The ultimate amplifier SMPS. What comes after LLC? Isolated single stage PFC.
Advantages:
- High power factor at mid to high load across full line range.
- Only one custom magnetic component.
- All bulk capacitance in secondary side. Only average audio power going through SMPS until very low frequencies.
- Simplicity. Lowest part count in power stage.
- ZVS ON achievable at mid to high load. At low load conduction is discontinuous and the FET turns on through an inductor. Then the IGBT turns on at close to zero volts.
- ZCS OFF achievable at low to mid load. The IGBT turn off into a capacitor, Ic falls as Vce rises, Ic takes longer to fall to zero as load increases.
- IGBT used in a way that results in minimum switching loss at any load.
- Very low idle dissipation achievable.
- Up to 2KW short term from E42 transformer and 3 TO-220 mid current IGBTs
Disadvantages:
- Complex control required, involving some MCU code, at both sides of isolation barrier.
- Achieving IGBT ZCS OFF up to higher load is in conflict with achieving low losses at low to mid load.
- Low power burst mode and high power current limiting result in small amounts of mechanical noise from magnetics and capacitors.
Simplified power stage schematic is attached, combining resonant operation, flyback, PFC and isolation. A few parts and tweaks are not drawn.
This video shows switching waveform across a power sweep: https://www.youtube.com/watch?v=ztISHByR88s
Advantages:
- High power factor at mid to high load across full line range.
- Only one custom magnetic component.
- All bulk capacitance in secondary side. Only average audio power going through SMPS until very low frequencies.
- Simplicity. Lowest part count in power stage.
- ZVS ON achievable at mid to high load. At low load conduction is discontinuous and the FET turns on through an inductor. Then the IGBT turns on at close to zero volts.
- ZCS OFF achievable at low to mid load. The IGBT turn off into a capacitor, Ic falls as Vce rises, Ic takes longer to fall to zero as load increases.
- IGBT used in a way that results in minimum switching loss at any load.
- Very low idle dissipation achievable.
- Up to 2KW short term from E42 transformer and 3 TO-220 mid current IGBTs
Disadvantages:
- Complex control required, involving some MCU code, at both sides of isolation barrier.
- Achieving IGBT ZCS OFF up to higher load is in conflict with achieving low losses at low to mid load.
- Low power burst mode and high power current limiting result in small amounts of mechanical noise from magnetics and capacitors.
Simplified power stage schematic is attached, combining resonant operation, flyback, PFC and isolation. A few parts and tweaks are not drawn.
This video shows switching waveform across a power sweep: https://www.youtube.com/watch?v=ztISHByR88s
Attachments
Last edited:
More than double the peak power going through the SMPS is required if bulk capacitors are in primary side. This results in about 2x losses in FETs, windings and any resistive element. The classic approach of bulk capacitors in secondary side still prevails, audio has crest factor, and even for a crest factor of 1 power at top of the wave is 2x average power of the wave.
Last edited:
Pretty impressive. Do you even use a HV active probe for this stuff? I own an incredibly expensive scope, Keysight MSOX 3104T and I don't have the guts to do this. HV active probes are so expensive, if you built your own HV probe do share.
AC_REF and I_REF I assume is doing the PFC brain work (with your PIC MCU), I "think" I've seem something like this in an MPPT/Inverter charge controller. I think you will have alot more fun with power electronics in the energy conversion space, audio seems to have reached its innovation ceiling.
Last edited:
I have a 1200V rated probe that shows fast edges better. 600V Hameg probes work OK for SMPS but fast edges can appear to be faster than actually are due to capacitor losses. Remember bulk capacitance is in secondary side. Isolation transformers and variac under the bench.
A brief description of PFC brain, not understandable for newbies or copiers of course:
Two switched integrators are used for control:
- "IV" for integration of instantaneous line voltage, it is reset and rearmed at end of GL=1. Gain is controlled by "power command", which comes from complex MCU calculations.
- "II" for integration of inductor current, it reset during GH=1, active rest of time. Fixed gain.
The converter has 4 states:
- GA=high: starts when IV exceeds a "minimum charge" threshold or a fixed dead timeout for minimum frequency control.
- GL=high: starts when SW is below a ZVS threshold, or GA=high timeout (a short fixed time for component protection).
- GH=high: starts when II exceeds IV, I_REF exceeds a threshold, or GL=high timeout. Time is fixed for minimum frequency control. I_REF threshold is a function of line voltage for overload control.
