Steve, I've enjoyed this series and its given me the impetus to solder up the 3 phase rectifier I designed a year or so ago (in an attempt to improve the efficiency of my wind generator) also with the LT 4320. I 'hit the wall' after the boards were delivered when I realised that I had no way of bench testing the board except one phase at a time! Duh!
I must say I am fascinated by that MOSFET-Diode-Bridge 'MDB' rectifier concept - nice work there Steve! And thanks for replying to my LCR meter query earlier. Regards.
I think you are on the right way about switched-mode experiment, starting with the core full bridge, get it working, adding active rectifying and so on...
Further to your comments (10:22 onwards) about using schottky diodes in the plain bridge, I would note that the situations where diode losses contribute more to inefficiency are those where the voltage is low - so schottky diodes ARE a viable option. Yes, at 240 VAC schottky isn't an option but 2 Volts of loss is OK because the loss is less than 1%. At 20 VAC that loss is 10%, so changing the schottky, maybe with a drop of 0.8 Volts, brings the losses down to 4%, a worthwhile saving. So, as always, the fine details of a design are important.
Voltage drop and efficiency may not be of equivalent importance depending on the design, despite them equating to the same thing. A mains powered application may not care about the voltage drop, but you might be concerned with the heat dissipation. You're right though, lower voltage applications may care about the voltage drop more.
Excellent video going over a topic not many people give much thought to. If you're planning to build your own SMPS an interesting video topic I'd love to see you do first would be diode selection in DC-DC switching converters. Though of course I expect you'd be wanting to use a bootstrapped NMOS to reduce losses at higher currents in your own design. The balancing act between reverse recovery charge, leakage current, voltage drop (and the rise in Vd with increasing forward currents) is actually pretty interesting once you start trying to optimize a design. Two diodes with similar looking characteristics on a datasheet can sometimes vary massively when it comes to the actual measured efficiency of a converter.
I think many who investigate or study electronics have the experience of setting up circuits using a generic bench DC power supply, which is fine, but instills the idea that a supply is designed and then a load is connected, and as such we test using a resistive load which is very simplistic in how power may be used. The primary task in delivering power is in understanding the characteristics of the load and then design the supply and transmission (wiring) to match, which falls under the relatively recent area of design concentration call power integrity (PI). So it is important to understand the dynamic characteristics of the load and how to design for its needs. If you want to understand power supply efficiencies in relation to design requirements just investigate the evolution of Personal Computer switching power supplies. The power/efficiency price point largely drives designs and produces common design trends in approach and in devices used like control ICs. One obvious trend for high efficiency is in addressing diode rectification loss, as you have illustrated in the video.
Thanks for the follow up Steve! The capacitor on the output of the LT4320 made a pretty sizeable difference, i think it may use the output voltage to drive the gate charge pumps and without the capacitor it couldn't do that at low ac voltages. Also thanks for the wide range of testing, i could see this really being useful in things like audio amplifiers etc where you have many supplies that need rectifying.
You could try a MUR640 in place of the 1N400x diodes, which is an ultra fast 4A, 600V diode, although the forward voltage can exceed 1V at 4A. But it works very well in SMPS applications.
My channel is mostly audio. For use in power amplifiers of say 50+watts at present I tend to use Schottky diodes and a conventional capacitor smoothing which does work well. After watching these videos, I'm wondering if there is an alternative that would make sense without costing a fortune. Power tends to be the issue.
Hi Michael, I think some of the folks on DIYaudio have been experimenting with synchronous bridge rectifiers. Schottky diodes are pretty good if they have the ratings you need, but I do like the 2x Schottky and 2x FET design. I might try this out with discrete components to see how it behaves.
Sorry to ask outside the ep topic, why your videos are so clear without glare or reflections, what sort of lighting or camera setting you using, if you don't mind
Nothing too special, I have an array of overhead lights and a Canon HF G series camera. I always avoid certain angles on shiny things to reduce reflections as far as possible though.
I would assume the LT4320 charge pump is operating all time, suplied from rectified side capacitor. Limiting factor for AC frequency is the chip's turn on/off strength for FET gate charges.
