A 65W Flyback SMPS - diyAudio
Go Back   Home > Forums > Amplifiers > Power Supplies

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 26th January 2010, 04:51 AM   #1
diyAudio Member
 
Join Date: Aug 2008
Location: Seattle, WA
Default A 65W Flyback SMPS

My First.
I would appreciate any suggestions/improvements. Please keep it a little on the simple side as I am still on the learning curve.

I have tested it to half power (couldn't go higher for lack of heat sinks). It is holding voltage but I have spikes of 500mV or more in the output. I would appreciate any suggestions you might have as to how to minimize them.

Here is the schematic:
http://www.digispeaker.com/files/p7power-c.pdf

It is built on a breadboard which is probably a problem:
http://www.digispeaker.com/files/psu_c_large.JPG

Here are the un-snubbed and snubbed Vds (gate pulse is in blue):
http://www.digispeaker.com/files/Vds_nosnub_large.JPG
http://www.digispeaker.com/files/vds_snub_large.JPG

Here are the un-snubbed and snubbed Ids (as measured across the sense resistor) (gate is blue trace):
http://www.digispeaker.com/files/Ids_nosnub_large.JPG
http://www.digispeaker.com/files/ids_snub_large.JPG

Here is what the output (blue trace) looks like (without the output inductor):
http://www.digispeaker.com/files/out_noL_large.JPG

And finally, here is what the output (blue trace) looks like with an inductor in the output:
http://www.digispeaker.com/files/out_L_large.JPG
  Reply With Quote
Old 27th January 2010, 06:55 PM   #2
diyAudio Member
 
Sch3mat1c's Avatar
 
Join Date: Jan 2003
Location: Milwaukee, WI
Send a message via ICQ to Sch3mat1c Send a message via AIM to Sch3mat1c
Schematic comments:

C1 and C5 are supposed to be placed to the right of the common mode choke. You don't need them at the line, since if you're trying to shunt noise at the line, you've...basically already failed.

Why the zener regulator starter? A resistor is acceptable. The UC384x series contain an internal zener for protection (30V, IIRC). If you insist, you could add a 1N4746 to clamp it lower.

50kHz seems low. I made a 30W flyback (just a few transistors, but same operating principle) which runs around 200kHz, using a higher primary inductance than you specify. Your duty cycle may be fairly low under rated conditions, leading to high ripple current. Waveforms will show if my guess is wrong.

The 500V MOSFET is not enough for the maximum 340VDC input! It should be double that. 800V is typical, and I'd say 600V minimum. If you'll only ever use 120VAC input, 400V is fine.

Your clamp snubber is quite heavy: 1k clamping ~100VDC is 100mA and 10W! Depending on leakage inductance, ~47k would be more typical (I used 56k 1W). The capacitor is also quite large, 0.01uF is more common. The dV/dt snubber seems about right, but I'd like to note that a simple RC snubber is more traditional, maybe around 470pF and 1k 1W. This would dampen the oscillations you see, which may be helpful, maybe not an issue.

Your filter capacitors are extremely small! For 3A, you should have more like five or ten times as much capacitance! The ESR of a 220uF capacitor alone is enough to produce the ripple you're seeing, and that means big heat in your capacitors, let alone the disproportionate ripple you're seeing.

Why does your output filter specify 2.2nH? That's going to do approximately nothing -- your capacitors alone have about ten times more ESL. Did you mean 2.2uH, or better still, 22uH? 22uH should do an excellent job.

Finally, the feedback circuit is quite typical, although I'm not sure how much the 17nF + 845k zero will accomplish, it seems too high an impedance to have a useful effect. Transient response testing will tell.

As for the waveforms:
Unsnubbed Vds shows classic avalanche just above Vds(max). It also lasts fairly long, which suggests you have an awful lot of leakage inductance in your transformer! Same for the snubbed waveform, it's delivering quite a long pulse into your snubber. Try rewinding it with two primary winding sections in series, you'll just about cut that in half.

Ids waveforms look quite typical, but they have a lot of dV/dt trash on them. This is probably due to probe technique. To measure this waveform:
http://webpages.charter.net/dawill/Images/RegBO2.jpg
I used a coax cable directly across the sense resistor. There is no HF+ hash on this waveform. You can see a negative bounce remains, but that's probably a real event when the bipolar transistor turns off (you're seeing negative base current).

