I did electronics training in the army, and this PSU circuit has some inconsistencies.
The primary one being that these snubber circuits that consist of a cap and a resistor are designed for one thing only, that is, the suppression of high frequency ringing that is caused between the capacitance of a switched off diode, and the leakage inductance of the transformer secondary.
The resistor part of the snubber circuit is designed to dampen the ringing caused at switch off, with the capacitor chosen only to reduce power dissipation in the resistor during the rest of the sine wave.
The capacitor passes the high frequency ringing through the resistor, which then dampens the ringing to a single pulse (ideally), thus limiting the RF noise going down the line to the main filter capacitors which at this frequency, have too high an inductance to effectively filter it. This high frequency noise is will cause intermodulation and will affect SN ratio, as well as increase the chance the chip amp will go into oscillation, especially as a lot of "audiophile" circuits either reduce or eliminate completely the zobel network at the output.
This snubber network is at once and completely seperate to the issue of higher capacitance main filter caps reducing the high end response, and causing "slow" supplies which dampen high end.
The only reason this occurs is that as you go up in capacitance value, you go up in esr and inductance of the cap, which limits the peak current capability of the supply to power the amp during high current peaks, and will also lessen the damping ability of the amplifier, as the inductance of the main caps will limit their ability to sink and source current.
So far, there are two completely seperate effects in place.
1. the damping of the high frequency ringing of the transformer secondary and the capacitor that is momentarily caused in a standard silicon diode during switch off. This is an LC network, with damping resistor R, where L is the secondary, and C is the capacitance of the rectifier.
2. the inductance of the main filter caps creating a current shaping network of C and L and R, with all of these originating primarily in the capacitor itself.
Using the principle of GIGO, that is, Garbage In, Garbage Out, there are multiple steps here to go through.
One. Reduce to as high as degree as possible any RFI/EMI interference going into the circuit. This includes using an rfi filter on the mains input and shielding the transformer. Twisting the leads in and out of the transformer will also help. As well, the leads into and out of the transformer (especially in a torroid) should be perpendicular to the core and straight for at least a few inches, this helps reduce stray magnetic field pickup from the core.
Two.
Type of rectifier, two schools of thought here, standard block silicon rectifiers, or high speed schottky diodes.
The cons and pros are as follows.
Standard Bridge: Slow speed, higher voltage and current ratings then high speed diodes. Cause HF noise on switch off. Can handle brief current surges far in excess of the rating.
Schottky: High speed, lower current and voltage ratings then standard diodes. More expensive. Not to capable of high current surges.
I personally prefer bridge. Cheaper, easier to mount onto a heatsink, and the four diodes in the bridge are matched because they are one silicon assembly. Its up to you here. If you use standard bridges, you need to read the next section on snubbers.
Snubbers: Only needed with standard "slow" silicon rectifiers. On switch off, when the diode is reversed biased, there is a momentary conduction in the opposite direction as the "holes" in the silicon are filled by electrons. Once filled, the diode is then an open circuit. During this filling, the diode creates HF noise as the electrons are shunted around. Once filled, there is a barrier region between the two P and N elements of a diode. This non conductive barrier region is effectively a capacitor. Its a capacitor, directly connected to an inductor. ( the transformer)
This LC circuit has a resonant frequency that is excited by the noise of the collapsing diode barrier. This noise is carried along the DC line to the rest of the circuit, and its often a very high frequency, 400khz or above that quite easily radiates into space as much as propagates along the conductors.
So, you must SNUB this oscillation. Putting a capacitance in parallel with the transformer secondary is BAD!!! DONT DO IT.
It effectively adds to the capacitance of the diodes, increasing the C of the LC circuit. This brings down the frequency of the oscilation, possibly into the high range of the sound band.
I say again, do not put a capacitor of any value across the secondary of a transformer.
Its a LC Tank circuit begging to be energised by any noise.
The only component that needs to be put after the secondary is the diode bridge.
This is where the snubbers come in. They are simply a resistor of a low value, 10-20 ohms, in series with a capacitor, 100nf is common. The cap is NOT part of the tank circuit of the diode/transformer due to the resistor. Its only job is to allow any HF noise created by the diode its attached to to pass through the resistor, where the resistor will then dissipate it as heat. Not much, a few microwatts, you can safely use a 1/2 watt resistor.