- Dead time: starts when GH=high time ends. Time is a function of output voltage, line peak voltage and line instantaneous voltage, calculated witn MCU.
Self oscillation and integrated charge control remove flyback inductor discontinuous-to-continuous duty-cycle slope non-linearity from transfer function, and also removes resonance between flyback inductor and HV-to-RV resonant capacitor. The circuit breaks into subharmonic oscillation when a too-low input impedance is demanded for a too-low input instantaneous voltage (input impedance becoming negative below self-oscillating loop crossover frequency, and lower input voltage results in lower open loop gain), which is the practical impedance/bandwidth limit of the converter, together with main inductor saturation around peak line voltage.
A brief description of PFC brain, not understandable for newbies or copiers of course:
Two switched integrators are used for control:
- "IV" for integration of instantaneous line voltage, it is reset and rearmed at end of GL=1. Gain is controlled by "power command", which comes from complex MCU calculations.
- "II" for integration of inductor current, it reset during GH=1, active rest of time. Fixed gain.
The converter has 4 states:
- GA=high: starts when IV exceeds a "minimum charge" threshold or a fixed dead timeout for minimum frequency control.
- GL=high: starts when SW is below a ZVS threshold, or GA=high timeout (a short fixed time for component protection).
- GH=high: starts when II exceeds IV, I_REF exceeds a threshold, or GL=high timeout. Time is fixed for minimum frequency control. I_REF threshold is a function of line voltage for overload control.
- Dead time: starts when GH=high time ends. Time is a function of output voltage, line peak voltage and line instantaneous voltage, calculated witn MCU.
Self oscillation and integrated charge control remove flyback inductor discontinuous-to-continuous duty-cycle slope non-linearity from transfer function, and also removes resonance between flyback inductor and HV-to-RV resonant capacitor. The circuit breaks into subharmonic oscillation when a too-low input impedance is demanded for a too-low input instantaneous voltage (input impedance becoming negative below self-oscillating loop crossover frequency, and lower input voltage results in lower open loop gain), which is the practical impedance/bandwidth limit of the converter, together with main inductor saturation around peak line voltage.
Last edited:
There are some MCUs in the PIC16F177x and PIC16F178x range that include almost the hardware needed (op-amps, DACs, comparators, PWM blocks, logic cells, etc.) to implement converters like this, and easier ones of course.
The other side needs a MCU to supervise any good amplifier anyway. Digital feedback, exchanging many signals with just 2 digital optos, plus a simple photocoupler for power up.
The other side needs a MCU to supervise any good amplifier anyway. Digital feedback, exchanging many signals with just 2 digital optos, plus a simple photocoupler for power up.
Last edited:
IGBT have lower conduction losses than MOSFET, and when the circuit takes care of switching losses IGBT provide highest efficiency, even at 130khz. This is the trend being taken by industry in home appliances (for example I got a 2KW induction cooker that uses just one TO-3P IGBT). There are a few IGBT types in the range of $1 to $1.5 that fit the purpose for resonant switchers, like FGP15N60UNDF, IKP20N65H5, STGP20V60DF or AOT15B60D. This is about 50% the cost of comparable MOSFET.
Great chips but the IO is a bit hair raising at times, I recently got caught with mysterious offsets on some analogue pins that turned out to be the internal pullups! I had stupidly assumed these would be disabled on an analogue pin
I see your point about IGBT's in ZV/CS at high frequencies, interesting because it's hard to get (expensive) high voltage low rdson mosfets, I work on solar stuff so not applicable on this site.
Power conversion leads to the same tricks in all fields, only that some tricks apply more to some fields than others. MOSFET are fast, but switching losses can account for 50% of total losses. When the circuit allows for resonant switching edges, or can be modified for resonant with little changes, a small capacitor can cut losses by about 50%.
Some day I want to make solar power conversion, with resonant tricks and optimum power point, like a compact system capable of working as charge regulator, inverter and UPS, with configurable ratios of surplus energy to provide to mains service or to batteries. I have ideas for that field too, and a roof asking for ~10x 130W panels.
Some day I want to make solar power conversion, with resonant tricks and optimum power point, like a compact system capable of working as charge regulator, inverter and UPS, with configurable ratios of surplus energy to provide to mains service or to batteries. I have ideas for that field too, and a roof asking for ~10x 130W panels.