I just have to ask: In all of your videos, there is a constant high frequency ringing in the background. I remember you had a video where you identified a ringing SMD capacitor, so I am surprised nobody ever noticed this sound in all of your videos. When I pause it, it clearly goes away, so it's not my ears. Not sure if this is your camera, your lights, your PC or what not. This constant sound makes it hard to listen to your videos with headphones on :-/ But I want to watch more, cause I am learning a lot and your videos are otherwise really great!
I suspect it's my server rack which has a lot of fans running. I add an audio filter, but there's probably still some noise creeping in so I'll see if I can tweak it.
Nice video. The efficiency measurement would be a little more accurate if you used a larger capacitor to minimize ripple. As it was, the energy efficiency of an ideal capacitor when being charged from a voltage source is given by 1/2 * (1 + Vmin / Vmax), which gives a significantly greater loss for the lower performance solutions.
To implement a design like the LT4320, you'd still need the charge pump. You can do it without any controller with 2x P and 2x N channel MOSFETs. I might make another test PCB to try out some discrete solutions like this.
That one annoying via... but if you look long enough, there's an alternative layout that would let you make that board single-sided. It requires running a track around the outside-facing edge of the power in connector.
Can you elaborate on the probe issue a bit more? Is it because the AC ground/neutral would be connected via earth ground to DC ground and thus the output would appear sinusoidal? I guess I trying to understand if this could have gone bang or not?
The best way to explain it would be to draw out the schematic. Since the ground lead on all of your scope probes are connected together, if you probe the AC and the DC side at the same time, there is a shorted diode. The scope and diodes probably won't go bang, but your electronics on the DC side of the bridge probably will.
It is a pity that the NMLU1210 is obsoleted, as it looked like a very tidy PCB space saving option for small retro transformer projects. Anyone found a replacement single chip alternative?
Using the EZ81 tube did the job better than any other diode I could find in one of my applications since the diodes would break over 100ma current with the factory design of that unit no matter if the diode used was rated 3 amps and 1000 volts it would break immediately, I also made a 3 fast recovery diodes chain with parallell resistors that worked even if these diodes was rated 200V each and 1 amp they worked better than one single diode. I think what happens is the peak voltage times the frequency is too much for the single diode hence a weak design of that power supply but the tube again shine over the silicone ;D
If you use slow (non ultra fast or shottky diodes like 1n4007) with high frequency (SMPS) or low voltage diodes (>1000V peak reverse voltage for 1n4007) that is expected result (diode's overheating and failing). Genuine 1n4007 (either 2pcs for output voltage
@@volodumurkalunyak4651 in my case there was no diode that could handle the task. No matter if it was a high voltage ultra fast diode or any other kind you can pick. They all fail without hesitation. As I described the EZ81 and a specially crafted chain diode was the only thing that could do the task without failing. There is no transformer. And no bridge rectifier. It is a pulse step-up voltage booster which have a working voltage of 390 volts DC with up to about 150ma current. The EZ81 is superior for this task and there won’t be any diode that gonna do this better because I did not have any failure after I put this in after the factory diodes broke or any other one for that matter. The problem is that no Schottky diode will handle that voltage anyway. And 4007 is too slow for the frequency. Hence the only working diode configuration was a chain with 3x200volt 20nano second recovery diodes with resistor balancing was the only working solution except the tube. So no a diode is not best for this task. The EZ81 is superior in this case. Plus the fact when a diode reach a failing point it does die permanently at this point in a matter of nano seconds. The tube on the other hand will generally be able to recover from temporary overload/over voltage conditions and still live. So generally diodes are good for most stuff today. But there are still some things only a tube can properly handle. Plus the tubes are cooler to look at
@@Pulverrostmannen 390V is a input voltage, output voltage? What is topology, ordinal boost converter? There are high frequency high voltage silicon-carbide diodes specifically made towards PFC application (ultra fast, 600V+). Cree c5d0517H will be overkill, yet still cheaper than EZ81. There can be anouther reason why diodes fail: output capacitor placement and imperdance.