The output waveforms contain understandable features: trash from induced RFI, and the actual waveform of interest, which seems to be quite closely the current waveform dropped across the capacitor's ESR (since ESL is small and C is relatively large at this frequency). The turn on dI/dt is very slow (i.e., it's not a straight rising edge!) which again suggests big leakage inductance.

Despite the large area of your layout, it doesn't seem to be a big problem. In general, you should make it as small as possible. I put my entire circuit on one row:
http://webpages.charter.net/dawill/Images/RegBO1.jpg
Your transformer and capacitor choices are the biggest problems -- improve them and you'll get much better performance.

Thanks for the excellent measurements, they were exactly as needed and straight to the point.

Tim
__________________
Seven Transistor Labs
Projects and Resources
  Reply With Quote
Old 27th January 2010, 10:48 PM   #3
diyAudio Member
 
Join Date: Aug 2008
Location: Seattle, WA
Thank you Sch3mat1c for sharing your experience. I greatly appreciate it. It is just me, Marty Brown and Abraham Pressman in my basement and those guys can be a little difficult to follow.

Quote:
C1 and C5 are supposed to be placed to the right of the common mode choke.
My bad. Thanx. One question in this area... am I supposed to tie earth to my output ground? I see this in some designs and not in others. ?? What is a correct treatment for earth gound?

Quote:
50kHz seems low.
Marty had something pretty close to this design so I thought I would start there. One question, you say increase the primary inductance? I thought it should be decreased as frequency is increased. ?? Your guess is correct, however. My pulses are a little short. Right now, for 50% duty cycle, I calculated my inductance as L = (Vinmin * duty) / (Ipeak * freq) = (127 * .5) / (2.8A * 50KHz) = 450uH. Wouldn't increasing the frequency decrease the inductance?

Quote:
The 500V MOSFET is not enough
Initially, I had 4A, 500V switches in there and I blew them all up before I realised what was happening. These 12A, 500V is all I had laying around. My next order will include the parts you suggest.

Quote:
Your clamp snubber is quite heavy
Again, another parts issue. I don't have any/many 1W to 3W resistors laying around here. I happened to have some 1K and 5K 50W dummy loads so that is what I am using until I get a new order placed. Thanks for the practical snubber estimates. I will give them a try.

Quote:
Your filter capacitors are extremely small!
Thanks. I will try increasing them when I get home from my day job. Thanks.

Quote:
Why does your output filter specify 2.2nH
Mistake in my schematic. It is actually 2.2uH. Again, it is just something I had laying around from another project. I will order some bigger ones.

Quote:
not sure how much the 17nF + 845k zero will accomplish
Thanks again. I will try to decrease it when I get home. I was just following along in Marty's book. I thought it seemed a large. I haven't gotten to the point where I am really whacking on the supply so I don't think it really is coming into play yet.

Quote:
awful lot of leakage inductance in your transformer
Well... yep. Turns out, I suck at winding transformers. I measured it at 30uH but it is probably higher. I don't own an inductance bridge, so I just built a little RCL circuit with high Q on my breadboard. I then scoped the voltage across the capacitor and the inductor. I had my scope sum the two traces. When the trace was flat (or as close as I could get), I recorded the frequency and then calculated the inductance from that. This seems to have worked for the primary inductance (around 450uH). Would this method work for something as small as the leakage inductance?

Quote:
coax cable directly across the sense resistor
Hmm... I guess I don't understand what you mean. Do you have a picture of what you are doing with the coaxial cable? To make the Ids measurements, I just connected my differential probe (probemaster 4232) across the sense resistor and then did the math. Do you own a current probe? They seem expensive. I own a Rigol DS1102E. Do you have a recommendation for a current probe for this scope?

One final question: Do you use Spice in your power supply design? I have generally avoided it in the past, but I see that it is easier to use these days. I guess I am asking if it is worth the time to set it all up. If I could get to a decent simulation, then I could try some of these experiments without having to wait for parts to come.

