ELSO,
That a very good application note.
Looks like Kieran O’Malley of ON Semiconductor also see the value of using tantalum caps. He does make an excellent point that selecting the correct size and ESR is still a trial and error method.
The Mallory THF and the Sprague 550D series offer the best ERS for their size. I have not seen a surface mount tantalum that matches the performance of these radials.
Also most engineers or unaware of these caps, their are specials and are not cheap. I have used them in switching power supplies, timing circuits, regulators and to reduce the noise in digital circuits.
That a very good application note.
Looks like Kieran O’Malley of ON Semiconductor also see the value of using tantalum caps. He does make an excellent point that selecting the correct size and ESR is still a trial and error method.
The Mallory THF and the Sprague 550D series offer the best ERS for their size. I have not seen a surface mount tantalum that matches the performance of these radials.
Also most engineers or unaware of these caps, their are specials and are not cheap. I have used them in switching power supplies, timing circuits, regulators and to reduce the noise in digital circuits.
Millwood,
Your right Kemet really does not cater to that maket. I know that
Sprague and Mallory do. Since Vishay purchase Sprague I'm not sure if the have those older products in their line up.
Sprague also has some really good Eletrolytics that or super low ESR. I have used the axial 604D in switch mode power supplies, their great. Also, the have radial eletrolytics 604D which have four terminals, 81D, 676D and 673D and the Tantalum 550D.
Your right Kemet really does not cater to that maket. I know that
Sprague and Mallory do. Since Vishay purchase Sprague I'm not sure if the have those older products in their line up.
Sprague also has some really good Eletrolytics that or super low ESR. I have used the axial 604D in switch mode power supplies, their great. Also, the have radial eletrolytics 604D which have four terminals, 81D, 676D and 673D and the Tantalum 550D.
ESR
Looking at ESR for the Mallory THF caps shows typical ESR values in the 50 to 100 milliohm range. These are pretty close to the range of values for the Oscon caps. be careful reading ESL numbers because trace inductance in the PCB can easily be several times that number.
The origional Sullzer Regulator artical used tantalums but it seems that few audio designs use them today. Pricing is part of the issue I am sure, but I see few of them in even high budget designs. It seems I have heard some talk of a different sonic signature but haven't played much with them. I will trade Jim some Oscons for some and report on the results one of these days. He has a faster scope so I will have to leave the digital supply noise comparisons to him. He also has an even bigger collection of exotic film caps than I do and will probably frisk me when leaving the next time I go over there since I mentioned this.
Looking at ESR for the Mallory THF caps shows typical ESR values in the 50 to 100 milliohm range. These are pretty close to the range of values for the Oscon caps. be careful reading ESL numbers because trace inductance in the PCB can easily be several times that number.
The origional Sullzer Regulator artical used tantalums but it seems that few audio designs use them today. Pricing is part of the issue I am sure, but I see few of them in even high budget designs. It seems I have heard some talk of a different sonic signature but haven't played much with them. I will trade Jim some Oscons for some and report on the results one of these days. He has a faster scope so I will have to leave the digital supply noise comparisons to him. He also has an even bigger collection of exotic film caps than I do and will probably frisk me when leaving the next time I go over there since I mentioned this.
fdegrove said:Hi,
Almost any AN on regulators will tell you you'll need to use cap(s) with some ESR or you risk having a nice oscillator.
Naturally it's much more fun spouting wrong info than reading application notes...
Cheers,
Naturally, it is much more fun making statements without any basis,
a more rigorous method is to take a look at those application notes and say something like "of the X application notes I have looked at, Y advise against the use of low ESR.
As a starter, I would offer that National has to say in their LM317 datasheet about the output cap:
"Although LM317 is stable with no output capacitance, like any feedback circuit, certain values of external capacitance can cause excessive ringing. This occurs with values between 500pf and 5000pf. A 1uf solid tantalum (or 25uf aluminum electrolytic) on the output swamps this effect and insures stability. An increase of the load capacitance larger than 10uf will merely improve the loop stability and output impedance".
it earlier had explained why a 25uf aluminum electrolytic is equivalent to a 1uf: because of the tantalum's low impedance at high frequencies.