As you can see, the R and C combination serve as the brakes on any possible oscillation caused by the diode reverse capacitance, and the inductance of the transformer.
This finishes the First part of this post, being the role of a snubber.
Now, to the aforementioned lack of the main filter caps to provide high current pulses to the amplifier.
The bigger the cap, the more inductance. The higher ESR. The less the amp has the ability to reproduce High Frequencies, which leads to the fake assumption that a smaller cap is better. Which is why some gainclone amps suggest 1000uf caps for main filter caps, or even less for the "audiophile".
Right assumption, wrong solution as most have found out that a cap of 1000uf is barely enough to provide bass, and can reduce dynamics and can also increase hum.
A small cap has lower ESR, and Lower self inductance.
The solution as shown in this thread is to NOT put a snubber network on the rear end of the PSU. All it will do is snub any HF noise thats made it down the circuit from the diode bridge.
Its a good solution, to the wrong problem, in the wrong place.
Put the snubber where it belongs. One per diode.
To solve the problem of higher inductance in bigger caps, you dont add an RC network. An rc network by definition has a specific rise and fall time. You want the quickest RISE and the quickest FALL to help provide sink and source to the rather messy load of a loudspeaker.
The simplest solution to this is to provide a low esr/low inductance cap close as possible to the chip amp power leads.
The caps that are in switch mode power supply secondaries are an ideal choice if you can find the right voltages. The closer the cap is to the amp, the less inductance there is, which means a quicker rise and fall time. Inductors RESIST current change. Capacitors RESIST voltage change. You want your amp to be supplied with current from a power supply that can easily supply 1 amp to 10 amps without a delay, and vice versa. The less inductance you have, the better. The less resistance you have the better.
This leads to the best solution, find a set of caps with the lowest esr and self inductance possible, and mount them as close to your chip amp as possible. Make the wires between the amp and your low esr cap as thick as possible, and as short as possible.
A good compromise is to add a set of smaller caps to your amp rails. Lets say you have 2 5600uf caps per rail. Add two 1000uf caps per rail. Maybe even a 470uf. The smaller the cap, the lower its inductance.
Think of it backwards.
When a chip amp looks at its power supply, it wants the power supply to be a very low impedance. When the chip wants current, it wants it NOW. not in a few microseconds, as you wait for your inductances to rise in magnetic flux to allow the current flow.
YOU WANT IT NOW.
This is the definition of a "FAST" power supply. It can go from 100 ma to 10 amps in a flash. This is quite simply achieved by following the rules of high frequency/high current design.
Thick Wires.
Short Wires.
Low inductance and low ESR capacitors.
To sum up.
Snubber networks, being the R and C combination, should go on your diodes. Close as possible, short as wires as possible.
Shield your psu. GIGO.
Low inductance. Use a set of high value capacitors to give your amp your bass, but have a set of lower value, low esr caps to provide that initial current burst while the main caps ramp up their current output.
Short Thick Wires.
Cheers.
The primary one being that these snubber circuits that consist of a cap and a resistor are designed for one thing only, that is, the suppression of high frequency ringing that is caused between the capacitance of a switched off diode, and the leakage inductance of the transformer secondary.
The resistor part of the snubber circuit is designed to dampen the ringing caused at switch off, with the capacitor chosen only to reduce power dissipation in the resistor during the rest of the sine wave.
The capacitor passes the high frequency ringing through the resistor, which then dampens the ringing to a single pulse (ideally), thus limiting the RF noise going down the line to the main filter capacitors which at this frequency, have too high an inductance to effectively filter it. This high frequency noise is will cause intermodulation and will affect SN ratio, as well as increase the chance the chip amp will go into oscillation, especially as a lot of "audiophile" circuits either reduce or eliminate completely the zobel network at the output.
This snubber network is at once and completely seperate to the issue of higher capacitance main filter caps reducing the high end response, and causing "slow" supplies which dampen high end.
The only reason this occurs is that as you go up in capacitance value, you go up in esr and inductance of the cap, which limits the peak current capability of the supply to power the amp during high current peaks, and will also lessen the damping ability of the amplifier, as the inductance of the main caps will limit their ability to sink and source current.
