Switching Power supply

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
What about a switching power supply for a LM3886 amplifier?
My idea is to make a switching power supply for two mono amplifiers based on LM3886 chip but the power supply is made in a different box from the two amplifiers.
How many possibilites I have to do a great job if I think to use a good output filter for power supply's output voltage and big capacitors for amplifier's input supply voltage?
I want to connect the two amplifiers to power supply with a shilded cable to avoid trasmission of high frequency noises and a big filter in the power supply to avoid trasmission of noises below 100kHz.
I know that it's probably a well-known stupid idea :clown: but I want to know if someone have tried this solution and eventually results obtained.
Thanks

Giuliano
 
it's been discussed on different and varying threads -- but DONT use big caps on the output of the SMPS -- a lot of heat is generated in these caps and usually 200uF or 330 uF are adequate. ripple is one thing, switching transients quite something different.

the output choke design is critical -- the value is inversely related to the current drawn -- your PC draws a relatively constant current from the SMPS, but in an amplifier the current will vary depending on the biasing. an smps has to be loaded at all times.
 
jackinnj said:
...an smps has to be loaded at all times.

Why did you say so?
In my opinion SMPS is able to work fine with light and heavy load.
Thanks for suggestions, infact probably I will use tantalum capacitor + MKT capacitors for SMPS output voltage to eliminate the ripple at high frequency in addiction to an big electrolitic capacitor (that has the function to be the bulk capacitor).
Thanks

Giuliano
 
allanon77 said:


Why did you say so?
In my opinion SMPS is able to work fine with light and heavy load.
Thanks for suggestions, infact probably I will use tantalum capacitor + MKT capacitors for SMPS output voltage to eliminate the ripple at high frequency in addiction to an big electrolitic capacitor (that has the function to be the bulk capacitor).
Thanks

Giuliano

because most of the energy in an SMPS is stored in the inductor --

is there need for a reservoir capacitor -- a SMPS has poor transient response compared to a linear supply -- you might find it better, although less efficient, to use a high quality, discrete linear post-regulator.
 
Your project is perfectly achievable but your appear to have very little knowledge and no SMPS experience

You will have to do some smaller and simpler SMPS projects in order to learn and gain the required experience. You will also require a >10Mhz dual oscilloscope and some home-made equipment to test transformer and inductor saturation characteristics

The output filter is usually just a second order low-pass filter. Its transfer characteristic depends on capacitance, capacitor ESR, inductance and inductor Rdc. Capacitance may be chosen to be small to reduce size and cost or to be huge to reduce output peak current requirements, but the value of the inductor and its Rdc have to be chosen to match the capacitor and get little overshoot [use a simulator to learn how each parameter affects the overall behavior of the filter]

Capacitors should be rated in excess of the ripple current requrements [you have to learn how to calculate it]. Also, inductor saturation current should be rated in excess of the peak current requirement [again you have to learn how to calculate it and how to design inductors]

Then, the entire system, including the 2nd order roll-off is enclosed into a feedback loop controlled by an op-amp, and you have to learn how to frequency-compensate this feedback loop for stability [the more peaky the filter, the more compensation required]

There are also more complex control techniques involving load current sensing and two nested control loops [op-amps] each with its own frequency compensation to make it stable [another subject to learn about]

Also you have to learn about transformer design and about optimizing switching device behavior to make it ringing-free. And proper PCB routing is like an art...

As you can see, there are several subjects involved in designing such an SMPS. Try simpler projects first
 
Doesn't that depend entirely on the construction of the SMPS? If it's built as a comparator type, ie you compare the regulated and filtered output to a reference voltage to control the switching instead of a fixed clock cycle, that should give excellent transient response as well as no fixed clock.
 
AGGEMAM said:
Doesn't that depend entirely on the construction of the SMPS? If it's built as a comparator type, ie you compare the regulated and filtered output to a reference voltage to control the switching instead of a fixed clock cycle, that should give excellent transient response as well as no fixed clock.

It takes several cycles for the loop to stabilize itself -- so the problem can long be in the rearview mirror before it is fixed (or the loop can go into a condition of never stabilizing -- as when you use a converter for a purpose for which it wasn't intended) -- the linear regulator burns the energy in about 1/10th the time (according to something I read from National Semi -- will dig up the source somewhere.)
 