I don't wanna mess with your thread but you know I spend more time slowing down mosfets to reduce EMI, in fact I give up some switching efficiency as a deliberate policy in that way. You got me really thinking about resonance and IGBT's now
Your certainly in the right part of the world for PV, here I am at 52deg lat but still good. 130W panels are a bit small, I got 265's at a good price, you can go up to 320+ but the $$/W gets worse again. You will have trouble on DIY sites anything to do with GTI's all the crazies come out the wood and tell you it's not allowed lols anyway a new direction for your talents. Don't forget water heating, no plumbing I don't use batteries, prefer my woodburner for space heating!
Your certainly in the right part of the world for PV, here I am at 52deg lat but still good. 130W panels are a bit small, I got 265's at a good price, you can go up to 320+ but the $$/W gets worse again. You will have trouble on DIY sites anything to do with GTI's all the crazies come out the wood and tell you it's not allowed lols anyway a new direction for your talents. Don't forget water heating, no plumbing
I don't use batteries, prefer my woodburner for space heating!
Resonant is different. A transistor can be turned off faster into a closely placed ceramic capacitor without increased EMI. In the circuit I showed the IGBTs are turned off into a capacitor in parallel with transformer primary. The FET is also turned off into a smaller capacitor to keep losses low, but this part is omitted from the drawing. The circuit was designed to obtain the desired level of conducted EMI. Radiated in small loops is not a problem as long as ringing is suppressed.
Something I meant to ask, are you programming it in C, assembler or a mixture ?
I find maths on the 16F in assembler a bit excruciating at times being only 8 bit, no multiplier and a very limited instruction set (but it's fast). An 18F would be much better but then you don't get the analogue peripherals......
I'm programming it in pseudo code similar to C but more relaxed, then manually compiling it to assembler, this allows to choose optimization strategy for each code section. For support tasks such as generating precalculated tables based on parameters I use the good old BASIC. For example, for changing current limit I have to edit and run a BAS file (with external editor), then re-compile and program the MCU, as a pre-calculated table is used.
What other brands call "a PFC power supply".
The following waveforms were obtained measuring voltage drop across a 0.1ohm shunt in series with mains supply. A 5KVA variac was used to get desired line voltage, and 3KVA isolation transformers were used to make it safe.
Low quality 120V mains was simulated by stepping 230V to 120V with variac before feeding 1:1 isolation transformers. Mid quality 120V mains was obtained with a setting close to 1:1 in variac and doing the 2:1 step-down with isolation transformers. Mid quality 230V mains was obtained with close to 1:1 setting in variac and 1:1 in isolation transformers.
This is what other brand calls "a PFC power supply". It's just a single stage rectifier with isolated output and wide input voltage range, implemented with a UCC2891/2/3/4 control IC not intended for PFC (fixed freq peak current control), waveform distortion is high as hell, harmonics are strong, worse than old diode bridge and caps, and instability arises at low quality mains. Steps in current waveform are observed at high-line high-load, maybe from a control loop with maybe only 16 discrete power levels possible. This is *cheap* design/copy (as Behringer) but a bit more rugged.
Pictures in order:
- line current @ 200W @ 120V low quality mains, onset of PFC regulations
- line current @ 350W @ 120V low quality mains falling to 115V
- line current @ 740W @ 120V low quality mains falling to 110V
- line current @ 860W @ 120V mid quality mains falling to 114V, stable overload
- line current @ 200W @ 230V mid quality mains, onset of PFC regulations
- line current @ 400W @ 230V mid quality mains falling to 226V
- line current @ 800W @ 230V mid quality mains falling to 221V
- line current @ 1100W @ 230V mid quality mains falling to 219V, hiccup in response to overload
The following waveforms were obtained measuring voltage drop across a 0.1ohm shunt in series with mains supply. A 5KVA variac was used to get desired line voltage, and 3KVA isolation transformers were used to make it safe.
Low quality 120V mains was simulated by stepping 230V to 120V with variac before feeding 1:1 isolation transformers. Mid quality 120V mains was obtained with a setting close to 1:1 in variac and doing the 2:1 step-down with isolation transformers. Mid quality 230V mains was obtained with close to 1:1 setting in variac and 1:1 in isolation transformers.
This is what other brand calls "a PFC power supply". It's just a single stage rectifier with isolated output and wide input voltage range, implemented with a UCC2891/2/3/4 control IC not intended for PFC (fixed freq peak current control), waveform distortion is high as hell, harmonics are strong, worse than old diode bridge and caps, and instability arises at low quality mains. Steps in current waveform are observed at high-line high-load, maybe from a control loop with maybe only 16 discrete power levels possible. This is *cheap* design/copy (as Behringer) but a bit more rugged.
