Building a SMPS for an audio amplifier - some pointers required
Here's the start of a long post... 
I've read a lot on SMPS design in the past 2 years, and this year i also acquired all tools needed to go from schematic/simulation to the real thing. Namely an oscilloscope, and more recently a SMD rework station. All i lack is an isolation transformer but i preferred to get the hot air station instead and i'm very glad i did that as my new controller board is all SMDs, mostly recovered from motherboards, and it's DANG SMALL!
Anyway any nasties will reflect into the secondary so i'm not worried by that.
I chose a rather unusual approach to a controller, because it just stuck into my brain for some reason. That is, i use a 555 timer. No no no, don't shout at me already. Instead of using the imprecise and inefficient way of using pin 5 for voltage regulation (one schematic on the internet claimed up to 150W output, yeah right... i built it myself and it delivered 15 watts at most with the switching transistor getting hot even without a load), an idea struck me at some point.
Long story short, it's been simulated, built as a low power 5v -> 12v converter that scored in at 89% efficiency at 3.5W out, and my latest adaptation of this controller thing is an off-line 2-transistor forward converter that i intend to power my subwoofer amp.
Now you'll be like... What? An offline supply controlled by a 555??? Well yes. My clever idea was as follows. I used the schematic that allows duty cycle control independent to frequency, so i can set maximum duty cycle with a trimpot (now this has become fixed as i made the move to a complete SMD board). Then, i used a LM311 comparator to toggle pin 4 (reset) and this is where the magic comes in. Basically what i'm getting is pulse skipping. Not the cleanest method of regulation, but it works well and it's quite efficient. A TL431 serves as reference for the comparator and a resistor divider sets the output voltage. Then the output of the 555 drives a small mosfet that switches the 15v rail via a 1uF cap into a base drive transformer recovered from an ATX power supply. Then the transformer drives a pair of MJE13007s. The controller board is powered from a separate power supply (i had a tiny mains transformer around anyway), which is regulated with a 7815. Thus no need for an opto for feedback, i simply wire the output voltage straight into the divider.
I have gotten to the point where i figured out the snubbers i need to keep my transistors from blowing. A couple weeks ago i managed to blow ALL the switching transistors i had lying around the house until i figured out the snubber, and eventually bought some new MJE13007s. They're just 80 cents a pop anyway.
I will be trying to do some non-dissipative snubbers next as those resistors get quite hot, but for now i'm happy that i'm not blowing transistors anymore.
I have tested the power supply up to 80W using a transformer wound in a really ugly way.
Today i started winding the "proper" transformer. I use an EI-33 core (have a bunch of those around, i don't think i have to tell you where i got them from), and i wound the primary with 5 strands of 28AWG in parallel, the secondary with 4 strands. I have 48 turns in my primary. I used the formula 7.75A/mm^2 to determine the number of strands. I'm looking at +/-70v out, 400W peak power (my subwoofer is 8 ohm 250W). I'm not expecting 400W continuous as this is audio where the 1/8 power rule stands true, and this will be an amp for home use and i live in an apartment. Also in its new box (the underside of my desk actually, more on that later), the sub is quite loud even with my 2*30W amp that i am using for my main speakers.
Switching frequency is 100kHz or somewhere close to that. The base drive waveform isn't exactly square, but the waveform on the secondary of the transformer is nice and fat. The influence of the snubbers is clearly seen, there is a slope on the falling edge and a bit of ringing on the way, but no spikes at all.
Okay here comes my first Q. My calculations (for worst case low line + full power conditions) ended up with 30 turns for each secondary. I only did one secondary today as my hands hurt already, and i plugged the transformer in leaving the secondary open. I use a 100W lightbulb to limit power for initial testing, it lit up dimly then slowly faded off. Scoping the secondary with a 10x probe reveals 340Vp-p or 170v peak. Indeed this is running at normal line, open-loop, at no load and unrectified, but isn't it a bit too much for a 70v nominal winding? The highest voltage power rectifier i have around is a MBR20100, i don't think that'll work here...
The second Q is about the control loop. I keep reading about error amplifier compensation, but all i have in my circuit is a comparator! And it works just fine without any kind of compensation. I guess compensation is for the error amp that would do the overcurrent protection. Well, if i don't want to bother with any of that and just use fuses, i won't need it right? I know what can happen if the voltage control loop fails, and i've blown enough switches to realize what happens in case of short circuit. But i think i can avoid that situation in my amp design. Because if the amp shorts, it's probably going to cost more than $1.6 to repair so replacing the MJE13007s along doesn't bother me.
That's it for now, here's my schematic.
A few notices are in order: Firstly - the inductance values in the sim are purely hypothetical, secondly, the comparator is a LM311, thirdly, i have 3 amp primary diodes in the real thing, and fourthly i am using MJE13007s not 13009s. Oh and i also forgot the little MOSFET that is between 555 and drive transformer. Other than that all the values are the same in the real circuit.
