Diode swapping: ordinary diodes Vs. schotky

[gootee]

I think is more about the PCB trace series inductance. The electrolithyc capacitors have little inductance compared to those long PCB (or straight wires). That internal inductance is the ESL (series with ESR) and is given by the leads mainly.
Once you have looong PCB between the capacitor combo and the fast IC (consumer), the series inductance will "isolate" the PS capacitor from the IC (in HF domain). There will be no ringing at the source, because fast transients cannot get from consumer all the way back there. That's why, like you said too, for fast IC's, the local capacitor is important.

I was going to also mention the trace inductance as a possible resonance component, but was in a hurry and a little lazy. However, the amount of inductance is only a factor in the resonant frequency, not in whether or not something will be resonant. So having MORE total inductance doesn't favor the traces or wires over the electrolytic. The traces or wires have ESR and ESL, both of which are incrementally distributed along them. The ESR of the traces or wires is relatively constant vs frequency and will tend to damp any resonance, whereas the electrolytic's ESR basically vanishes at high frequencies. And it is very close to and in parallel with the low-ESR film cap in question. So they could resonate easily and well. And they don't necessarily need to be near the load, or anything else, to be excited well-enough to ring.

So I'd guess that both the trace inductance and the electrolytic's inductance could ring with a film cap, especially with heavy excitation. But as I explained above, the electrolytic will probably make a more-effective resonant circuit (higher Q).

I do agree that electrolytic (and basically all other) capacitors don't have much if any internal inductance except due to the distance between their leads and their lead lengths. (By the way, for those who believe that film caps must have high inductance because they are basically rolled-up foil, they usually are not that inductive, because one entire "side" of each roll is usually connected, not just the end of the roll.)

 

[gootee]

Oh, also, after doing a little more reading, I find that high-spreed rectifiers are much less noisy than low-speed ones, in terms of spewing RF. But the high-speed ones should also be the "soft recovery" type.

 

[RCruz]

Most of what you're doing here sounds more-or-less OK, as in at least not necessarily harmful and maybe even helpful. But be advised that paralleling an electrolytic with a very-low-loss type of a much smaller cap, such as any film type or an NPO/C0G ceramic, especially with less than say 1% of the electrolytic's capacitance (but could still happen with more), could form a resonance (in combination with the parasitic inductance of the electrolytic), which could excite high-frequency ringing, which is "a bad thing".

You could put 0.1 Ohms or more in series with the small cap, to create a snubber, but that's not really needed unless there's already some ringing going on and would also probably defeat the original intent. Or you might get lucky. Just be aware that it could also cause problems.

If you have the space available, a far better improvement possibility would probably be a large polypropylene film cap, having at least 1% of the capacitance of all the existing paralleled electrolytics combined.

Also, ultra-fast diodes will tend to create some high-frequency ringing, too. I don't see how they could help and would probably prefer going the other way, to soft-recovery types. The big caps are what should be supplying the power to the system, not the rectifiers. You want the rectification effects to be invisible, if possible. The caps are there to smooth their effects so making faster pulses there doesn't make sense. I could be wrong but that's how I see it at first glance.

And note that Schottkys usually have about half as much, or maybe less, for their forward voltage drop. So they're probably mainly just increasing your power supply rail voltages a little bit.

Putting a snubber network across each of your rectifier diodes could be a good improvement. The essential part of a snubber is the resistor. A small cap is put in series just so the resistor doesn't have to dissipate so much power, from the "normal" frequencies that are meant to be there. A snubber is meant to quell any high-frequency ringing. They can also be useful in terminating digital lines, etc.

OK, I'll tell you one secret: As far as the power supply quality is concerned, the main smoothing caps, by the rectifiers, are probably less important, for tweaking, than the decoupling caps that should be at each point of load (i.e. at each active device's power/gnd pins (or sometimes between their + and - rails). Those caps are the ones that supply the fast transient demands for current, and are also the ones more designers probably get wrong.

More decoupling capacitance will usually buy you both more-accurate or faster transient response, and, less disturbance of the power supply rail voltage by the transient current demands (which generate nasty voltages if they have to be pulled through the parasitic inductance and resistance of the power supply rail conductors, which usually can't be done fast-enough anyway).

BUT, note that a capacitor that is physically larger will also have more intrinsic parasitic inductance, which you don't want there. So maybe the best way, especially for electrolytics,would be to parallel two or more smaller caps that add up to the capacitance you want. That lowers both the parasitic inductance (ESL) and the parasitic resistance (ESR). It can be difficult, physically, since you also don't want long leads on them. Adding one or two in parallel but soldered on the bottom side of a board might be easy, in some cases. BUT, if you can use more capacitance but still keep the same case size, or even just the same lead spacing, then I'd go with doing it that way.

