The hash choke suggestion on my part was to help see the real noise on the scope, when I did not understand that the BW limit button would do the same thing for the original poster, or to suppress RF or EMI coming from other sources. I did not mean it for filtering diode switching noise, if there is any, or to suggest it as a good alternative for snubbers. Sorry for any misdirection.
My (limited) experience with noise from bridge diodes suggests that it is preferrable to not generate it in the first place, especially in low power cases. Meaning, pick a schottky diode that will not spray HF all over the place when it turns off, if that does not work then snubbers, and only after that try to filter it out.
Also, depending on what type of noise, and where it is coming from, a couple cheap ferrite beads may do as well, and they fit into tight places better than axial hash chokes with inch-long cores.
PM
My (limited) experience with noise from bridge diodes suggests that it is preferrable to not generate it in the first place, especially in low power cases. Meaning, pick a schottky diode that will not spray HF all over the place when it turns off, if that does not work then snubbers, and only after that try to filter it out.
Also, depending on what type of noise, and where it is coming from, a couple cheap ferrite beads may do as well, and they fit into tight places better than axial hash chokes with inch-long cores.
PM
jwb said:
It's only an asumption since I'm not able to confirm this. That's why I take some countermeasures here.
I'm using Fast recovery diodes 1N4936 rated at 1A 400V. And the draw is low for the application.
It seems I'm somewhat paranoid after reading about this.
Thanks for your help and opinion.
PMiczek said:
I'll be using ferrite beads at the last portion of the supply to the stage plus a bigger ferrite at the end of the unbilical cable to the CDP, guess that should take care of any remaining garbage.
🙂
but I don't like chokes, since they are reactive to HF garbage but also to the not-so-constant current that your dac/amp/whatever you want draws
Line regulation v. noise
I don't think linear regulators are really made for that (HF filter). They do best with DC and low frequency ripple, normal line and load changes etc.
At higher frequencies the output of a typical 3-pin regulator can have much greater noise on it than most people expect. The number most people look at first when selecting a regulator is the 60-80db ripple rejection ratio. This is only valid around 120Hz (normal rectifier ripple, NOT for bridge switching noise).
Had to look at my National Semi LM137 datasheet for this:
Using the LM137 as an example, typical ripple rejection first drops slowly, (from 60 to 50 db between 100Hz and 1KHz), and then more quickly, by 20 db per decade between 1K and 100KHz. Between 100KHz and 1MHz it is only about 10db.
Output impedance rises from 10 milliohms to 1 ohm over the same range.
A 10u bypass at the adjustment pin, the much touted improvement in the ripple rejection ratio, is only 10db better over most of that range, and has almost no effect at 1MHz, which is why some people add the film cap as the original poster did.
Line transient response is only 50-60%, so the size of a spike at the output resulting from a step change of 1 volt at the regulator input is around 500-600 millivolts.
Most of these numbers are only valid at 25 degrees C, and can get dramatically worse (especially the output noise) as the reg heats up, which is why it looks like some regulators have overkill heatsinks.
Btw, the 25C is not ambient, but 25C T-j (junction temperature).
To make it more interesting, in this datasheet, output noise is specified as RMS at f <= 10KHz in percent of V-out and at 25C T-j, which makes the number smaller and tells you nothing about how much high frequency temperature induced noise there will be when the reg loads down.
This is all in the back part of the regulator datasheets where the graphs are. Normally it is not worth reading unless you are an engineer, but this may be one exception.
PM
I don't think linear regulators are really made for that (HF filter). They do best with DC and low frequency ripple, normal line and load changes etc.
At higher frequencies the output of a typical 3-pin regulator can have much greater noise on it than most people expect. The number most people look at first when selecting a regulator is the 60-80db ripple rejection ratio. This is only valid around 120Hz (normal rectifier ripple, NOT for bridge switching noise).
Had to look at my National Semi LM137 datasheet for this:
Using the LM137 as an example, typical ripple rejection first drops slowly, (from 60 to 50 db between 100Hz and 1KHz), and then more quickly, by 20 db per decade between 1K and 100KHz. Between 100KHz and 1MHz it is only about 10db.
Output impedance rises from 10 milliohms to 1 ohm over the same range.
A 10u bypass at the adjustment pin, the much touted improvement in the ripple rejection ratio, is only 10db better over most of that range, and has almost no effect at 1MHz, which is why some people add the film cap as the original poster did.
Line transient response is only 50-60%, so the size of a spike at the output resulting from a step change of 1 volt at the regulator input is around 500-600 millivolts.
Most of these numbers are only valid at 25 degrees C, and can get dramatically worse (especially the output noise) as the reg heats up, which is why it looks like some regulators have overkill heatsinks.
