The capacitor on a regulator output is only to dampen/attenuate the regulator, and I have found that they don’t sound as good with ceramic caps. Many of the regulators are made to be applied to miniature products so the data sheet will suggest a ceramic part. I would too if I wanted to stay in business selling mass quantities to cell phone manufacturers...
Have been on the hunt for a great regulator output cap lately myself.
Have been on the hunt for a great regulator output cap lately myself.
The codeword of 'ceramic capacitor' on a regulator data sheet really means that you don't have to use a solid Tantalum capacitor for stability, which was required for older regulators form the 70s, especially negative regulators. The funky ESR of a solid Ta cap was used to stabilize the regulator's control loop, and you couldn't use just any capacitor of sufficient size. With modern 'ceramic capable' regulators, you can use pretty much any cap, with as low of an ESR as you'd like, and the regulator will still be stable.
So, if you have the room, I think a modern film cap would probably be the cleanest option for a few µF of capacitance, probably a metallized stacked polypropylene.
Another suggestion I'd make is to use only the minimum bypass capacitance that you can get away with, and let the regulator do the rest. Extra bypass capacitance just adds more potentially troublesome current paths to the PCB, and unless you get everything right, this isn't helpful.
So, if you have the room, I think a modern film cap would probably be the cleanest option for a few µF of capacitance, probably a metallized stacked polypropylene.
Another suggestion I'd make is to use only the minimum bypass capacitance that you can get away with, and let the regulator do the rest. Extra bypass capacitance just adds more potentially troublesome current paths to the PCB, and unless you get everything right, this isn't helpful.
I agree about large parts not always helping things, is all a balance, inductance kills details it seems, or maybe it’s really the current like you suggested. Many applications have the output cap as the only one after the reg, besides the tiny bypass.
I can see how Ta parts would be recommended in that application with the flattish impedance curve. Just not low enough for some of the newer regs it seems though, as you mentioned. I had one recently on a reg but soon replaced it with a standard electrolytic, was surprised at how much better it works/sounds after.
I can see how Ta parts would be recommended in that application with the flattish impedance curve. Just not low enough for some of the newer regs it seems though, as you mentioned. I had one recently on a reg but soon replaced it with a standard electrolytic, was surprised at how much better it works/sounds after.
Hi sumotan,
Well, regulators are basically like a reference voltage and an amplifier. Just like your audio amplifier, these work better and absolutely require a low impedance from the power side, the input. Because there can be a lag between a sudden current spike and the actual ability to supply it, an output capacitor fills that gap. It also reduces the output impedance at higher frequencies, which is something your circuit needs to work better or at all.
It isn't practical or desired to design a power supply with high AC impedance, and this is really what you are suggesting would be beneficial. The regulator's active circuitry needs a low AC impedance supply side to work at it's best. Most will oscillate if that need isn't satisfied. This is also true of the circuit you are powering from the regulator. So the capacitor on the output of the regulator isn't to damp it (it would have a series resistor too if it was used for that). That capacitor is there for your circuit mostly. Again, you need a low AC impedance just like the regulator does.
-Chris
Well, regulators are basically like a reference voltage and an amplifier. Just like your audio amplifier, these work better and absolutely require a low impedance from the power side, the input. Because there can be a lag between a sudden current spike and the actual ability to supply it, an output capacitor fills that gap. It also reduces the output impedance at higher frequencies, which is something your circuit needs to work better or at all.
It isn't practical or desired to design a power supply with high AC impedance, and this is really what you are suggesting would be beneficial. The regulator's active circuitry needs a low AC impedance supply side to work at it's best. Most will oscillate if that need isn't satisfied. This is also true of the circuit you are powering from the regulator. So the capacitor on the output of the regulator isn't to damp it (it would have a series resistor too if it was used for that). That capacitor is there for your circuit mostly. Again, you need a low AC impedance just like the regulator does.
-Chris
One data that was added in the '90s as awfully important for regulators to be used in audio was impedance.
To me it was Walt Jung that pointed that importance in several The Audio Amateur issues as a major influence in audio quality.
