Linear PSU Design DIY
Jan 1, 2014 18:47:02 GMT
Post by pcourtney1 on Jan 1, 2014 18:47:02 GMT
Sandy et al, HAPPY NEW YEAR
my brother has just given me back my Squeezebox Touch (he lost the cheap power supply that came with) , so I have been researching what would be a good DIY replacement and came across this very good article by John Swenson, and wanted to sanity check it first with the forum - if I may
here is what caught my eye
forums.slimdevices.com/showthread.php?82648-Linear-Power-Supplies/page7
home.comcast.net/~johnswenson1/stereo/SB_5V.GIF
I've done a huge amount of PSU testing, design and building in regards to the Squeezebox Touch, and what I'm presenting here is not just conjecture, its been tested many times.
My conclusions are that a large percentage of improvements with PSU design for the Touch are related to high frequency noise sent BACK into the mains and picked up by other components. Power sent into the Touch itself has little impact on what goes on inside the Touch (not zero, but quite small). The differences in mains injected noise is vastly greater than any changes in the power actually delivered to the components inside the Touch.
The focus of good PSU design should be on decreasing what gets sent back down the AC line, not making the absolute lowest noise, or lowest impedance feed to the Touch. Many of the attempts at getting the "best" power to the Touch also increase the noise sent down the line.
For a linear supply there are mainly three things that contribute to noise sent back down the line:
1) reverse recovery noise when the diodes switch
2) transformer ringing and
3) current spikes when the diodes conduct.
Using Schottky diodes can get rid of the reverse recovery noise, transformer ringing can be dealt with with an RC network across the secondary which damps the resonance. This damping of the transformer ringing is very effective but almost never done. If you check out 100 different linear PSU's there will be a very high probability that NONE of them have damped the transformer resonance. Its pathetic that the higher the "quality" of the transformer the worse the resonance is, so actually using one of these expensive "high end" transformers is one of the worst things you can do. (unless you damp it)
The "current spikes" are the hard part, these are caused by using a traditional PSU design of transformer, diodes and a big cap. The diodes only conduct when the voltage from the transformer is greater than the voltage on the cap. Thus the power coming from the transformer is in short high current spikes. For example a common design for the 5V 2A for the Touch will actually have 20A spikes coming from the transformer. These 20A spikes have all kinds of high frequency components which get sent right back through the transformer and into the mains. They also excite the above mentioned transformer resonance causing the transformer to ring like a bell, even if you use Schottky diodes.
The DC signal coming out of such a design is a sawtooth wave. This also has large amounts of high frequency components. Unfortunately most rectifier designs have very little input rejection at high frequencies, they do great at low frequencies, but at high frequencies they are poor. This is one reason that discrete regulators have been used in audiophile designs for so long, they can have much better high frequency input rejection, which is needed to correctly handle the sawtooth.
There is a nice effective solution called a choke. Not the little high frequency things designed to run at 50KHz and up, but big heavy things which have significant inductance at 120Hz. If properly designed they allow continuous conduction through the transformer and rectifier, eliminating the high current spikes. Another advantage is that the output waveform is a pure sine wave, no high frequency harmonics.
Put the choke filter, transformer damping and Schottky diodes together and you have a supply that injects almost nothing back into the AC line and delivers a clean pure sine wave to the rectifier, and with this design you don't NEED a complicated expensive regulator.
I've put together a design using these principles, the schematic is at:
home.comcast.net/~johnswenson1/stereo/SB_5V.GIF
All the parts should be available at Mouser or Digi-Key. I'm sure there are other distributors in other parts of the world where you can get these parts. You should be able to get all the parts for $75 or so. Give it a try I think you will like this design.
Message sent to John Swenson on the slimdevices forum
Hi John, I have few questions:
1. What AC voltage should deliver the transformer and how big should it be ?
2. R1 - 1W is it enough ?
3. Can I use instead of LT1048 one superteddyreg and if yes, is C5 and C6 necessary ? Or should i just forget str and use LT1048 ?
4. When I am building it and measuring, what voltage should it be after choke and after reg ?
Answers:
1, this design uses 10VAC out of the transformer. A little larger won't hurt, it just means the the regulator has to dissipate more power. How big depends on whether you will be running a USB disk drive off it. If not using a bus powered drive 25VA and up is fine. If you want to run a drive off this I would go for 50VA.
2. I'm using a 1/2W carbon composition resistor.
3. Sure you can use other regulators if you want, just make sure you have enough voltage for the regulator drop out. This design was tuned for the dropout of the LT1084, (I don't remember off the top of my head, you can look it up). If the regulator you want to use has a a greater drop out voltage you will have to up the voltage on the transformer. This design is very easy on transformers so you do not need to significantly over rate them.