Btw, the 25C is not ambient, but 25C T-j (junction temperature).
To make it more interesting, in this datasheet, output noise is specified as RMS at f <= 10KHz in percent of V-out and at 25C T-j, which makes the number smaller and tells you nothing about how much high frequency temperature induced noise there will be when the reg loads down.
This is all in the back part of the regulator datasheets where the graphs are. Normally it is not worth reading unless you are an engineer, but this may be one exception.
PM
Re: Line regulation v. noise
This is all in the back part of the regulator datasheets where the graphs are. Normally it is not worth reading unless you are an engineer, but this may be one exception.
-----------------------------------------------
Trouble is, as is for opamps, vendors are selective about what they put in the data sheets. These are often not comparable amongst types or between vendors.
Electronics data is area where normalisation and standardisation are lacking. Why does the LT1085 family have completely different pin arrangements? - to stop direct comparison and replacement?
After 30 or more years of 3 terminal regulators, we are still talking about discrete circuits as being much superior.
This is all in the back part of the regulator datasheets where the graphs are. Normally it is not worth reading unless you are an engineer, but this may be one exception.
-----------------------------------------------
Trouble is, as is for opamps, vendors are selective about what they put in the data sheets. These are often not comparable amongst types or between vendors.
Electronics data is area where normalisation and standardisation are lacking. Why does the LT1085 family have completely different pin arrangements? - to stop direct comparison and replacement?
After 30 or more years of 3 terminal regulators, we are still talking about discrete circuits as being much superior.
Re: Line regulation v. noise
For the novices in the audience, could you remind us why we care what's going on up there? Is this primarily a concern for DACs, clocks, and such?
PMiczek said:
For the novices in the audience, could you remind us why we care what's going on up there? Is this primarily a concern for DACs, clocks, and such?
Some DACs appear to be affected by high frequency noise, some are not.
I have never seen objective measurements of this similar to power supply rejection numbers for op amps, but there seems to be enough evidence from people with good ears to believe this is so, just like it is generally accepted that good audio op amps need to be a lot "faster" (higher slew rate) than the upper end of the audible audio range.
New DACs include functions which border on digital signal processing, re- and upsampling, digital filtering at high clock rates etc. This is probably where the effects come from, and part of the reason why some people prefer non-oversampling DACs.
The digital clock and data at the input of the DAC is on the order of 1-3 MHz (44.1KHz to 96KHz x 16bits x 2 channels), and if the rising and falling edges of the signals become distorted by noise of a certain frequency, the timing of the output samples from the DAC can be affected, and the noise frequency or a harmonic can show up at the output along with the audio. This depends very much on the DAC and on the listener. Outboard reclocking devices can help remove some of the jitter coming from the CD transport, but can't do a thing for noise in the DAC, its digital receiver, or its power supply.
It is also worth noting that any outboard jitter filter or reclocking device will have two effects. It will reduce jitter at some range of frequencies, but it will ADD its own jitter to what is left. Clock kits on the other hand do not suffer from this because you are replacing the original clock or crystal, not adding another layer of logic.
You cannot HEAR high frequency noise, but you can sometimes tell the effects indirectly, from increased harmonic distortion, higher background noise, smearing of the audio signal, and loss of what people call "definition" in higher frequencies. Components which come after the DAC or the CDP can hide the effects, but not remove them alltogether.
Personally I can sometimes see effects on a scope that I cannot hear, as well as hear things I cannot measure. The only reliable way to tell is if you can both hear and measure the effect and if the same thing can be duplicated by someone else, using DIFFERENT gear. Otherwise, you may be measuring noise from a nearby radio station, TV set, battery charger, computer, UPS, or the supply of your instrument.
This is all very subjective (not to mention off topic relative to the original post).
I have never seen objective measurements of this similar to power supply rejection numbers for op amps, but there seems to be enough evidence from people with good ears to believe this is so, just like it is generally accepted that good audio op amps need to be a lot "faster" (higher slew rate) than the upper end of the audible audio range.
New DACs include functions which border on digital signal processing, re- and upsampling, digital filtering at high clock rates etc. This is probably where the effects come from, and part of the reason why some people prefer non-oversampling DACs.
The digital clock and data at the input of the DAC is on the order of 1-3 MHz (44.1KHz to 96KHz x 16bits x 2 channels), and if the rising and falling edges of the signals become distorted by noise of a certain frequency, the timing of the output samples from the DAC can be affected, and the noise frequency or a harmonic can show up at the output along with the audio. This depends very much on the DAC and on the listener. Outboard reclocking devices can help remove some of the jitter coming from the CD transport, but can't do a thing for noise in the DAC, its digital receiver, or its power supply.
