We have to examine the maths and physics of pfc digression in the 1kW domain: Continuous/average mode running f= 60KHz; required core volume would be around 73cm3 i.e E65 core size. Boost diode losses around 7W. Inductor 300uH at 25% ripple.
Looking at 100KHz operation the core size would come down to around 44cm3 i.e E55. However boost diode and switching mos losses rise to 10W % 35W for Mosfet. Inductor value around 180uH. Clearly with rising freq, switching losses begin to mount.
A magnetic snubber would improve efficiency at 88V.
Theres alot to reiterate. At 1kW a lowRds value of mosfet is quite important. The sense resistor value roughly 20milli-ohm 3W is going to be troublesome. (A mosfet could be used in lieu). Current transformers add complications i.e require extra knowledge. The EMI noise is also an issue and is related to ripple current in the inductor. Choosing a higher value inductor lowers this proportionally but the size goes up.
Another issue is to run both DC/DC and pfc at the same frequency; this then becomes too low for the DC/DC converter when one can realistically see a running freq over 100Khz.
With the efficiency figure I previously mentioned may seem pessimistic. However the pfc inductor value and core volume is important...that is any peak current that might saturate the core would be devastating for both switching devices. At a kW a fail would be quite dramatic.
AN overall efficiency of just 70% for both DC/DC and pfc combined
running at 1KW o/p power was suggest designinging for 1400W.....that would be the case for Telecom......but for music rating your peak condition is unlikely to be reached. As a designer I would go for the expected peak figure. Only my opin.
The hardest part is to <see a spec what one wants> and eventually arrive at a compromise sep and build round it. There is also the aspect of the build time. The other poss is just build the DC/DC part and swallow the grim cap input power factor. I'm against this.
I can look at the required input cap values; but this cap-input policy is fast disappearing and I'm against this.........Note recent energy legistation in CA.
EVA has seen the audio end through, again as she mentions the european mains volts sort of rescues the pfc.
I hope you can see what happens between steady music conditions and actual worst case peak power.
Cold water on ideas ? still luke warm.
richj
Looking at 100KHz operation the core size would come down to around 44cm3 i.e E55. However boost diode and switching mos losses rise to 10W % 35W for Mosfet. Inductor value around 180uH. Clearly with rising freq, switching losses begin to mount.
A magnetic snubber would improve efficiency at 88V.
Theres alot to reiterate. At 1kW a lowRds value of mosfet is quite important. The sense resistor value roughly 20milli-ohm 3W is going to be troublesome. (A mosfet could be used in lieu). Current transformers add complications i.e require extra knowledge. The EMI noise is also an issue and is related to ripple current in the inductor. Choosing a higher value inductor lowers this proportionally but the size goes up.
Another issue is to run both DC/DC and pfc at the same frequency; this then becomes too low for the DC/DC converter when one can realistically see a running freq over 100Khz.
With the efficiency figure I previously mentioned may seem pessimistic. However the pfc inductor value and core volume is important...that is any peak current that might saturate the core would be devastating for both switching devices. At a kW a fail would be quite dramatic.
AN overall efficiency of just 70% for both DC/DC and pfc combined
running at 1KW o/p power was suggest designinging for 1400W.....that would be the case for Telecom......but for music rating your peak condition is unlikely to be reached. As a designer I would go for the expected peak figure. Only my opin.
The hardest part is to <see a spec what one wants> and eventually arrive at a compromise sep and build round it. There is also the aspect of the build time. The other poss is just build the DC/DC part and swallow the grim cap input power factor. I'm against this.
I can look at the required input cap values; but this cap-input policy is fast disappearing and I'm against this.........Note recent energy legistation in CA.
EVA has seen the audio end through, again as she mentions the european mains volts sort of rescues the pfc.
I hope you can see what happens between steady music conditions and actual worst case peak power.