4. You want the output to be 5V, it doesn't have to be super exact. 5.1 V will be fine, don't go much below. 4.98 is fine, but 4.8 might be too low. Don't go above 5.2 V. As to the output of the choke, that depends on the dropout of the regulator in use. If you are going for 5V out and your regulator needs 2V dropout, the output of the choke should NEVER go below 7V. Give your self a little margin.
my brother has just given me back my Squeezebox Touch (he lost the cheap power supply that came with) , so I have been researching what would be a good DIY replacement and came across this very good article by John Swenson, and wanted to sanity check it first with the forum - if I may
here is what caught my eye
forums.slimdevices.com/showthread.php?82648-Linear-Power-Supplies/page7
home.comcast.net/~johnswenson1/stereo/SB_5V.GIF
I've done a huge amount of PSU testing, design and building in regards to the Squeezebox Touch, and what I'm presenting here is not just conjecture, its been tested many times.
My conclusions are that a large percentage of improvements with PSU design for the Touch are related to high frequency noise sent BACK into the mains and picked up by other components. Power sent into the Touch itself has little impact on what goes on inside the Touch (not zero, but quite small). The differences in mains injected noise is vastly greater than any changes in the power actually delivered to the components inside the Touch.
The focus of good PSU design should be on decreasing what gets sent back down the AC line, not making the absolute lowest noise, or lowest impedance feed to the Touch. Many of the attempts at getting the "best" power to the Touch also increase the noise sent down the line.
For a linear supply there are mainly three things that contribute to noise sent back down the line:
1) reverse recovery noise when the diodes switch
2) transformer ringing and
3) current spikes when the diodes conduct.
Using Schottky diodes can get rid of the reverse recovery noise, transformer ringing can be dealt with with an RC network across the secondary which damps the resonance. This damping of the transformer ringing is very effective but almost never done. If you check out 100 different linear PSU's there will be a very high probability that NONE of them have damped the transformer resonance. Its pathetic that the higher the "quality" of the transformer the worse the resonance is, so actually using one of these expensive "high end" transformers is one of the worst things you can do. (unless you damp it)
The "current spikes" are the hard part, these are caused by using a traditional PSU design of transformer, diodes and a big cap. The diodes only conduct when the voltage from the transformer is greater than the voltage on the cap. Thus the power coming from the transformer is in short high current spikes. For example a common design for the 5V 2A for the Touch will actually have 20A spikes coming from the transformer. These 20A spikes have all kinds of high frequency components which get sent right back through the transformer and into the mains. They also excite the above mentioned transformer resonance causing the transformer to ring like a bell, even if you use Schottky diodes.
The DC signal coming out of such a design is a sawtooth wave. This also has large amounts of high frequency components. Unfortunately most rectifier designs have very little input rejection at high frequencies, they do great at low frequencies, but at high frequencies they are poor. This is one reason that discrete regulators have been used in audiophile designs for so long, they can have much better high frequency input rejection, which is needed to correctly handle the sawtooth.
There is a nice effective solution called a choke. Not the little high frequency things designed to run at 50KHz and up, but big heavy things which have significant inductance at 120Hz. If properly designed they allow continuous conduction through the transformer and rectifier, eliminating the high current spikes. Another advantage is that the output waveform is a pure sine wave, no high frequency harmonics.
Put the choke filter, transformer damping and Schottky diodes together and you have a supply that injects almost nothing back into the AC line and delivers a clean pure sine wave to the rectifier, and with this design you don't NEED a complicated expensive regulator.
I've put together a design using these principles, the schematic is at:
home.comcast.net/~johnswenson1/stereo/SB_5V.GIF
All the parts should be available at Mouser or Digi-Key. I'm sure there are other distributors in other parts of the world where you can get these parts. You should be able to get all the parts for $75 or so. Give it a try I think you will like this design.
Message sent to John Swenson on the slimdevices forum
Hi John, I have few questions:
1. What AC voltage should deliver the transformer and how big should it be ?
2. R1 - 1W is it enough ?
3. Can I use instead of LT1048 one superteddyreg and if yes, is C5 and C6 necessary ? Or should i just forget str and use LT1048 ?
4. When I am building it and measuring, what voltage should it be after choke and after reg ?
Answers:
1, this design uses 10VAC out of the transformer. A little larger won't hurt, it just means the the regulator has to dissipate more power. How big depends on whether you will be running a USB disk drive off it. If not using a bus powered drive 25VA and up is fine. If you want to run a drive off this I would go for 50VA.
2. I'm using a 1/2W carbon composition resistor.
3. Sure you can use other regulators if you want, just make sure you have enough voltage for the regulator drop out. This design was tuned for the dropout of the LT1084, (I don't remember off the top of my head, you can look it up). If the regulator you want to use has a a greater drop out voltage you will have to up the voltage on the transformer. This design is very easy on transformers so you do not need to significantly over rate them.
4. You want the output to be 5V, it doesn't have to be super exact. 5.1 V will be fine, don't go much below. 4.98 is fine, but 4.8 might be too low. Don't go above 5.2 V. As to the output of the choke, that depends on the dropout of the regulator in use. If you are going for 5V out and your regulator needs 2V dropout, the output of the choke should NEVER go below 7V. Give your self a little margin.