Generic snubber values: fixed-/variable-voltage regs

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I'm seeking generic snubber values for common 3-pin regs:

Fixed-voltage (e.g. 7805, 7812, LM340-12, etc.):
Some one once suggested: 33uF normal electrolytic from output pin to ground. Parallel to the 33uF: 0.47R + 10uF electrolytic (use a good-quality cap here).
(suggest more here)

Variable-voltage (e.g. 317):
(suggest here)

Thx,
-hm

P.S. I don't have a lot of experience with snubbers nor do I have access to a 'scope. I think both are probably necessary to fine-tune snubber values. This is why I'm asking for "generic" suggestions.
 
hollowman said:
I'm seeking generic snubber values for common 3-pin regs:

Fixed-voltage (e.g. 7805, 7812, LM340-12, etc.):
Some one once suggested: 33uF normal electrolytic from output pin to ground. Parallel to the 33uF: 0.47R + 10uF electrolytic (use a good-quality cap here).
(suggest more here)

Variable-voltage (e.g. 317):
(suggest here)

Thx,
-hm

P.S. I don't have a lot of experience with snubbers nor do I have access to a 'scope. I think both are probably necessary to fine-tune snubber values. This is why I'm asking for "generic" suggestions.

I think you should definitely start with the input and output capacitors that are recommended by National Semiconductor, in the datasheets for those devices, and in any related Application Notes. They are available free from http://www.national.com .

There is a very good thread about what you are wanting, here:

http://www.diyaudio.com/forums/showthread.php?s=&threadid=106648&perpage=25&pagenumber=1

If you can find the ESL and ESR specs (AT a frequency of interest) for an electrolytic that you want to bypass with a smaller cap, you can look at post #17 in the thread referenced above.

If you do need or want to use an actual snubber network (usually to try to cancel some excess inductance, possibly that of a large electrolytic, or, maybe more likely, a wire or PCB trace, or, of course, a transformer winding), you might be able to estimate the parameters needed to design the snubber, and then use the papers on snubber design, from the links below:

There is a good paper about snubber design on the Cornell Dubilier website, at http://www.cde.com , in the technical papers section:

http://www.cornell-dubilier.com/design.pdf

There are some other good ones on line, too:

http://www.hagtech.com/pdf/snubber.pdf

http://www.ridleyengineering.com/snubber.htm

http://www.maxim-ic.com/appnotes.cfm/appnote_number/3835

http://archive.chipcenter.com/circuitcellar/november00/c1100rp58.htm

It might be difficult, without an oscilloscope. But you can often get fairly comparable views and results with a circuit simulator, such as LTspice, free from http://www.linear.com , especially if you also model the parasitics of components and traces/wires.
 
These regulators are stable without any snubber, although output capacitor ESR may not be too low.

Check this out: http://www.diyaudio.com/forums/showthread.php?s=&threadid=114444

The purpose of RC and RLC snubber networks is to exhibit a low and resistive impedance across a particular frequency band. This comes very handy either to reduce the HF gain of an amplifier stage (whose output is in the form of current) to a known value, or to damp parasitistic RLC resonances like the ones you get in the power paths of SMPS and class D amplifiers.
 
Eva's point is important, those regs (and let alone low-drop types) quite often dislike an OSCON or a 10uF ceramic or other bigger ultra-low ESR stuff directly at their output. Such should be isolated with a small R, say 0.22R or 0.47R. The small increase in load-dependency of output voltage due to that series R is normally of no concern.

Decent layout and circuit design practices (local series LC/RC decoupling), cascading two reguators, reasonable quiescent ballast (>=20mA) and some other "tweaks" (especially for the 317/337) will typically give more improvements that snubber vs. no snubber.

And today's electrolytics are good enough not to worry about it, btw.