'Poor transient responde in SMPS' is just a myth created from a lack of knowledge and by people too lazy to learn to analyze the actual circuit behavior

SMPS just behave as 2nd order filters driven from a programmable current source whose output depends both on input voltage and on the driving signal from the error amplifier [the exact relationship is topology-dependent]

There is actually a crossover frequency, at wich the control loop action is reducing the actual output impedance of the 2nd filter to half the expected value. Above the crossover frequency the impedance of the capacitor dominates, below the crossover frequency the action of the control loop dominates. This crossover frequency is a design parameter free to play with, it depends on filter design and error-amp compensation

So all SMPS at high frequencies behave just like the output filter capacitors driven from a 'slow' current source and show the same HF transient response as these capacitors

Typical crossover frequencies may range from 50Hz to 5Khz [strongly depending on output filter size]

And to correct jackinnj statements :

Major energy storage happens *allways* on capacitors. SMPS filter inductors usually store 5 to 20 times less energy than its associated filter capacitors. Also, maximum inductor energy storage happens only at full load [E=.5*L*I^2], while capacitor energy storage is *constant* [E=.5*C*V^2]. This figures out to be pretty obvious when you design some output filters and inductors. Four 2200uF 25V capacitors charged to 15V use little space and store 1 Joule. How big is an inductor capable of storing 1 Joule or more energy?

Answer : Bigger than the entire SMPS, maybe bigger than two or three SMPSs each with 1 Joule energy storage on output capacitors. To better understand think that the usual 5cm diameter yellow-white iron-powder toroid core stores only 8 mili Joules before saturation
 
Eva said:
Your project is perfectly achievable but your appear to have very little knowledge and no SMPS experience

Sorry but why do you think so? because I don't talk about every SMPS's rules ... ;) ... thank you for your suggestion I will talk with my boss about my lacks ... ;)
I'm a electronic designer, young, but however I designed some SMPS in the last years.

Anyway ... I thank you for everything you said. You gave many useful information for new designers.

My dubts are about the results of the use of an well-designed SMPS with an audio chip to obtain, finally, a good audio result.
I haven't big problems with SMPS design but every audiophile I heard say the same thing: "the only supply for audio amplifiers is bridge+huge capacitors". So I thought: "Is it a problem of poor SMPS designers or it's really impossible to obtain great results with SMPS for audio amplifiers?
I'm interested with some personal experience of some audiophiles in SMPS supply for good audio amplifier.
Thanks to all.

P.S. I agree with you, I think that it's possible to design a SMPS with good transient response; my dubts are about the possibility to have a low noise (comparable to "bridge+huge capacitors").

Giuliano
 
allanon77 said:


Sorry but why do you think so? because I don't talk about every SMPS's rules ... ;) ... thank you for your suggestion I will talk with my boss about my lacks ... ;)
I'm a electronic designer, young, but however I designed some SMPS in the last years.

Anyway ... I thank you for everything you said. You gave many useful information for new designers....
etc

X Eva
Sorry for my irony! :)
after my post I read other your posts and I have to say that you effectly know much about SMPS (more than I ;) ).
My question is: have you ever designed a SMPS for audio application? have you measured the noise ... oops ... ripple :D at high frequency, with what results?
Do you agree with my idea to make a SMPS in a different box with a shielded cable to supply the amplifiers?
And I read (other post) that you avoid paralleling of different capacitors because that causes ringing oscillation, but if I place between the capacitors a small resistor probably this could be a solution at the problem; secondly probably this kind of oscillation have a high frequency content (MHz) because I suppose they are generated by parasitic inductance and the capacitances so a further reduction of these oscillation could be the shielded cable (I talked before).
I thank you in advance.
Regards

Giuliano
 
jackinnj, I am talking about a clockless or self oscellating circuit so speaking of delay cycles is in principle not the case. However you do have a point but it entirely depends on the design. You have to have a feedback loop that is several times faster than the switching circuit for one. The real problem is the floating equivalent switching frequency in the MHz range which makes use of transformers pretty much useless.

allanon77, though long cables will decrease stray magnetic fields in the PS and the amp from influencing eachother. It also increases the radio transmitted ripple and if the cable is long enough, this might affect all your household appliances. Sure you can use shielded cable but no shielding is perfect.
 
blasteriz said:
Now I am designing amplifier with linear power supply, because SMPS makes too much noise in supply and everywhere else.

I thought that Eva had made the point quite obvious that noise in SMPS's is something you can design your way out off. But it's not only in the schematic design but even more so in the PCB design. And the latter being a more trial and error part or sheer luck than anything else really.
 
