I'm working on a smps design but have some questions
The supply needs to generate:
+40V
-40V
+5V
+3.3V
low voltage supply (5V) for PWM
The +/-40V will supply an audio amp with around 450Watts, the rest of the power is shared by the +5 and 3.3V windings.
Supply is going to be full-bridge type, zero volt switching.
Question 1:
I think it's more important to regulate the +/-40 supply. There's going to be 2 windings for this. Does it make sense to have the feedback circuit to the pwm measure across the +/-40, checking for 80V or is it better to measure just + or - 40V to ground?
Question 2:
Cross regulation isn't very clear to me. But, does it make sense to have 2 smps: one for the +/-40V and another for the low voltage stuff?
Question 3:
Assuming I wind all the voltages on a single core. I think that it would go something like (from inner to outer):
primary, pwm 5V, +40, -40, +5, +3.3
How hard is it to fit, and what are the consequences of this many windings on a single core?
Question 4:
I'm pretty sure a bobbin will be used for this. Also, in order to meet safety requirments, either safety margins need to be used with another barrier from primary to secondary, or I could use triple insulated wire. Any thoughts there? Is that stuff expensive? Easier to work with than margins? I think only one layer needs to be wound with it, yes? Although if the primary is triple insulated, I suppose the 5V pwm voltage winding needs to be also (given the stack I listed) so that all the remaining windings can be normal magnet wire. Is that correct?
The supply needs to generate:
+40V
-40V
+5V
+3.3V
low voltage supply (5V) for PWM
The +/-40V will supply an audio amp with around 450Watts, the rest of the power is shared by the +5 and 3.3V windings.
Supply is going to be full-bridge type, zero volt switching.
Question 1:
I think it's more important to regulate the +/-40 supply. There's going to be 2 windings for this. Does it make sense to have the feedback circuit to the pwm measure across the +/-40, checking for 80V or is it better to measure just + or - 40V to ground?
Question 2:
Cross regulation isn't very clear to me. But, does it make sense to have 2 smps: one for the +/-40V and another for the low voltage stuff?
Question 3:
Assuming I wind all the voltages on a single core. I think that it would go something like (from inner to outer):
primary, pwm 5V, +40, -40, +5, +3.3
How hard is it to fit, and what are the consequences of this many windings on a single core?
Question 4:
I'm pretty sure a bobbin will be used for this. Also, in order to meet safety requirments, either safety margins need to be used with another barrier from primary to secondary, or I could use triple insulated wire. Any thoughts there? Is that stuff expensive? Easier to work with than margins? I think only one layer needs to be wound with it, yes? Although if the primary is triple insulated, I suppose the 5V pwm voltage winding needs to be also (given the stack I listed) so that all the remaining windings can be normal magnet wire. Is that correct?
For a 450W supply, one solution is to produce +/-40V from a two-transistor forward converter, with the low-voltage supplies produced from dc-dc buck regulators off the 40V output.
The dynamic loading on the +/- 40V outputs will see a large current swing (unless you're feeding a class-A amp) so it's not wise to try to get all the outputs from the main stage: you'll find serious cross-regulation problems.
The dynamic loading on the +/- 40V outputs will see a large current swing (unless you're feeding a class-A amp) so it's not wise to try to get all the outputs from the main stage: you'll find serious cross-regulation problems.
First, you can use a sitching chip/regulator capable of bipolar measurement -- the LT1683 -- but it is SSOP20 and about $9.00 at Digikey. i think it is rather overkill.
you must use mylar tape between the windings --
you could use push-pull -- the transformer doesn't retain energy --you want it passing the energy through to the load -- take a look at this PDF by National Semiconductor -- they walk you through the mathematics of the transformer:
http://www.national.com/appinfo/power/files/LM5030_Push-Pull-12V-60W-2001i-new-4.pdf
i've linked a lot of SMPS articles on my this page of my website:
http://www.tech-diy.com/smps.htm
you must use mylar tape between the windings --
you could use push-pull -- the transformer doesn't retain energy --you want it passing the energy through to the load -- take a look at this PDF by National Semiconductor -- they walk you through the mathematics of the transformer:
http://www.national.com/appinfo/power/files/LM5030_Push-Pull-12V-60W-2001i-new-4.pdf
i've linked a lot of SMPS articles on my this page of my website:
http://www.tech-diy.com/smps.htm
Ouroboros said:
Yeah, I agree with you - cross reg stuff - which partly prompted this thread. BTW, amplifier is Class-D type, 6 or possibly eight channels worth.
This option of another follow on stage came to mind a while ago, but I completely forgot about it - until now. I suppose I could live with the efficiency loss, but need to think about that. All the digital logic will be running, let's say, around 30 watts - the total isn't yet clear, but am working to nail that down. My supply is going to be PFC'd, so efficiency total may not be all that wonderful -> .9 * .9 = 0.81. Then add another switcher for the low voltage stuff, another factor of 0.9, maybe 72% total efficiency there - hmm. At least the heat gets distributed around
It's a similar situation to one I have here at work on a 250W PSU I designed for a large voice alarm / fire alarm panel. That has an L4981 PFC stage, followed by the 2-transistor forward convertor isolating stge producing the main 46V dc output (to some Class-D audio amps), with 24V dc and 3.3V dc buck regulators to feed the electronics of the system, together with two separate switch-mode battery chargers feeding a pair of 24V 26A-hr backup battery packs, all supplied by dc-dc buck stages from the 46V.
