Hi,
like Bruno has already told you, additional bypassing of reservoir caps has no technical merit. Class D supply current has harmonics of the switching frequency that go up in range to several tens of megahertz. Proposed circuit called "snubber" with it's inenvitable wiring inductance is efectively a relatively high Q resonant circuit and it can nicely resonate with harmonics of the switching frequency (which is variable in self oscillating designs).
A nice example of this is the UcD400 module itself. It's input power supply connector is localy decoupled with ceramic capacitors. With inductance of cca 70cm of twisted 2.5mm2 wire this capacitor resonates nicely around 1MHz. The last thing you may want to add is some more uncontrolled resonance.
If we now leave technical matters aside, I will not say that "snubber" may not have a desirable sonic effect. Some people even hear improvements from GigaHertz tweeters and stuff like that.
Best regards,
Jaka Racman
like Bruno has already told you, additional bypassing of reservoir caps has no technical merit. Class D supply current has harmonics of the switching frequency that go up in range to several tens of megahertz. Proposed circuit called "snubber" with it's inenvitable wiring inductance is efectively a relatively high Q resonant circuit and it can nicely resonate with harmonics of the switching frequency (which is variable in self oscillating designs).
A nice example of this is the UcD400 module itself. It's input power supply connector is localy decoupled with ceramic capacitors. With inductance of cca 70cm of twisted 2.5mm2 wire this capacitor resonates nicely around 1MHz. The last thing you may want to add is some more uncontrolled resonance.
If we now leave technical matters aside, I will not say that "snubber" may not have a desirable sonic effect. Some people even hear improvements from GigaHertz tweeters and stuff like that.
Best regards,
Jaka Racman
Hi Jaka,
Yes I remember that and that's why I wouldn't mess with the module itself.
At this point I'm only half tempted to try it, first I'd like to try a supply with nothing more than schottky rectifiers, not even snubbed themselves.
If you have another word for this snubber besides "zobel" I'd like to know it.
Thanks,
Chris
Yes I remember that and that's why I wouldn't mess with the module itself.
At this point I'm only half tempted to try it, first I'd like to try a supply with nothing more than schottky rectifiers, not even snubbed themselves.
If you have another word for this snubber besides "zobel" I'd like to know it.
Thanks,
Chris
Hi Chris,
I can certainly not call them snubbers. I have been using snubbers in SMPS designs for years and their function is to damp the resonance, not to introduce it. I do object to calling proposed circuit snubber in context of switch mode technology (class D and SMPS). But you have many ideas how to call them otherwise in Carlos's sig.😉
Best regards,
Jaka Racman
I can certainly not call them snubbers. I have been using snubbers in SMPS designs for years and their function is to damp the resonance, not to introduce it. I do object to calling proposed circuit snubber in context of switch mode technology (class D and SMPS). But you have many ideas how to call them otherwise in Carlos's sig.😉
Best regards,
Jaka Racman
Hi Jaka,
I believe the term snubber in this application was coined by Carlos.
How relevant a word is it? As you point out it's not the usual function of a snubber but that term does belong to multiple functions, such as voltage damping, slew rate limiting of voltage or current, and overshoot clamps, I've also seen many more complex ones as I'm sure you have even more so.
It seems they all work by controlling circuit reactances, which I think is how it works here, and so the term might apply, as does zobel.
I asked if any others might apply to facilitate research of it.
Whatever it's called if it introduces resonance you broke it! 🙂
Bruno did say it may have an affect as well placed before the amp, I wouldn't attempt to modify the PCB this way, only to flatten the bandwidth of the supply itself. Many having done it reported very positive results.
Regards,
Chris
I believe the term snubber in this application was coined by Carlos.
How relevant a word is it? As you point out it's not the usual function of a snubber but that term does belong to multiple functions, such as voltage damping, slew rate limiting of voltage or current, and overshoot clamps, I've also seen many more complex ones as I'm sure you have even more so.
It seems they all work by controlling circuit reactances, which I think is how it works here, and so the term might apply, as does zobel.
I asked if any others might apply to facilitate research of it.
Whatever it's called if it introduces resonance you broke it! 🙂
Bruno did say it may have an affect as well placed before the amp, I wouldn't attempt to modify the PCB this way, only to flatten the bandwidth of the supply itself. Many having done it reported very positive results.