So far, there are two completely seperate effects in place.
1. the damping of the high frequency ringing of the transformer secondary and the capacitor that is momentarily caused in a standard silicon diode during switch off. This is an LC network, with damping resistor R, where L is the secondary, and C is the capacitance of the rectifier.
2. the inductance of the main filter caps creating a current shaping network of C and L and R, with all of these originating primarily in the capacitor itself.
Using the principle of GIGO, that is, Garbage In, Garbage Out, there are multiple steps here to go through.
One. Reduce to as high as degree as possible any RFI/EMI interference going into the circuit. This includes using an rfi filter on the mains input and shielding the transformer. Twisting the leads in and out of the transformer will also help. As well, the leads into and out of the transformer (especially in a torroid) should be perpendicular to the core and straight for at least a few inches, this helps reduce stray magnetic field pickup from the core.
Two.
Type of rectifier, two schools of thought here, standard block silicon rectifiers, or high speed schottky diodes.
The cons and pros are as follows.
Standard Bridge: Slow speed, higher voltage and current ratings then high speed diodes. Cause HF noise on switch off. Can handle brief current surges far in excess of the rating.
Schottky: High speed, lower current and voltage ratings then standard diodes. More expensive. Not to capable of high current surges.
I personally prefer bridge. Cheaper, easier to mount onto a heatsink, and the four diodes in the bridge are matched because they are one silicon assembly. Its up to you here. If you use standard bridges, you need to read the next section on snubbers.
Snubbers: Only needed with standard "slow" silicon rectifiers. On switch off, when the diode is reversed biased, there is a momentary conduction in the opposite direction as the "holes" in the silicon are filled by electrons. Once filled, the diode is then an open circuit. During this filling, the diode creates HF noise as the electrons are shunted around. Once filled, there is a barrier region between the two P and N elements of a diode. This non conductive barrier region is effectively a capacitor. Its a capacitor, directly connected to an inductor. ( the transformer)
This LC circuit has a resonant frequency that is excited by the noise of the collapsing diode barrier. This noise is carried along the DC line to the rest of the circuit, and its often a very high frequency, 400khz or above that quite easily radiates into space as much as propagates along the conductors.
So, you must SNUB this oscillation. Putting a capacitance in parallel with the transformer secondary is BAD!!! DONT DO IT.
It effectively adds to the capacitance of the diodes, increasing the C of the LC circuit. This brings down the frequency of the oscilation, possibly into the high range of the sound band.
I say again, do not put a capacitor of any value across the secondary of a transformer.
Its a LC Tank circuit begging to be energised by any noise.
The only component that needs to be put after the secondary is the diode bridge.
This is where the snubbers come in. They are simply a resistor of a low value, 10-20 ohms, in series with a capacitor, 100nf is common. The cap is NOT part of the tank circuit of the diode/transformer due to the resistor. Its only job is to allow any HF noise created by the diode its attached to to pass through the resistor, where the resistor will then dissipate it as heat. Not much, a few microwatts, you can safely use a 1/2 watt resistor.
As you can see, the R and C combination serve as the brakes on any possible oscillation caused by the diode reverse capacitance, and the inductance of the transformer.
This finishes the First part of this post, being the role of a snubber.
Now, to the aforementioned lack of the main filter caps to provide high current pulses to the amplifier.
The bigger the cap, the more inductance. The higher ESR. The less the amp has the ability to reproduce High Frequencies, which leads to the fake assumption that a smaller cap is better. Which is why some gainclone amps suggest 1000uf caps for main filter caps, or even less for the "audiophile".
Right assumption, wrong solution as most have found out that a cap of 1000uf is barely enough to provide bass, and can reduce dynamics and can also increase hum.
A small cap has lower ESR, and Lower self inductance.
The solution as shown in this thread is to NOT put a snubber network on the rear end of the PSU. All it will do is snub any HF noise thats made it down the circuit from the diode bridge.
Its a good solution, to the wrong problem, in the wrong place.
Put the snubber where it belongs. One per diode.
To solve the problem of higher inductance in bigger caps, you dont add an RC network. An rc network by definition has a specific rise and fall time. You want the quickest RISE and the quickest FALL to help provide sink and source to the rather messy load of a loudspeaker.