You'll find that the PFC stage will give very high efficiency. It provides me with 400V dc from the normal UK 230V mains. Fortunately I don't have to be able to work from 115V mains. It will, but the efficiency drops when the PFC has to provide 400V dc from 115V supplies.
ST Micro provide ggod app notes for the L4981 chip.
You'll find that the PFC stage will give very high efficiency. It provides me with 400V dc from the normal UK 230V mains. Fortunately I don't have to be able to work from 115V mains. It will, but the efficiency drops when the PFC has to provide 400V dc from 115V supplies.
ST Micro provide ggod app notes for the L4981 chip.
First, I would probably use a small flyback for the +5V and +3.3V outputs, probably a self-oscillating one (simple, reliable and inherently overload-proof), since these small windings would cause a lot of trouble if they had to be placed in the main transformer and output inductor. Also, even when putting your best efforts on it, cross-regulation wouldn't probably be good enough to avoid undervoltage and overvoltage conditions to the logic chips, so post-regulation would be required. Furthermore, having a separate small flyback for +5V and +3.3V allows to implement standby and shutdown features in a quite easy way by just turning-off the main converter.
I would wind the main transformer this way: half primary | bifilar secondaries | half primary, with the primary halves connected in series. Output inductor should be also bifilar (one for both outputs), and one or more stacked iron-powder cores may be well suited for that. Doing it this way, you can even sense only one of the +40V outputs, or you can sense both in series as you mention, it will be quite stable anyway provided good coupling and matching between both 40V outputs.
Triple insulated wire seems to be expensive and very hard to find, I have not been able to find any practical source for it in small quantities yet. But its main drawback is size: it's much thicker than standard wire, so I would regard it as only practical for small turn counts in low-voltage outputs.
To wind such a power transformer, I would probably employ something like a ETD39 core (first guess of required size) with a vertical bobbin, leaving 5mm margin between the windings and the edge of the coil in the side of the pins and 2.5mm in the opposite side. Pins from the primaries and secondaries should be in opposite sides of the coil former and the three usual layers of mylar tape should be placed between primary and secondary winding layers. You may stack two sets of bifilar secondaries connected in paralell if required.
p.s. I would enjoy designing and building such a PSU if anybody were willing to pay for it
I would wind the main transformer this way: half primary | bifilar secondaries | half primary, with the primary halves connected in series. Output inductor should be also bifilar (one for both outputs), and one or more stacked iron-powder cores may be well suited for that. Doing it this way, you can even sense only one of the +40V outputs, or you can sense both in series as you mention, it will be quite stable anyway provided good coupling and matching between both 40V outputs.
Triple insulated wire seems to be expensive and very hard to find, I have not been able to find any practical source for it in small quantities yet. But its main drawback is size: it's much thicker than standard wire, so I would regard it as only practical for small turn counts in low-voltage outputs.
To wind such a power transformer, I would probably employ something like a ETD39 core (first guess of required size) with a vertical bobbin, leaving 5mm margin between the windings and the edge of the coil in the side of the pins and 2.5mm in the opposite side. Pins from the primaries and secondaries should be in opposite sides of the coil former and the three usual layers of mylar tape should be placed between primary and secondary winding layers. You may stack two sets of bifilar secondaries connected in paralell if required.
p.s. I would enjoy designing and building such a PSU if anybody were willing to pay for it
Hi Eva!
Where would you connect the primary of this? To the PFC? To the +40V? It should work well off the +40 supply, I think , and won't need to meet safety. Also, I'm not familiar with any self-oscillating circuits. Can you point me at a data sheet?
I need a little help on the bifilar windings, since I don't know how to set that up.
Can you point me to a good source for the bobbins, cores, and clips? Not sure where to get them.
Oh, you didn't mention where to put the low voltage winding for the pwm? Any thoughts?
Eva said:
Where would you connect the primary of this? To the PFC? To the +40V? It should work well off the +40 supply, I think , and won't need to meet safety. Also, I'm not familiar with any self-oscillating circuits. Can you point me at a data sheet?
I need a little help on the bifilar windings, since I don't know how to set that up.
Can you point me to a good source for the bobbins, cores, and clips? Not sure where to get them.
Oh, you didn't mention where to put the low voltage winding for the pwm? Any thoughts?
gearheadgene said:
someone will have nightmares -- but I have done bifilar windings this way -- get the approximate length of wire you need, fold in half, attach one end to a drill chuck, the other to a fixed point and wind it slowly -- works !!!!
good source of cores and bobbins -- ham radio guys use Amidon -- but even more use computer power supply transformers -- you take the main transformer off the PCB, wrap in aluminum foil, bake in a 400 degree oven for 45 minutes and gently pry apart making sure not to crack the core or bobbin -- this works beautifully and will give you a core capable of 300 watts -- you can stack two or more cores for more juice.