Cold water on ideas ? still luke warm.
richj
Eva, sounds like you have some solid hueristic knowledge of audio. I'm not following your numbers, though. If the amp is 100W rms, then wouldn't the peak power be 141 watts (100 * root 2)? Then a 6 channel system would need 846 watts peak. How did you come up with 1200 watts worst case?
You confirmed my gut feeling that the real world average power consumption is going to be significantly less - I just don't have the experience to back that up. I used to do some audio work for telephone systems, and I recall looking at the crest factor - but don't remember the details except that peak to average was a big number.
So getting back to your description, I would need to supply 200 watts average with enough capacitance/storage to handle the transient peaks?
You confirmed my gut feeling that the real world average power consumption is going to be significantly less - I just don't have the experience to back that up. I used to do some audio work for telephone systems, and I recall looking at the crest factor - but don't remember the details except that peak to average was a big number.
So getting back to your description, I would need to supply 200 watts average with enough capacitance/storage to handle the transient peaks?
gearheadgene said:
Exactly. Note that this requires careful pulse by pulse and average current limiting in the PFC section, and a lot of capacitance in the PFC bus. However, a compromise between output power and capacitance may be found, I would recommend 400W or 500W for the PFC as these numbers are somewhat easier to reach than 1000W. There are even 500W evaluation boards for sale, I think.
(my son wanted to see that 
Can you elaborate on the pulse by pulse limiting? What type of regulator would you recommend for this power level - half-bridge, 2 transistor forward, flyback, other?
I am going to look to see what eval boards are available. As I said before, On-semi has a 1KW pfc available as eval board but am not sure if they also have a 500W dc-dc eval board. Let me check. If anyone knows of some, feel free to chime in.
Pulse by pulse current limiting is achieved by comparing the instantaneous inductor current with a certain threshold value and turning off the switch for the rest of the pulse if that value is reached. Most control ICs have that comparator and logic built-in and it's very useful in preventing disasters and in allowing the PFC to operate even with a slight degree of main inductor saturation, squeezing its power output to the maximum.
But pulse by pulse limiting causes chaos instability (period doubling artifacts) when operating in continuous mode with duty cycles above 50%, this is not destructive but it creates audible noise in the inductor. That is why it's highly desirable to add average current limiting with a thresold slightly lower than the one used in peak current limiting, in such a way that peak current limiting is only activated in transient or abnormal conditions and average limiting dominates.
Average current limiting is not built-in in most control ICs, but it's easily achieved by clamping the multiplier output (connected to current amplifier input) because it controls how much current has to flow through the inductor.
In my 2KW prototype I use a L4981A control IC and I do average current limiting with a LM358 low-cost op-amp and two transistors arranged as a current mirror, that steals current from the multiplier output when required. The result is a current waveform perfectly clipped just before inductor saturation.
The DC-DC converter is the simplest part, I think that an unregulated half bridge with two 10A IGBTs would be fine (note that there are SMPS IGBTs intended to provide fast switching at the expense of higher conduction losses, and standard slow IGBTs with the lowest conduction losses, intended for low frequency switching. You should avoid the latter as their turn-off loses are too high to be practical in SMPS).
PD: There are dozens of PFC control ICs and most of their manufacturers have comprehensive application notes available in their websites that explain PFC design criteria and procedures. You should download them and study them if you haven't done it yet.
But pulse by pulse limiting causes chaos instability (period doubling artifacts) when operating in continuous mode with duty cycles above 50%, this is not destructive but it creates audible noise in the inductor. That is why it's highly desirable to add average current limiting with a thresold slightly lower than the one used in peak current limiting, in such a way that peak current limiting is only activated in transient or abnormal conditions and average limiting dominates.
Average current limiting is not built-in in most control ICs, but it's easily achieved by clamping the multiplier output (connected to current amplifier input) because it controls how much current has to flow through the inductor.