@@volodumurkalunyak4651 390 volts are the output voltage. It is a boost converter type of circuit which is factory made. The factory diode (a US3M Ultra fast glass passivated 1KV 3Amp 75ns SMD diode) this fails at about 200volt with 100ma output power. With my EZ81 it been tested to 390 volts and 200ma as well as with my special designed 600 volt chain diode as well and then it will not break. However this test have been performed with much MUCH beefier diodes as well with same voltage rating and fast recovery but they fail anyway and it is expensive to test all kinds anyway but i shot all types I had on hand. I could still use the chain diode I made myself and it will not fail as all other diodes do. ( STTH102RL 200V 1A 20NS ) which is a switching and flyback suited diode. 3 such diodes in series with 3 parallel high value resistors and it also works as a replacement for the stock diode. However I like the tube because it’s used with a tube amp anyway so it fits with the rest of the stuff. But I have made some measures and I am pretty sure the boost converter is rather poorly made when it comes to the design but I have not taken the transformer apart to see if it has a flyback design or not. But it has a primary and secondary winding separated from each other with no auxiliary winding but I am pretty sure it will have an air gap using a flyback design because the high frequency and that it can easily generate more than 1KV if the feedback is disabled. My guess is that the peak reverse voltage is way more than the output voltage and the measure insanely high frequency ringing on the secondary is enough to put any given diode into reverse breakdown or excessive heat due to reverse currents during reverse polarity. And also since it is a secondary winding pulse mode design voltage pump it also lack a snubber network making the ringing and peak voltage even worse. So yeah. The tube can handle this without even breaking a sweat. And only my other 600volt special diode can as well. Everything else will break. EVEN a RFUH30 twin 600V 15amps 30NS put in parallel for 30amps. Not even seconds. So what explains my other 600 volt diode can handle it? Maybe headroom. Maybe the resistors in parallel or maybe even the lower recovery time of 20NS. I did not go further than getting to a working solution with my EZ81 which have been running for way beyond 1000 hours now still going strong and I even put in the most used and blackened tube I could find as a test. It just works
@@Pulverrostmannen those all diodes fail from excesive reverse voltage. You have a flyback topology converter (gapped iron core, transformer with only 2 windings, no extra magnetics), that requires output diode rated at least Vpri*(Nsec/Npri) + Vsec. Vpri*turns ratio is a reflected voltage on a secondsry coil. Output diode shouldn't fail from high frequency ringing as it closes before ringing starts. Primary switch breakdown voltage requirement is following Vpri + Vsec*(Npri/Nsec)+Vovershoot. Converter might use a forward topology, that requires snubbers at primary side and output voltage is directly proportional to input or 2 output diodes (from each side of output transformer coil to input of output inductor) are required. If output diode is soldered wrong, that turns unregulated forward converter into flyback or wise versa.
Replacing lt4320 with four mosfets is not so trivial because mosfets, when open, conduct current in both ways. Usually people use a opamp/comparator accross source and drain (with some hysteris) to make it conducting only one way. Such devices called "ideal diode controller" or something. There are many solutions on the market. It's relatively easy to make one with opamp, but watchout opamp's offset voltage: www.eeweb.com/use-a-self-powered-op-amp-to-create-low-leakage-rectifier/ . As on on-semi part, I've designed a similar device using discrete parts, you can see it here: www.eevblog.com/forum/projects/semi-active-bridge-rectifier/ . Why I didn't use lt4320? It wasn't easily available to me, not very cheap and doesn't work below 9V. There are alternative ICs with similar function, but there are not many of them. One of alternatives is TEA2208T , but I haven't tried it yet.
Really excited for the coming switched-mode AC-DC power supply tutorial(s)!
Me too!!
Likewise!
Hell yeah for the AC-DC project
Steve, I've enjoyed this series and its given me the impetus to solder up the 3 phase rectifier I designed a year or so ago (in an attempt to improve the efficiency of my wind generator) also with the LT 4320. I 'hit the wall' after the boards were delivered when I realised that I had no way of bench testing the board except one phase at a time! Duh!
I really like the format and content of your videos. +1 for designing PCBs to compare designs/flux/whatever.
I must say I am fascinated by that MOSFET-Diode-Bridge 'MDB' rectifier concept - nice work there Steve! And thanks for replying to my LCR meter query earlier.
Regards.
Looking forward to seeing the classic SDG take on those switched-mode power supplies!!
I think you are on the right way about switched-mode experiment, starting with the core full bridge, get it working, adding active rectifying and so on...