You still have to be careful to not add low-loss (e.g. film) types of small capacitances indiscriminately, next to electrolytics, for decoupling, since they could cause unwanted high-frequency resonances. If you have a good oscilloscope, though, you could know pretty quickly if a particular value was OK. But, basically, the physical size determines the speed, not the capacitance (speed is different than resonant frequency). So probably use the most capacitance you can, for a given case size that will fit (but smaller case size is better, too), if you're doing non-electrolytics. Where higher currents are possible, add an electrolytic or two in parallel with (or to replace) whatever electrolytic is already there. In all cases, having the absolute minimum physical connection length is critical, so no axial-lead electrolytics are allowed, for example (and paralleling smaller caps is usually better than using one big one).

One possible problem would be for a fast/digital device where the original designer was very good and took all of the parasitics into account and used exactly the right decoupling caps. Changing them might undo some good design work. On the other hand, if they didn't tweak the decoupling caps properly to begin with, and you can make the LC resonances (as seen by the power pins) match the clock frequency and possibly also the frequency corresponding to the risetime the pin wants the current to have, you might dramatically lower the noise on the power rails, while also letting the chip take perfect-mouthful-sized bites of current. (You'd probably need a good oscilloscope, at least, to do that, though.) You can usually tell if they took the trouble to tweak them by looking at the values. If they just used 0.1 uF or .01 uF, possibly with a 10 uF electrolytic, then they were probably just guessing.

Have fun.

Tom

Thank you sir.

 

[cbdb]

There is very little accurate info in this thread on diodes!

Schottkys are for fast turn off (not on), which is only useful in circuits that switch the diode on and off continuously (like in PS charging ). Most other diodes are on all the time or off until they need to protect something. Replacing these is either useless, or a bad idea. (or a really bad idea if they are used to set bias in current sources or output stages).You must look at each diode in circuit to see its function. This typical wholesale swapping of components without the knowledge of what they actually do is a bad idea, but still rampant in these forums. And its more proof that people hear what they want to.

PS Power supply cap charging pulses are not fast, they are 60hz (as the voltage rises above the cap voltage (at a 60hz speed) the diode slowly turns on and then follows the 60 hz till it starts to drop again). This is a very common misconception that causes all kinds of nonexistant concerns.

 

[gootee]

There is very little accurate info in this thread on diodes!

Schottkys are for fast turn off (not on), which is only useful in circuits that switch the diode on and off continuously (like in PS charging ). Most other diodes are on all the time or off until they need to protect something. Replacing these is either useless, or a bad idea. (or a really bad idea if they are used to set bias in current sources or output stages).You must look at each diode in circuit to see its function. This typical wholesale swapping of components without the knowledge of what they actually do is a bad idea, but still rampant in these forums. And its more proof that people hear what they want to.

PS Power supply cap charging pulses are not fast, they are 60hz (as the voltage rises above the cap voltage (at a 60hz speed) the diode slowly turns on and then follows the 60 hz till it starts to drop again). This is a very common misconception that causes all kinds of nonexistant concerns.

I believe it is you who has the misconception.

Power supply charging current pulses are not 60 Hz, nor "60 Hz speed", and they are relatively fast, in terms of rise and fall time, and are also of relatively short duration. Their repetiton rate is 120 Hz, with full-wave rectification of 60 Hz mains. But they are not the same shape as the rectified sine. They do not "follow" its shape. They are not sinusoidal. When the rectified sine's voltage rises above the smoothing caps' ripple voltage by enough, the diode suddenly turns on and the amplitude of the current rises very quickly from zero to a maximum (which depends on capacitor size and load current and parasitics) as current rushes into the low impedance of the smoothing capacitors to charge them and the caps' voltage rises until the diode suddenly turns off and the current falls very quickly back to zero, which typically all happens within 2 or 3 milliseconds, unless power factor correction circuitry is used. See Figure 9 at Unregulated Power Supply Design .

 

[magnoman]

I think that Schottky diodes are fast in turn off (not having stored charge the circuit needs only to act on the diodes capacitance).
I wouldn't expect much of this to matter for the still rather slow pulses from a 60Hz source.

Hope this helps
-Antonio

 

[tvrgeek]

I used Hexafred diodes in a power amp. It improved the 60 Hz by 8 dB. BUT, this is an unregulated supply. It is hard for me to think the rectifiers could make much difference with a low power regulated supply. It's cheap, so give it a try.

