Hi All,
I'm thinking about designing a SMPS full digital (I2S in) PurePath amplifier which will be an active 2 channel. Being used in a car for a compo set. So power will be at around 80..100W not to much I guess
Offcourse there will be one PCB for left and right.
But I have some doubt at certain components in the SMPS section. Hope you guys could help be out here.
My choice for a transformer core was set on a Ferroxcube ETD29 3C90. After alot of reading I think this core should be large enough for the power requirement. My aim is running at 50kHz, Np at 4x4 turns 1mm wire and Ns at 15 turns 0,8mm. I only need one +40V supply at around 2...2,5A so a full bridge rectifier is going to be used. Could someone verify I am safe as for the Bmax rating of this core with maybe a Max Vin of 15V?
The choice of the fets are the IRF2907, somewhere on this forum they a supposed to made in heaven or something Specs look good also
After reading that the PurePath powerstages need some better regulated powersupply due to a moderate low PSRR I came to the conclusion that I need feedback in this SMPS. So more reading..... In some threads the SG3525 was regarded as a bad choice for this arrangement and a better choice would be the UC2825 with a current mode feedback loop. But offcourse the secondary needs te be isolated for audio, so a current sense transformer is needed I think. Need some light on this issue also, if some one could get me started on this then that would be great.
All this be put on a double layer 70µm PCB with mostly SMD components. So some EMI RFI issues will be hopefully avoided within a certain degree.
Input and output chokes, snubbers and capacitors will be on my discovery next. Consider that as a HINT
Any help would be great and I hope I did some of my homework the correct way.
I'm thinking about designing a SMPS full digital (I2S in) PurePath amplifier which will be an active 2 channel. Being used in a car for a compo set. So power will be at around 80..100W not to much I guess
Offcourse there will be one PCB for left and right.But I have some doubt at certain components in the SMPS section. Hope you guys could help be out here.
My choice for a transformer core was set on a Ferroxcube ETD29 3C90. After alot of reading I think this core should be large enough for the power requirement. My aim is running at 50kHz, Np at 4x4 turns 1mm wire and Ns at 15 turns 0,8mm. I only need one +40V supply at around 2...2,5A so a full bridge rectifier is going to be used. Could someone verify I am safe as for the Bmax rating of this core with maybe a Max Vin of 15V?
The choice of the fets are the IRF2907, somewhere on this forum they a supposed to made in heaven or something
Specs look good also After reading that the PurePath powerstages need some better regulated powersupply due to a moderate low PSRR I came to the conclusion that I need feedback in this SMPS. So more reading..... In some threads the SG3525 was regarded as a bad choice for this arrangement and a better choice would be the UC2825 with a current mode feedback loop. But offcourse the secondary needs te be isolated for audio, so a current sense transformer is needed I think. Need some light on this issue also, if some one could get me started on this then that would be great.
All this be put on a double layer 70µm PCB with mostly SMD components. So some EMI RFI issues will be hopefully avoided within a certain degree.
Input and output chokes, snubbers and capacitors will be on my discovery next. Consider that as a HINT
Any help would be great and I hope I did some of my homework the correct way.
Hi Trilithium,
I have done some calculations and your ETD29 core seems too small, at least with your current number of primary turns.
I assume that your input voltage range is 10..15V so the PWM controller will
be at full duty cycle at Vin=10V.
At 50kHz (100kHz oscillator frequency) your maximum on time is ton=9.5us (at Vin=10V and assuming your controller is limiting the duty cycle at max 95%)
Your Bmax is Bmax=Vin*ton/(N*Ae)=10V*9.5us/(4*76mm^2)=312mT.
This is really too high, not for the core saturation but for the core losses.
Lets calculate your core losses at 312mT and 50kHz. From 3C90 material specs the core loss density at 300mT and 50kHz are 500kW/m3.
Your core losses are Pcore=500kW/m3*Ve=2.7W.
The transformer temperature rise not counting the copper losses is
Trise=(Pcore/Sa)^0.833. The surface area (Sa) of the ETD29 core is
2150mm^2 so your temperature rise will be 56°C... Much too high, because you have to take into account also the copper losses which are load dependent.
Consider the following stuff:
1) Increase the number of turns to keep Bmax<200mT.
2) Select a bigger core or a core with different shape. RM or PQ core offers
very large Ae at the expense of a small winding area. Take a look on RM12, it has 146mm^2 Ae in a footprint comparable to ETD29. Of course the winding area is much smaller than ETD29.
3) Increase your switching frequency, but take care than core losses are frequency dependent. On top of that going too high with the frequency will increase switching losses and EMI problems.
Your mosfets are ok, probably a little bit overrated. At those power level in a 12V push-pull you can go also with IRF540 or similar mosfets (much less expensive). Consider that this SMPS is for audio application so it will never give you 100W continuosly. Your continuous rating will be in the range of 20-25W.
If you need to regulate your smps you don't need a current transformer. To sense the primary current just use a shunt resistor connected between the two mosfets source and gnd. To regulated the output preserving the galvanic isolation use an optocoupler and a TL431 shunt regulator on the secondary side.
Good choice to use current mode control. Chips like SG3525 use voltage mode. Current mode control avoid transformer saturation due to brigde imbalance.
ciao
-marco
I have done some calculations and your ETD29 core seems too small, at least with your current number of primary turns.