Thanks for sharing Steve, I found it really interesting.
Great video again. Thanks.
If you have time do AC-DC with active bridge or bridge less totem pole PFC.
FULL BRIDGE REKTIFIER!
Further to your comments (10:22 onwards) about using schottky diodes in the plain bridge, I would note that the situations where diode losses contribute more to inefficiency are those where the voltage is low - so schottky diodes ARE a viable option. Yes, at 240 VAC schottky isn't an option but 2 Volts of loss is OK because the loss is less than 1%. At 20 VAC that loss is 10%, so changing the schottky, maybe with a drop of 0.8 Volts, brings the losses down to 4%, a worthwhile saving. So, as always, the fine details of a design are important.
Voltage drop and efficiency may not be of equivalent importance depending on the design, despite them equating to the same thing. A mains powered application may not care about the voltage drop, but you might be concerned with the heat dissipation. You're right though, lower voltage applications may care about the voltage drop more.
Excellent video going over a topic not many people give much thought to. If you're planning to build your own SMPS an interesting video topic I'd love to see you do first would be diode selection in DC-DC switching converters. Though of course I expect you'd be wanting to use a bootstrapped NMOS to reduce losses at higher currents in your own design.
The balancing act between reverse recovery charge, leakage current, voltage drop (and the rise in Vd with increasing forward currents) is actually pretty interesting once you start trying to optimize a design. Two diodes with similar looking characteristics on a datasheet can sometimes vary massively when it comes to the actual measured efficiency of a converter.
LT4320 is over $10 each!
I would love to see a video on your new converter.
Doing a discrete version of the LT4320 would be fun, adaptable for high voltage or wattage...
Great video again. Would love to see the switch mode power supply video!
I think many who investigate or study electronics have the experience of setting up circuits using a generic bench DC power supply, which is fine, but instills the idea that a supply is designed and then a load is connected, and as such we test using a resistive load which is very simplistic in how power may be used. The primary task in delivering power is in understanding the characteristics of the load and then design the supply and transmission (wiring) to match, which falls under the relatively recent area of design concentration call power integrity (PI). So it is important to understand the dynamic characteristics of the load and how to design for its needs.
If you want to understand power supply efficiencies in relation to design requirements just investigate the evolution of Personal Computer switching power supplies. The power/efficiency price point largely drives designs and produces common design trends in approach and in devices used like control ICs. One obvious trend for high efficiency is in addressing diode rectification loss, as you have illustrated in the video.
Thanks for the follow up Steve!
The capacitor on the output of the LT4320 made a pretty sizeable difference, i think it may use the output voltage to drive the gate charge pumps and without the capacitor it couldn't do that at low ac voltages.
Also thanks for the wide range of testing, i could see this really being useful in things like audio amplifiers etc where you have many supplies that need rectifying.
You could try a MUR640 in place of the 1N400x diodes, which is an ultra fast 4A, 600V diode, although the forward voltage can exceed 1V at 4A. But it works very well in SMPS applications.
My channel is mostly audio. For use in power amplifiers of say 50+watts at present I tend to use Schottky diodes and a conventional capacitor smoothing which does work well. After watching these videos, I'm wondering if there is an alternative that would make sense without costing a fortune. Power tends to be the issue.
Hi Michael, I think some of the folks on DIYaudio have been experimenting with synchronous bridge rectifiers. Schottky diodes are pretty good if they have the ratings you need, but I do like the 2x Schottky and 2x FET design. I might try this out with discrete components to see how it behaves.
Keen as, well done!
Sorry to ask outside the ep topic, why your videos are so clear without glare or reflections, what sort of lighting or camera setting you using, if you don't mind
Nothing too special, I have an array of overhead lights and a Canon HF G series camera. I always avoid certain angles on shiny things to reduce reflections as far as possible though.
I would assume the LT4320 charge pump is operating all time, suplied from rectified side capacitor. Limiting factor for AC frequency is the chip's turn on/off strength for FET gate charges.
I just have to ask: In all of your videos, there is a constant high frequency ringing in the background. I remember you had a video where you identified a ringing SMD capacitor, so I am surprised nobody ever noticed this sound in all of your videos. When I pause it, it clearly goes away, so it's not my ears. Not sure if this is your camera, your lights, your PC or what not. This constant sound makes it hard to listen to your videos with headphones on :-/ But I want to watch more, cause I am learning a lot and your videos are otherwise really great!