Noise is for those who have no interest in understanding what happens inside a SMPS [or for those that talk a lot about signals they have never seen on an oscilloscope]

'Noise' is not the right term for SMPS since it means a 'random not-deterministic signal' and SMPS allways output and radiate deterministic and periodic parasitistic signals

Those signals are the sum of a main ripple component [a sqare or triangle wave at the switching frequency] and some parasitistic ringing components [at a few Mhz] caused by RLC parasitistic resonances between windings, diodes, transistors and PCB tracks

Ripple by itself is easily filtered and happens at too low frequencies to be sustantially radiated

Ringing is harder to filter because it is easily radiated due to its higher frequency nature. Also, ringing versus efficiency is a compromise since ringing may be reduced or killed just by slowing down switching transients [more switching losses], increasing PCB and transformer DCR [to damp parasitistic RLC at the expense of more dissipation], using inherently slower high voltage switching devices like bipolar transistors or IGBTs [not prone to ringing as opposed to MOSFETs], etc...

You may build a working and reliable SMPS but, how much does it ring at switching transients?. Taming ringing requires a lot of trial and error
 
want to reduce noise -- try some of the slew-controlled chips from Linear Technology --

the cheap fix is to use a linear post-regulator -- Linear has an application note combining the LT3439 switcher chip with the LT1761 and LT1964 ultra-low noise linear regulators.

i have some Sorensen telcom switching supplies which have "noise" in the uVs -- totally discrete (and they are ancient!)
 
Eva said:
Noise is for those who have no interest in understanding what happens inside a SMPS [or for those that talk a lot about signals they have never seen on an oscilloscope]

'Noise' is not the right term for SMPS since it means a 'random not-deterministic signal' and SMPS allways output and radiate deterministic and periodic parasitistic signals

Those signals are the sum of a main ripple component [a sqare or triangle wave at the switching frequency] and some parasitistic ringing components [at a few Mhz] caused by RLC parasitistic resonances between windings, diodes, transistors and PCB tracks

I agree with your assertions but probably "noise" is an uncorrect but simple way to understand what you are talking about: "unwanted signals". But it's true that if you talk about "noise" could be source of misunderstandings because you put toghether different kind of signals with different characteristics (and wirth different way to limit). However normally when I say "noise" I know what I am talking about.


Ripple by itself is easily filtered and happens at too low frequencies to be sustantially radiated


Yes but probably for audio application it's the more important to limit ... isn't it true?


Ringing is harder to filter because it is easily radiated due to its higher frequency nature. Also, ringing versus efficiency is a compromise since ringing may be reduced or killed just by slowing down switching transients [more switching losses], increasing PCB and transformer DCR [to damp parasitistic RLC at the expense of more dissipation], using inherently slower high voltage switching devices like bipolar transistors or IGBTs [not prone to ringing as opposed to MOSFETs], etc...

You may build a working and reliable SMPS but, how much does it ring at switching transients?. Taming ringing requires a lot of trial and error


But if you use BJT you don't have the possibility to use more than one component for a switch, this is not a big limitation (for dissipation, size, etc...).
In the past I studied a new design of a Boost PFC (1kW) and I choosed the power switch (frequency choosed 50kHz output voltage 400V, wide range application); At the end I had to choose between 2XIRFB20N50K e 2XIRGB20B60 so I estimated the power losses of the switches (I had some limitations: case not bigger than TO-220 and dissipator with Rth=1.7C/W).
The winner was: IRFB20N50K. What do you think about? Did I made some mistakes?
Thank a lot

Giuliano


X Jackinnj

Yes your solution to use at the end a linear regulator, like what you talked about, could work but probably not for audio amplifier with some watts (that is my case). Thanks.
 
allanon77 said:

X Jackinnj

Yes your solution to use at the end a linear regulator, like what you talked about, could work but probably not for audio amplifier with some watts (that is my case). Thanks.

yes, the linear regulator burns watts. but DIYr's don't have to advertise the fact that their products are "green".

for the moment, Linear's site seems to be down, but I note that they do have a low noise gate driver, and that the 1A chip like the 3439 can be used to drive other devices. it is really worthwhile to take a look at the PDF or the Application Notes for these devices.

the LT3439 has come down in price -- $7 each -- pretty stemmy compared to a TL494 or SG3525 -- but you probably save in terms the magnetics and snubbers, if you want a low noise power supply.

one thing which both National and Linear take great effort to point out is that some of the current pathway's conduct a lot more than one would suspect (and for very brief periods of time) -- there are nanohenries everywhere!
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.