 

[underwurlde]

BUT, note that a capacitor that is physically larger will also have more intrinsic parasitic inductance, which you don't want there. So maybe the best way, especially for electrolytics,would be to parallel two or more smaller caps that add up to the capacitance you want. That lowers both the parasitic inductance (ESL) and the parasitic resistance (ESR). It can be difficult, physically, since you also don't want long leads on them. Adding one or two in parallel but soldered on the bottom side of a board might be easy, in some cases.

My thinking exactly (a whole bunch of smaller capacitors such as Rubycon ZLs bunched together - PCB layout consisting of a top solid plane to (+) and bottom layer solid plane to (-), but something I keep quiet about for fear of ridicule especially when I see the likes of Naim and their HiCap power supplies:

An externally hosted image should be here but it was not working when we last tested it.

It just seems counter intuitive to my thinking - super-duper big massively expensive capacitors and then... long leads wired to them? Hell, who am I to argue with Naim though. Back to being quiet I think.

A

 

[stephensank]

I used Hexafred diodes in a power amp. It improved the 60 Hz by 8 dB. BUT, this is an unregulated supply. It is hard for me to think the rectifiers could make much difference with a low power regulated supply. It's cheap, so give it a try.

You should not have seen a reduction in 60 or 120Hz noise level going to soft recovery diodes(hexfreds, etc.), unless your original rectification was almost impossibly bad. As has been described above, the point of using soft recoveries is about *not adding* any high frequency noise to the supply. Standard silicon diodes have a lot of ringing as they stop & start conducting with each wave cycle, generating HF noise, which even the best electrolytic caps are not very good(i.e., NEVER 100% effective) at filtering out. So, what should have been noticeable when you changed to them, if the amp is otherwise reasonably well done, is smoother high end and perhaps "blacker" background(lack of low level grunge). This effect is actually FAR more evident in preamps & other lower level circuits.

 

[gootee]

My thinking exactly (a whole bunch of smaller capacitors such as Rubycon ZLs bunched together - PCB layout consisting of a top solid plane to (+) and bottom layer solid plane to (-), but something I keep quiet about for fear of ridicule especially when I see the likes of Naim and their HiCap power supplies:

An externally hosted image should be here but it was not working when we last tested it.

It just seems counter intuitive to my thinking - super-duper big massively expensive capacitors and then... long leads wired to them? Hell, who am I to argue with Naim though. Back to being quiet I think.

A

Your thinking is excellent but (Wow) your wiring looks gorgeous!

(If there's no hum don't wory about it but if there is you might consider flipping the extra "corners" up so that the wires stay close together for longer, to make the enclosed loop area smaller.)

 

[drowranger]

I have read that...

Oh, also, after doing a little more reading, I find that high-spreed rectifiers are much less noisy than low-speed ones, in terms of spewing RF. But the high-speed ones should also be the "soft recovery" type.

try putting capacitor film type usually .01uf or .1uf in parallel and soldered it as close as possible to the rectifiers or onto its leads.

 

[gootee]

try putting capacitor film type usually .01uf or .1uf in parallel and soldered it as close as possible to the rectifiers or onto its leads.

You can do much better than that by adding a series resistor, to make it into a snubber network.

Here is the the post with the practical method for choosing the optimal snubber component values:

http://www.diyaudio.com/forums/powe...lm-caps-electrolytic-caps-30.html#post2828689

 

[drowranger]

ok I will

You can do much better than that by adding a series resistor, to make it into a snubber network.

Here is the the post with the practical method for choosing the optimal snubber component values:

http://www.diyaudio.com/forums/powe...lm-caps-electrolytic-caps-30.html#post2828689

thanks,
joel

 

[tvrgeek]

You should not have seen a reduction in 60 or 120Hz noise level going to soft recovery diodes(hexfreds, etc.), unless your original rectification was almost impossibly bad. As has been described above, the point of using soft recoveries is about *not adding* any high frequency noise to the supply. Standard silicon diodes have a lot of ringing as they stop & start conducting with each wave cycle, generating HF noise, which even the best electrolytic caps are not very good(i.e., NEVER 100% effective) at filtering out. So, what should have been noticeable when you changed to them, if the amp is otherwise reasonably well done, is smoother high end and perhaps "blacker" background(lack of low level grunge). This effect is actually FAR more evident in preamps & other lower level circuits.

I did measure considerable overall reduction of noise on the rails. But lower 120 Hz and up as well. I guess if this is so more evident in low level stages, then it means the regulators are that poor. Believable. Hard to characterize the amp in total as the process ended in different outputs and a total change in compensation circuits. It has considerably wider BW and 4 fewer caps in the circuit.