It is also worth noting that any outboard jitter filter or reclocking device will have two effects. It will reduce jitter at some range of frequencies, but it will ADD its own jitter to what is left. Clock kits on the other hand do not suffer from this because you are replacing the original clock or crystal, not adding another layer of logic.
You cannot HEAR high frequency noise, but you can sometimes tell the effects indirectly, from increased harmonic distortion, higher background noise, smearing of the audio signal, and loss of what people call "definition" in higher frequencies. Components which come after the DAC or the CDP can hide the effects, but not remove them alltogether.
Personally I can sometimes see effects on a scope that I cannot hear, as well as hear things I cannot measure. The only reliable way to tell is if you can both hear and measure the effect and if the same thing can be duplicated by someone else, using DIFFERENT gear. Otherwise, you may be measuring noise from a nearby radio station, TV set, battery charger, computer, UPS, or the supply of your instrument.
This is all very subjective (not to mention off topic relative to the original post).
PMiczek said:
Nice explanation from Milwaukee, and nice city it is also!
Would you think the Jung regs are suitable for DAC applications?
Power supplies are often at first neglected in D/A and A/D designs, and later often blamed for problems they have nothing to do with.
My experience with Jung-type regulators is limited, and if I were given a choice on which circuit I could make work better in a do-it-yourself context, I would pick a 3-pin or 5-pin integrated regulator supply, along with a local regulator, pretty much as described at the beginning of this thread, and without the oversized filter inductors. (well, I DID try to make a case for oversized heatsinks earlier... 😱)
Testing circuits with noise specs well below 100 microvolts in the presence of 3-5V digital logic (digital audio receiver), diode switching noise (bridge rectifiers), and 100MHz pulse transformers (S/PDIF coax input has rise and fall times on the order of 10-20 nanoseconds) is very tricky. Usually it requires special scope probes (forget the typical 10x probe), ground isolation, and RF screen rooms. In other words $$$, and there better not be a fluorescent light ballast or a computer power supply in the room with you at the time. Below that level, I can only estimate noise, and I have to do measurements on weekends with an empty office, and with overhead lights and computers turned off.
(There are few things as ridiculous as an engineer with bifocals looking for a test point with a mini mag light... 🙂)
Consider that a typical CDP-equivalent analog out is 2-2.25V RMS, or about 6V p-p, and 100uV is about 96db down from that. About the same magnitude as a one bit error on a 16-bit DAC output.
The post that started this thread was about a pre-reg, followed by a local regulator, presumably near the DAC supply pins.
The 317/337 combo with the filter caps and adjustment pin decoupling described above should work well for that purpose, with or w/o the CLC and CRC filters.
Jung regulators and other types using an op amp or discrete components as an error amplifier can have great performance, but they require a lot of parts. Among other things, they depend on bootstrapping and a pre-regulator to work right, so this is not exactly an apples-to-apples comparison with a 317/337 circuit.
To realize their potential, you have to have a really good understanding of ALL of the operating parameters of your circuit, the DAC (or instrumentation amplifier or whatever). not just the basic line, load, and noise specs, AND a good layout. If not, the theoretically better performance of the Jung circuit wil be swamped by other factors and lost.
One of the other diy members has a really good, pro-Jung presentation and circuit on his web site:
http://www.alw.audio.dsl.pipex.com/jung_schematic.htm
and there is also a way-cool supply design on Per Anders web site, complete with pics, schematics, and pcb for sale:
http://home.swipnet.se/~w-50719/hifi/
PM
My experience with Jung-type regulators is limited, and if I were given a choice on which circuit I could make work better in a do-it-yourself context, I would pick a 3-pin or 5-pin integrated regulator supply, along with a local regulator, pretty much as described at the beginning of this thread, and without the oversized filter inductors. (well, I DID try to make a case for oversized heatsinks earlier... 😱)
Testing circuits with noise specs well below 100 microvolts in the presence of 3-5V digital logic (digital audio receiver), diode switching noise (bridge rectifiers), and 100MHz pulse transformers (S/PDIF coax input has rise and fall times on the order of 10-20 nanoseconds) is very tricky. Usually it requires special scope probes (forget the typical 10x probe), ground isolation, and RF screen rooms. In other words $$$, and there better not be a fluorescent light ballast or a computer power supply in the room with you at the time. Below that level, I can only estimate noise, and I have to do measurements on weekends with an empty office, and with overhead lights and computers turned off.