The simplest solution to this is to provide a low esr/low inductance cap close as possible to the chip amp power leads.
The caps that are in switch mode power supply secondaries are an ideal choice if you can find the right voltages. The closer the cap is to the amp, the less inductance there is, which means a quicker rise and fall time. Inductors RESIST current change. Capacitors RESIST voltage change. You want your amp to be supplied with current from a power supply that can easily supply 1 amp to 10 amps without a delay, and vice versa. The less inductance you have, the better. The less resistance you have the better.
This leads to the best solution, find a set of caps with the lowest esr and self inductance possible, and mount them as close to your chip amp as possible. Make the wires between the amp and your low esr cap as thick as possible, and as short as possible.
A good compromise is to add a set of smaller caps to your amp rails. Lets say you have 2 5600uf caps per rail. Add two 1000uf caps per rail. Maybe even a 470uf. The smaller the cap, the lower its inductance.
Think of it backwards.
When a chip amp looks at its power supply, it wants the power supply to be a very low impedance. When the chip wants current, it wants it NOW. not in a few microseconds, as you wait for your inductances to rise in magnetic flux to allow the current flow.
YOU WANT IT NOW.
This is the definition of a "FAST" power supply. It can go from 100 ma to 10 amps in a flash. This is quite simply achieved by following the rules of high frequency/high current design.
Thick Wires.
Short Wires.
Low inductance and low ESR capacitors.
To sum up.
Snubber networks, being the R and C combination, should go on your diodes. Close as possible, short as wires as possible.
Shield your psu. GIGO.
Low inductance. Use a set of high value capacitors to give your amp your bass, but have a set of lower value, low esr caps to provide that initial current burst while the main caps ramp up their current output.
Short Thick Wires.
Cheers.
Thanks! Electronics has been a hobby of mine since i was a a kid, apparently born with a screwdriver in my hand haha
Im in the process of using that philosophy in building my own amp, 5 channels + 1 sub. Have 4 toroids ready, need another 1, and a few more bridges and caps. Should be fun!
Oh, and an addition to that above long post of mine.
When working with 4 ohm speakers, its even MORE important, especially with lower voltage power supplies, that provide even more current.
My five channel amp is using lm3875s into 4 ohm speakers, so im using the lower voltage rail, but higher current. So far, im using two transformers for the 5 channel amp, each transformer has two windings of 18 volts AC at 8.75 amps, so in the end i will end up with around +- 24 volts for my chips, with a current capability of around 17 amps for each rail. I am still thinking about that not being enough for 5 mono amp modules, so another two transformers may be in order, which will give me around 35 amps per rail, which with 5 modules is 7 amps. The LM3875 SPIKE protection kicks in at just under 4 amps at that supply rail, so i will try it at 2 transformers for the moment, and see how much it sags during loud passages. The amps wont be doing the bass, thats what the sub is for, so i dont expect a lot of sag in the supply.
Another two toroids will be a bit of overkill, but this is DIY!!
I am also looking at possibly watercooling the amp chips, using my experience with computer watercooling.
The main psu for the 5 channel block will have a voltage/current meter on it, just to see how it goes. Should be an exciting project!!
Im in the process of using that philosophy in building my own amp, 5 channels + 1 sub. Have 4 toroids ready, need another 1, and a few more bridges and caps. Should be fun!
Oh, and an addition to that above long post of mine.
When working with 4 ohm speakers, its even MORE important, especially with lower voltage power supplies, that provide even more current.
My five channel amp is using lm3875s into 4 ohm speakers, so im using the lower voltage rail, but higher current. So far, im using two transformers for the 5 channel amp, each transformer has two windings of 18 volts AC at 8.75 amps, so in the end i will end up with around +- 24 volts for my chips, with a current capability of around 17 amps for each rail. I am still thinking about that not being enough for 5 mono amp modules, so another two transformers may be in order, which will give me around 35 amps per rail, which with 5 modules is 7 amps. The LM3875 SPIKE protection kicks in at just under 4 amps at that supply rail, so i will try it at 2 transformers for the moment, and see how much it sags during loud passages. The amps wont be doing the bass, thats what the sub is for, so i dont expect a lot of sag in the supply.
