i am sorry to say it, but there is no way that with vin=230VAC and P(out) = 500W that you are getting 91% efficency.
I used to work for a huge lighting company, and used to spend weeks doing nothing but efficiency measurements on switch-mode 5W LED Drivers with Vin = 32V.
The maximum efficiency achieved was when the LED string voltage was high (about 24V)......This maximum efficiency was 91%.
With lower LED string voltages, the efficiency was around 78%.
-These were low current SMPS's......with very low Vf Schottkys and low RDSon FETs.
So there is just absolutely no way in the world that you are getting 91%+ efficiency with an isolated BCM flyback at 500W off the mains.
You will be lucky to be getting seventy something percent.
Are you measuring instruments calibrated?
I used to work for a huge lighting company, and used to spend weeks doing nothing but efficiency measurements on switch-mode 5W LED Drivers with Vin = 32V.
The maximum efficiency achieved was when the LED string voltage was high (about 24V)......This maximum efficiency was 91%.
With lower LED string voltages, the efficiency was around 78%.
-These were low current SMPS's......with very low Vf Schottkys and low RDSon FETs.
So there is just absolutely no way in the world that you are getting 91%+ efficiency with an isolated BCM flyback at 500W off the mains.
You will be lucky to be getting seventy something percent.
Are you measuring instruments calibrated?
Build it, measure it and you will see...
Let's calculate the total losses of my SMPS at 500W/230VAC.
1) MOSFETS
Type: STW25NM50N --> Rds_on=0.11ohm (not bad ah...)
Ipri_RMS=3.19A --> Pcond=2*Ipri_RMS^2*Rds_on=2.24W
By now I neglect switching loss, lets assume they are 2W
2) Diodes
Type: MBR20200CT --> Vf=0.85V Rd=20mR
Isec_RMS=7.6A, Isec_AVG=5A
Pcond = 2*(Isec_AVG*Vf+Isec_RMS^2*Rd)=10.8W
Pswitch is negligible
3) Input bridge
Assuming an input PF=0.55 and a GBU8J bridge (Vf=1.2V Rd=70mR)
I_IN_AVG=1.54A / I_IN_RMS=4.25A
Pbridge=2*(I_IN_AVG*Vf+Rd*I_IN_RMS^2)=4.37W
4) Transformer
Primary DC resistance=0.112ohm
Secondary DC resistance=0.015ohm
Assume Fr ratio=3 (very very bad...)
Primary AC resistance=3*0.112ohm=0.336ohm
Secondary AC resistance=3*0.015ohm=0.045ohm
Primary copper loss=Ip_RMS^2*RACp=3.4W
Secondary copper loss=2*Isec_RMS^2*RACs=5.2W
Core losses:
-Fsw@230VAC/500W=50kHz
-deltaB=150mT
Applying Steinmetz for 3C90 material: Pcore=3.4W
Total transformer losses=3.4+5.2+3.4=12W
Total losses of major components: Ploss=2.2W+2W+10.8W+4.37W+12W=31.3W
Calculated efficiency=500/(500+31.3)=94.1%
Now I have neglected the power loss in capacitors ESR, in the input EMI filter,
the power needed by the PWM but anyway..... to reach 70% efficiency there are still around 183W that are burned somewhere but I can not notice it..... strange.
Hey if you are speaking about 5W low voltage stuff it is clear that the efficiency can not be so high. 90% efficiency on 5W output means wasting no more than 0.5W.
It is easier to get high efficiency from big stuff rather than from small stuff. The worst combination will be low voltage (input or output) and high current...
Now I hope you believe me, if not that's your problem, not mine
ciao
-marco
Let's calculate the total losses of my SMPS at 500W/230VAC.
1) MOSFETS
Type: STW25NM50N --> Rds_on=0.11ohm (not bad ah...)
Ipri_RMS=3.19A --> Pcond=2*Ipri_RMS^2*Rds_on=2.24W
By now I neglect switching loss, lets assume they are 2W
2) Diodes
Type: MBR20200CT --> Vf=0.85V Rd=20mR
Isec_RMS=7.6A, Isec_AVG=5A
Pcond = 2*(Isec_AVG*Vf+Isec_RMS^2*Rd)=10.8W
Pswitch is negligible
3) Input bridge
Assuming an input PF=0.55 and a GBU8J bridge (Vf=1.2V Rd=70mR)
I_IN_AVG=1.54A / I_IN_RMS=4.25A
Pbridge=2*(I_IN_AVG*Vf+Rd*I_IN_RMS^2)=4.37W
4) Transformer
Primary DC resistance=0.112ohm
Secondary DC resistance=0.015ohm
Assume Fr ratio=3 (very very bad...)