- Klaus
 
Importance of Snubbers Questioned

Thx for everyone's useful feedback on this topic.
Originally posted by KSTR ...
Decent layout and circuit design practices (local series LC/RC decoupling), cascading two reguators, reasonable quiescent ballast (>=20mA) and some other "tweaks" (especially for the 317/337) will typically give more improvements that snubber vs. no snubber.

Definitely agree about the importance of layout. Can you elaborate a bit more on:

- cascading two reguators (e.g., perhaps something like this for DACs?)
- reasonable quiescent ballast (>=20mA)
- other "tweaks" (especially for the 317/337)

(Links to important, topical threads/posts on these issues would be fine).

Thx again!
-hm
 
Snubbering vs. bypasing

gootee said:
For adjustable regulators, bypassing the adjust pin with, for example, a small electrolytic to ground can dramatically lower the output ripple + noise.
Thx for that tip! Just so my concepts and nomenclature correct -- and for the sake of correctly searching the web for more info on this topic -- is what you suggest "snubbering", "bypassing", and/or some mixture?
 
Re: Importance of Snubbers Questioned

Hi,
hollowman said:
- cascading two reguators (e.g., perhaps something like this for DACs?)
- reasonable quiescent ballast (>=20mA)
- other "tweaks" (especially for the 317/337)

Cascading is simple, just two complete regulators one after the other, say 7815-->7812. There are more elaborate schemes, see datasheets/AppNotes for these parts (and the LM317), like "floating pre-regulators" (only useful for circuits with variable output voltage). Care must be taken that both regulators always have enough voltage to work upon (the voltage "headroom"), usually 2V-3V for the common types is recommendend.

Also, the ballast can be as simple as a resistor, or combine it with a Zener and LED for a "PWR GOOD" display. If your circuit has enough guaranteed minimum quiescent current then you don't need then ballast, of course. I found that at least 50mA standing current greatly improves the regulation, as does a headroom of 3Volts minimum.

A good and simple tweak, besides the ADJ-pin bypass for 3x7, see here:
http://www.diyaudio.com/forums/showthread.php?postid=356154#post356154

Another similar method, also usable for "fixed" regs (which aren't fixed, only two pre-configured resistors are built in), is to elevate the REF/GND-pin with a reference voltage, obtained from a LM329 or TL431 or similar. For example, a 7805+LM329 gives a far better 12V-reg as a plain 7812 (because the internal node, where the two resistors meet, cannot be bypassed and therefore loopgain gets less with increased output voltage). The advantage of elevation vs. bypass is that elevation has no backpower issues. The principle is the same: don't divide down the error voltage (at least for AC).

See LM317/LM317HV datasheets and AppNotes on this:
http://www.national.com/JS/searchDocument.do?textfield=LM317

- Klaus
 
gootee said:
For adjustable regulators, bypassing the adjust pin with, for example, a small electrolytic to ground can dramatically lower the output ripple + noise.

Are you referring to the 10uF electro cap from the last schema from Datasheet pg. 15? This schema also suggests the use of a diode (D1).

Also --and addition to the suggestion above -- how about a 1uF (Vout to gnd), as shown here:
An externally hosted image should be here but it was not working when we last tested it.
 
There are more elaborate schemes, see datasheets/AppNotes for these parts (and the LM317), like "floating pre-regulators" (only useful for circuits with variable output voltage).
Actually the tracking preregulator as shown in the old LM117 / 317 datasheet works really well where higher performance is required even for fixed voltage outputs.

It separates the issues of line and load regulation - like a cascode; and performs better than the same two regs just connected in series.
 
Three-terminal super regs

AndrewT said:
yes that 10uF improves the 317/337 performance considerably. Without it the 7812/5/8 series is better.
Try increasing this cap all the way to 220uF. The bigger values must have the protection diode added.

Thx for the tip. I didn't know one could take the value of that cap THAT high.

What about the performance of some the so-called industrial regs? E.g. "LT" (Linear Technology), such as LT108x series? They cost more, can handle more current (I think) but I've heard lots of folks say they are not worth using -- at least for audio -- due to noise.

