From memory yes. As current goes up ripple increases. But I suspect that is simply due to the fact that the ripple reduction is a constant ratio, and increased current draw results in increased ripple on the input, and a proportionally attenuated ripple on the output..
BTW I measured the voltage hot (after running for several months and I had + and - 9.96V (nominally 10V). I then turned the unit off for a couple of hours and measuring immediately after startup I had + and - 10.12V after a minute or so (heatsinks on the LM317's starting to get warm) down to + - 10.06V I'm not sure if this is due to the BC560C's or whether it is normal behaviour for an LM317 as it's temp rises.
Tony.
BTW I measured the voltage hot (after running for several months and I had + and - 9.96V (nominally 10V). I then turned the unit off for a couple of hours and measuring immediately after startup I had + and - 10.12V after a minute or so (heatsinks on the LM317's starting to get warm) down to + - 10.06V I'm not sure if this is due to the BC560C's or whether it is normal behaviour for an LM317 as it's temp rises.
Tony.
Cold to Hot Tempco
But even so 10.12V (cold) down to 9.96V (very thoroughly warmed up) is only +1.6% when cold.
That is a small error that most circuits will simply ignore.
If one requires a very precise voltage one will select a more sophisticated regulator that guarantees by design to achieve the required tempco.
That's obviously a tempco thing.
But even so 10.12V (cold) down to 9.96V (very thoroughly warmed up) is only +1.6% when cold.
That is a small error that most circuits will simply ignore.
If one requires a very precise voltage one will select a more sophisticated regulator that guarantees by design to achieve the required tempco.
Last edited:
If one were to adjust the smoothing capacitance to maintain similar "ripple" voltage at the input of the regulator, would there still be a difference in "noise" excluding hum (@ 100/120Hz) on the output as the output current is changed?
Good advice, and getting the correct ESR caps at the output is something that even professional designers seem to overlook with this ubiquitous component.
One of our customers did just that. Years and years go by, no problem. But then a batch of regs gave problems with oscillations. Obviously with something so simple, it must be the batch of regulators at fault... right? Ehrm, no, the fault actually was caused by insufficient ESR of the used tantalums at the output of the regs as well as close to the load. It was this particular batch that was somehow more sensitive to the design flaw of the circuit and exposed it.
This also meant that all those products made in the previous years were marginally stable at best. And indeed, one of these could temporarily be made to oscillate with a short squirt of freeze spray right on the regulator.
And the solution was exactly as wintermute suggested. With a 0.22 Ohms resistor in series, all was well again...
Oh, and indeed, the oscillation frequency was well within the audible band.
Edit: OnSemi gives some useful tips on the implementation of the external components on page 7 of this datasheet.
Edit 2: TS also mentioned the LM1086ADJ. In its datasheet, there's quite some information on the stable operation regarding ESR, worth a read.
Last edited:
Before you go through the trouble of building a linear power supply, you should ask yourself what problems you are trying to solve, and is it likely that an external regulated supply will fix the problems? It's entirely possible that your external power supply is fine, and the issue is with internally generated noise, thermal regulation, local supply bypassing, etc.
The first thing to do is look at your switching power supply on an oscilloscope, if you have or can borrow one. I've tested some switching PSes that are horrible- spikey, ringing, and a broad spectrum of noise. Others I've tested are dead flat out to 60MHz. If you have one of the latter, you may want to look at improving the regulation inside the DAC. I assume that it has +/- 5V regulators, and something stiffer for the analog section.
If you are looking for better clarity and an expanded soundstage, local small-value bypass caps next to each digital chip and op amp will help tremendously. With a system of decent resolution, even a casual observer can usually hear a difference when you add 1uF bypass caps, and most people can hear the difference when you use a 1uF/0.1uF/0.01uF combo. (I have no idea why filtering out noise at frequencies several orders of magnitude higher than we can hear has an audible effect, but it does.)
If you want punchier sound, tighter bass, the cheap and easy thing to do is add large-value electrolytic capacitors to your power supply. Even with 2 9400uF caps on the input of a 12V power supply I built a couple weeks ago, the ripple was still evident on my 'scope. If you add too much capacitance, you have to start worrying about inrush current and the like, but 10,000-15,000uF isn't excessive and the improvement can be dramatic.
I agree with some of the other comments about the need for a little resistance in series with the electrolytic caps can be crucial, but in general, more capacitance and lower ESR are better up until the point that it starts to ring.
If you do construct the external power supply, pay attention to the basics. Ensure that the voltage drop across the 317 is less than 12V, and don't run it at it's maximum current output. Make sure your heat sink is adequate to keep the regulator at 25-50 degrees C if possible, at certainly no more than 75 degrees C. And keep that 0.33uF cap in there- it will most likely make an audible difference.