Of course, what those National engineers know about their products?
millwood said:[snip]a more rigorous method is to take a look at those application notes and say something like "of the X application notes I have looked at, Y advise against the use of low ESR.
As a starter, I would offer that National has to say in their LM317 datasheet about the output cap:
"Although LM317 is stable with no output capacitance, like any feedback circuit, certain values of external capacitance can cause excessive ringing. This occurs with values between 500pf and 5000pf. A 1uf solid tantalum (or 25uf aluminum electrolytic) on the output swamps this effect and insures stability. An increase of the load capacitance larger than 10uf will merely improve the loop stability and output impedance".
it earlier had explained why a 25uf aluminum electrolytic is equivalent to a 1uf: because of the tantalum's low impedance at high frequencies.
Of course, what those National engineers know about their products?
Well, Millwood, I noiticed that your quotes, although you give the impression that they support your idea that ESR is not needed, don't address ESR. So, I'm at a loss what this is supposed to mean.
Since you seem to have unbounded trust in Application Engineers, how about this:
http://www.analog.com/UploadedFiles...541217020958034433123764LDO_BrochureFINAL.pdf
I just cut out the first mention of ESR in the attached jpg. Have fun.
Jan Didden
Attachments
janneman said:
Jan, you couldn't have been more wrong in your impression of my position on ESR. You are advised to read earlier posts of mine where I clearly stated my position.
My attempt at presenting the National paragraph is to highlight the fact that we shall not rap scientific theories into supporting our conclusions (aka the high esr/ esl combination).
Not to mention a certain statement made earlier by someone with regards to the number of application notes on low ESR caps is not supported by National.
I could not get why are you all mill the wind with ESR.
We need a zero in loop gain, and we get it either by finite ESR or by other means.
From AN-1148: "National Semiconductor does have LDO regulators like the LP2985 and LP2989 which are specifically designed to work with extremely low ESR capacitors like surface-mount ceramics. This type of capacitor can have ESR values as low
as 5-10 m, which will cause most typical LDO regulators to oscillate.
To make the LP2985 stable with such low ESR values, an internal zero is built in which takes the place of the ESR zero previously provided by the Tantalum output capacitor. The effect of this is to shift the stable ESR range downward. A typical LDO with no added internal zero might be stable ESR range from about 100 m to 5 (well suited for Tantalums but not ceramics). The stable range for the LP2985 extends down to 3 m, and has an upper limit of about 500 m so it can be used with ceramics."
We need a zero in loop gain, and we get it either by finite ESR or by other means.
From AN-1148: "National Semiconductor does have LDO regulators like the LP2985 and LP2989 which are specifically designed to work with extremely low ESR capacitors like surface-mount ceramics. This type of capacitor can have ESR values as low
as 5-10 m, which will cause most typical LDO regulators to oscillate.
To make the LP2985 stable with such low ESR values, an internal zero is built in which takes the place of the ESR zero previously provided by the Tantalum output capacitor. The effect of this is to shift the stable ESR range downward. A typical LDO with no added internal zero might be stable ESR range from about 100 m to 5 (well suited for Tantalums but not ceramics). The stable range for the LP2985 extends down to 3 m, and has an upper limit of about 500 m so it can be used with ceramics."
The app note references being quoted seem to all apply to the monolithic low dropout regulator, where the pass transistor is operated in common emitter mode and is a lateral PNP for the positive regulator case. This is really a somewhat different problem than the super reg topology. For the low dropout case, the open loop output impedance without the load capacitor is on the order of tens of kOhms, while for the super reg topology the open loop output impedance without the capacitor is on the order of tenths of an Ohm. Also the low dropout regs have a second pole at a few hundred kHz due to the lateral PNP, while the super reg uses a pass transistor deliberately chosen for high ft, avoiding this issue. The analysis of the low dropout regulators seems to neglect the series inductance of the load capacitor as well. The OnSemi app note quoted earlier shows in figure 5 a unity loop gain frequency of 100 kHz or so. We're looking at numbers more like 1-10 MHz for the super reg topology. So whille many similar considerations apply between the two topologies with respect to the load capacitor, it would be a mistake, I think, to make a blanket assumption that what works well for one will automatically work well for the other. We have a factor of 10-100 difference in bandwidth between low dropout and super reg to consider, for one.
