The power input to the zero volts line should be to the junction between C6 & C11, not to some remote "star".
The output to the next stage should be from the junction of C7+R3+C8+C9 and from the junction of C12+R8+C13+C14
These two junctions need to meet at a theoretical zero volts node.
The output to the next stage should be from the junction of C7+R3+C8+C9 and from the junction of C12+R8+C13+C14
These two junctions need to meet at a theoretical zero volts node.
Why the 10R?? That kills all the low impedance that is provided by the 317. That''s working backwards.
Jan
Jan
No.
The optimal approach is that illustrated by Peufeu in post #16; why it works out this way is something I separately wrote-up here: Using 3-pin regulators off-piste: part 3
The idea of 'low impedance OR low noise' as a binary choice is unhelpful. If you want the best-possible-implementation of a simple/cheap regulator, the approach to achieve it is given in this thread (and the lessons extend to any kind of active, feedback-based regulation.)
If all you want you want is something 'ultralow noise' but don't care for regulation - and a 10 ohm series impedance is no sort of regulation - then there are many other easier /cheaper/simpler ways to do that, too, starting with the 'cap -multiplier' = other elaborations of an 'RC filter with pass-element' approach.
The optimal approach is that illustrated by Peufeu in post #16; why it works out this way is something I separately wrote-up here: Using 3-pin regulators off-piste: part 3
The idea of 'low impedance OR low noise' as a binary choice is unhelpful. If you want the best-possible-implementation of a simple/cheap regulator, the approach to achieve it is given in this thread (and the lessons extend to any kind of active, feedback-based regulation.)
If all you want you want is something 'ultralow noise' but don't care for regulation - and a 10 ohm series impedance is no sort of regulation - then there are many other easier /cheaper/simpler ways to do that, too, starting with the 'cap -multiplier' = other elaborations of an 'RC filter with pass-element' approach.
In order to optimise something, you need to have a target/goal; how on earth can anyone know whether any approach is optimal or suboptimal when we haven't a clue what ctrlx's requirements are? Of course oscillating regulators are suboptimal for about any application you can think of, but I can't possibly say whether the RC filter helps or harms.
Yes, if you are going to do CRC then pre reg (provided you keep at least 5V in to out differential) is IMO the way to go. Then you should still get lower noise, and the output impedance of your PS is still low.
Tests I did here http://www.diyaudio.com/forums/blogs/wintermute/574-yarps-finally-some-progress.html Note that that powersupply was an excercise in overkill. The final version had 4700uF caps not 10,000uF (which is still overkill). But the scope traces should give an idea of the effectiveness of the CRCRC before the reg. Unfortunately I didn't do what would have been the most important measurment (which I think I have done since) which was to show output of the reg with a single cap and with the full crcrc network. I may have done some measurements in my LM317 experiments thread...
Tony.
Tests I did here http://www.diyaudio.com/forums/blogs/wintermute/574-yarps-finally-some-progress.html Note that that powersupply was an excercise in overkill. The final version had 4700uF caps not 10,000uF (which is still overkill). But the scope traces should give an idea of the effectiveness of the CRCRC before the reg. Unfortunately I didn't do what would have been the most important measurment (which I think I have done since) which was to show output of the reg with a single cap and with the full crcrc network. I may have done some measurements in my LM317 experiments thread...
Tony.
I would've never guessed cap selection and implementation would be so confusing for these regs.
I'm beginning to question whether or not to even use them. If they oscillate that easily, I wouldn't know it since I don't own a scope.
I'm beginning to question whether or not to even use them. If they oscillate that easily, I wouldn't know it since I don't own a scope.
Tony, I have been reading your posts about this reg with great interest while building a Linear PSU for my DAC.
Do you have more details of your measuring setup with the FFT and the soundcard capture (the one George inspired to you to try)?
Cheers.
Nothing confusing really. You just need to understand that not all caps are equal electronically. And that using whatever cap is popular with moders these days is not always a good idea.
Oscillation prevention appears very often as a motif in electronics design, so getting comfortable with it will help you in the long run.
About lm317, the datasheet is your friend. If you want something extra, try some other highly regarded source like Rod Elliott's projects.
I have also found the article below really helpful
Using 3-pin regulators off-piste: part 1
Read all 4 parts and the tnt cross-references and it will all become clearer.
You will also see some nice ways to make the reg even better, with just a few leds.