Primary AC resistance=3*0.112ohm=0.336ohm
Secondary AC resistance=3*0.015ohm=0.045ohm
Primary copper loss=Ip_RMS^2*RACp=3.4W
Secondary copper loss=2*Isec_RMS^2*RACs=5.2W
Core losses:
-Fsw@230VAC/500W=50kHz
-deltaB=150mT
Applying Steinmetz for 3C90 material: Pcore=3.4W
Total transformer losses=3.4+5.2+3.4=12W
Total losses of major components: Ploss=2.2W+2W+10.8W+4.37W+12W=31.3W
Calculated efficiency=500/(500+31.3)=94.1%
Now I have neglected the power loss in capacitors ESR, in the input EMI filter,
the power needed by the PWM but anyway..... to reach 70% efficiency there are still around 183W that are burned somewhere but I can not notice it..... strange.
Hey if you are speaking about 5W low voltage stuff it is clear that the efficiency can not be so high. 90% efficiency on 5W output means wasting no more than 0.5W.
It is easier to get high efficiency from big stuff rather than from small stuff. The worst combination will be low voltage (input or output) and high current...
Now I hope you believe me, if not that's your problem, not mine
ciao
-marco
Just for reference:
check out the Meanwell website Switching Power Supply - Mean Well Switching Power Supply Manufacturer
and download the specs of a product called SDR-120.
This is a 24V/5A SMPS with PFC and univeral mains input.
The efficiency they claim is >91%, I have measured more than 92%.
The structure is a BCM boost PFC and a BCM flyback with synchronous
rectifier.
If they can reach >91% with PFC and at only 24V output why I can not
reach 93% without PFC, with 100V output and europen only input range?
check out the Meanwell website Switching Power Supply - Mean Well Switching Power Supply Manufacturer
and download the specs of a product called SDR-120.
This is a 24V/5A SMPS with PFC and univeral mains input.
The efficiency they claim is >91%, I have measured more than 92%.
The structure is a BCM boost PFC and a BCM flyback with synchronous
rectifier.
If they can reach >91% with PFC and at only 24V output why I can not
reach 93% without PFC, with 100V output and europen only input range?
Looks pretty good!
I would've went with a half bridge forward converter at 500W, but as long as the capacitors handle the ripple, flyback is still acceptable. BCM switching is nice, I made a PFC with a BCM controller before. Makes the waveforms nice and clean and predictable.
Is the double EMI filter on the input really necessary? Do you have any rough estimates on the EMI this thing makes? I wouldn't think it would be too bad -- schottky diodes don't exhibit snap recovery, and it's hard to drive 25A MOSFETs with enough current, through a drive transformer, to generate significant harmonics. I wouldn't be surprised if you can knock off one of those stages, or both, and get away with a mere filtered input jack, maybe with the addition of a few Y-type ceramic capacitors to nail down the remaining RFI.
As for efficiency, that's reasonable. I love high voltages because you can pull so many watts from just a few amperes. 230VAC is pretty good, but it gets real impressive if you have 400 or 480V three phase available! Not that you'd be able to listen to a 480V-supplied amplifier for very long...
Oh, one more thing: did you consider IGBTs? I didn't notice if the operating frequency would be suitable (IGBTs are usually useful under 100kHz). A pair of 10-20A rated IGBTs at 600 or 1200V should do the same job with even lower conduction losses (2-3V drop, instead of whatever Rds(on) drops).
Tim
I would've went with a half bridge forward converter at 500W, but as long as the capacitors handle the ripple, flyback is still acceptable. BCM switching is nice, I made a PFC with a BCM controller before. Makes the waveforms nice and clean and predictable.
Is the double EMI filter on the input really necessary? Do you have any rough estimates on the EMI this thing makes? I wouldn't think it would be too bad -- schottky diodes don't exhibit snap recovery, and it's hard to drive 25A MOSFETs with enough current, through a drive transformer, to generate significant harmonics. I wouldn't be surprised if you can knock off one of those stages, or both, and get away with a mere filtered input jack, maybe with the addition of a few Y-type ceramic capacitors to nail down the remaining RFI.
As for efficiency, that's reasonable. I love high voltages because you can pull so many watts from just a few amperes. 230VAC is pretty good, but it gets real impressive if you have 400 or 480V three phase available! Not that you'd be able to listen to a 480V-supplied amplifier for very long...
Oh, one more thing: did you consider IGBTs? I didn't notice if the operating frequency would be suitable (IGBTs are usually useful under 100kHz). A pair of 10-20A rated IGBTs at 600 or 1200V should do the same job with even lower conduction losses (2-3V drop, instead of whatever Rds(on) drops).
Tim
Hi Sch3matic,
The double stage EMI filter is necessary to pass the EMC.
With this filter I am compliant to EN55022 Class B requirement
for conducted emissions.
The two stages are not equals: the first common mode choke (15mH) filters
mainly low frequencies but it is not as effective between 1MHz and 5MHz.
Here is where the second common mode choke (150uH) works at best.