Another two toroids will be a bit of overkill, but this is DIY!!
I am also looking at possibly watercooling the amp chips, using my experience with computer watercooling.
The main psu for the 5 channel block will have a voltage/current meter on it, just to see how it goes. Should be an exciting project!!
No probs. I thought i would put it all in one post that makes sense, because on these forums so far there are a lot misconceptions about the DC side of amplifiers.
DC is easy! AC, not so easy thats for sure.
There are a lot of misconceptions about things i would love to clear up, as there are a lot of things done to amps because "someone else told me it sounds good" and after you have modified something, your brain goes "this sounds better, because i just spent x dollars, or did y amount of hours of work"
DC is easy! AC, not so easy thats for sure.
There are a lot of misconceptions about things i would love to clear up, as there are a lot of things done to amps because "someone else told me it sounds good" and after you have modified something, your brain goes "this sounds better, because i just spent x dollars, or did y amount of hours of work"
Hi Rainwulf,
A cooper wires like these will be enough to make PSU for 4780? They are 5mm of thick and 2mm of width.
If I understand you correctly, you recommend to put snubbers across every diode in PSU (if it's slow bridge)?
Do you know a snubber values most suitable for chipamps?
Could you advise a low inductance and low ESR capacitors?
Thanks.
A cooper wires like these will be enough to make PSU for 4780? They are 5mm of thick and 2mm of width.
If I understand you correctly, you recommend to put snubbers across every diode in PSU (if it's slow bridge)?
Do you know a snubber values most suitable for chipamps?
Could you advise a low inductance and low ESR capacitors?
Thanks.
For a snubber, a common value is 100ohm in series with a 100nf cap (0.1uf) across each diode.
That copper bar should be perfect. Overkill.
The SPIKE protection will limit the current to a max of 16 amps for a 0.2ms load at full peak in stereo. Use the bar as a earth and across the caps. Run your wires from those bars.
Whats your supply voltage at peak?
And dont forget to run nice thick wires to the amp module. They are capable of carrying a lot mroe then 16 amps of course but thick wires will have very low inductance.
I would love to hear your opinions of the amp once you have finished it.
As for low inductance/low esr caps, the best and cheapest way is to get a series of lower value caps in parallel. Like 1000uf, 4-5 a side.
Even if they themselves arent low esr/inductance, a lot of them in parallel will reduce the value. The common main audio quality caps like elna and blackgates should have no problems, but even cheap caps, as long as you have enough in parallel you should be ok.
Just think to yourself, your amp wants immediate current, so the closer the main filter caps are to the amp the better, and the thicker the wire the better. There is a point of diminishing returns, just look at how thick the wire is that goes into the chip itself.
I will post some pics soon of my own chip amp PSU that im building. Will be in a couple of days.
That copper bar should be perfect. Overkill.
The SPIKE protection will limit the current to a max of 16 amps for a 0.2ms load at full peak in stereo. Use the bar as a earth and across the caps. Run your wires from those bars.
Whats your supply voltage at peak?
And dont forget to run nice thick wires to the amp module. They are capable of carrying a lot mroe then 16 amps of course but thick wires will have very low inductance.
I would love to hear your opinions of the amp once you have finished it.
As for low inductance/low esr caps, the best and cheapest way is to get a series of lower value caps in parallel. Like 1000uf, 4-5 a side.
Even if they themselves arent low esr/inductance, a lot of them in parallel will reduce the value. The common main audio quality caps like elna and blackgates should have no problems, but even cheap caps, as long as you have enough in parallel you should be ok.
Just think to yourself, your amp wants immediate current, so the closer the main filter caps are to the amp the better, and the thicker the wire the better. There is a point of diminishing returns, just look at how thick the wire is that goes into the chip itself.
I will post some pics soon of my own chip amp PSU that im building. Will be in a couple of days.
Does a regulated PSU stand up to the demands of an amp ? The most common regulated PSU here being based on LM338 or the LT1083..A national application note on PSU design however suggests these lack the bandwidth for use in audio ..how exactly is that ?..if anybody has an idea..
