Hi,
What is it about SMPSs that make them unsuitable for capacitive loads.
Similarly what makes the older trafo based linear power supply more suitable for it?
Thanks and Regards,
WA
What is it about SMPSs that make them unsuitable for capacitive loads.
Similarly what makes the older trafo based linear power supply more suitable for it?
Thanks and Regards,
WA
Well, they (SMPS) have at least a primary current limit that trips in a short amount of time. Since you have a power supply that refreshes its output in tens of microseconds rather than milliseconds, why do you need to follow it with a huge cap bank? If the power supply is rated for the max peak current of the amplifier, local bypassing at board level should be sufficient.
They do? How so? I mean, your average SMPS has a fat capacitor at its output. Why would it have issues with additonal capacitance? It will just go into current limit until it is charged.
As for your linear PSU, adding capacitance increases peak current in your rectifier diodes. They may not like that.
As for your linear PSU, adding capacitance increases peak current in your rectifier diodes. They may not like that.
1. If it really would go into current limit and start "pumping" it can create all sorts of nasty peaks/EMI/noises at its output. Maybe it switches off completely and make substantial pauses.
2. SMPS have an optimized loop, if you alter that with additional capacitances, you basically alter the loop and again, all sorts of funny things can happen. Maybe it goes up in flames. Who knows?
I would strongly suggest consulting the datasheet before you add capacitance and/or precisely measure if it still works the same with additional capacitance.
One of the omitted/undocumented features of many an SMPS and why this is can easily be understood when one understands how these devices operate. The output caps as used by design are certainly not chosen randomly. Everyone can try out extra large value capacitance and see what happens (don't do this inside a home and take safety precautions). Adding random capacitance just does not work out with the average SMPS.
Linear PSU's also don't like large value capacitors at their outputs in general. And why would one do that? Large value capacitors are used to filter the voltage when rectified and reduce ripple voltage before the regulator.
Linear PSU's also don't like large value capacitors at their outputs in general. And why would one do that? Large value capacitors are used to filter the voltage when rectified and reduce ripple voltage before the regulator.
Last edited:
That´s not very gently put but if you want, yes.
Most of the SMPS you see around here don´t have an application note from the manufacturer; many do not even have a basic datasheet so that´s why I think it´s important to say it out loud!
Especially beginners think it´s safe to get an SMPS to avoid working with 120/240VAC and then do stupid things to their crxxp SMPS which can turn out to be just as dangerous.
A bulk cap of unlimited size can produce a short circuit of unlimited duration. SMPS are short circuit proof by shutting down after a certain short circuit duration, then try to restart, shut down again etc etc. This is called hiccup mode and absolutely common.
This way the loading capacitance is restricted, the max value often stated in the specs.
SMPS are indeed different to linear PSU in that 50/60Hz rectification takes place on the primary side and the bulk storage caps are on the primary side. Increasing caps on the secondary is counterproductive in that case as stress is increased and loop stability will be affected. The old recipes from linear PSU are NOT applicable to SMPS. Calling them crappy btw just shows the lack of understanding.
This way the loading capacitance is restricted, the max value often stated in the specs.
SMPS are indeed different to linear PSU in that 50/60Hz rectification takes place on the primary side and the bulk storage caps are on the primary side. Increasing caps on the secondary is counterproductive in that case as stress is increased and loop stability will be affected. The old recipes from linear PSU are NOT applicable to SMPS. Calling them crappy btw just shows the lack of understanding.
You are right if you consider the cheap stuff you can buy on the Asian bazaar. But if you buy quality brand stuff there is certainly a datasheet with specs.
Most DIYaudio people are not aware of a correct understanding of stability.
Regulated power supply stability depends of the output capacitor, wether it is linear or switch mode.
Messing with additional caps directly at the output can only make trouble in the design.
Caps for bypassing onboard powered devices are ok because they are not directly at the PSU output, il is assumed there is enough impedance ( wire inductance ) to sort of separate these caps.
Do not invent fancy stuff at PSU outputs, use them in the usual way, that is how they are designed.
Regulated power supply stability depends of the output capacitor, wether it is linear or switch mode.
Messing with additional caps directly at the output can only make trouble in the design.
Caps for bypassing onboard powered devices are ok because they are not directly at the PSU output, il is assumed there is enough impedance ( wire inductance ) to sort of separate these caps.
Do not invent fancy stuff at PSU outputs, use them in the usual way, that is how they are designed.
What I consider I have just written. Just read the sentence again;-)
As far as I can see that is mostly stuff from Ebay, Aliexpress & Connex.
No datasheets or specs, no website, maybe it´s not even clear who manufactures these.
Meanwell is another chunk used around here. They have datasheets, specs, are thoroughly tested and incredibly priced for what they are. No, these are not crappy even if they fail with 1000´s of µF on their outputs because that´s not the intention. I´d guess though they kinda hiccup their way through any capacitance thrown at them. But this can get noisy.
At work we use Vicor for qualification tests. These work at almost 100% load, 24/7 for many months. Clean reliable power but expensive. Also pretty hard to make unstable.
ATX computer supplies are an option. 15mV of ripple on 60A of current is impressive, as is a 12 year warranty. They are big and chunky though.