I've read a lot on SMPS design in the past 2 years, and this year i also acquired all tools needed to go from schematic/simulation to the real thing. Namely an oscilloscope, and more recently a SMD rework station. All i lack is an isolation transformer but i preferred to get the hot air station instead and i'm very glad i did that as my new controller board is all SMDs, mostly recovered from motherboards, and it's DANG SMALL!
Anyway any nasties will reflect into the secondary so i'm not worried by that.I chose a rather unusual approach to a controller, because it just stuck into my brain for some reason. That is, i use a 555 timer. No no no, don't shout at me already. Instead of using the imprecise and inefficient way of using pin 5 for voltage regulation (one schematic on the internet claimed up to 150W output, yeah right... i built it myself and it delivered 15 watts at most with the switching transistor getting hot even without a load), an idea struck me at some point.
Long story short, it's been simulated, built as a low power 5v -> 12v converter that scored in at 89% efficiency at 3.5W out, and my latest adaptation of this controller thing is an off-line 2-transistor forward converter that i intend to power my subwoofer amp.
Now you'll be like... What? An offline supply controlled by a 555??? Well yes. My clever idea was as follows. I used the schematic that allows duty cycle control independent to frequency, so i can set maximum duty cycle with a trimpot (now this has become fixed as i made the move to a complete SMD board). Then, i used a LM311 comparator to toggle pin 4 (reset) and this is where the magic comes in. Basically what i'm getting is pulse skipping. Not the cleanest method of regulation, but it works well and it's quite efficient. A TL431 serves as reference for the comparator and a resistor divider sets the output voltage. Then the output of the 555 drives a small mosfet that switches the 15v rail via a 1uF cap into a base drive transformer recovered from an ATX power supply. Then the transformer drives a pair of MJE13007s. The controller board is powered from a separate power supply (i had a tiny mains transformer around anyway), which is regulated with a 7815. Thus no need for an opto for feedback, i simply wire the output voltage straight into the divider.
I have gotten to the point where i figured out the snubbers i need to keep my transistors from blowing. A couple weeks ago i managed to blow ALL the switching transistors i had lying around the house until i figured out the snubber, and eventually bought some new MJE13007s. They're just 80 cents a pop anyway.
I will be trying to do some non-dissipative snubbers next as those resistors get quite hot, but for now i'm happy that i'm not blowing transistors anymore.I have tested the power supply up to 80W using a transformer wound in a really ugly way.
Today i started winding the "proper" transformer. I use an EI-33 core (have a bunch of those around, i don't think i have to tell you where i got them from), and i wound the primary with 5 strands of 28AWG in parallel, the secondary with 4 strands. I have 48 turns in my primary. I used the formula 7.75A/mm^2 to determine the number of strands. I'm looking at +/-70v out, 400W peak power (my subwoofer is 8 ohm 250W). I'm not expecting 400W continuous as this is audio where the 1/8 power rule stands true, and this will be an amp for home use and i live in an apartment. Also in its new box (the underside of my desk actually, more on that later), the sub is quite loud even with my 2*30W amp that i am using for my main speakers.
Switching frequency is 100kHz or somewhere close to that. The base drive waveform isn't exactly square, but the waveform on the secondary of the transformer is nice and fat. The influence of the snubbers is clearly seen, there is a slope on the falling edge and a bit of ringing on the way, but no spikes at all.
Okay here comes my first Q. My calculations (for worst case low line + full power conditions) ended up with 30 turns for each secondary. I only did one secondary today as my hands hurt already, and i plugged the transformer in leaving the secondary open. I use a 100W lightbulb to limit power for initial testing, it lit up dimly then slowly faded off. Scoping the secondary with a 10x probe reveals 340Vp-p or 170v peak. Indeed this is running at normal line, open-loop, at no load and unrectified, but isn't it a bit too much for a 70v nominal winding? The highest voltage power rectifier i have around is a MBR20100, i don't think that'll work here...
The second Q is about the control loop. I keep reading about error amplifier compensation, but all i have in my circuit is a comparator! And it works just fine without any kind of compensation. I guess compensation is for the error amp that would do the overcurrent protection. Well, if i don't want to bother with any of that and just use fuses, i won't need it right? I know what can happen if the voltage control loop fails, and i've blown enough switches to realize what happens in case of short circuit. But i think i can avoid that situation in my amp design. Because if the amp shorts, it's probably going to cost more than $1.6 to repair so replacing the MJE13007s along doesn't bother me.
That's it for now, here's my schematic.
A few notices are in order: Firstly - the inductance values in the sim are purely hypothetical, secondly, the comparator is a LM311, thirdly, i have 3 amp primary diodes in the real thing, and fourthly i am using MJE13007s not 13009s. Oh and i also forgot the little MOSFET that is between 555 and drive transformer. Other than that all the values are the same in the real circuit.
Thus what my mosfet was actually doing was shorting the 15v supply to ground... DUH!
Along with 2 other classmates of mine from highschool, will be fun especially with those guys around.
Comment