OS-CON
Fred,
I looked at the spec's on the Sayno OS-CON cap's, look's good. I'll have to try some and compair the performance to the Sprague 550D and the Mallory THF caps.
http://www.secc.co.jp/english/index.html
Fred,
I looked at the spec's on the Sayno OS-CON cap's, look's good. I'll have to try some and compair the performance to the Sprague 550D and the Mallory THF caps.
http://www.secc.co.jp/english/index.html
>it would be a mistake, I think, to make a blanket assumption that what works well for one will automatically work well for the other
both, LDO reg and super reg, have three poles
dominant pole associated with error amp
power pole associated with the power transistor and drive circuitry,
load pole associated with Rout and Cload
>while the super reg uses a pass transistor deliberately chosen for high ft, avoiding this issue
it will be in the range 1-10 MHz
>while for the super reg topology the open loop output impedance without the capacitor is on the order of tenths of an Ohm
0.1 Ohm and 10 uF = 160 kHz
both, LDO reg and super reg, have three poles
dominant pole associated with error amp
power pole associated with the power transistor and drive circuitry,
load pole associated with Rout and Cload
>while the super reg uses a pass transistor deliberately chosen for high ft, avoiding this issue
it will be in the range 1-10 MHz
>while for the super reg topology the open loop output impedance without the capacitor is on the order of tenths of an Ohm
0.1 Ohm and 10 uF = 160 kHz
AD797 ouch
Andy,
After getting the DAC up and running, it has six super regulators I found the issues with the AD797.
1. The regulator did not oscillate with out a load.
2. With the load 3 out of the 6 regulators oscillated. Sockets are great. The oscillation was somewhere around 10Mhz and about 60 mv, ouch riding on the DC.
The cap I used are dipped run of the mill tantalums not the low ESR flavor.
3. As the data sheet states the AD797 the phase will approach drop passes 0 degrees phase margin at gain of less than about 18. It interesting that half the AD797 did not oscillate.
So dropped in some AD825’s and some OP27’s to see how they did, of neither one broke into oscillation.
Andy,
After getting the DAC up and running, it has six super regulators I found the issues with the AD797.
1. The regulator did not oscillate with out a load.
2. With the load 3 out of the 6 regulators oscillated. Sockets are great. The oscillation was somewhere around 10Mhz and about 60 mv, ouch riding on the DC.
The cap I used are dipped run of the mill tantalums not the low ESR flavor.
3. As the data sheet states the AD797 the phase will approach drop passes 0 degrees phase margin at gain of less than about 18. It interesting that half the AD797 did not oscillate.
So dropped in some AD825’s and some OP27’s to see how they did, of neither one broke into oscillation.
Re: AD797 ouch
Having a 100 MHz gain-bandwidth product in the op-amp is a tough one! I think the ~30 MHz GBP of the AD825 is more reasonable. I'm doing this from memory, but I seem to recall that the phase characteristics of the AD825 were very civilized at high frequencies. In my sims of a high-voltage incarnation of the super reg, the AD825 had the best transient performance of anything I tried. I wanted to get max line rejection of 120 Hz, so I ended up with an OP-37 with the feedback loop AC gain set to 5. Transient response wasn't as good as the AD825 but the line rejection at 120 Hz was 40 dB better than the AD825 due to the increased low-frequency loop gain.
Good luck with the project!
jewilson said:
Having a 100 MHz gain-bandwidth product in the op-amp is a tough one! I think the ~30 MHz GBP of the AD825 is more reasonable. I'm doing this from memory, but I seem to recall that the phase characteristics of the AD825 were very civilized at high frequencies. In my sims of a high-voltage incarnation of the super reg, the AD825 had the best transient performance of anything I tried. I wanted to get max line rejection of 120 Hz, so I ended up with an OP-37 with the feedback loop AC gain set to 5. Transient response wasn't as good as the AD825 but the line rejection at 120 Hz was 40 dB better than the AD825 due to the increased low-frequency loop gain.
Good luck with the project!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.