I assume that your input voltage range is 10..15V so the PWM controller will
be at full duty cycle at Vin=10V.
At 50kHz (100kHz oscillator frequency) your maximum on time is ton=9.5us (at Vin=10V and assuming your controller is limiting the duty cycle at max 95%)
Your Bmax is Bmax=Vin*ton/(N*Ae)=10V*9.5us/(4*76mm^2)=312mT.
This is really too high, not for the core saturation but for the core losses.
Lets calculate your core losses at 312mT and 50kHz. From 3C90 material specs the core loss density at 300mT and 50kHz are 500kW/m3.
Your core losses are Pcore=500kW/m3*Ve=2.7W.
The transformer temperature rise not counting the copper losses is
Trise=(Pcore/Sa)^0.833. The surface area (Sa) of the ETD29 core is
2150mm^2 so your temperature rise will be 56°C... Much too high, because you have to take into account also the copper losses which are load dependent.
Consider the following stuff:
1) Increase the number of turns to keep Bmax<200mT.
2) Select a bigger core or a core with different shape. RM or PQ core offers
very large Ae at the expense of a small winding area. Take a look on RM12, it has 146mm^2 Ae in a footprint comparable to ETD29. Of course the winding area is much smaller than ETD29.
3) Increase your switching frequency, but take care than core losses are frequency dependent. On top of that going too high with the frequency will increase switching losses and EMI problems.
Your mosfets are ok, probably a little bit overrated. At those power level in a 12V push-pull you can go also with IRF540 or similar mosfets (much less expensive). Consider that this SMPS is for audio application so it will never give you 100W continuosly. Your continuous rating will be in the range of 20-25W.
If you need to regulate your smps you don't need a current transformer. To sense the primary current just use a shunt resistor connected between the two mosfets source and gnd. To regulated the output preserving the galvanic isolation use an optocoupler and a TL431 shunt regulator on the secondary side.
Good choice to use current mode control. Chips like SG3525 use voltage mode. Current mode control avoid transformer saturation due to brigde imbalance.
ciao
-marco
Thanks for the great reply.
So I did some calculations of my own with these formulas and came up with a 6 turns primary with a ETD29 and ETD34 . The RM12 seems very small with the amount of wire needed, but I have not yet calucated the nescessary winding surface yet.
I'm not sure where to find the Sa of the ETD34 core but I did a cross multiply with the ETD29 and came to about 3050mm^2. Looks good.
With the ETD29 I came to a Bmax of 208mT and a Trise of 26°C. Bmax still a bit high but Trise looks a lot better.
This took me to the ETD34 and gave a Bmax of 163mT and a Trise of 22°C. This all sound reasonable I think, right?
Price of the mosfets isn't really an issue, maybe I'll do a dive in the parts bin.
As for the 20-25W, I did consider this. But also the fact that volume is logarithmic. A bit more is preferred just to be sure. Not sure how a 25W design can handle peaks though...
The regulation sounds good. I'll have a look into that shunt direction.
Regards.
So I did some calculations of my own with these formulas and came up with a 6 turns primary with a ETD29 and ETD34 . The RM12 seems very small with the amount of wire needed, but I have not yet calucated the nescessary winding surface yet.
I'm not sure where to find the Sa of the ETD34 core but I did a cross multiply with the ETD29 and came to about 3050mm^2. Looks good.
With the ETD29 I came to a Bmax of 208mT and a Trise of 26°C. Bmax still a bit high but Trise looks a lot better.
This took me to the ETD34 and gave a Bmax of 163mT and a Trise of 22°C. This all sound reasonable I think, right?
Price of the mosfets isn't really an issue, maybe I'll do a dive in the parts bin.
As for the 20-25W, I did consider this. But also the fact that volume is logarithmic. A bit more is preferred just to be sure. Not sure how a 25W design can handle peaks though...
The regulation sounds good. I'll have a look into that shunt direction.
Regards.
Hi Trilithium,
These numbers looks much better than the previous.
Even with ETD29 208mT is reasonable at 50kHz.
If you can fit all your wire on ETD29 use this core; there is no need
for ETD34.
Don't oversize your wires gauges, it is an audio amplifier and you don't need
to pull 100W continuosly from your SMPS.
For the MOSFETS: IRF2907 has huge Qg (180nC) and you don't need its very low Rds_on of 3.5mohm at those power levels. Due to its huge Qg you will
need a very strong gate driver to drive it properly. Probably the integrate driver of your PWM controller is not enough and you need an external gate driver (extra cost and PCB area).
Go for a smaller mosfet, with lower Qg and higher Rds_on you will have only advantages.
ciao
-marco
These numbers looks much better than the previous.
Even with ETD29 208mT is reasonable at 50kHz.
If you can fit all your wire on ETD29 use this core; there is no need
for ETD34.
Don't oversize your wires gauges, it is an audio amplifier and you don't need
to pull 100W continuosly from your SMPS.
For the MOSFETS: IRF2907 has huge Qg (180nC) and you don't need its very low Rds_on of 3.5mohm at those power levels. Due to its huge Qg you will
need a very strong gate driver to drive it properly. Probably the integrate driver of your PWM controller is not enough and you need an external gate driver (extra cost and PCB area).
Go for a smaller mosfet, with lower Qg and higher Rds_on you will have only advantages.
ciao
-marco
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.