I also love high voltages, making SMPS running from 3-phase 400V or 480V gives always impressive results in terms of efficiency.
Here in Switzerland we have the 3phase 400VAC available in every home so if you want you can build your SMPS running directly from it with impressive performances.
I did not considered IGBTs at all. Not because they are not good but because I have always used MOSFETs with good results and I have done very few works with IGBTs.
Anyway in this SMPS the two mosfets have an Rds_ON of about 0.11ohm so it means that af full peak current of around 11A they lose about 1.2V each.
Using an IGBT with a VCEsat < 1V means using a rather slow type, forget about Warp-speed IGBTs and the like. Maybe I can gain something in conduction loss but what about switching losses?
My switching frequency increases up to around 180kHz at moderate load, I don't think that a low VCEsat IGBT can keep up at these frequencies.
Thank you
-marco
The double stage EMI filter is necessary to pass the EMC.
With this filter I am compliant to EN55022 Class B requirement
for conducted emissions.
The two stages are not equals: the first common mode choke (15mH) filters
mainly low frequencies but it is not as effective between 1MHz and 5MHz.
Here is where the second common mode choke (150uH) works at best.
I also love high voltages, making SMPS running from 3-phase 400V or 480V gives always impressive results in terms of efficiency.
Here in Switzerland we have the 3phase 400VAC available in every home so if you want you can build your SMPS running directly from it with impressive performances.
I did not considered IGBTs at all. Not because they are not good but because I have always used MOSFETs with good results and I have done very few works with IGBTs.
Anyway in this SMPS the two mosfets have an Rds_ON of about 0.11ohm so it means that af full peak current of around 11A they lose about 1.2V each.
Using an IGBT with a VCEsat < 1V means using a rather slow type, forget about Warp-speed IGBTs and the like. Maybe I can gain something in conduction loss but what about switching losses?
My switching frequency increases up to around 180kHz at moderate load, I don't think that a low VCEsat IGBT can keep up at these frequencies.
Thank you
-marco
The other tradeoff is switching speed; if you are able to push enough current through the drive transformer to switch those beefy MOSFETs, then you will certainly win both in frequency, switching loss and conduction loss. If, for instance, a smaller or cheaper drive transformer were required, IGBTs would likewise be necessary for the reduced gate capacitance.
Tim
Tim
...it seems like this is turning into an ABSOLUTE "break-through post".......that long sort after post which really , genuinely demonstrates that flybacks (even BCM ones) can VERY SATISFACTORILY be used for audio at 500W.....even without a PFC...........However, its not the ultimate breakthrough as you are only venturing as far as European mains, and not going down to 90VAC (even with a voltage doubler link)
As we know, in audio, the size of the power supply doesn't really matter....since at least guitarists like their amplifiers to be big chunks on the stage.......can you imagine Axel Rose with some tiny little cube for an amplifier?...no
So far we see that the only opposition to this BCM flyback is the output capacitor ripple current.
....Personally i wouldnt see that as a problem, since with unlimited solution size, i can just whack in loads of paralleled elctrolytics.
I was expecting to see extremely high switching losses, and lots of loss due to the leakage in the transformer..and lots of hysteresis loss....and i was expecting the transformer to be bigger than ETC44.
But i think the revolution is beginning, and the LLC resonant SMPS is turning out to be a hoax(?), since you can get 91%+ efficiency with a 500W flyback, with no PFC, and with BCM operation and 16 Amp peak diode current.
Who needs resonant LLC for audio now?
Also, with your fantastic efficiency, why don't you just go for single switch flyback?
It would mean a higher voltage fet is used, with slightly higher rdson, but just use a bigger heatsink?
As we know, in audio, the size of the power supply doesn't really matter....since at least guitarists like their amplifiers to be big chunks on the stage.......can you imagine Axel Rose with some tiny little cube for an amplifier?...no
So far we see that the only opposition to this BCM flyback is the output capacitor ripple current.
....Personally i wouldnt see that as a problem, since with unlimited solution size, i can just whack in loads of paralleled elctrolytics.
I was expecting to see extremely high switching losses, and lots of loss due to the leakage in the transformer..and lots of hysteresis loss....and i was expecting the transformer to be bigger than ETC44.
But i think the revolution is beginning, and the LLC resonant SMPS is turning out to be a hoax(?), since you can get 91%+ efficiency with a 500W flyback, with no PFC, and with BCM operation and 16 Amp peak diode current.
Who needs resonant LLC for audio now?
Also, with your fantastic efficiency, why don't you just go for single switch flyback?
It would mean a higher voltage fet is used, with slightly higher rdson, but just use a bigger heatsink?
Last edited:
oops sorry , i see your core is EER42/42/20
http://www.tdk.com.tw/ctl_pdf/ctl_2.1.5.pdf
It has an Ae of 240mm^2.
-But their is no Amin given in its datasheet.
-this is bad news for you as its "Amin" that counts, not "Ae".