I suspect it's my server rack which has a lot of fans running. I add an audio filter, but there's probably still some noise creeping in so I'll see if I can tweak it.
Question on the MOSFET/Bridge Rectifier. Is this design superior to the crystal radio receiver, also called a crystal set?
Nice video. The efficiency measurement would be a little more accurate if you used a larger capacitor to minimize ripple. As it was, the energy efficiency of an ideal capacitor when being charged from a voltage source is given by 1/2 * (1 + Vmin / Vmax), which gives a significantly greater loss for the lower performance solutions.
Wonder if using a cheap microcontroller with a built in H-bridge driver (for the MOSfets) would be a cheap yet very flexible solution?
To implement a design like the LT4320, you'd still need the charge pump. You can do it without any controller with 2x P and 2x N channel MOSFETs. I might make another test PCB to try out some discrete solutions like this.
That one annoying via... but if you look long enough, there's an alternative layout that would let you make that board single-sided. It requires running a track around the outside-facing edge of the power in connector.
interesting test. 👍👍
Can you elaborate on the probe issue a bit more? Is it because the AC ground/neutral would be connected via earth ground to DC ground and thus the output would appear sinusoidal? I guess I trying to understand if this could have gone bang or not?
The best way to explain it would be to draw out the schematic. Since the ground lead on all of your scope probes are connected together, if you probe the AC and the DC side at the same time, there is a shorted diode. The scope and diodes probably won't go bang, but your electronics on the DC side of the bridge probably will.
EEVBlog has a video on this: ua-cam.com/video/xaELqAo4kkQ/v-deo.html
What if you put 10nf ceramic capacitors on each of those diodes?
How about if the standard Si MOSFETs were replaced with gallium nitride MOSFETs?
I'm wanting to make a ac to DC switch mode as well!
Wath will be the next soldering iron ? Olso very Nice video.
It's probably the Ersa I-Con before Hakko
@@sdgelectronics olso you are good teacher .
It is a pity that the NMLU1210 is obsoleted, as it looked like a very tidy PCB space saving option for small retro transformer projects. Anyone found a replacement single chip alternative?
I'm on the hunt for a replacement. I'll also test out using discrete components, but it'll lose out on the size advantage.
Using the EZ81 tube did the job better than any other diode I could find in one of my applications since the diodes would break over 100ma current with the factory design of that unit no matter if the diode used was rated 3 amps and 1000 volts it would break immediately, I also made a 3 fast recovery diodes chain with parallell resistors that worked even if these diodes was rated 200V each and 1 amp they worked better than one single diode. I think what happens is the peak voltage times the frequency is too much for the single diode hence a weak design of that power supply but the tube again shine over the silicone ;D
If you use slow (non ultra fast or shottky diodes like 1n4007) with high frequency (SMPS) or low voltage diodes (>1000V peak reverse voltage for 1n4007) that is expected result (diode's overheating and failing). Genuine 1n4007 (either 2pcs for output voltage
@@volodumurkalunyak4651 in my case there was no diode that could handle the task. No matter if it was a high voltage ultra fast diode or any other kind you can pick. They all fail without hesitation. As I described the EZ81 and a specially crafted chain diode was the only thing that could do the task without failing.
There is no transformer. And no bridge rectifier.
It is a pulse step-up voltage booster which have a working voltage of 390 volts DC with up to about 150ma current.
The EZ81 is superior for this task and there won’t be any diode that gonna do this better because I did not have any failure after I put this in after the factory diodes broke or any other one for that matter.
The problem is that no Schottky diode will handle that voltage anyway. And 4007 is too slow for the frequency. Hence the only working diode configuration was a chain with 3x200volt 20nano second recovery diodes with resistor balancing was the only working solution except the tube.
So no a diode is not best for this task. The EZ81 is superior in this case.
Plus the fact when a diode reach a failing point it does die permanently at this point in a matter of nano seconds.
The tube on the other hand will generally be able to recover from temporary overload/over voltage conditions and still live.
