I'm designing a +-18 VDC power supply for a line level project. Basically, a simple CT transformer, rectifier pair, filter caps and a pair of LM317's for voltage regulation as has been done thousands of times.
While browsing I ran across a PS similar to the one I'm planning also using the LM317. What caught my attention is that it uses a 36 VCT 15 VA transformer, but can produce up to +-20 VDC. I've gone over the LM317 data sheets, but don't see where the LM317 can provide a higher DC output than the input. The transformer I've selected is 24 VCT 15 VA based on conventional PS design guidelines. Is there any advantage of using the 36 VCT over the 48 VCT transformer?
Next, a couple of design questions:
I've read where some designers, including John Curl, use 2.2 uF film caps in parallel to the electrolytic filter caps. They are easy and cheap enough to add, but what is the value of bypassing the filters?
The PS will be on a separate PCB from the rest of the project. Is there a rule of thumb as to how large to make the traces? or, should I make them as large as PCB area will permit?
Thanks in advance for your input
While browsing I ran across a PS similar to the one I'm planning also using the LM317. What caught my attention is that it uses a 36 VCT 15 VA transformer, but can produce up to +-20 VDC. I've gone over the LM317 data sheets, but don't see where the LM317 can provide a higher DC output than the input. The transformer I've selected is 24 VCT 15 VA based on conventional PS design guidelines. Is there any advantage of using the 36 VCT over the 48 VCT transformer?
Next, a couple of design questions:
I've read where some designers, including John Curl, use 2.2 uF film caps in parallel to the electrolytic filter caps. They are easy and cheap enough to add, but what is the value of bypassing the filters?
The PS will be on a separate PCB from the rest of the project. Is there a rule of thumb as to how large to make the traces? or, should I make them as large as PCB area will permit?
Thanks in advance for your input
36V CT would be 18VAC rms per secondary half. Feed that to the rectifier and filter cap and you get around (1.4)(18 - 0.7V diode drop) = 24.2VDC across the cap since it charges up to nearly the peak of the waveform, and then into your LM317. That input voltage lets the LM317 drop around 4V, a good number for that part and taking line voltage fluctuations into account, giving you 20VDC regulated out of the chip if the resistors are set for 20V.
A typo here?
A typo here?
Last edited:
48VCT would result in a lot more power dissipation. For half the secondary, (24V - 0.7)(1.4) = 32.6VDC into the regulator. If you wanted 18VDC out, that would be 32.6V - 18V = 14.6V across the chip. If you were pulling 300mA, that would be (0.3A)(14.6V) = 4.4W, vs. (0.3A)(4V) = 1.2W for the 36VCT transformer.
36VCT would probably be the better choice if you are shooting for something in the +/-18VDC to +/-20VDC range.
The usual reason for bypassing big filter caps with smaller caps is lower impedance at a given frequency with the smaller 0.1uF - 1uF caps (due to lower internal series inductance), hence a better ability to pass higher frequencies. Your DC power supply looks more like a short to ground for ac across the audio frequency range (a good thing), looking back into the supply from the output port.
I'll have to leave your trace/layout question for someone more current than me.
36VCT would probably be the better choice if you are shooting for something in the +/-18VDC to +/-20VDC range.
The usual reason for bypassing big filter caps with smaller caps is lower impedance at a given frequency with the smaller 0.1uF - 1uF caps (due to lower internal series inductance), hence a better ability to pass higher frequencies. Your DC power supply looks more like a short to ground for ac across the audio frequency range (a good thing), looking back into the supply from the output port.
I'll have to leave your trace/layout question for someone more current than me.
Last edited:
Here is some interesting reading on capacitors on Rod Elliot's site:
Capacitor Characteristics
He makes a good case that increased inductance in larger power supply filter caps really isn't internal as much as from the leads in bad designs. From section 4.0 in that article:
"... a mere 10mm of lead length (6nH) creates a series resonant circuit at close to 2MHz. Increase the capacitance to 10,000uF, and it is now 20kHz"
You may get some opinions here that bypassing the power supply filter caps is not needed, if you do good layout work and keep the traces and leads short. I usually bypass the filters, but that is just me.
Capacitor Characteristics
He makes a good case that increased inductance in larger power supply filter caps really isn't internal as much as from the leads in bad designs. From section 4.0 in that article:
"... a mere 10mm of lead length (6nH) creates a series resonant circuit at close to 2MHz. Increase the capacitance to 10,000uF, and it is now 20kHz"
You may get some opinions here that bypassing the power supply filter caps is not needed, if you do good layout work and keep the traces and leads short. I usually bypass the filters, but that is just me.
Last edited:
From the sims I've done the best ripple rejection with the LM317 seems to be when the In-out differential is 5V+ (going higher than 5V doesn't give any additional benefit, and will just result in more power dissipation in the regulator).
I tried to verify this on my breadboarded circuit, but my scope resolution was not good enough. I'm going to try a trick I saw another diy audio member do http://www.diyaudio.com/forums/power-supplies/184068-psu-rc-multistage-filtering.html which is to record the rails of the PS and use an fft to view the harmonics, that may verify/debunk the results of my sims.