Regards,
Chris
Hi Chris,
maybe it would be good to read excellent presentation by JosephK in http://www.diyaudio.com/forums/showthread.php?postid=547988#post547988 .
Best regards,
Jaka Racman
maybe it would be good to read excellent presentation by JosephK in http://www.diyaudio.com/forums/showthread.php?postid=547988#post547988 .
Best regards,
Jaka Racman
Jaka Racman said:
From that thread, it seem that the so called "snubbering" is supposed to create a flat soure impedance/phase for high frequency power transmission. It looks good, but the design would vary with cap characteristics and type of power supply. If the power supply is regulated, then the "snubber" compensates for the high frequency quite well.
If that is its purpose.......as I assume it is.........then it is a zobel. It cancels out the self-inductance a large cap by placing an RC in parallel.
Snubbers are used on bridges, to damp out ringing. Just like Jaka said.
Jocko
Snubbers are used on bridges, to damp out ringing. Just like Jaka said.
Jocko
According to TI in their app note slup100 an RC network is the most common form of rail "snubbing". But it is late and I may have that allllllll mixed up right now.
Joseph_K did excellent work but never came to figure out why Carlito's way seems to have any effect.
I believe it's a simple case of phase lead, be it linear or unregulated supply.
Also the reason the impedance is so low is because snubbers dampen critically when their impedance is equal to the source impedance, this puts us in the sub ohm region.
Regards,
Chris
Joseph_K did excellent work but never came to figure out why Carlito's way seems to have any effect.
I believe it's a simple case of phase lead, be it linear or unregulated supply.
Also the reason the impedance is so low is because snubbers dampen critically when their impedance is equal to the source impedance, this puts us in the sub ohm region.
Regards,
Chris
Jocko Homo said:
Well, a zobel is compensating the load, the "snubber" seems to compensate the source, I'd really like to create some models to see if there are some trends that can be followed.
soongsc said:
All the same I've heard of zobel's on inputs and all over. If "Zobel" only used it on the output then... who's going to patent this? 🙂
classd4sure said:
You're right, those graphs were gain graphs not impedance graphs. Ignor my previouse post.
This means that theoretically, as the demand for current from the amp varies, it actually resonates with the PSU, and the snubber is supposed to damp it out.
I kind of wonder whether SMPS with class D amps would benefit from a snubber.
soongsc said:
I believe snubber as the term applies in the mentioned TI application note is likely more for RFI decoupling and also to keep parasitic resonances down in multi component decoupling networks, like we have here.
In this case it's very much the same resonances we're damping, and I believe the phase lead introduced by the network is responsible for some improvement, but more importantly it's the termination of the supply at a constant impedance over a widebande, which would give your amps output filter a near ideal voltage source (ideal voltage source also=zero impedance) from DC well into the RF range I'm assuming.
If that is in fact the beneficial factor of it, I'd imagine it would help most in improving any perceived lag in slew rate .... because even though the slew rate of the amp isn't a problem it will be still be limited to that of the supply. If in the case like UCD you have an amp that itself has a flat frequency response it could be very worthwhile doing, in ordre to more less complete the package.
Hi,
I guess there is a lot of misunderstanding here. Electrolytic capacitor has a nice bathtub like impedance curve with flat bottom over one or even two decades with NO RESONANCES at all. See attached picture from Epcos datasheet . I do have HP4192 impedance analyzer (5Hz to 13MHz) ind I can assure you that other electrolytics look the same. Typical electrolytic cap will have inductance around 20nH which limits it's impedance as we go higher with frequency.
20nH is approximately inductance of a square centimeter large loop of wire. Good foil cap will have inductance 5nH or more (stacked foil design, wound are around 10-15nH) when measured directly in impedance analyzer fixture and 20nH or more when mounted DIY style. Inductance of combined RC series circuit with 1W resistor is higher than that of the electrolyitic cap itself. Wiring from the electrolyitic cap to the module will have inductance around 200-500nH (10 to 25 times inductance of the electrolytic alone). Now tell me how those "snubbers" flatten impedance of the PS as seen by the module.
Best regards,
Jaka Racman
I guess there is a lot of misunderstanding here. Electrolytic capacitor has a nice bathtub like impedance curve with flat bottom over one or even two decades with NO RESONANCES at all. See attached picture from Epcos datasheet . I do have HP4192 impedance analyzer (5Hz to 13MHz) ind I can assure you that other electrolytics look the same. Typical electrolytic cap will have inductance around 20nH which limits it's impedance as we go higher with frequency.