In my 2KW prototype I use a L4981A control IC and I do average current limiting with a LM358 low-cost op-amp and two transistors arranged as a current mirror, that steals current from the multiplier output when required. The result is a current waveform perfectly clipped just before inductor saturation.
The DC-DC converter is the simplest part, I think that an unregulated half bridge with two 10A IGBTs would be fine (note that there are SMPS IGBTs intended to provide fast switching at the expense of higher conduction losses, and standard slow IGBTs with the lowest conduction losses, intended for low frequency switching. You should avoid the latter as their turn-off loses are too high to be practical in SMPS).
PD: There are dozens of PFC control ICs and most of their manufacturers have comprehensive application notes available in their websites that explain PFC design criteria and procedures. You should download them and study them if you haven't done it yet.
Kenshin said:
Slope compensation is not practical for current limiting purposes, particularly in continuos-mode PFC converters, whose duty cycle continuously sweeps between 10% and 90%. I tried it and the required amount of slope compensation in order to get stability was huge, the current limit threshold was sweeping from 10A at the beginning of the cycle to 20A at the end. Note that inductor current ripple was only 2 to 3A peak to peak, so an error of 10A in current limit threshold is absolutely unacceptable.
The inductor was rated 650uH 20A and Fsw was 45Khz for that prototype. It was definitely required to have the peak current limiting threshold always set at 20A in order to obtain the expected 2KW output with input voltages as low as 160V AC.
I'm not so aware of EMI reduction effects due to chaos, it just makes the circuit to jump between Fsw, Fsw/2, Fsw/3, etc... in a randomly basis, but base period doesn't change. Note thet Fsw/n shares most of the harmonics of Fsw. On the other hand, average EMI reduction is usually achieved by periodically sweeping Fsw between two values half an octave apart (for example 45 to 63Khz), so Fsw*n harmonics get also swept and evenly spread over the RF spectrum.
Kenshin said:
It also means:
4: An inaccurate observation, effect, or result, especially one resulting from the technology used in scientific investigation or from experimental error: The apparent pattern in the data was an artifact of the collection method. (From Amerigan Heritage Dictionary).
2 : a product of artificial character (as in a scientific test) due usually to extraneous (as human) agency. (From Merriam-Webster Dictionary).
Altough what I meant to say was just 'unexpected, undesirable, unpredictable, very complex, hard to take rid of and hard to understand period-doubling effects'. Sorry.
gearheadgene said:
umm.....looking at the above comment.........it's becoming glaringly obvious that the subject is getting over complicated.....for someone starting......
--> I would try and get hold of some pfc demoboards........for a start. To set everything up as EVa suggests one needs to be pretty addicted at smps electronic benchwork and above all knowledge.
richj
Eva,
So you are suggesting to let the pfc handle the bulk regulation? How would the dc-dc part not be regulated? Maybe it's running without feedback? I definitely want to be isolated from the mains.
I'm going to dig into the cycle by cycle limiting that you are talking about, and will ask more questions then.
You talked about average mode pfc, but what about the critical conduction mode? I was attracted to it because it may have lower overall emi because of the frequency dithering - yet, how would that same feature affect the noise on the output? Something else to think about, is how the switching artifacts will affect the audio. Any insight there?
Rich: Although the thread is taking a couple of twists, I think it's still on track. Sure is generating a lot of views, isn't it?
gene
So you are suggesting to let the pfc handle the bulk regulation? How would the dc-dc part not be regulated? Maybe it's running without feedback? I definitely want to be isolated from the mains.
I'm going to dig into the cycle by cycle limiting that you are talking about, and will ask more questions then.
You talked about average mode pfc, but what about the critical conduction mode? I was attracted to it because it may have lower overall emi because of the frequency dithering - yet, how would that same feature affect the noise on the output? Something else to think about, is how the switching artifacts will affect the audio. Any insight there?