I believe at 500W , this core would not be big enough.
What is your Bmax at 500W and vin = 187VAC?
Also, have you got K-Type thermocouple temperature measurements for all semiconductors and diodes?
It is my suspicion that you are running your semiconductors way above an acceptable temperature, and that your SMPS is going to have a short operational life.
If i am wrong, then i'll take my hat off, as you will be the pioneer of the cheapest and best Audio SMPS ever......you will revolutionise SMPS's for Audio use....seriously!
http://www.tdk.com.tw/ctl_pdf/ctl_2.1.5.pdf
It has an Ae of 240mm^2.
-But their is no Amin given in its datasheet.
-this is bad news for you as its "Amin" that counts, not "Ae".
I believe at 500W , this core would not be big enough.
What is your Bmax at 500W and vin = 187VAC?
Also, have you got K-Type thermocouple temperature measurements for all semiconductors and diodes?
It is my suspicion that you are running your semiconductors way above an acceptable temperature, and that your SMPS is going to have a short operational life.
If i am wrong, then i'll take my hat off, as you will be the pioneer of the cheapest and best Audio SMPS ever......you will revolutionise SMPS's for Audio use....seriously!
Last edited:
I am surprised this post is not attracting more attention.
If a BCM Flyback with Vin =230VAC can be 92.5% efficient, then how efficient is an LLC resonant converter under the same circumstances.....or a full-bridge.?
.....I would have thought very close to 100% efficient?
Something just doesn't add up here.
If a BCM Flyback with Vin =230VAC can be 92.5% efficient, then how efficient is an LLC resonant converter under the same circumstances.....or a full-bridge.?
.....I would have thought very close to 100% efficient?
Something just doesn't add up here.
Quit hating and start building... i was able to pull 650W at 87% efficiency in basic half bridge, no resonant no nothing, straight PWM with good ol' TL494, from a pair of 13009s in TO-247 (that's right, BJTs!) and at 300W efficiency was 93%. Switching frequency 66kHz, output +/-60v. Transformer made from two EI-33 cores commonly encountered in low-end ATX supplies, using only the "E"s. Bobbin case was made by breaking off the top of one case and the bottom of another, then superglued the two halves together to make a double height bobbin. 
The heatsink was from an ATX PSU, admittedly it was one of the beefier types but still not huge, and the fan was a 80mm 2400rpm one with temperature control, and it didn't reach maximum speed (it was set to reach max at 70C) even though my testing was done in the summer with 29C room temp. Output rectifiers were BYW29-200 and the secondary heatsink didn't even get warm.
It is not difficult to obtain high efficiency at high voltages, especially when you're only using the output rectifiers at half their current rating or less.
The heatsink was from an ATX PSU, admittedly it was one of the beefier types but still not huge, and the fan was a 80mm 2400rpm one with temperature control, and it didn't reach maximum speed (it was set to reach max at 70C) even though my testing was done in the summer with 29C room temp. Output rectifiers were BYW29-200 and the secondary heatsink didn't even get warm.
It is not difficult to obtain high efficiency at high voltages, especially when you're only using the output rectifiers at half their current rating or less.
Last edited:
Transformer datails:
Core EER42/42/20 --> Ae=240mm^2
Amin is not indicated but who cares??? In those geometries Amin is practically
equal to Ae and with all the simplifications you have to do to calculate Bmax
this subtle difference worth nothing.
Primary inductance= 250uH
Primary turns= 32
Bmax=Ipeak*Lpri/(Npri*Ae)
Bmax @ 500W / 230V = 300mT
Bmax @ 500W / 187V = 350mT
I know it is quite high at 187V but even if the transfomer starts saturating there is a second level OCP inside the L6565 that cut off the primary mosfets without damage.
Btw IT IS FOR AUDIO USE!!! WHO WILL USE IT @187VAC / 500W CONTINUOUS OUTPUT??
Probably the first thing that blow putting a 500W sinewave into 4ohm in not my SMPS but the class-D amplifier or the speakers themselves.
It is probably the 10th time I am saying it!
I have not taken thermal measurement at 500W, I have taken them at 200W load and the tranformer is at around 75°C with Tamb=25°C.
Semiconductors are pretty cold, expect the diode bridge.
Btw: have you seen which kind of mosfet am I using? I have used the same to make a 2.4kW converter with good efficiency, 500W even in BMC flyback is nothing for them.
Output capacitors ripple: the calculated value is 5.7Arms
I am using 2x Nippon Chemicon KZE 1000uF/63V per rail + 2x Nippon Chemicon KZE 470uF/63V per rail after the output common mode choke.
The caps are rated:
1000uF/63V NCC KZE: 2.9Arms @ 105°C
470uF/63V NCC KZE: 2Arms @ 105°C
so a total of 7.8Arms allowed at 105°C, they see only 5.7Arms.