You will also see that C5 and C6 are really useful.
When you feel like reading more, have a look at the Analog Devices application note about decoupling too. You will then see that as Jan already said about those bypasses, those FM/FC caps really belong close to your load.
Oscillation prevention appears very often as a motif in electronics design, so getting comfortable with it will help you in the long run.
About lm317, the datasheet is your friend. If you want something extra, try some other highly regarded source like Rod Elliott's projects.
I have also found the article below really helpful
Using 3-pin regulators off-piste: part 1
Read all 4 parts and the tnt cross-references and it will all become clearer.
You will also see some nice ways to make the reg even better, with just a few leds.
You will also see that C5 and C6 are really useful.
When you feel like reading more, have a look at the Analog Devices application note about decoupling too. You will then see that as Jan already said about those bypasses, those FM/FC caps really belong close to your load.
Hi YashN, I probably don't have any more info than what I posted in the threads. The measurements one went a bit dead in the end (I wanted to make the final version so I could hook it up to my DC-B1, and that used all my parts that I was experimenting with) , I really would like to do some more experimenting at some point, possibly with a circuit specifically for that purpose.
I did build an instrumentation amp (that George very kindly sent me the parts for) but I think I botched something as I was not able to get it working properly, I revisited it recently and nearly smoked something... I need to run it off proper + - rails rather than with a resistive divider to give pseudo + - rails.....
I now have a USB sound card Focusrite 2i2 and have done a breadboard poc of a balanced to SE converter to use with it. I believe using that it should reduce noise even further (and running from laptop which is not mains connected has to help too).
Effectively the measurements are done by connecting the rail of the PS (via a cap) to the input of an opamp which then feeds the input to the PC soundcard. My little pc preamp basically has 0.1X 1 X and 10X gain (from memory). I never did the proper calibration to get absolute voltage values, but I believe the method has good potential for examining the PSU noise via fft at least within the limits of the sound card.
I should re-iterate what I said in one of the other threads. I'm not an EE, just a guy with enough knowledge to be dangerous, who has read a lot of datasheets and done a LOT of simulation, and a fair amount of testing and measurement of actual LM317 circuits.
Tony.
I did build an instrumentation amp (that George very kindly sent me the parts for) but I think I botched something as I was not able to get it working properly, I revisited it recently and nearly smoked something... I need to run it off proper + - rails rather than with a resistive divider to give pseudo + - rails.....
I now have a USB sound card Focusrite 2i2 and have done a breadboard poc of a balanced to SE converter to use with it. I believe using that it should reduce noise even further (and running from laptop which is not mains connected has to help too).
Effectively the measurements are done by connecting the rail of the PS (via a cap) to the input of an opamp which then feeds the input to the PC soundcard. My little pc preamp basically has 0.1X 1 X and 10X gain (from memory). I never did the proper calibration to get absolute voltage values, but I believe the method has good potential for examining the PSU noise via fft at least within the limits of the sound card.
I should re-iterate what I said in one of the other threads. I'm not an EE, just a guy with enough knowledge to be dangerous, who has read a lot of datasheets and done a LOT of simulation, and a fair amount of testing and measurement of actual LM317 circuits.
Tony.
Last edited:
ammel68, don't forget about the 'noise source' end of your power supply - the basic old secondary winding, diodes and first filter cap. Poor layout, poor part selection, poor connection points, and poor noise egress mitigation could allow noise to bypass some of the noise attenuation you were hoping from the regulation stage, and could impose more noise than necessary on the input of the regulator for starters.
It is common now to at least use fast recovery diodes, such as UF4007. A smaller current rating for such a diode is a benefit, as it usually minimises reverse recovery and off-state capacitance.
Restricting the layout of transformer and rectifier and first filter cap minimises the loop area for rectification noise currents. If your power transformer has substantial leakage inductance, then a tuned snubber across the secondary may help alleviate diode turn-off noise glitches appearing elsewhere in the audio circuitry. If you can use a power transformer with at least one electrostatic shield then that may be a benefit. Choosing a radial, low ESR electrolytic for the first filter cap, and not too large that it forces longer wiring distances, is a good path to take. And don't forget to to use the cap terminals as the take-off connection points for subsequent circuitry (ie. the regulator).
It is common now to at least use fast recovery diodes, such as UF4007. A smaller current rating for such a diode is a benefit, as it usually minimises reverse recovery and off-state capacitance.