The first thing to do is look at your switching power supply on an oscilloscope, if you have or can borrow one. I've tested some switching PSes that are horrible- spikey, ringing, and a broad spectrum of noise. Others I've tested are dead flat out to 60MHz. If you have one of the latter, you may want to look at improving the regulation inside the DAC. I assume that it has +/- 5V regulators, and something stiffer for the analog section.
If you are looking for better clarity and an expanded soundstage, local small-value bypass caps next to each digital chip and op amp will help tremendously. With a system of decent resolution, even a casual observer can usually hear a difference when you add 1uF bypass caps, and most people can hear the difference when you use a 1uF/0.1uF/0.01uF combo. (I have no idea why filtering out noise at frequencies several orders of magnitude higher than we can hear has an audible effect, but it does.)
If you want punchier sound, tighter bass, the cheap and easy thing to do is add large-value electrolytic capacitors to your power supply. Even with 2 9400uF caps on the input of a 12V power supply I built a couple weeks ago, the ripple was still evident on my 'scope. If you add too much capacitance, you have to start worrying about inrush current and the like, but 10,000-15,000uF isn't excessive and the improvement can be dramatic.
I agree with some of the other comments about the need for a little resistance in series with the electrolytic caps can be crucial, but in general, more capacitance and lower ESR are better up until the point that it starts to ring.
If you do construct the external power supply, pay attention to the basics. Ensure that the voltage drop across the 317 is less than 12V, and don't run it at it's maximum current output. Make sure your heat sink is adequate to keep the regulator at 25-50 degrees C if possible, at certainly no more than 75 degrees C. And keep that 0.33uF cap in there- it will most likely make an audible difference.
This may help to answer the question Andrew. It's from the original natsemi datasheet (note TI have released a revised datasheet which I've also just been looking at that seems to have more specs than before).
This also shows graphically one of the things that I had read. That if your current draw is very low that it's a good Idea to add a load resistor to increase the load to at least 100mA for best performance.
It seems the sweet spot is between about 100mA and 500mA for maximum ripple rejection, at least with the original natsemi device anyway (based on this graph). Second attachment is from the new TI datasheet (which looks a lot more warts and all)!
Tony.
This also shows graphically one of the things that I had read. That if your current draw is very low that it's a good Idea to add a load resistor to increase the load to at least 100mA for best performance.
It seems the sweet spot is between about 100mA and 500mA for maximum ripple rejection, at least with the original natsemi device anyway (based on this graph). Second attachment is from the new TI datasheet (which looks a lot more warts and all)!
Tony.
Attachments
Last edited:
I tend to use a Cadj and generally adopt a much higher value than 10uF, usually 100uF or 220uF
That really gets the hum rejection way up.
I always ensure a 10mA minimum current by using the lower value out to adj resistor, 100r or 120r.
I have never tried adding a dummy load across the output to force the 317 into passing more current, in attempts to improve hum attenuation that is already excellent.
This LF performance has always been good. It's the HF that the 317 and most others is reputed to not be good at and it's for this area of performance that the datasheets go very quiet.
I too don't know if noise changes with current loading, but to me that is a different issue from hum attenuation.
That really gets the hum rejection way up.
I always ensure a 10mA minimum current by using the lower value out to adj resistor, 100r or 120r.
I have never tried adding a dummy load across the output to force the 317 into passing more current, in attempts to improve hum attenuation that is already excellent.
This LF performance has always been good. It's the HF that the 317 and most others is reputed to not be good at and it's for this area of performance that the datasheets go very quiet.
I too don't know if noise changes with current loading, but to me that is a different issue from hum attenuation.
Last edited:
I had intended to post these graphs as well, but then I saw AndrewT's exclusion of ripple to his question. I'm not sure that Ripple Rejection vs Current is the figure he's after.
Perhaps noise generated by the voltage regulator itself vs current?
TI has a graph of Ripple Rejection vs Frequency in its datasheet on page 5 which clearly shows that the RR is excellent up to about 10 kHz and then sharply drops to a mere -20 dB at 1 MHz. They don't even specify above that, but it stands to reason it gets even worse. Basically, for HF it's as if it's not even there...
Thougt so...
Last edited:
AND the noise that passes straight through because HF attenuation is terrible.
That's where a rCRC comes in to help. Even better rC(L+R)C to precede the reg.
Last edited:
Hmmm I see I hadn't quite grasped what Andrew was asking. Broad Spectrum noise and what of that is potentially passing through (so might be on the input of the reg) and what is coming from the reg itself (and how increase in current effects that)! I've only really looked at the ripple WRT using CRCRC on the input, and effects of different current draw. I didn't really pay attention to absolute noise floor (that I can remember, though I may have some measurements that could shed some light, I'll see if I can find anything).
Tony.
Tony.
- Status
- Not open for further replies.