Pictures in order:
- line current @ 200W @ 120V low quality mains, onset of PFC regulations
- line current @ 350W @ 120V low quality mains falling to 115V
- line current @ 740W @ 120V low quality mains falling to 110V
- line current @ 860W @ 120V mid quality mains falling to 114V, stable overload
- line current @ 200W @ 230V mid quality mains, onset of PFC regulations
- line current @ 400W @ 230V mid quality mains falling to 226V
- line current @ 800W @ 230V mid quality mains falling to 221V
- line current @ 1100W @ 230V mid quality mains falling to 219V, hiccup in response to overload
Attachments
-
pas_line_current_7_1100W_in_230V_mid_quality_falling_219V_hiccup_limting.jpg -
pas_line_current_6_800W_in_230V_mid_quality_falling_221V.jpg -
pas_line_current_5_400W_in_230V_mid_quality_falling_226V.jpg -
pas_line_current_4_200W_in_230V_mid_quality_onset_PFC_regs.jpg -
pas_line_current_3_860W_in_120V_mid_quality_falling_114V.jpg -
pas_line_current_2_740W_in_120V_low_quality_falling_110V.jpg -
pas_line_current_1_350W_in_120V_low_quality_falling_115V.jpg -
pas_line_current_0_200W_in_120V_low_quality_onset_PFC_regs.jpg
What I call a "single stage PFC power supply".
These images were obtained from a prototype of the circuit and control approach explained at the beginning of the thread. The result is very low harmonics and phase shift, nearly sine current waveform at all conditions, also on low quality line. Incidentally my implementation uses same size power stage as "other brand", but optimized with the help of hundreds of hours of engineering and a custom control technique I developed. This results in 2x the part count, being the difference mostly in the amount of small parts.
Pictures in order:
- line current @ 200W @ 120V low quality mains, onset of PFC regulations
- line current @ 350W @ 120V low quality mains falling to 115V
- line current @ 700W @ 120V low quality mains falling to 110V
- line current @ 1400W @ 120V low quality mains falling to 100V, stable overload (PSU)
- line current @ 200W @ 230V mid quality mains, onset of PFC regulations
- line current @ 400W @ 230V mid quality mains falling to 226V
- line current @ 800W @ 230V mid quality mains falling to 221V
- line current @ 1600W @ 230V mid quality mains falling to 213V, stable overload (amplifier)
- line current @ 2KW @ 235V mid quality mains falling to 215V, had to reduce amplifier dummy load impedance to overload power supply at 230V
These images were obtained from a prototype of the circuit and control approach explained at the beginning of the thread. The result is very low harmonics and phase shift, nearly sine current waveform at all conditions, also on low quality line. Incidentally my implementation uses same size power stage as "other brand", but optimized with the help of hundreds of hours of engineering and a custom control technique I developed. This results in 2x the part count, being the difference mostly in the amount of small parts.
Pictures in order:
- line current @ 200W @ 120V low quality mains, onset of PFC regulations
- line current @ 350W @ 120V low quality mains falling to 115V
- line current @ 700W @ 120V low quality mains falling to 110V
- line current @ 1400W @ 120V low quality mains falling to 100V, stable overload (PSU)
- line current @ 200W @ 230V mid quality mains, onset of PFC regulations
- line current @ 400W @ 230V mid quality mains falling to 226V
- line current @ 800W @ 230V mid quality mains falling to 221V
- line current @ 1600W @ 230V mid quality mains falling to 213V, stable overload (amplifier)
- line current @ 2KW @ 235V mid quality mains falling to 215V, had to reduce amplifier dummy load impedance to overload power supply at 230V
Attachments
-
sma_line_current_8_2KW_in_235V_mid_quality_falling_215V.jpg -
sma_line_current_7_1600W_in_230V_mid_quality_falling_213V.jpg -
sma_line_current_6_800W_in_230V_mid_quality_falling_221V.jpg -
sma_line_current_5_400W_in_230V_mid_quality_falling_226V.jpg -
sma_line_current_4_200W_in_230V_mid_quality_onset_PFC_regs.jpg -
sma_line_current_3_1400W_in_120V_low_quality_falling_100V.jpg -
sma_line_current_2_700W_in_120V_low_quality_falling_110V.jpg -
sma_line_current_1_350W_in_120V_low_quality_falling_115V.jpg -
sma_line_current_0_200W_in_120V_low_quality_onset_PFC_regs.jpg
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.