Very few regulators back then had an impedance curve, and on the impedance comparison tests they made in TAA only the LM317 and LM337 had such info on the official datasheet.
That seems to be the norm still today, no output impedance curves, and if you have a look at the recently mentioned ADP7142 regulator you may confirm that.
So how do recent regulators compare to the Super Regulator, for instance, using the latest curves Walt Jung provided?
Anatech, how can you make yourself sure the DC you are feeding a regulator with is the lowest impedance you can provide?
To me it was Walt Jung that pointed that importance in several The Audio Amateur issues as a major influence in audio quality.
Very few regulators back then had an impedance curve, and on the impedance comparison tests they made in TAA only the LM317 and LM337 had such info on the official datasheet.
That seems to be the norm still today, no output impedance curves, and if you have a look at the recently mentioned ADP7142 regulator you may confirm that.
So how do recent regulators compare to the Super Regulator, for instance, using the latest curves Walt Jung provided?
Anatech, how can you make yourself sure the DC you are feeding a regulator with is the lowest impedance you can provide?
Hi Carl,
That is simple. No need to focus on making the impedance the lowest you can provide. Just by following the recommendations in the datasheet and any papers that cover this will suffice. This is a situation where their is a maximum source impedance and dropping far below that will improve performance, but with diminishing returns on efforts. So you design your circuit to provide a reasonable impedance looking back into the supply from the regulator. Add some margin for ageing and everything should be good for years to come.
Some of the newer regulators can become unstable with very low ESR capacitors, so a case of knowing your parts - and reading the data sheet.
-Chris
That is simple. No need to focus on making the impedance the lowest you can provide. Just by following the recommendations in the datasheet and any papers that cover this will suffice. This is a situation where their is a maximum source impedance and dropping far below that will improve performance, but with diminishing returns on efforts. So you design your circuit to provide a reasonable impedance looking back into the supply from the regulator. Add some margin for ageing and everything should be good for years to come.
Some of the newer regulators can become unstable with very low ESR capacitors, so a case of knowing your parts - and reading the data sheet.
-Chris
Hi Chris,
Even if my name is Carlos, I got used to be called Carl.
I mention that about the DC input impedance, because RC or LC filtering techniques after the diode bridge should affect impedance, wouldn't it?
Even so they are used on many high quality regulators.
What about the output impedance curves, so we can compare chip regulators to discrete supplies such as the Super Regulator?
And of course about using low ESR capacitors at the regulator's output. If you do, you should add a small series resistor to the output bypass cap.
Even if my name is Carlos, I got used to be called Carl.
I mention that about the DC input impedance, because RC or LC filtering techniques after the diode bridge should affect impedance, wouldn't it?
Even so they are used on many high quality regulators.
What about the output impedance curves, so we can compare chip regulators to discrete supplies such as the Super Regulator?
And of course about using low ESR capacitors at the regulator's output. If you do, you should add a small series resistor to the output bypass cap.
Hi Carl,
I'll do my best to remember your preferred name. I can't promise anything.
-Chris
I'll do my best to remember your preferred name. I can't promise anything.
Maybe a little, but the most important factor would be the final filter. Having a high impedance between the power supply and input AC is a very good thing. That's one reason why the super regulators work so well. I routinely put resistors between the diode bridge and the input filter capacitor. Near the regulator you would want some light bulk storage and then some bypass capacitor(s). Therefore the regulator does see a very low AC impedance regardless of what I did in the rectification - filtering stage.
-Chris
What about means for comparing Super Regulator specs to modern chip regulators, and how is this affected by the capacitors that are being discussed here?
The idea of using a high quality film capacitor seems like a very good one, but some regulators need more output capacitance, don't they?
There's a comparative subjective test on the web, it shouldn't be hard to find, where several capacitor types and values are used, the results being placed on a graphic as a quality note, not as a curve.
On that case, the DUT regulator was a 3X7, but it could be used with many other types.
The influence of the center pin bypass, setting the output voltage,was also evaluated.
The bypass capacitors used, both on center pin and output, went from nothing to about
100uF, if I remember well. Capacitor brands, such as BG or Silmic were also tried.