On top of that on my amp board I have 2 Nichicon 2200uF/63V per rail
rated 3.2Arms @ 105°C
I think that I have enough margin for ripple considering also the it is:
FOR AUDIO USE!!
Extension to universal mains 85-265VAC:
Of course it is possible, there are 3 solutions:
1) Change the transfomer turn ratio by a factor of about 2, but:
- needs a bigger core
- needs higher voltage output diodes --> forget Schottky
- needs a bigger bridge, well dissipated, the input current will be around
10Arms @ 85VAC
- need bigger input EMI filter chokes to withstand the input current
- need a lot of capacitance at the input to withstand the 100Hz ripple
--> very bad/ugly/crazy solution with low efficiency
(but for sure higher than 80%)
2) Keep quite everything as it is and add a voltage doubler input stage, but:
- need A LOT of capacitance in the voltage doubler, probably something
around 2200uF/200V to withstand the huge 100Hz ripple current at low
input voltage. The capacitors are not high capacitance because you need
the capacitance but to withstand the ripple current. The more
capacitance you add the worst becomes the PF and the higher becomes
the ripple current. Its like a cat chasing its tail....
- needs a bigger bridge, well dissipated, the input current will be around
10Arms @ 85VAC
- need bigger input EMI filter chokes to withstand the input current
--> efficiency won't suffer that much, I expect close to 90% but it
becomes very big, bulky and expensive, I don't like it
3) Add an active boost PFC regulated at 400V:
- the turn ratio can be decreased a little bit with advantage on current
and voltage stress on the semiconductors. Maybe it will be possible to
use and MBR20150 (150V) and gain something also there in term of
efficiency .
- you can use smaller MOSFETs and save someting or keep things as they
are and gain in efficiency
- Primary bulk cap will be small, the ripple in it will be smaller than now and
there are absolutely no contraints on the holdup time because:
IT IS FOR AUDIO USE!
- PFC can be a BCM boost or better an CCM average current controlled
boost.
- A well done PFC can have very high efficiency, for sure >92% @ 120VAC
>95% @ 230VAC
- Bridge and EMI filter can remain the same because of the near 1 PF the
input current at 120VAC will be quite the same as it is now at 230VAC
- It is the solution I prefer but sorry, now I have not time to invest in it, if
someone of you can do the job will be appreciated.
@ KV3Audio
of course I can post plots of output ripple voltage but now I have no time
to dedicate to it. By the end of the week I will post the plots of ripple,
dynamic behavior, switching waveforms and efficiency measurement for
those who don't believe me.
Efficiency @ eem2am:
I tell you what I have made:
- LLC converter at 48V/2A output with BCM boost PFC @ 400V
PFC efficiency @ 230VAC --> 97%
LLC efficiency --> 95%
total efficiency @ 230VAC --> 92%
- Phase shift converter 24V/100A with CCM buck PFC @ 385V
note in the phase shift there is a current doubler syncronous rectifier and
an active ORing diode
PFC efficiecy @ 3ph/400VAC --> 98%
Phase shift efficiency --> 94%
total efficiency @ 3ph/400VAC --> 92%
- BCM flyback converter 48V/10A with BCM boost PFC @ 750V
PFC efficiency @ 400VAC --> 97%
BCM flyback efficiency --> 95%
total efficiency @ 400VAC --> 92%
Yes you are right using modern techiques (and Th3 Un1Que is the example) you can achieve very high efficiencies.
But who care about crazy efficiencies when it is FOR AUDIO USE?
I hope now it is clear.
ciao
-marco
Core EER42/42/20 --> Ae=240mm^2
Amin is not indicated but who cares??? In those geometries Amin is practically
equal to Ae and with all the simplifications you have to do to calculate Bmax
this subtle difference worth nothing.
Primary inductance= 250uH
Primary turns= 32
Bmax=Ipeak*Lpri/(Npri*Ae)
Bmax @ 500W / 230V = 300mT
Bmax @ 500W / 187V = 350mT
I know it is quite high at 187V but even if the transfomer starts saturating there is a second level OCP inside the L6565 that cut off the primary mosfets without damage.
Btw IT IS FOR AUDIO USE!!! WHO WILL USE IT @187VAC / 500W CONTINUOUS OUTPUT??
Probably the first thing that blow putting a 500W sinewave into 4ohm in not my SMPS but the class-D amplifier or the speakers themselves.
It is probably the 10th time I am saying it!
I have not taken thermal measurement at 500W, I have taken them at 200W load and the tranformer is at around 75°C with Tamb=25°C.
Semiconductors are pretty cold, expect the diode bridge.
Btw: have you seen which kind of mosfet am I using? I have used the same to make a 2.4kW converter with good efficiency, 500W even in BMC flyback is nothing for them.
Output capacitors ripple: the calculated value is 5.7Arms
I am using 2x Nippon Chemicon KZE 1000uF/63V per rail + 2x Nippon Chemicon KZE 470uF/63V per rail after the output common mode choke.
