary.
But yes, the coupled inductor works in both DCM and CCM.
Discontinuous mode inductors are going to have a high frequency, high voltage resonant associated with them, which may be annoying. For your converter, it's probably in the 250 kHz-350 kHz range.
Fast transient response is a nice goal, but there are some fundamental limitations to worry about. First, with a linear type voltage mode controller, is the loop bandwidth. You don't want to set it too high or else you'll have too much gain at the switching frequency. To have adequate margin, you will want the unity gain bandwidth to be no more than 25% of the switching frequency, and may find that the practical upper limit for your circuit is 20%. For voltage mode control, the power stage transfer function changes at the continuous/discontinuous mode boundary.
But yes, the coupled inductor works in both DCM and CCM.
OK Thanks,
My switching frequency is 130KHz.
The resonance sounds worrying......though i wonder if i may be free from worry as the load is a Class D amplifier and its output filter may get rid of such high frequency?
I have set the feedback compensation to be able to handle the continuous mode case.
I always associate discontinuous mode with greater stability...like in the flyback where its much easier to compensate, and has higher transient response than continuous mode.
My switching frequency is 130KHz.
The resonance sounds worrying......though i wonder if i may be free from worry as the load is a Class D amplifier and its output filter may get rid of such high frequency?
I have set the feedback compensation to be able to handle the continuous mode case.
I always associate discontinuous mode with greater stability...like in the flyback where its much easier to compensate, and has higher transient response than continuous mode.
Last edited:
The CCM/DCM modes will greatly affect your loop bandwidth. The compensation for CCM should be stable during DCM but the small signal unity gain bandwidth will be low (couple of kilohertz or less). Have you run any simulations to see how the power supply will respond to large load steps that transition between DCM and CCM?
The resonance thing for DCM may not be a big deal, but at light load, energy stored in the ringing L-C tank circuit is still there when the next switching cycle occurs, and you will see a subharmonic ripple in the output voltage.
If you have access to a spectrum analyzer, I recommend that you look at the output noise spectrum at various load conditions to make sure there is nothing troublesome. Most SMPS analog circuits are not designed for low noise, and I typically see broadband noise up to about ten kilohertz, with amplitude as high as 300 uV(RMS) in a 300 Hz bandwidth.
The DCM energy storage/transfer will typically show up as fairly distinct spurs that pop up at fractions of Fs, varying with small changes in the load. It also shows up on an oscilloscope as "jitter" at the OFF switching instant of the PWM. If you treat it as a jitter problem it can be really frustrating trying to fix it.
Interesting project, eem2am!
The resonance thing for DCM may not be a big deal, but at light load, energy stored in the ringing L-C tank circuit is still there when the next switching cycle occurs, and you will see a subharmonic ripple in the output voltage.
If you have access to a spectrum analyzer, I recommend that you look at the output noise spectrum at various load conditions to make sure there is nothing troublesome. Most SMPS analog circuits are not designed for low noise, and I typically see broadband noise up to about ten kilohertz, with amplitude as high as 300 uV(RMS) in a 300 Hz bandwidth.
The DCM energy storage/transfer will typically show up as fairly distinct spurs that pop up at fractions of Fs, varying with small changes in the load. It also shows up on an oscilloscope as "jitter" at the OFF switching instant of the PWM. If you treat it as a jitter problem it can be really frustrating trying to fix it.
Interesting project, eem2am!
Thanks Neophyte, - sorry because i did not say that this is supplying a class d amp.......... hence i am not too concerned of noise , since the class d amp filter will filter that away for me.
It is intriguing that the coupled inductor is not behaving like a transformer, where the flux of the pri and sec cancel each other out....in the coupled inductors, there fluxes are in the same direction..............this means that each winding is seeing more flux per amp of current than it flows.............
...and this in turn means that each coil has more inductance, the greater the current that flows in the other coil....
so our coupled inductor friends are actually widely varying inductances.?
It must be so ?
...because the flux does not know which coil produced it....neither does either coil know that the other coil is there alongside it....
When both coils are flowing current, each coil sees a big flux develop in the core...because of the existence of the flux from the current of the opposite coil...........so each coil is a bigger inductance, than when not coupled....and a varying inductance, -varying depending on the amount of current flowing in the other coil.
So our coupled inductor friends are in fact varying inductances?
when both are flowing current...each coil is actually trying to push current the other way, in the opposite coil?
It is intriguing that the coupled inductor is not behaving like a transformer, where the flux of the pri and sec cancel each other out....in the coupled inductors, there fluxes are in the same direction..............this means that each winding is seeing more flux per amp of current than it flows.............
...and this in turn means that each coil has more inductance, the greater the current that flows in the other coil....
so our coupled inductor friends are actually widely varying inductances.?
It must be so ?
...because the flux does not know which coil produced it....neither does either coil know that the other coil is there alongside it....
When both coils are flowing current, each coil sees a big flux develop in the core...because of the existence of the flux from the current of the opposite coil...........so each coil is a bigger inductance, than when not coupled....and a varying inductance, -varying depending on the amount of current flowing in the other coil.
So our coupled inductor friends are in fact varying inductances?
when both are flowing current...each coil is actually trying to push current the other way, in the opposite coil?
- Status
- Not open for further replies.