Thanks again Sch3mat1c.
  Reply With Quote
Old 28th January 2010, 06:49 AM   #4
diyAudio Member
 
Sch3mat1c's Avatar
 
Join Date: Jan 2003
Location: Milwaukee, WI
Send a message via ICQ to Sch3mat1c Send a message via AIM to Sch3mat1c
Quote:
Originally Posted by tylerjbrooks View Post
Thank you Sch3mat1c for sharing your experience. I greatly appreciate it. It is just me, Marty Brown and Abraham Pressman in my basement and those guys can be a little difficult to follow.
You're welcome. I'm currently entering the plateau of SMPS knowledge and therefore eager to demonstrate my knowledge (or have it corrected if it turns out false..), so come and get it.

Quote:
One question in this area... am I supposed to tie earth to my output ground? I see this in some designs and not in others. ?? What is a correct treatment for earth gound?
If you need an isolated output, use a Y-type capacitor between primary and secondary grounds. This shunts high frequency trash which is coupled by primary-to-secondary capacitance and switching dV/dt, giving it a path to return by (instead of being conducted through the input and output leads). If you don't need an isolated output, then earth it, shorting out that EMC loop.

Quote:
Marty had something pretty close to this design so I thought I would start there. One question, you say increase the primary inductance? I thought it should be decreased as frequency is increased. ?? Your guess is correct, however. My pulses are a little short. Right now, for 50% duty cycle, I calculated my inductance as L = (Vinmin * duty) / (Ipeak * freq) = (127 * .5) / (2.8A * 50KHz) = 450uH. Wouldn't increasing the frequency decrease the inductance?
For constant power, inductance is inversely proportional to frequency, yes. The proportionality is how much power you need (the factors of Vmin, Ipeak and typical duty cycle), which according to your waveform, you've overestimated somewhat -- at least under this load condition. If you are well under load, this may well be the desired waveform.

The collector waveform I got, under a medium load condition, is this:
http://webpages.charter.net/dawill/Images/RegBO3.jpg
Notice that particular transformer also had way too much leakage inductance. I fixed that by splitting the primary into two halves, with the secondary wound inbetween. Now the waveforms are beautiful. (Capacitance is higher, so the ringing has a higher Q requiring a heavier snubber to dampen, and EMC issues are potentially increased.)

Now, at this load condition, you can see after the flyback pulse, voltage kind of slows down, then it turns on again. When very lightly loaded, that just peters out to flatness for a while, before kicking on again (this design is effectively variable off-time).

Quote:
Well... yep. Turns out, I suck at winding transformers. I measured it at 30uH but it is probably higher. I don't own an inductance bridge, so I just built a little RCL circuit with high Q on my breadboard. I then scoped the voltage across the capacitor and the inductor. I had my scope sum the two traces. When the trace was flat (or as close as I could get), I recorded the frequency and then calculated the inductance from that. This seems to have worked for the primary inductance (around 450uH). Would this method work for something as small as the leakage inductance?
Gee, I would simply short the secondary (or the biggest secondaries) and measure the primary like it's an inductor, parallel resonant, and measure resonant frequency. L = 1 / (4*pi^2*F^2*C).

Quote:
Hmm... I guess I don't understand what you mean. Do you have a picture of what you are doing with the coaxial cable? To make the Ids measurements, I just connected my differential probe (probemaster 4232) across the sense resistor and then did the math. Do you own a current probe? They seem expensive. I own a Rigol DS1102E. Do you have a recommendation for a current probe for this scope?
For this measurement, I terminated one end of a BNC cable, ran the other to a spare BNC-on-leads, and ran the leads (twisted pair style) down to the breadboard.
http://webpages.charter.net/dawill/Images/RegBO1.jpg
See the red/black twisted pair coming up the middle, and connecting across the tan resistor which is in series with the power transistor emitter.

In your case, EMI might be affecting the probe itself. I'm too cheap for dif probes, so I just use wires. A few ferrite beads will make a fine differential measurement at HF.

Quote:
One final question: Do you use Spice in your power supply design? I have generally avoided it in the past, but I see that it is easier to use these days. I guess I am asking if it is worth the time to set it all up. If I could get to a decent simulation, then I could try some of these experiments without having to wait for parts to come.
The key is setting up a proper model of the physical circuit, and making sure it simulates it correctly. This circuit has so far eluded my simulation ability:
http://webpages.charter.net/dawill/Images/RegBO.gif
The simulation breaks somewhere with "timestep too small" -- it's trying to adjust parameters on the fly to prevent divide by zero for instance, and whatever it's trying, it's not succeeding.