One thing that surprised was that, subjectively, using no caps rated quite high in audio perception. Bass, mids and treble were considered more real and natural.
The idea of using a high quality film capacitor seems like a very good one, but some regulators need more output capacitance, don't they?
There's a comparative subjective test on the web, it shouldn't be hard to find, where several capacitor types and values are used, the results being placed on a graphic as a quality note, not as a curve.
On that case, the DUT regulator was a 3X7, but it could be used with many other types.
The influence of the center pin bypass, setting the output voltage,was also evaluated.
The bypass capacitors used, both on center pin and output, went from nothing to about
100uF, if I remember well. Capacitor brands, such as BG or Silmic were also tried.
One thing that surprised was that, subjectively, using no caps rated quite high in audio perception. Bass, mids and treble were considered more real and natural.
Hi Carl,
Careful about this. I'm using measurements and only listen to the successful results. As soon as you say "sounds better", all bets are off. This is a black and white measurement thing where regulators are concerned. The subjective thing you would have to listen for is a reduction of noise. Listening in mono, or just one channel, prevents you from being drawn into the effects of the music.
It's so easy to become distracted by the music and that is hard to guard against. Keeping the source material in mono allows you to more easily evaluate any changes that were made.
-Chris
Careful about this. I'm using measurements and only listen to the successful results. As soon as you say "sounds better", all bets are off. This is a black and white measurement thing where regulators are concerned. The subjective thing you would have to listen for is a reduction of noise. Listening in mono, or just one channel, prevents you from being drawn into the effects of the music.
It's so easy to become distracted by the music and that is hard to guard against. Keeping the source material in mono allows you to more easily evaluate any changes that were made.
-Chris
Chris,
Did I say "sound better"? I think I did not. I prefer to employ the term "different" or "differences", which is something you will probably listen. Whether it's best or worst is completely subjective and personal, and will probably differ from one person to another.
It's a bit different, in my opinion, to say more natural or real. Sometimes the problem is that many people do not know how a real acoustic instrument sounds, or an unamplified voice. If you do have such reference, then I think it is valid to use them.
My background is in professional location audio, as I worked in location audio for films for close to 40 years. And my personal task was to use my memory on how the actual sound source sounded and put into tape, analog at the time.
The accuracy of that chain was primordial, and you get to pick very fine differences and alterations, sometimes correcting.
Measurements, sorry to say, do not get to that accuracy and many times hardly represent something that can be measured. Maybe that's why you need objective and subjective evaluations, BOTH.
Did I say "sound better"? I think I did not. I prefer to employ the term "different" or "differences", which is something you will probably listen. Whether it's best or worst is completely subjective and personal, and will probably differ from one person to another.
It's a bit different, in my opinion, to say more natural or real. Sometimes the problem is that many people do not know how a real acoustic instrument sounds, or an unamplified voice. If you do have such reference, then I think it is valid to use them.
My background is in professional location audio, as I worked in location audio for films for close to 40 years. And my personal task was to use my memory on how the actual sound source sounded and put into tape, analog at the time.
The accuracy of that chain was primordial, and you get to pick very fine differences and alterations, sometimes correcting.
Measurements, sorry to say, do not get to that accuracy and many times hardly represent something that can be measured. Maybe that's why you need objective and subjective evaluations, BOTH.
Hi Carl,
Sorry then, my interpretation of any phrase with the word "subjectively" can be equated to "sounds better". What does "subjectively" mean to you? I'd like to speak the same language to you to avoid further miscommunication.
I'm on record as both measuring and listening both. I listen more to the the circuit I'm designing than the power supply because the supply quality is more easily quantified with test instruments. Anything that you can hear can now be measured. But where it comes down to a power supply regulator, it boils down to noise level and any additional hum or buzz that is introduced. When an audio circuit is unhappy with the power supply it will generally be presented as noise level between the harmonic peaks. It often turns out that harmonics to the test signal can sometimes be reduced with a better power supply as well.
The quality of the power supply is the first step in the direction of a good or bad audio circuit depending on whether the power supply operates properly in my opinion. There isn't one best way to achieve this. You can come at the problem from several different directions and be successful. Just as there isn't a "best capacitor" you can use. It is helpful if the audio circuit has some level of power supply rejection ratio.