20nH is approximately inductance of a square centimeter large loop of wire. Good foil cap will have inductance 5nH or more (stacked foil design, wound are around 10-15nH) when measured directly in impedance analyzer fixture and 20nH or more when mounted DIY style. Inductance of combined RC series circuit with 1W resistor is higher than that of the electrolyitic cap itself. Wiring from the electrolyitic cap to the module will have inductance around 200-500nH (10 to 25 times inductance of the electrolytic alone). Now tell me how those "snubbers" flatten impedance of the PS as seen by the module.
Best regards,
Jaka Racman
Attachments
Jaka Racman said:
This is the impedance between the power rail and ground when used in a power supply. So less capability to regulate outside the flat part of the curve. The lower frequency side can probably be handled by a regulated power supply, the high frequency side traditionally is handled by small caps in parallel close to the current drain compponent. I tend to think the phase of the current and the current onset/turnoff characteristics as the load varies with music.
Jocko Homo said:
Exactly.
As Phred said, "snubberizing is welcome and encouraged". 😀
Where is that guy?
Snubber?
As Jacka said, my footer gives several nomenclature options. 😀
Anyway, I tend to call a zobel to an R//L series element on the output of an amp.
On speaker crossovers they call the R/C a zobel.
It's not very consensual.
I think it's more appropriate to call this a snubber, and that's why I did.
Like TI calls it:
As Jacka said, my footer gives several nomenclature options. 😀
Anyway, I tend to call a zobel to an R//L series element on the output of an amp.
On speaker crossovers they call the R/C a zobel.
It's not very consensual.
I think it's more appropriate to call this a snubber, and that's why I did.
Like TI calls it:
Attachments
Jaka Racman said:
I think it is a matter of perception. If you perceive the “Zobel/Snubber” as an augmentation of the power supplies source impedance it would not look to be very beneficial but if you look at it as a termination to that .5uh lead inductance with the high switching rate on the other end you can see how it would reduce the inductive ringing.
Roger
the important thing is that it ' s easy to try these snubbers
i've tried them on ucd180v.1 and Panasonic saxr45 ( equibit ) and yes , it works and surprisingly , not only the treble are better but the midrange and bass too , the effect is very audible , like when you change your power caps
alain
i've tried them on ucd180v.1 and Panasonic saxr45 ( equibit ) and yes , it works and surprisingly , not only the treble are better but the midrange and bass too , the effect is very audible , like when you change your power caps
alain
Au contraire mon amis.
The important thing is understanding why it works. The fun thing is then experiencing it. (I haven't yet).
Now that people can have an understanding as to why it could be beneficial they'll be more interested in trying it.
Roger I think you got it pretty well dead on, other than to specificy it be a switching application, the same little trick seems to work in regulated supplies, and raw dc supplies for a variety of classes.
I kind of view it as tugging on the reactances in order to provide a linear response at a wider bandwidth. It would compensate for ringing, RFI picked up over, as well inductance of the PSU wire as well, why not put one on each side of the wire and really treat it like a transmission line and see if the sound of the wire vanishes?
Could be fun.
Carlos thanks for sharing your TI paper, I too think it is best called a snubber over a zobel since a zobel is with a very specific purpose in mind, he didn't see the full usefulness of it perhaps?
Regards,
Chris
The important thing is understanding why it works. The fun thing is then experiencing it. (I haven't yet).
Now that people can have an understanding as to why it could be beneficial they'll be more interested in trying it.
Roger I think you got it pretty well dead on, other than to specificy it be a switching application, the same little trick seems to work in regulated supplies, and raw dc supplies for a variety of classes.
I kind of view it as tugging on the reactances in order to provide a linear response at a wider bandwidth. It would compensate for ringing, RFI picked up over, as well inductance of the PSU wire as well, why not put one on each side of the wire and really treat it like a transmission line and see if the sound of the wire vanishes?
Could be fun.
Carlos thanks for sharing your TI paper, I too think it is best called a snubber over a zobel since a zobel is with a very specific purpose in mind, he didn't see the full usefulness of it perhaps?
Regards,
Chris
- Status
- Not open for further replies.