Simpler circuits tend to work better. This circuit:
Click the image to open in full size.
simulates exactly as I designed it.

A circuit like your own should simulate fine. Be sure to include factors like leakage inductance (usually entered through coupling factor), parasitic capacitance, series and parallel resistance, ESR and ESL.

Proper intuition can cover all the bases fairly reliably. Bob Widlar was notorious for entering meetings with breadboard in hand, exclaiming "according to the simulator, this circuit can not work!" For mortals working with simple power supplies, it's easy to take stock of the first-order "not-really-there-but-are" parts, like ESR, ESL, DCR, parasitic C and so on. By approximating waveforms with squares and triangles, you can simply use the fundamental definitions of these components: I = C * dV/dt and V = L * dI/dt.

EMC depends on second and higher order squirrelies, like the mere distribution of turns on your transformer, and PCB layout, trace length and spacing, etc. It should be no surprise that EMC is difficult to track in person, and even more difficult to simulate!

Tim
__________________
Seven Transistor Labs
Projects and Resources
  Reply With Quote
Old 28th January 2010, 07:50 AM   #5
diyAudio Member
 
Join Date: Aug 2008
Location: Seattle, WA
I increased the inductance by taking a little of my gap away.
I increased the resistance in my clamp to 15K (biggest dummy load I had).
I bumped the output capacitance to ~2800uF.
I replaced the output inductor with 22uH.

I flipped the switch... slowly ramped up my AC supply to 110VAC... and I had a beautiful Vds!

I was just about to hit the capture button on my scope when my AC supply pegged and my fuse blew.

A few hours later and I am still not sure what happened. I have replaced/tested many parts. It feels like my transformer has melted... but it seems to checkout ok (inductance looks right. voltage ratios between windings look right...).

Well... I have fried something and it isn't obvious. At this point, I have things torn up enough that I will probably take your suggestion and do a new layout.
  Reply With Quote
Old 29th January 2010, 01:41 AM   #6
diyAudio Member
 
Sch3mat1c's Avatar
 
Join Date: Jan 2003
Location: Milwaukee, WI
Send a message via ICQ to Sch3mat1c Send a message via AIM to Sch3mat1c
You should be monitoring drain voltage and source current at all times. Those tell you everything about the most critical component, the power transistor. If current is ever higher than design maximum, something is wrong. If the current waveform is rising, then rising faster before turning off, that means your transformer is saturating, and something is wrong. If your drain waveform isn't coming down below 10V while on, something is wrong (mind that measurement methods matter, because the AC component is large -- you can measure Vds(sat) more accurately with a resistor and zener clamp, allowing a lower voltage setting on the scope).

If it was avalanching again, it could've exploded fairly well. The transistor is the only big thing that's going across the supply rails, it's a good bet it's dead. Transistors tend to fail shorted to all three, so your current sense resistor and transformer primary probably got the full brunt of the supply caps discharging (plus whatever came through the fuse in the next couple of miliseconds). The 3844 might be smoke too

Tim
__________________
Seven Transistor Labs
Projects and Resources
  Reply With Quote
Old 21st February 2010, 12:38 AM   #7
diyAudio Member
 
Join Date: Aug 2008
Location: Seattle, WA
Default Sucess

Hey Sch3mat1c.
I just wanted you to know that I have been experimenting with this design and had some success with it. The long delay in a response was due to many experiments and some real-life issues getting in the way.

I ordered more parts and spent a lot of time winding transformers. Turns out, how you wind these things makes a big difference. As you suggested, splitting the primary winding between the inner and outer layers reduced the leakage inductance. Putting the windings in series helps a lot. Doubling the winding and connecting them in parallel helps even more (but uses a lot of wire). Also, Marty Brown's book (section 3.5.9) suggested a progressive style of winding as opposed to a straight style. This also reduced the leakage inductance.