I suspect the fact that all datasheets for the LM317 specify in-out differential as being 5V for their ripple rejection curves is a good indicator that it is the sweet spot though.
Also you say you are doing +- with a pair of LM317's (I'm assuming not an LM317/LM337). I'm doing this too, but was under the impression you can only do it with dual secondaries and separate rectifiers for + and - not a center tapped transformer. The reason being that the output of one reg is tied to the 0V (0V being positive) but there is a drop over the reg of a few volts, so the other side will be at a different potential and cannot share a common path to the transformer with the other reg circuit...
Tony.
I tried to verify this on my breadboarded circuit, but my scope resolution was not good enough. I'm going to try a trick I saw another diy audio member do http://www.diyaudio.com/forums/power-supplies/184068-psu-rc-multistage-filtering.html which is to record the rails of the PS and use an fft to view the harmonics, that may verify/debunk the results of my sims.
I suspect the fact that all datasheets for the LM317 specify in-out differential as being 5V for their ripple rejection curves is a good indicator that it is the sweet spot though.
Also you say you are doing +- with a pair of LM317's (I'm assuming not an LM317/LM337). I'm doing this too, but was under the impression you can only do it with dual secondaries and separate rectifiers for + and - not a center tapped transformer. The reason being that the output of one reg is tied to the 0V (0V being positive) but there is a drop over the reg of a few volts, so the other side will be at a different potential and cannot share a common path to the transformer with the other reg circuit...
Tony.
Hi Tony,
I've flipped and flopped over dual LM317 and LM317/LM337. I have both, and will probably breadboard both for grins before I'm done. Do you have any idea if one design has any advantages over the other?
Thanks for pointing out the voltage drop issue on the dual LM317 design. It was one of the questions I almost asked.
The transformer does have dual secondaries, so I can use it either way. I'll hang onto it to make a variable lab PS and purchase a 36V/18V dual secondary for this project.
I appreciate all the input - I'm getting back into electronics after ~20 years, and can't believe how much I've forgotten.
I've flipped and flopped over dual LM317 and LM317/LM337. I have both, and will probably breadboard both for grins before I'm done. Do you have any idea if one design has any advantages over the other?
Thanks for pointing out the voltage drop issue on the dual LM317 design. It was one of the questions I almost asked.
The transformer does have dual secondaries, so I can use it either way. I'll hang onto it to make a variable lab PS and purchase a 36V/18V dual secondary for this project.
I appreciate all the input - I'm getting back into electronics after ~20 years, and can't believe how much I've forgotten.
Hi Paul, my decision to go with dual LM317 over LM317/LM337 was mainly based on anecdotal observations from others, and partly because I wanted to do something different.
I've read quite a few times that the performance of the LM337 is inferior to the LM317, I don't know if this is actually true or not. Probably the thing that swayed me the most was a subjective evaluation someone did, which said that the sound was better when the +ve and -ve rails were basically identical (apart from the obvious voltage difference). This wasn't an observation made with LM3x7's and I think the poster had done extensive tests to get +ve and -ve rails with the same output impedance and noise levels. I decided the best way to try and achieve this symmetry was to use two LM317's. Whether or not it makes any difference at all doesn't really concern me, there is a bit more expense involved in going with the dual LM317's due to extra bridge and caps, but in the scheme of things I figured that it was at leas plausible it could affect performance so was worth perusing, even though it is highly unlikely that I would be able to perceive a difference.
I've gone completely overkill with my circuit and after building the prototype have realized it was even more overkill than I initially thought, but it was a learning project for me so that is what matters
Tony.
I've read quite a few times that the performance of the LM337 is inferior to the LM317, I don't know if this is actually true or not. Probably the thing that swayed me the most was a subjective evaluation someone did, which said that the sound was better when the +ve and -ve rails were basically identical (apart from the obvious voltage difference). This wasn't an observation made with LM3x7's and I think the poster had done extensive tests to get +ve and -ve rails with the same output impedance and noise levels. I decided the best way to try and achieve this symmetry was to use two LM317's. Whether or not it makes any difference at all doesn't really concern me, there is a bit more expense involved in going with the dual LM317's due to extra bridge and caps, but in the scheme of things I figured that it was at leas plausible it could affect performance so was worth perusing, even though it is highly unlikely that I would be able to perceive a difference.
I've gone completely overkill with my circuit and after building the prototype have realized it was even more overkill than I initially thought, but it was a learning project for me so that is what matters
Tony.
Last edited:
Since you have the dual secondaries and could go the dual LM317 route, you might want to take a look at the LT3080, one of the more modern versions of the LM317. Beats the LM317 specs by 10x in a couple of areas, plus the LT3080 is an LDO, with a dropout voltage across it of only around 400mV:
LT3080 - Adjustable1.1A Single Resistor Low Dropout Regulator - Linear Technology
Digi-Key - LT3080ET#PBF-ND (Manufacturer - LT3080ET#PBF)
A few things to look out for vs. the LM317:
* There is a minimum output cap requirement of 2.2uF for stability since it is an LDO.