-Chris
Sorry then, my interpretation of any phrase with the word "subjectively" can be equated to "sounds better". What does "subjectively" mean to you? I'd like to speak the same language to you to avoid further miscommunication.
I'm on record as both measuring and listening both. I listen more to the the circuit I'm designing than the power supply because the supply quality is more easily quantified with test instruments. Anything that you can hear can now be measured. But where it comes down to a power supply regulator, it boils down to noise level and any additional hum or buzz that is introduced. When an audio circuit is unhappy with the power supply it will generally be presented as noise level between the harmonic peaks. It often turns out that harmonics to the test signal can sometimes be reduced with a better power supply as well.
The quality of the power supply is the first step in the direction of a good or bad audio circuit depending on whether the power supply operates properly in my opinion. There isn't one best way to achieve this. You can come at the problem from several different directions and be successful. Just as there isn't a "best capacitor" you can use. It is helpful if the audio circuit has some level of power supply rejection ratio.
-Chris
Hi Chris,
I don't see how "subjective" can be equated to "sounds better". If you get to better or worst terms, which I oppose because they explain very little, both are subjective evaluations.
When I say subjective is the opposite of measuring, which many claim as objective but I don't think it's really so. Many electronics engineers deny the existence of subjective evaluations, as being the product of imagination and other personal responses.
I certainly do not agree with you when you say that anything you can hear can be measured. We didn't yet get there, and there are many things you can listen that have no way of being measured, at least not one I know of.
The human ear, on the other side, can not hear many things that can be measured, like higher frequencies.
Listening is a very complex act, a very complex situation, which I haven't yet seen how to convert into accurate measurings.
Subjective for me means what you find specific to the vibrations you are listening to, both direct from the source and its reflections. Then how you pick all that with a microphone and how you can identify the sound you recorded compared to the original source after the audio chain is completed. For me the only words that should be used are more real or more natural. In any case they most certainly sound different from the original sound.
The word "different" is more accurate to describe a sound you hear when you do not know the original source. You complete what's missing with your imagination, your memory of that sound if you ever listened to it.
One problem with many modern world people is that a lot of them have never listened to an actual acoustic instrument with no amplification involved. I find it very sad and poor when I go to a good acoustic hall and they are using electronic amplification.
Those differences people will listen to will also depend upon the transparency of the playing system. In my opinion, as it might be difficult to define a "transparent playing system", a trained ear can listen to incredible minor changes. But for the trained to listen to those minor changes, the playing system should be as transparent as possible.
It's probably an egg/chicken situation, but it's something I have lived. Due to that constant real audio source/recorded source confrontation, you get to train your ear so it won't be fooled by imagination or other distortions.
At the same time, training your ear is something anyone can do or be taught on how to do it. Usually by teaching on how to pay attention to the parts an audio event is made of, particularly the interaction with reflections or lack of them.
I don't see how "subjective" can be equated to "sounds better". If you get to better or worst terms, which I oppose because they explain very little, both are subjective evaluations.
When I say subjective is the opposite of measuring, which many claim as objective but I don't think it's really so. Many electronics engineers deny the existence of subjective evaluations, as being the product of imagination and other personal responses.
I certainly do not agree with you when you say that anything you can hear can be measured. We didn't yet get there, and there are many things you can listen that have no way of being measured, at least not one I know of.
The human ear, on the other side, can not hear many things that can be measured, like higher frequencies.
Listening is a very complex act, a very complex situation, which I haven't yet seen how to convert into accurate measurings.
Subjective for me means what you find specific to the vibrations you are listening to, both direct from the source and its reflections. Then how you pick all that with a microphone and how you can identify the sound you recorded compared to the original source after the audio chain is completed. For me the only words that should be used are more real or more natural. In any case they most certainly sound different from the original sound.
The word "different" is more accurate to describe a sound you hear when you do not know the original source. You complete what's missing with your imagination, your memory of that sound if you ever listened to it.
One problem with many modern world people is that a lot of them have never listened to an actual acoustic instrument with no amplification involved. I find it very sad and poor when I go to a good acoustic hall and they are using electronic amplification.