I increased the inductance to more than 1mH (I started at 450uH) and that helped me get the spikes under control (so the controller wouldn't 'rattle' - false triggers). Increasing the inductance and lowering the leakage inductance, as you suggested, made a world of difference. After that, I could use some more traditional values in my snubbers and they would burn a lot cooler.

Thanx again for your help. Here are a few pictures of the result.
New Layout:
http://www.digispeaker.com/files/psu_d_large.JPG
Vds and Ids:
http://www.digispeaker.com/files/vds_ids_large.JPG
Vds and +5V Output:
http://www.digispeaker.com/files/vds_output_large.JPG
  Reply With Quote
Old 23rd July 2010, 06:05 PM   #8
diyAudio Member
 
Join Date: Jan 2010
hi tylerjbrooks,

thanks for sharing. As i am studying on how a smps works and then i found your post. it's interesting.

I would like to ask u that how u measure the waveform on the primary side of your smps? Do u use any isolator-transformer or you're using the add function of the scope?

I have a DIY smps which i bought it. I decided to start to have a real actual seeing on how actually a smps works.Since i have no isolator transformer i decided to use the add function of my scope to have a look on the real waveform. I can see some signals on my scope but i'm not sure is it the correct waveform i'm looking for, for the switching part.

Do u try to measure the waveform on the primary side?How u do it by the way or maybe can u share some of the waveforms?

thank you.
  Reply With Quote
Old 25th July 2010, 04:38 AM   #9
diyAudio Member
 
Sch3mat1c's Avatar
 
Join Date: Jan 2003
Location: Milwaukee, WI
Send a message via ICQ to Sch3mat1c Send a message via AIM to Sch3mat1c
I use a 120V isolation transformer when working on offline circuits. Not only is this a safety item, it allows you to ground the probe to circuit common. An ordinary 10x probe is sufficient to measure the primary waveform.

Tyler: looks good. Still a lot of crunch on turn-on, which could be reduced with a dI/dt snubber (maybe a ferrite bead on the FET drain). Mostly, it's probably just measuring error; to really get a good signal, you practically need a coax cable soldered directly to the shunt resistor. And you have to make sure the shunt is noninductive, keeping in mind that an inch of wire is already 10nH or so! Measuring after the current-sense R+C helps, too. And good old fashioned ground loop, which can be a big problem on breadboards.

Tim
__________________
Seven Transistor Labs
Projects and Resources
  Reply With Quote
Old 25th July 2010, 05:20 PM   #10
diyAudio Member
 
Join Date: Aug 2008
Location: Seattle, WA
motbuddy...
I use a 2A AC power supply when I work on these things. I have an old BK Percision 1653 that lets me limit the current. Much safer. Also, I use a differential probe from (ProbeMaster that lets me probe around without worrying about the ground.

Do you have access to Spice? I started using it heavily right after I finished this experiment. It is a pretty useful learning tool. It allows you to rapidly try out different topologies safely and quickly. I suggest you give it a try.

Sch3mat1c...
Thank you for your help. I learned a lot with this experiment. I was able to get a pretty reasonable response from my breadboard. I never did make a PCB for this design. As noted above, right after this experiment, I started using Spice to model my analog stuff. This turned out to be a good move as it let me try different things quickly and safely. However, I realize there is nothing like actually doing it on the bench. Just winding transformers is an art!

I am currently working on a 'sinewave dimmer'. These are a kind of light dimmer that work like an autotransformer. I have uploaded a schematic that includes a sinewave dimmer topology (along with triac and igbt based designs) if you are interested. I would appreciate any thoughts you might have... but I realize this is way off topic for this board.
Attached Files
File Type: pdf dimmer_three.pdf (29.8 KB, 191 views)
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
self oscillating flyback smps blackmamba Power Supplies 15 19th October 2011 07:51 AM
LM2575 for flyback? lumanauw Power Supplies 6 9th October 2010 07:02 PM
110W Flyback chimi Power Supplies 6 27th October 2008 12:38 PM
SMPS, flyback, or vibrator - how to get 280v off of some batteries? Spasticteapot Power Supplies 3 14th July 2007 02:16 AM
How does a VAC Renaissance 70/70 get 65W from 4 300Bs? Saurav Tubes / Valves 20 24th September 2003 12:51 AM


New To Site? Need Help?

All times are GMT. The time now is 02:04 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2