So generally diodes are good for most stuff today. But there are still some things only a tube can properly handle.
Plus the tubes are cooler to look at
@@Pulverrostmannen 390V is a input voltage, output voltage? What is topology, ordinal boost converter? There are high frequency high voltage silicon-carbide diodes specifically made towards PFC application (ultra fast, 600V+).
Cree c5d0517H will be overkill, yet still cheaper than EZ81.
There can be anouther reason why diodes fail: output capacitor placement and imperdance.
@@volodumurkalunyak4651 390 volts are the output voltage. It is a boost converter type of circuit which is factory made. The factory diode (a US3M Ultra fast glass passivated 1KV 3Amp 75ns SMD diode) this fails at about 200volt with 100ma output power. With my EZ81 it been tested to 390 volts and 200ma as well as with my special designed 600 volt chain diode as well and then it will not break. However this test have been performed with much MUCH beefier diodes as well with same voltage rating and fast recovery but they fail anyway and it is expensive to test all kinds anyway but i shot all types I had on hand.
I could still use the chain diode I made myself and it will not fail as all other diodes do.
( STTH102RL 200V 1A 20NS ) which is a switching and flyback suited diode.
3 such diodes in series with 3 parallel high value resistors and it also works as a replacement for the stock diode.
However I like the tube because it’s used with a tube amp anyway so it fits with the rest of the stuff.
But I have made some measures and I am pretty sure the boost converter is rather poorly made when it comes to the design but I have not taken the transformer apart to see if it has a flyback design or not.
But it has a primary and secondary winding separated from each other with no auxiliary winding but I am pretty sure it will have an air gap using a flyback design because the high frequency and that it can easily generate more than 1KV if the feedback is disabled.
My guess is that the peak reverse voltage is way more than the output voltage and the measure insanely high frequency ringing on the secondary is enough to put any given diode into reverse breakdown or excessive heat due to reverse currents during reverse polarity.
And also since it is a secondary winding pulse mode design voltage pump it also lack a snubber network making the ringing and peak voltage even worse.
So yeah. The tube can handle this without even breaking a sweat. And only my other 600volt special diode can as well. Everything else will break. EVEN a RFUH30 twin 600V 15amps 30NS put in parallel for 30amps. Not even seconds.
So what explains my other 600 volt diode can handle it? Maybe headroom. Maybe the resistors in parallel or maybe even the lower recovery time of 20NS.
I did not go further than getting to a working solution with my EZ81 which have been running for way beyond 1000 hours now still going strong and I even put in the most used and blackened tube I could find as a test. It just works
@@Pulverrostmannen those all diodes fail from excesive reverse voltage. You have a flyback topology converter (gapped iron core, transformer with only 2 windings, no extra magnetics), that requires output diode rated at least Vpri*(Nsec/Npri) + Vsec. Vpri*turns ratio is a reflected voltage on a secondsry coil. Output diode shouldn't fail from high frequency ringing as it closes before ringing starts. Primary switch breakdown voltage requirement is following Vpri + Vsec*(Npri/Nsec)+Vovershoot.
Converter might use a forward topology, that requires snubbers at primary side and output voltage is directly proportional to input or 2 output diodes (from each side of output transformer coil to input of output inductor) are required. If output diode is soldered wrong, that turns unregulated forward converter into flyback or wise versa.
Just a sugestion can be a Hakko product .
GSM1 Can handle 200ma max normal temperature
Replacing lt4320 with four mosfets is not so trivial because mosfets, when open, conduct current in both ways. Usually people use a opamp/comparator accross source and drain (with some hysteris) to make it conducting only one way. Such devices called "ideal diode controller" or something. There are many solutions on the market. It's relatively easy to make one with opamp, but watchout opamp's offset voltage: www.eeweb.com/use-a-self-powered-op-amp-to-create-low-leakage-rectifier/ .
As on on-semi part, I've designed a similar device using discrete parts, you can see it here: www.eevblog.com/forum/projects/semi-active-bridge-rectifier/ . Why I didn't use lt4320? It wasn't easily available to me, not very cheap and doesn't work below 9V. There are alternative ICs with similar function, but there are not many of them. One of alternatives is TEA2208T , but I haven't tried it yet.
Very interesting thanks for sharing!