(There are few things as ridiculous as an engineer with bifocals looking for a test point with a mini mag light... 🙂)
Consider that a typical CDP-equivalent analog out is 2-2.25V RMS, or about 6V p-p, and 100uV is about 96db down from that. About the same magnitude as a one bit error on a 16-bit DAC output.
The post that started this thread was about a pre-reg, followed by a local regulator, presumably near the DAC supply pins.
The 317/337 combo with the filter caps and adjustment pin decoupling described above should work well for that purpose, with or w/o the CLC and CRC filters.
Jung regulators and other types using an op amp or discrete components as an error amplifier can have great performance, but they require a lot of parts. Among other things, they depend on bootstrapping and a pre-regulator to work right, so this is not exactly an apples-to-apples comparison with a 317/337 circuit.
To realize their potential, you have to have a really good understanding of ALL of the operating parameters of your circuit, the DAC (or instrumentation amplifier or whatever). not just the basic line, load, and noise specs, AND a good layout. If not, the theoretically better performance of the Jung circuit wil be swamped by other factors and lost.
One of the other diy members has a really good, pro-Jung presentation and circuit on his web site:
http://www.alw.audio.dsl.pipex.com/jung_schematic.htm
and there is also a way-cool supply design on Per Anders web site, complete with pics, schematics, and pcb for sale:
http://home.swipnet.se/~w-50719/hifi/
PM
----------------------------jwb said:
The 1085 family only has 3 terminals. On 5 terminal ones, how often is the enable terminal used in audio?
PMiczek said:
With modern parts we are normally aiming for 140dB, not 96dB, and we will certainly have failed if we are getting a lot of single-bit errors. By that logic, why even worry about signal integrity, let along signal-to-noise?
Yes 140dB is questionable, but the point (for me) of doing my own audio is to make something that isn't provided by the market, do it better than the market, or do it as well as the market but for less money. If I only aimed for 96dB S/N I might as well buy a discman and call it a day.
fmak said:
Beats me, but that wasn't the terminal I was referring to. A three terminal regulator lacks the important SENSE lead.
Beats me, but that wasn't the terminal I was referring to. A three terminal regulator lacks the important SENSE lead. [/B][/QUOTE]
--------------------------------------------
A sense terminal may or may not be a good thing. I have been trying out the 20 uV LT1762/3 series. The SMT t terminal unit is extremely sensitive to the sense signal and presence/lack of of a large ground plane. On the contrary, the fixed voltage 40 uV T0 220 with a grounded heat sink works like a treat (LT19xx)
--------------------------------------------
A sense terminal may or may not be a good thing. I have been trying out the 20 uV LT1762/3 series. The SMT t terminal unit is extremely sensitive to the sense signal and presence/lack of of a large ground plane. On the contrary, the fixed voltage 40 uV T0 220 with a grounded heat sink works like a treat (LT19xx)
jung superregs
Hi passgear,
if I can just drop my pennys worth in, you really owe it to yourself to try the jung superregs as produced by Andy Weekes. He has a website at http://www.alw.audio.dsl.pipex.com/ with heaps of information even if you decide not to buy. I've tried lm337/317 and the lt1086 regs with full decoupling, in dual, single, dual tracking modes, sulzer regs etc with all manner of psu combinations and frankly the audible differences were negligible in comparison. His regs dump all over the competition from a huge height and you can hear the differences from the first 2 notes. I currently have a couple powering the clock in my cpd and the improvement is frankly awesome. As soon as possible I will have a fist full more scattered in cdp and preamp- The reports and info so far indicate that gains are just as great elsewhere. And don't kid yourself that something similar will sound as good. This circuit is really tuned and his latest mod of a single specific cap literally doubles the performance sound quality wise.
cheers
Ced
A total convert to superregulation
Hi passgear,
if I can just drop my pennys worth in, you really owe it to yourself to try the jung superregs as produced by Andy Weekes. He has a website at http://www.alw.audio.dsl.pipex.com/ with heaps of information even if you decide not to buy. I've tried lm337/317 and the lt1086 regs with full decoupling, in dual, single, dual tracking modes, sulzer regs etc with all manner of psu combinations and frankly the audible differences were negligible in comparison. His regs dump all over the competition from a huge height and you can hear the differences from the first 2 notes. I currently have a couple powering the clock in my cpd and the improvement is frankly awesome. As soon as possible I will have a fist full more scattered in cdp and preamp- The reports and info so far indicate that gains are just as great elsewhere. And don't kid yourself that something similar will sound as good. This circuit is really tuned and his latest mod of a single specific cap literally doubles the performance sound quality wise.
cheers
Ced
A total convert to superregulation
- Status
- Not open for further replies.