Also the application note suggests some ideas contrary to rainwulfs..it has 100n caps placed in paralell against trafo secondaries...and has only 100n across each diode no snubber..
http://www.national.com/an/AN/AN-1849.pdf#page=1
Also the application note suggests some ideas contrary to rainwulfs..it has 100n caps placed in paralell against trafo secondaries...and has only 100n across each diode no snubber..
http://www.national.com/an/AN/AN-1849.pdf#page=1
There is a lot of design notes that specify just caps across diodes. In fact i used to do the same thing myself, until i did some research and found out that just caps by themselves across the diodes actually add to the diode capacitance, which makes perfect sense.
Two caps in parallel add together. The resistor is added to actually dampen the oscillations.
As for the cap across the secondary, read the following.
http://en.wikipedia.org/wiki/LC_circuit
By putting a cap across an inductor, you are creating a perfect LC tank circuit, being fed by noise from the mains, and from noisy rectifier diodes. This oscillation is a lot higher then the diodes can themselves rectify, because they are too slow, and get passed right along the DC chain direct to the amp module.
As for regulated power supplies, i cant help you there. I would personally think if the power supply can supply the max current needed by the amp, it should be fine. It may have something to do with the reaction time of the regulator module. Maybe, when the amp requests a large current surge, the regulator module takes too long to respond.
I actually got the information about tank circuits of a link posted in these forums about DC power supply design. Cant find it now can i... damn.
Two caps in parallel add together. The resistor is added to actually dampen the oscillations.
As for the cap across the secondary, read the following.
http://en.wikipedia.org/wiki/LC_circuit
By putting a cap across an inductor, you are creating a perfect LC tank circuit, being fed by noise from the mains, and from noisy rectifier diodes. This oscillation is a lot higher then the diodes can themselves rectify, because they are too slow, and get passed right along the DC chain direct to the amp module.
As for regulated power supplies, i cant help you there. I would personally think if the power supply can supply the max current needed by the amp, it should be fine. It may have something to do with the reaction time of the regulator module. Maybe, when the amp requests a large current surge, the regulator module takes too long to respond.
I actually got the information about tank circuits of a link posted in these forums about DC power supply design. Cant find it now can i... damn.
Perhaps they thought a low value cap across the secondary will limit noise coming through the transformer.
Its kind of strange. Its a much better idea to limit the noise as it comes IN to the transformer, because the transformer then limits noise ( due to its inductance and design at 50/60hz) and because any noise will also get transformed down just as the actual voltage is.
Try it with, and without and see what happens. I might even do it myself and use my scope to check out the noise/oscillation if any at the bridge.
My personal philosophy is to reduce noise at its source. So that means a RFI filter at the mains input, and snubbers across the actual diodes.
The big huge core of a toroid will do a lot to reduce noise.
Its kind of strange. Its a much better idea to limit the noise as it comes IN to the transformer, because the transformer then limits noise ( due to its inductance and design at 50/60hz) and because any noise will also get transformed down just as the actual voltage is.
Try it with, and without and see what happens. I might even do it myself and use my scope to check out the noise/oscillation if any at the bridge.
My personal philosophy is to reduce noise at its source. So that means a RFI filter at the mains input, and snubbers across the actual diodes.
The big huge core of a toroid will do a lot to reduce noise.
Hi,
there can be a lot of variation in the value of both components in the snubber. These are very dependant on the parameters of the transformer.
Accurate measuring of the transformer and modeling of the damping effect of the snubber or measuring the reaction of the completed PSU to fast changing loads are required to find suitable component values for the snubber.
there can be a lot of variation in the value of both components in the snubber. These are very dependant on the parameters of the transformer.
Accurate measuring of the transformer and modeling of the damping effect of the snubber or measuring the reaction of the completed PSU to fast changing loads are required to find suitable component values for the snubber.
The resistor in the snubber isnt part of an RC time circuit. Its only role is to dampen any high frequency oscillations that the capacitor lets through.
If the resistor is too low, there wont be enough dampening, and if its too high, there will be none. There will be a sweet spot that only a digital capture scope can let you find, but a good value after research on the internet is the 100r 100nf value that goes with standard silicon recitifier diodes.
If the resistor is too low, there wont be enough dampening, and if its too high, there will be none. There will be a sweet spot that only a digital capture scope can let you find, but a good value after research on the internet is the 100r 100nf value that goes with standard silicon recitifier diodes.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.