Those differences people will listen to will also depend upon the transparency of the playing system. In my opinion, as it might be difficult to define a "transparent playing system", a trained ear can listen to incredible minor changes. But for the trained to listen to those minor changes, the playing system should be as transparent as possible.
It's probably an egg/chicken situation, but it's something I have lived. Due to that constant real audio source/recorded source confrontation, you get to train your ear so it won't be fooled by imagination or other distortions.
At the same time, training your ear is something anyone can do or be taught on how to do it. Usually by teaching on how to pay attention to the parts an audio event is made of, particularly the interaction with reflections or lack of them.
Hi Carl,
I love live instruments. A symphony is a treat for me.
My latest instrument is an RTX-6001. It looks at the spectrum from electronic equipment up to approx 96 KHz to a depth of 24 bits. I'm looking at the residuals from the AC supply and any test tone fed to the equipment. With averaging I can "see" down past -120 dB. Finally I can see past my earlier limitations with my work. So far I've been able to see various signals and noise types that colour the sound. I'll look at the output using a 40 MHz spectrum analyser as well to make sure there isn't anything misbehaving up there (HP 3585A). I used to have to judge performance using a single number, THD + noise (HP 339A). But, I backed that up with an oscilloscope on the monitor output, and later the 3585A and still later, an HP 35665A signal analyser. Using the additional instrumentation allowed me to see a lot more information about the operation of the audio equipment and I learned a lot over the years doing this. Finally, with the RTX-6001 I now have a noise floor lower than the equipment I am working on. Now I can see the various things people are complaining about more clearly.
So yes, if you can hear a problem, I can measure it. Carl, test equipment has come a great long way in the past few decades. If you are able to use and interpret this new test equipment you can truly see what is going on. However, what goes on between the ears of the observer will probably always be a complete mystery. One I can't worry about. All I care about is that the audio equipment passes a signal without modifying it too much. It's all we can control. The rest of the journey is beyond our reach and when I am presented with some equipment with a problem, that becomes the start and end of my concerns. Just that piece of equipment.
Trying to control or affect more than that is a fool's errand. So within the boundaries of that piece of equipment, I can see anything you can hear from that piece of audio gear.
We were talking about power supplies before you dragged in variables far out of the scope of the discussion. Can we limit the discussion to the original scope then?
-Chris
I love live instruments. A symphony is a treat for me.
My latest instrument is an RTX-6001. It looks at the spectrum from electronic equipment up to approx 96 KHz to a depth of 24 bits. I'm looking at the residuals from the AC supply and any test tone fed to the equipment. With averaging I can "see" down past -120 dB. Finally I can see past my earlier limitations with my work. So far I've been able to see various signals and noise types that colour the sound. I'll look at the output using a 40 MHz spectrum analyser as well to make sure there isn't anything misbehaving up there (HP 3585A). I used to have to judge performance using a single number, THD + noise (HP 339A). But, I backed that up with an oscilloscope on the monitor output, and later the 3585A and still later, an HP 35665A signal analyser. Using the additional instrumentation allowed me to see a lot more information about the operation of the audio equipment and I learned a lot over the years doing this. Finally, with the RTX-6001 I now have a noise floor lower than the equipment I am working on. Now I can see the various things people are complaining about more clearly.
So yes, if you can hear a problem, I can measure it. Carl, test equipment has come a great long way in the past few decades. If you are able to use and interpret this new test equipment you can truly see what is going on. However, what goes on between the ears of the observer will probably always be a complete mystery. One I can't worry about. All I care about is that the audio equipment passes a signal without modifying it too much. It's all we can control. The rest of the journey is beyond our reach and when I am presented with some equipment with a problem, that becomes the start and end of my concerns. Just that piece of equipment.
Trying to control or affect more than that is a fool's errand. So within the boundaries of that piece of equipment, I can see anything you can hear from that piece of audio gear.
We were talking about power supplies before you dragged in variables far out of the scope of the discussion. Can we limit the discussion to the original scope then?
-Chris
Certainly, and adding the power supplies to this discussion seems very important. Not just the capacitors.