Restricting the layout of transformer and rectifier and first filter cap minimises the loop area for rectification noise currents. If your power transformer has substantial leakage inductance, then a tuned snubber across the secondary may help alleviate diode turn-off noise glitches appearing elsewhere in the audio circuitry. If you can use a power transformer with at least one electrostatic shield then that may be a benefit. Choosing a radial, low ESR electrolytic for the first filter cap, and not too large that it forces longer wiring distances, is a good path to take. And don't forget to to use the cap terminals as the take-off connection points for subsequent circuitry (ie. the regulator).
Last edited:
Thanks for your input trobbins.
So many things to consider. So the standard ole' 1N4xxx diodes aren't such a good choice anymore?
I thought about ordering some SB180 Schottkys from Mouser to try.
Salas recommends them for one of his supplies.
I would like to layout my own PCB, but with your comments about poor layout and connection points I'm not so sure without actually seeing a well laid out board first.
Ultimately, all I'm trying to do is come up with a good +-12VDC PS to power a simple op-amp line stage circuit.
Perhaps I need to look at other(or better) options other than a LM317/337 based supply?
I'm certainly open to suggestions.
I've built a few PSU's using the LM317/337 with no issue's at all. I'm not as experienced as other forum member's, but wonder if your all making this too complicated for what the OP wanted.
On the LM317AHV datasheet it says Cin is "required when regulator is located an appreciable distance from power supply filter." And Cout is "is not needed for stability, however, it does improve transient response." I've used a 10u electrolytic as Cout .
So for playing with opamp's I made +15v 0v -15v PSU using a centre tapped mains tfmr, 4 x 1N4007 for the bridge, caps of around 4700u for smoothing/reservoir LM317/337 + 220r as R1, 5k pot as R2 10u cap as Cout. all made from bit's mostly out the junk box. Used copper clad board with the tracks cut out with a dremmel and laid out pretty much as the schematic is. Think I used Rod Elliot's schematic see here - Power Supply for Preamps . Job done.
A.
On the LM317AHV datasheet it says Cin is "required when regulator is located an appreciable distance from power supply filter." And Cout is "is not needed for stability, however, it does improve transient response." I've used a 10u electrolytic as Cout .
So for playing with opamp's I made +15v 0v -15v PSU using a centre tapped mains tfmr, 4 x 1N4007 for the bridge, caps of around 4700u for smoothing/reservoir LM317/337 + 220r as R1, 5k pot as R2 10u cap as Cout. all made from bit's mostly out the junk box. Used copper clad board with the tracks cut out with a dremmel and laid out pretty much as the schematic is. Think I used Rod Elliot's schematic see here - Power Supply for Preamps . Job done.
A.
Thanks a lot, Tony. The way you explored it was certainly very methodical and helpful.
Last edited:
ammel68, the mur220's you mentioned in the first post look like a good choice. I'm currently making a DAC kit (that I bought 5 years ago) and it has BYV27-150's which have very similar specs to the mur220's.
I had been using BYV32E-200's but they are overkill for low current applications.
Tony.
I had been using BYV32E-200's but they are overkill for low current applications.
Tony.
The 2A device will have more reverse recovery, and has a higher capacitance, than the 1A UF4007 range. Given the likely load current is well below 1A, there appears to be no good reason, and a pedantic disadvantage, to using a 2A rated device.
But the peak current through the diode is many times larger than the load current. So it would probably be a good idea to check max values, which of course much depend on the reservoir capacitor value as well. Larger C gives larger peak current.
Jan
Jan
Larger C causes higher peak surge current at power on but has little influence on steady state repetitive peak current. Inrush limiting circuitry can greatly reduce peak current at power on. See Linear Power Supply Design for details.
The LM317 is stable with 100 nF but the LM337 will oscillate. It needs 1 uF tantalum or 10 uF normal cap. I have tested this myself. The datasheet says that also.
I'm not talking about inrush. The total power transferred has to be put into the reservoir caps during the time the diodes conduct, which typically is between 20% to 10% of the mains period. If you have a large cap, the voltage droop is smaller and the conduction angle for the diodes gets shorter. The shorter the conduction angle, the higher the peak current to transfer the required power.
For a very rough approximation, if the DC load is 1A, and the conduction of the diodes is 10% of the period, necessarily the diode peak current will be > 10A.
Jan
- Status
- Not open for further replies.