The caps are rated:
1000uF/63V NCC KZE: 2.9Arms @ 105°C
470uF/63V NCC KZE: 2Arms @ 105°C
so a total of 7.8Arms allowed at 105°C, they see only 5.7Arms.
On top of that on my amp board I have 2 Nichicon 2200uF/63V per rail
rated 3.2Arms @ 105°C
I think that I have enough margin for ripple considering also the it is:
FOR AUDIO USE!!
Extension to universal mains 85-265VAC:
Of course it is possible, there are 3 solutions:
1) Change the transfomer turn ratio by a factor of about 2, but:
- needs a bigger core
- needs higher voltage output diodes --> forget Schottky
- needs a bigger bridge, well dissipated, the input current will be around
10Arms @ 85VAC
- need bigger input EMI filter chokes to withstand the input current
- need a lot of capacitance at the input to withstand the 100Hz ripple
--> very bad/ugly/crazy solution with low efficiency
(but for sure higher than 80%)
2) Keep quite everything as it is and add a voltage doubler input stage, but:
- need A LOT of capacitance in the voltage doubler, probably something
around 2200uF/200V to withstand the huge 100Hz ripple current at low
input voltage. The capacitors are not high capacitance because you need
the capacitance but to withstand the ripple current. The more
capacitance you add the worst becomes the PF and the higher becomes
the ripple current. Its like a cat chasing its tail....
- needs a bigger bridge, well dissipated, the input current will be around
10Arms @ 85VAC
- need bigger input EMI filter chokes to withstand the input current
--> efficiency won't suffer that much, I expect close to 90% but it
becomes very big, bulky and expensive, I don't like it
3) Add an active boost PFC regulated at 400V:
- the turn ratio can be decreased a little bit with advantage on current
and voltage stress on the semiconductors. Maybe it will be possible to
use and MBR20150 (150V) and gain something also there in term of
efficiency .
- you can use smaller MOSFETs and save someting or keep things as they
are and gain in efficiency
- Primary bulk cap will be small, the ripple in it will be smaller than now and
there are absolutely no contraints on the holdup time because:
IT IS FOR AUDIO USE!
- PFC can be a BCM boost or better an CCM average current controlled
boost.
- A well done PFC can have very high efficiency, for sure >92% @ 120VAC
>95% @ 230VAC
- Bridge and EMI filter can remain the same because of the near 1 PF the
input current at 120VAC will be quite the same as it is now at 230VAC
- It is the solution I prefer but sorry, now I have not time to invest in it, if
someone of you can do the job will be appreciated.
@ KV3Audio
of course I can post plots of output ripple voltage but now I have no time
to dedicate to it. By the end of the week I will post the plots of ripple,
dynamic behavior, switching waveforms and efficiency measurement for
those who don't believe me.
Efficiency @ eem2am:
I tell you what I have made:
- LLC converter at 48V/2A output with BCM boost PFC @ 400V
PFC efficiency @ 230VAC --> 97%
LLC efficiency --> 95%
total efficiency @ 230VAC --> 92%
- Phase shift converter 24V/100A with CCM buck PFC @ 385V
note in the phase shift there is a current doubler syncronous rectifier and
an active ORing diode
PFC efficiecy @ 3ph/400VAC --> 98%
Phase shift efficiency --> 94%
total efficiency @ 3ph/400VAC --> 92%
- BCM flyback converter 48V/10A with BCM boost PFC @ 750V
PFC efficiency @ 400VAC --> 97%
BCM flyback efficiency --> 95%
total efficiency @ 400VAC --> 92%
Yes you are right using modern techiques (and Th3 Un1Que is the example) you can achieve very high efficiencies.
But who care about crazy efficiencies when it is FOR AUDIO USE?
I hope now it is clear.
ciao
-marco
Actually, my point was that even using techniques from 30 years ago you can still achieve high efficiency if the parts have room to breathe.
I have little experience with flyback supplies, i only built a small 50W 12v -> 400v flyback, but i remember Crown did something silly in one of their big amps, a 5kW flyback IIRC. Uses a stupid large core, but still, they showed it's doable. By far the biggest nuisance for me in high voltage supplies is the large output inductor required, and for now i'm sticking to unregulated, but of course there's the disadvantage of large capacitors. I can live with that though. The flyback is attractive from this point of view - all the benefits of an output inductor without actually needing one. I have an ETD49 lying around, i'll see what i can do with that next year.
But IMO unregulated SMPS + PFC front-end is the way to go for audio, and indeed this is the case in many high-end amplifiers. Nothing says "powerful" better than a huge bank of caps.