I've used those inline "brick" LED supplies for tube headphone amplifiers and phono stages. They are perfectly fine unless you buy a complete POS or expect to use it at 100% power.
I've used those inline "brick" LED supplies for tube headphone amplifiers and phono stages. They are perfectly fine unless you buy a complete POS or expect to use it at 100% power.
With big caps on output of the SMPS you will get lazy system, maybe adding more caps on load side will help, what I mean is when power supply is in one case and in another there are amp modules, it will be power bank for short "large power" request from amps.
Every SMPS work on "high" frequency and will fast deliver power to the load, so fewer capacitive load will enable more current to flow with "fat" pulse instead of more "short" pulses. Do not ask why I wrote that, but it is so. Music is not constant load and "ask" from SMPS controller to not raise the voltage when there is no need for power.
Adding some virtual load or passive load will help to minimize SMPS noise, but will increase consumption.
Every SMPS work on "high" frequency and will fast deliver power to the load, so fewer capacitive load will enable more current to flow with "fat" pulse instead of more "short" pulses. Do not ask why I wrote that, but it is so. Music is not constant load and "ask" from SMPS controller to not raise the voltage when there is no need for power.
Adding some virtual load or passive load will help to minimize SMPS noise, but will increase consumption.
It's not a SMPS issue. Linear regulators are the same. Look at the data sheet for LM317. Typical application shows 1.0uF at the output and they tell you it improves transient response but it's not needed.
A regulator of any kind cannot regulate with an arbitrarily large reservoir capacitor at the output. For the obvious reasons.
A regulator of any kind cannot regulate with an arbitrarily large reservoir capacitor at the output. For the obvious reasons.
I agree from the stability standpoint. The output cap of a switcher directly influences the feedback loop and it's not necessarily in a "more is better" manner. More like it is what it needs to be, for the specific design's phase margin to instability.
Of course the designer realizes it's going to be hooked to something, so they allow for additional capacitance of the device it's going to be powering. But they're not expecting to power Raspberry Pi Super Capacitor multi Farad levels of audiophoolery... More like what a normal PC motherboard might add additionally.
Now these SMPS on fleabay that are intended to power LED strings? Who knows where they sit in terms of a well done feedback control loop. For all anyone knows, as long as the lights appear steady - good enuff! Could be oscillating at a frequency you cant see.
The only way to know for sure is to test yourself. If you have a function generator with offset and amplitude controls, you can make one with a N channel FET and a non-inductive resistor, driven directly - should be OK for audio transients. You'll need to provide a minimum load to the PSU, so the FET only handles the transient current. And an oscilloscope to see what's happening -
That minimum load effects stability also. Too little and the loop cant regulate. SMPS Manufacturers dont like to include that on board, because the heat has to go somewhere and it makes efficiency look bad. So they depend on the application to provide it. 10% is common, so what happens when your class D goes quiet powered by a 20A SMPS?
An undocumented, knock-off, dirt cheap 20A SMPS? Probably doesnt have a minimum load built in, like a SMPS bench supply might -
Of course the designer realizes it's going to be hooked to something, so they allow for additional capacitance of the device it's going to be powering. But they're not expecting to power Raspberry Pi Super Capacitor multi Farad levels of audiophoolery... More like what a normal PC motherboard might add additionally.
Now these SMPS on fleabay that are intended to power LED strings? Who knows where they sit in terms of a well done feedback control loop. For all anyone knows, as long as the lights appear steady - good enuff! Could be oscillating at a frequency you cant see.
The only way to know for sure is to test yourself. If you have a function generator with offset and amplitude controls, you can make one with a N channel FET and a non-inductive resistor, driven directly - should be OK for audio transients. You'll need to provide a minimum load to the PSU, so the FET only handles the transient current. And an oscilloscope to see what's happening -
That minimum load effects stability also. Too little and the loop cant regulate. SMPS Manufacturers dont like to include that on board, because the heat has to go somewhere and it makes efficiency look bad. So they depend on the application to provide it. 10% is common, so what happens when your class D goes quiet powered by a 20A SMPS?
An undocumented, knock-off, dirt cheap 20A SMPS? Probably doesnt have a minimum load built in, like a SMPS bench supply might -
Last edited:
Finally after 14 posts there is a competent response regarding small signal stability and how adding anything but a resistive load may affect small signal regulation.
The other aspect is large signal disturbances, such as at power on, where many base level converters will protect themselves if the output extends in to current limit (as per a significant capacitance load or a cold filament type load).
Both of those performance aspects don't make the smps crap - it is the purchaser who naively bought the smps and expected it to do certain things that has the crap attitude when the smps does something they didn't anticipate.
The other aspect is large signal disturbances, such as at power on, where many base level converters will protect themselves if the output extends in to current limit (as per a significant capacitance load or a cold filament type load).
Both of those performance aspects don't make the smps crap - it is the purchaser who naively bought the smps and expected it to do certain things that has the crap attitude when the smps does something they didn't anticipate.
Not all smps are regulated. Hypex used unregulated, the output is proportional to the mains voltage.
These behave similar to linear supplies, and capacitive loading is possible
These behave similar to linear supplies, and capacitive loading is possible
Bansuri, you may want to review your comment about Hypex, and describe in detail what is and isn't regulated with such a converter, and also what happens for output over-current conditions.