Rich: Although the thread is taking a couple of twists, I think it's still on track. Sure is generating a lot of views, isn't it?
gene
Eva said:
Chaos does not necessarily means random jump between Fsw, Fsw/2, Fsw/3, etc... it only means the dynamics of a definite system become randomized. Some chaotic system can produce a wide, flat spectrum without any clear spikes.
This is perhaps why there's many papers discussing chaotic boost / flyback converter (even though the base sampleing period is fixed). But if it's practical/useful to a DIY SMPS for audio is unknown.
A interesting DIY use of chaos is electronic mosquito repeller: it generates frequencies between 9.5~12KHz to mimic "male mosquito sound".
I think the real artifacts about subharmonic oscillations is it may increase the current ripples. =>Smaller average current for the same peak current. This could hold back the output power.
Maybe a hysteresis (bang-bang) control of inductor current may be appreciated.
Put a current sensing resistor series with the inductor, and use opti-coupler to send current sensing signal to low-side?
Looks like the thread has run out of gas - Just thought I'd summarize things:
1. Great discussion, that led me to believe 500Watts in this application (audio amp) will be sufficient - maybe should rename the thread?
2. Will use the reference design from ST Micro using L4981A as a baseline PFC. It's 500 watts.
3. From PFC, am going to build DC-DC converter, probably half-bridge, 500 watts to convert to 35 volts.
4. I am interested in using PWM from Linear Tech (maybe LTC3721) because they provide simulation models. That lets me fool around with the design using LT SwitcherCad prior to building it.
I'll keep this updated as things move forward.
gene
1. Great discussion, that led me to believe 500Watts in this application (audio amp) will be sufficient - maybe should rename the thread?
2. Will use the reference design from ST Micro using L4981A as a baseline PFC. It's 500 watts.
3. From PFC, am going to build DC-DC converter, probably half-bridge, 500 watts to convert to 35 volts.
4. I am interested in using PWM from Linear Tech (maybe LTC3721) because they provide simulation models. That lets me fool around with the design using LT SwitcherCad prior to building it.
I'll keep this updated as things move forward.
gene
Elsewhere under Power Supply design I've posted my 760W PFC schematic based on the L4981, which you could adapt to your needs. You'd probably want to increase the caps on the 400V rail and this will lead to control circuit value changes.
As a main converter I would recommend the use of a phase shifted ZVS bridge for this application. I would certainly not use flyback or forward converter. I've posted the schematic of my 700W ZVS converter elsewhere on this forum. But if you're a SMPS 'freshman' , I'd steer clear of this and do a simple unregulated bridge of MOSfets or IGBTs as someone else has already suggested.
Good luck and prepare for many big bangs and disappointments. There's nothing like blowing something up to force one to learn how not to do it again!! 
As a main converter I would recommend the use of a phase shifted ZVS bridge for this application. I would certainly not use flyback or forward converter. I've posted the schematic of my 700W ZVS converter elsewhere on this forum. But if you're a SMPS 'freshman' , I'd steer clear of this and do a simple unregulated bridge of MOSfets or IGBTs as someone else has already suggested.
Good luck and prepare for many big bangs and disappointments. There's nothing like blowing something up to force one to learn how not to do it again!!
John Hope said:
As a main converter I would recommend the use of a phase shifted ZVS bridge for this application. I would certainly not use flyback or forward converter. I've posted the schematic of my 700W ZVS converter elsewhere on this forum. But if you're a SMPS 'freshman' , I'd steer clear of this and do a simple unregulated bridge of MOSfets or IGBTs as someone else has already suggested.
Good luck and prepare for many big bangs and disappointments. There's nothing like blowing something up to force one to learn how not to do it again!!
At the risk of sounding unknowing - I really don't know what you and Eva mean by an unregulated bridge. Can you tell me more about that?
Yes, I'm expecting some bangs. Has anyone started a thread on the safety stuff? It might be a nice thing to list out all the problems one can expect - what will go up in smoke, and how best to prepare.
gene
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.