I have little experience with flyback supplies, i only built a small 50W 12v -> 400v flyback, but i remember Crown did something silly in one of their big amps, a 5kW flyback IIRC. Uses a stupid large core, but still, they showed it's doable. By far the biggest nuisance for me in high voltage supplies is the large output inductor required, and for now i'm sticking to unregulated, but of course there's the disadvantage of large capacitors. I can live with that though. The flyback is attractive from this point of view - all the benefits of an output inductor without actually needing one. I have an ETD49 lying around, i'll see what i can do with that next year.
But IMO unregulated SMPS + PFC front-end is the way to go for audio, and indeed this is the case in many high-end amplifiers. Nothing says "powerful" better than a huge bank of caps.
Hi Th3 uN1Qu3,
yes I told exacltly the same: if you using Stone Age parts are able to reach reasonably high efficiency why it should not be possible with modern parts?
An ETD49 can be fine for this project you just need some more turns (smaller Ae compared to EER42) but much wider winding area.
PFC+unregulated forward is basically the simulation of the good old transformer with rectifier and huge caps bank.
From a structure like that you can teoretically have as much power as you want unless the PFC keeps up.... or you blow something.
You prefer to go unregulated just for saving the output inductor or there are other reasons?
@eem2am: Labgruppen FP14000, 14kW of output power!
Do you really think that its SMPS can deliver such power CONTINUOSLY maybe also at the lowest input voltage without exploding?
yes I told exacltly the same: if you using Stone Age parts are able to reach reasonably high efficiency why it should not be possible with modern parts?
An ETD49 can be fine for this project you just need some more turns (smaller Ae compared to EER42) but much wider winding area.
PFC+unregulated forward is basically the simulation of the good old transformer with rectifier and huge caps bank.
From a structure like that you can teoretically have as much power as you want unless the PFC keeps up.... or you blow something.
You prefer to go unregulated just for saving the output inductor or there are other reasons?
@eem2am: Labgruppen FP14000, 14kW of output power!
Do you really think that its SMPS can deliver such power CONTINUOSLY maybe also at the lowest input voltage without exploding?
There's also the fact that an audio amplifier is a pretty difficult load for a regulated supply, and i don't like having a minimum load wasting power all the time to keep the supply stable. I don't have all the equipment needed to see what's happening in a SMPS feedback loop, so i have to stick to the calculations and some trial-and-error, which makes things a bit more difficult than they could be.
OK Thanks,
I hear you say "its for audio", but if you search the web, you will not find any source who can actually say what is the time domain power consumption of a class D amplifier.
i have tried to find this before but failed...eg
http://www.diyaudio.com/forums/power-supplies/195442-power-flow-class-d-guitar-amplifier.html
...i do smps for class D but the class D is not yet built and somebody else do it. i do not know what for example is the maximum time duration of peak power.
unless you can exactly state what is the exact worst case time domain power consumption of the class D amplifier then the jury is out on this one, and the class D amplifier is guilty of continuous peak power draw until prooven innocent of this.
As i said,, for worst-case music signal (and who on earth knows what that really is) the time domain power draw is needed before we can under-rate components
I hear you say "its for audio", but if you search the web, you will not find any source who can actually say what is the time domain power consumption of a class D amplifier.
i have tried to find this before but failed...eg
http://www.diyaudio.com/forums/power-supplies/195442-power-flow-class-d-guitar-amplifier.html
...i do smps for class D but the class D is not yet built and somebody else do it. i do not know what for example is the maximum time duration of peak power.
unless you can exactly state what is the exact worst case time domain power consumption of the class D amplifier then the jury is out on this one, and the class D amplifier is guilty of continuous peak power draw until prooven innocent of this.
As i said,, for worst-case music signal (and who on earth knows what that really is) the time domain power draw is needed before we can under-rate components
Also, i hear you say your leakage inductance is 3uH.
This is 1.2%.
This is very good for a transformer with such a big gap.
Did you use a Frequency analyser to measure this or a simple cheap LCR meter?
Really, your leakage needs to be measured with a frequency analyser, and the measurement needs to be taken at that switching frequency which corresponds to your maximum power.
This is 1.2%.
This is very good for a transformer with such a big gap.
Did you use a Frequency analyser to measure this or a simple cheap LCR meter?
Really, your leakage needs to be measured with a frequency analyser, and the measurement needs to be taken at that switching frequency which corresponds to your maximum power.
Hi,
leakage inductance is measured with a LCR meter able to test at 100kHz.
The leakage was measured @ 100kHz / 1V level
It is the inductance measured at the primary with all the winding shorted, including bias and aux windings.
My switching frequency at full power is around 50kHz.
1.2% is not that bad but condider that it is sandwich wound. The gap lenght does not influence too much the leakage, its it really the geometry of the windings that counts.
On top of that in a two switch flyback the leakage inductance is not as important as in a single switch flyback. The majority of the energy stored in the leakage inductance is recycled back to the primary capacitor bank thanks to the freewheeling diodes and does not need to be dissipated in a lossy RCD as in a conventional flyback.