I'll also be experimenting w/ Martin's tracking pre-reg suggestion shortly.

For me, a key goal is compressing the physical layout of all these designs down to the footprint of a std-317-sized reg. Ala Tent- or LCAudio- or Audiocom-like "super regulators". Are there any DIY plans for these yet?
 
Check for LD1084 (mouser.com), a much-cheaper knockoff of LT1084, a 5-amp adjustable reg (But you can get free samples of the Linear parts...).

One thing you might have to watch out for, with a large adj pin bypass cap, is the input-to-output differential climbing too high, too fast, during startup, since the output is very-much "soft start"ed by having a large bypass cap, there. i.e. The input might get too far ahead of the output, and violate the in-out differential spec (during startup only, especially for higher output voltages). Installing a discrete soft-start for the input can circumvent that. See http://www.fullnet.com/gooteesp.htm , for examples of that.

With FIXED regulators, another thing you can do is to feed back only the AC part of the output , to an inverting opamp amplifier whose output takes the place of the ground for the reg's ground pin. This can actively cancel variations in output voltage, even if caused by an active load. Good Stuff. [Use a separate conductor for the feedback, all the way fom the load (i.e. "remote sense").]
 
10uF on ALL PINS of adj. regulators

gootee said:
For adjustable regulators, bypassing the adjust pin with, for example, a small electrolytic to ground can dramatically lower the output ripple + noise.
gootee/anyone:

Can that small-value cap be a tantalum (e.g. 10uF) -- and would *that* be superior to a good , same-size/same-voltage electro you orig. suggested?

Also, would it help if that 10uF tant was placed on *all* pins (i.e. input, output and adjust pins)?
 
Re: 10uF on ALL PINS of adj. regulators

hollowman said:

gootee/anyone:

Can that small-value cap be a tantalum (e.g. 10uF) -- and would *that* be superior to a good , same-size/same-voltage electro you orig. suggested?

Also, would it help if that 10uF tant was placed on *all* pins (i.e. input, output and adjust pins)?


Hi Hollowman,

Actually, this is all covered in the LT1083/84/85 datasheet, from linear.com . See page 7, in the 'Applications Information' section entitled 'Stability', near the bottom of the page, and continuing onto page 8.

As stated there, the input needs at least 10 uF (type not specified), and the output needs at least a 10 uF tantalum or a 50 uF aluminum, OR, if the adjust pin is bypassed with 20 uF (type not specified), then the output needs at least a 22 uF tantalum or 150 uF aluminum. (I'm guessing that the difference in the stated minimum uF values between tantalum and aluminum is probably because tantalums typically have higher ESR (Equivalent Series Resistance) per uF and there is a minimum ESR vs C requirement for the output cap, for stability.) Actually, they also say that the minimum 22uF tantalum or 150 uF aluminum on the output covers ALL cases of bypassing the adjust pin.

The datasheet also states that output capacitors of 100 uF or higher are typically used, and that even higher capacitance values can be used on the output, without limit, and that any higher C values, there, will further improve both the transient response and the stability. (But I have heard that for some regulator types, it's best to use only up to about 150 uF right at the output, and put any larger caps, if used, right at the load, if the load is more than a couple of inches from the regulator's output. That should work better, considering the inductance and resistance of intervening wires or PCB traces.)

The input cap is also often paralleled with a small-value film cap, such as 0.1 uF polyester or polypropylene, et al, as is the adjust-pin cap. But watch out for high-frequency resonances that might be created by using such low-ESR small-value parallel caps, that way.

If the regulator follows the large filter caps in a power supply, I would probably still use the 10uF input cap, especially if the regulator is more than a couple of inches away from the large caps.