* Has a minimum load requirement of 0.5mA, so you would want to throw something like an 20k resistor on the output.
* 1.1A vs 1.5A for the LM317, but can be paralleled
LT3080 - Adjustable1.1A Single Resistor Low Dropout Regulator - Linear Technology
Digi-Key - LT3080ET#PBF-ND (Manufacturer - LT3080ET#PBF)
A few things to look out for vs. the LM317:
* There is a minimum output cap requirement of 2.2uF for stability since it is an LDO.
* Has a minimum load requirement of 0.5mA, so you would want to throw something like an 20k resistor on the output.
* 1.1A vs 1.5A for the LM317, but can be paralleled
Very intriguing IC, more expensive, but looks to be worth the difference.
I wonder if/how larger Cout values positively affect the performance? Looks like I'll be rooting through the site for references tonight...
I was planning on a LED power indicator for each voltage, so that should suffice (as long as the LED doesn't somehow prevent the load from being sensed at power up)
I only need 500 mA, so this should work just fine.
Looks like I have to add a few on my next parts order
I wonder if/how larger Cout values positively affect the performance? Looks like I'll be rooting through the site for references tonight...
I was planning on a LED power indicator for each voltage, so that should suffice (as long as the LED doesn't somehow prevent the load from being sensed at power up)
I only need 500 mA, so this should work just fine.
Looks like I have to add a few on my next parts order
The LT3080 is definitely more pricey! You could get a whole handful of LM317s for the cost of two 3080s.
If the amp you are powering has the typical 0.1uF (ceramic or film, HF bypass) + 10-20uF electro caps (reservoir) on the various chips rail-to-rail or rail-to-ground, you may already be over that 2.2uF on the output. If it were me I would go ahead and add the 2.2uF anyway, just to be sure, since it would be in parallel with the load bypass caps.
The LED-for-load is a great idea.
If the amp you are powering has the typical 0.1uF (ceramic or film, HF bypass) + 10-20uF electro caps (reservoir) on the various chips rail-to-rail or rail-to-ground, you may already be over that 2.2uF on the output. If it were me I would go ahead and add the 2.2uF anyway, just to be sure, since it would be in parallel with the load bypass caps.
The LED-for-load is a great idea.
I'm not totally sold on the LT3080 for this application. I'm not very familiar with the part though. I generally only use LDOs where the circuit demands it (or perhaps more truthfully, when I'm opposed to changing the ps enough to not demand it). I'd give it's use a twice-over, not a once-over, to ensure there are no thermal issues at 500mA.
the LT3080 certainly looks interesting, seems it has very low noise (wish I had an apples with apples comparison of the noise specs between the LM317 and this)!
Ripple rejection doesn't appear to be quite as good as the LM317 but if the noise is indeed lower it would probably not be an issue. The datasheet claims the chip doesn't need a heatsink, so 500ma (1/2 rated capacity) shouldn't be an issue you would think....
Tony.
Ripple rejection doesn't appear to be quite as good as the LM317 but if the noise is indeed lower it would probably not be an issue. The datasheet claims the chip doesn't need a heatsink, so 500ma (1/2 rated capacity) shouldn't be an issue you would think....
Tony.
LM317 experiments and measurements
They say it doesn't need a heatsink, but they also tout the ability to parallel chips for increased current and "heat spread". At ~$4.50 a pop, that gets ridiculous real quick. From what I gathered, the Pd is <2W. Compared to the 20W claimed for the LM317. Maybe I just got the SMT versions' power spec.
They say it doesn't need a heatsink, but they also tout the ability to parallel chips for increased current and "heat spread". At ~$4.50 a pop, that gets ridiculous real quick. From what I gathered, the Pd is <2W. Compared to the 20W claimed for the LM317. Maybe I just got the SMT versions' power spec.
They imply that by saying that it can be used in "many applications requiring high current, adjustability to zero and no heat sink". They also tout the 350 mV drop, but that's in a configuration that uses a reference voltage, otherwise it's 1.35 V. I don't plan on using the reference voltage config, so that increased power dissipation is certainly going to generate additional heat.
Also, at the price, I think it's prudent to use heat sinks anyway. It can't hurt, it might help, and it looks cooler.
Also, at the price, I think it's prudent to use heat sinks anyway. It can't hurt, it might help, and it looks cooler.
Good point about the power dissipation specs - or lack of them! From this article the 1W - 2W appears to be the surface mount version, with the TO-220 package providing "higher power dissipation", but I have yet to find a spec for it in any of the LT literature. That is a little spooky.
Linear PR: LT3080 - Surface Mount 1.1A LDO is Easily Paralleled for High I(OUT) without Hot Spots
Given that, maybe the good old LM317 is the safest way to go, especially if you have a couple in the parts box right now.
Linear PR: LT3080 - Surface Mount 1.1A LDO is Easily Paralleled for High I(OUT) without Hot Spots
Given that, maybe the good old LM317 is the safest way to go, especially if you have a couple in the parts box right now.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.