Higher leakage means higher losses in the freewheeling diodes due to their longer conduction time.
This is another advantage of this structure compared to single switch flyback.
ciao
leakage inductance is measured with a LCR meter able to test at 100kHz.
The leakage was measured @ 100kHz / 1V level
It is the inductance measured at the primary with all the winding shorted, including bias and aux windings.
My switching frequency at full power is around 50kHz.
1.2% is not that bad but condider that it is sandwich wound. The gap lenght does not influence too much the leakage, its it really the geometry of the windings that counts.
On top of that in a two switch flyback the leakage inductance is not as important as in a single switch flyback. The majority of the energy stored in the leakage inductance is recycled back to the primary capacitor bank thanks to the freewheeling diodes and does not need to be dissipated in a lossy RCD as in a conventional flyback.
Higher leakage means higher losses in the freewheeling diodes due to their longer conduction time.
This is another advantage of this structure compared to single switch flyback.
ciao
I remember we agreed that even with trafo gate drive for top and bottom fet, the chances of the leakage being the same for top and bottom etc etc was unlikely.
We agreed that it was most likely that the FETs would not switch on at the same instant.
I just ran a simulation in LTspice of a 320W two transistor flyback in CCM with vin = 390VDC.
When i program the FETs to switch at the same instant, the fet dissipation is 6W in each FET.
When i delay the switching of the top fet so that it switches 150ns behind the bottom fet its a different story.
I get 12W dissipation in the bottom fet and 10W dissipation in the top FET.
With such a wide variation, how are you supposed to size your heatsink.?
You would have to use a big heatsink to take account of the worst case.
Ther ugly truth of switching losses is rearing its head for the two switch flyback.
We agreed that it was most likely that the FETs would not switch on at the same instant.
I just ran a simulation in LTspice of a 320W two transistor flyback in CCM with vin = 390VDC.
When i program the FETs to switch at the same instant, the fet dissipation is 6W in each FET.
When i delay the switching of the top fet so that it switches 150ns behind the bottom fet its a different story.
I get 12W dissipation in the bottom fet and 10W dissipation in the top FET.
With such a wide variation, how are you supposed to size your heatsink.?
You would have to use a big heatsink to take account of the worst case.
Ther ugly truth of switching losses is rearing its head for the two switch flyback.
The two fets will never switch at exacly the same instant.
This causes that the reflected voltage is not shared evenly between the two fets. Due to this fact the switching losses in the two fets will not be exactly the same but the sum of the two will be always the same.
Don't rely too much on simulations for switching losses, the mosfet models are usually oversimplified and the results are very different from the reality.
On top of that why a 2 switch flyback should have so lower switching losses
compared to a single switch?
A rough estimation of switching losses for turn off is Ploss=Vds*Idpeak/2*Fsw*toverlap, where Vds in the peak drain voltage, Idpeak the peak drain current at turn-off, Fsw is the switching frequency and toverlap is the time when the mosfet in its linear region and the voltage is rising and the current is falling. This has to be measured and the gate drive shuold be optimized for that. Forget simulating it, it is a waste of time.
A 2sw flyback has lower Vds but all the other parameter are roughly the same and on top of that the fets are 2 and not one --> 2*Ploss.
A BCM operation (1sw or 2sw) reduce turn on losses to very small values but turn off values are always the same.
To reduce turn off losses you can put a turn off snubber in parallel to your fet
or simply a properly sized film capacitor in parallel to the fet to reduce dv/dt.
A parallel capacitor (in CCM) decrease the turn off losses somewhat but increases turn on losses, some tradeoff is needed.
ciao
-marco
This causes that the reflected voltage is not shared evenly between the two fets. Due to this fact the switching losses in the two fets will not be exactly the same but the sum of the two will be always the same.
Don't rely too much on simulations for switching losses, the mosfet models are usually oversimplified and the results are very different from the reality.
On top of that why a 2 switch flyback should have so lower switching losses
compared to a single switch?
A rough estimation of switching losses for turn off is Ploss=Vds*Idpeak/2*Fsw*toverlap, where Vds in the peak drain voltage, Idpeak the peak drain current at turn-off, Fsw is the switching frequency and toverlap is the time when the mosfet in its linear region and the voltage is rising and the current is falling. This has to be measured and the gate drive shuold be optimized for that. Forget simulating it, it is a waste of time.
A 2sw flyback has lower Vds but all the other parameter are roughly the same and on top of that the fets are 2 and not one --> 2*Ploss.
A BCM operation (1sw or 2sw) reduce turn on losses to very small values but turn off values are always the same.
To reduce turn off losses you can put a turn off snubber in parallel to your fet
or simply a properly sized film capacitor in parallel to the fet to reduce dv/dt.
A parallel capacitor (in CCM) decrease the turn off losses somewhat but increases turn on losses, some tradeoff is needed.
ciao
-marco
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.