Regarding your question about the type of the bypass cap for the adjust pin: It will help a whole lot, regardless of whether it's tantalum or aluminum. But I think I would want to use the one with the lowest ESR at 120 Hz (or 100 Hz, for 50 Hz mains), or maybe whichever type is cheaper. If I were buying new ones, for that, I guess I'd go with aluminum, which is probably better on both counts (I'm guessing). But if you already have some 10 uF tantalums, maybe you could try one, or try paralleling two of them, there. Either way will be excellent, probably.
 
Re: Re: 10uF on ALL PINS of adj. regulators

gootee said:
Actually, this is all covered in the LT1083/84/85 datasheet, from linear.com . See page 7, in the 'Applications Information' section entitled 'Stability', near the bottom of the page, and continuing onto page 8.

As stated there, the input needs at least 10 uF (type not specified), and the output needs at least a 10 uF tantalum or a 50 uF aluminum, OR, if the adjust pin is bypassed with 20 uF (type not specified), then the output needs at least a 22 uF tantalum or 150 uF aluminum. (I'm guessing that the difference in the stated minimum uF values between tantalum and aluminum is probably because tantalums typically have higher ESR (Equivalent Series Resistance) per uF and there is a minimum ESR vs C requirement for the output cap, for stability.) Actually, they also say that the minimum 22uF tantalum or 150 uF aluminum on the output covers ALL cases of bypassing the adjust pin.

The datasheet also states that output capacitors of 100 uF or higher are typically used, and that even higher capacitance values can be used on the output, without limit, and that any higher C values, there, will further improve both the transient response and the stability. (But I have heard that for some regulator types, it's best to use only up to about 150 uF right at the output, and put any larger caps, if used, right at the load, if the load is more than a couple of inches from the regulator's output. That should work better, considering the inductance and resistance of intervening wires or PCB traces.)

The input cap is also often paralleled with a small-value film cap, such as 0.1 uF polyester or polypropylene, et al, as is the adjust-pin cap. But watch out for high-frequency resonances that might be created by using such low-ESR small-value parallel caps, that way.

If the regulator follows the large filter caps in a power supply, I would probably still use the 10uF input cap, especially if the regulator is more than a couple of inches away from the large caps.

Regarding your question about the type of the bypass cap for the adjust pin: It will help a whole lot, regardless of whether it's tantalum or aluminum. But I think I would want to use the one with the lowest ESR at 120 Hz (or 100 Hz, for 50 Hz mains), or maybe whichever type is cheaper. If I were buying new ones, for that, I guess I'd go with aluminum, which is probably better on both counts (I'm guessing). But if you already have some 10 uF tantalums, maybe you could try one, or try paralleling two of them, there. Either way will be excellent, probably.

I had dismissed the LT108x series because of somewhat-reputed accounts of these regs being very noisy for audio apps.

IAC, National's datasheet for their LM117-317 regs notes that...

The adjustment terminal can be bypassed to ground on the LM117 to improve ripple rejection. This bypass capacitor prevents ripple from being amplified as the output voltage is increased. With a 10 uF bypass capacitor 80dB ripple rejection is obtainable at any output level. Increases over 10 uF do not appreciably improve the ripple rejection at frequencies above 120Hz. If the bypass capacitor is used, it is sometimes necessary to include protection diodes to prevent the capacitor from discharging through internal low current paths and damaging the device.

In general, the best type of capacitors to use is solid tantalum. Solid tantalum capacitors have low impedance even at high frequencies. Depending upon capacitor construction, it takes about 25 uF in aluminum electrolytic to equal 1uF solid tantalum at high frequencies. Ceramic capacitors are also good at high frequencies; but some types have a large decrease in capacitance at frequencies around 0.5 MHz. For this reason, 0.01 uF disc may seem to work better than a 0.1 uF disc as a bypass.

I’ve also heard -- though not universally -- that the output cap should have a value on the lower end of the manuf’s suggested range (and esp with tants). I believe Naim audio uses this strategy: i.e. 10uF tant’s on each 317’s output. Maybe this is why their electronics have a rep for having excellent dynamics, rhythm and pace as well as solid, defined bass?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.