Eva said:p.s. I would enjoy designing and building such a PSU if anybody were willing to pay for it
Excuse me for polluting the topic, but you seem to be unreachable outside the forum.
We (my company) might have a project coming up which demands a reliable, extended temperature range power supply. Guesstimate is around 250W, 230VAC only, 2-3 DC output voltages of which one will be +5VDC for the digital stuff. Must comply to a shitload of regulations, and can probably not be bought due to mechanical issues.
My first offline SMPS ever (85-265VAC in, 16-32VDC in, builtin UPS, 48VDC+5VDC out) which I had to design and build because nobody else could do it at that moment (in the land of the blind, one-eye is king) works well and is awaiting certification, but I am still far, far away from becoming a seasoned SMPS designer. Now, I know what is going to happen next: you did it once, you can do it again. But personally I would prefer to outsource that part to someone with more experience.
Would you be willing to pick up that part of the design or provide some consultancy if it crosses my path? I do understand that your labour and prototypes must be paid for
gearheadgene:
In order to learn a bit about self oscillating flybacks, I recommend you studying the standby +5V section of seveal low-cost ATX PC power supplies. You will find a lot of ideas there.
Concerning the +5V supply for the big PWM, you can get it from the primary side of the oscillating flyback, as it creates a 10..20V low current supply for free (optocoupler supply).
I had some self-oscillating schematics that I took from ATX PSUs to study them some years ago, but they are in another computer that suffered a hard drive crash and it can no longer connect to internet, so I can't publish them until I repair it.
Check also MC44608 flyback control IC from On-Semi, it's routinely used in the PSUs of low-cost TVs as it has a whole lot of features packed in a DIP-8 package, while requiring very few external components. It even has a standby mode featuring very low losses.
In order to learn a bit about self oscillating flybacks, I recommend you studying the standby +5V section of seveal low-cost ATX PC power supplies. You will find a lot of ideas there.
Concerning the +5V supply for the big PWM, you can get it from the primary side of the oscillating flyback, as it creates a 10..20V low current supply for free (optocoupler supply).
I had some self-oscillating schematics that I took from ATX PSUs to study them some years ago, but they are in another computer that suffered a hard drive crash and it can no longer connect to internet, so I can't publish them until I repair it.
Check also MC44608 flyback control IC from On-Semi, it's routinely used in the PSUs of low-cost TVs as it has a whole lot of features packed in a DIP-8 package, while requiring very few external components. It even has a standby mode featuring very low losses.
Eva,
Thanks for the lead to ON semi, I'll check on it. In the mean time, just had a few questions regarding the bifilar windings you mentioned earlier. Say I go with the method you suggested - sandwich the bifilar +/-40V secondary between 2 half windings of the eprimary. Intuitivly, I believe this method will make both the + and - side match well. I'm not sure what splitting the primary does for you, though. At a minimum, it requires the mylar tape in 2 places and requires extra margins - making the construction a bit tougher, yes? In a garage operation, maybe this is ok, but what about production. Seems like a big adder to the cost. What's your feeling for a more traditional stack, e.g. all the primary windings, then the secondary (bifilar or otherwise).
btw, you can call me gene.
Thanks for the lead to ON semi, I'll check on it. In the mean time, just had a few questions regarding the bifilar windings you mentioned earlier. Say I go with the method you suggested - sandwich the bifilar +/-40V secondary between 2 half windings of the eprimary. Intuitivly, I believe this method will make both the + and - side match well. I'm not sure what splitting the primary does for you, though. At a minimum, it requires the mylar tape in 2 places and requires extra margins - making the construction a bit tougher, yes? In a garage operation, maybe this is ok, but what about production. Seems like a big adder to the cost. What's your feeling for a more traditional stack, e.g. all the primary windings, then the secondary (bifilar or otherwise).
btw, you can call me gene.
The ARRL headquarters is in Newington -- they have a lab there, probably a bunch of folks who can point you in the direction of obtaining the stuff you need for your SMPS. (There is also a "how to" in the ARRL handbook and you can look over the trafo winding section while you are there.)
the ARRL website www.arrl.org also has a page devoted to swap meets in every state -- great sources for parts.
jack
the ARRL website www.arrl.org also has a page devoted to swap meets in every state -- great sources for parts.
jack
What's your opinion on deriving power for the digital logic from the + or - 40VDC supply? Actually, now that I think about it, even if that doesn't load down the supply too much, it could add switching noise to the rails and that feeds my audio amps (yuck). Maybe not the best idea. But tempting, nonetheless (easier to go from +40 to 5 non-isolated, than it is from 400 to 5 isolated from a magnetics and compliance perspective)
Know of any reference designs for +5 and +3.3 VDC offline smps that uses PFC front end (+400VDC input)?
gene
Know of any reference designs for +5 and +3.3 VDC offline smps that uses PFC front end (+400VDC input)?
gene
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.