Sch3mat1c said:
A few months back I did some tests with opamps, chip amps and a basic discrete design -- the flyback converter generally into another order of magnitude.
The noise isn't easily filtered -- it's not ripple, it's noise.
Further -- don't forget that when you combine the switcher frequency with the audio signal (however they are coupled) you get positive and negative sidebands, and the lower sideband can easily reach into the audio range. The lower sideband can mix again etc., etc.
well, everyone who likes to listen to music is going to get a rude awakening when the power companies put "internet over power line".
flyback smps and the noise issue
I don't dispute the validity of your tests. An SMPS is noisy by nature, and flybacks are noisier than forward topologies, making them the worst.
Remember, with discrete parts, the layout on the board is very critical. The highly integrated TOP Switch, or any equivalent (such as ST Micro's "Viper") is much less noisy due to the small area of the circuit, particularly the parts connected to the switching node. In addition, these IC controllers (TOP Switch) use frequency modulation on the switching frequency, resulting in a spread spectrum noise distribution, which is much easier to filter. Post L-C filters should feature small-valued inductors (around 1.0 uh) with high self-resonant freq, and low dcr. Caps should be X7R ceramic, or low esr electrolytics bypassed with X7R ceramics. With ceramic caps only, an additional R-C snubber may be needed to dampen the high-Q formed by "L" & "C". In addition, these chips give you the option of connecting them for forward converter operation, resulting in even lower noise than the flyback. Power Integrations has an application note on their site, for forward converter operation.
I was only trying to make the poster aware of the options available in this day and age. The forward converter would definitely be a better choice, if noise is still a problem, but requires a filter inductor for each output, plus an additional rectifier for each output, as well as an additional primary winding to reset the core flux. The cross-regulation on forward topologies (including push-pull, half- and full-bridge) is not as good as the flyback. It can be improved by magnetically coupling the output inductors. Just thought I'd share this to give an idea of what's involved. Best regards.
I don't dispute the validity of your tests. An SMPS is noisy by nature, and flybacks are noisier than forward topologies, making them the worst.
Remember, with discrete parts, the layout on the board is very critical. The highly integrated TOP Switch, or any equivalent (such as ST Micro's "Viper") is much less noisy due to the small area of the circuit, particularly the parts connected to the switching node. In addition, these IC controllers (TOP Switch) use frequency modulation on the switching frequency, resulting in a spread spectrum noise distribution, which is much easier to filter. Post L-C filters should feature small-valued inductors (around 1.0 uh) with high self-resonant freq, and low dcr. Caps should be X7R ceramic, or low esr electrolytics bypassed with X7R ceramics. With ceramic caps only, an additional R-C snubber may be needed to dampen the high-Q formed by "L" & "C". In addition, these chips give you the option of connecting them for forward converter operation, resulting in even lower noise than the flyback. Power Integrations has an application note on their site, for forward converter operation.
I was only trying to make the poster aware of the options available in this day and age. The forward converter would definitely be a better choice, if noise is still a problem, but requires a filter inductor for each output, plus an additional rectifier for each output, as well as an additional primary winding to reset the core flux. The cross-regulation on forward topologies (including push-pull, half- and full-bridge) is not as good as the flyback. It can be improved by magnetically coupling the output inductors. Just thought I'd share this to give an idea of what's involved. Best regards.
Well gee, 100uV is too low to care about unless maybe you're powering a phono amp.
Still, can you explain to me what generates it, and why it's impossible to remove?
I like that 7414-based circuit, short and sweet... I'll check my IC pile (gasp!) to see if I have any in there, might have to build that one. Er.. will 200V work for the MOSFET? That's the best I seem to have right now...
OBTW half-wave choke input won't work. It needs a second diode to clamp the inductor, IIRC. Cap input would be easier.
Tim
Still, can you explain to me what generates it, and why it's impossible to remove?
I like that 7414-based circuit, short and sweet... I'll check my IC pile (gasp!) to see if I have any in there, might have to build that one. Er.. will 200V work for the MOSFET? That's the best I seem to have right now...
OBTW half-wave choke input won't work. It needs a second diode to clamp the inductor, IIRC. Cap input would be easier.
Tim
The 100uV is from the LT1533 ultra low noise switcher. Most SMPS are a 100 times worse, at least.
Here's another simple idea if efficiency is not a great concern, use a LM1875/LM3875/.... type power amplifier chip with sine wave signal to drive your ferrite xfmr. No noise except sine wave then.
(Well, you could use a P-P tube amplifier too, but kind of defeats the purpose then.)
Here's another simple idea if efficiency is not a great concern, use a LM1875/LM3875/.... type power amplifier chip with sine wave signal to drive your ferrite xfmr. No noise except sine wave then.
(Well, you could use a P-P tube amplifier too, but kind of defeats the purpose then.)
if you can't solve a problem - make it bigger!
Perhaps enlarging the scope of the problem will ultimately provide an even better solution. Since you want the high voltage for a tube amplifier anyway, try the Berning output stage. Glass Audio No 1, 2000. Solves the tube output xfmr and the HV supply all in one shot. Saves a lot of weight too.
I would suggest modifying the Berning design to use slew rate limited switching to avoid noise. Even better, use a two phase setup instead of the original single phase switching. By using two ferrite xfmrs, two switch pairs, and two sets of commutating diodes for the tubes, each phase only need cover 90+ degrees instead of 180 degree switching. That way, you can really slew rate limit the switching (and without any efficiency penulty either) and the audio signal always has a solid path thru the network, no need to filter switching gaps out of the audio output then. Its an ideal solution. If you have qualms about the SS devices in the audio path, the two phase design is formally equivalent to a diode and an always turned on Mosfet in series with the tube plate, not much sonic effect from that.
By the way, the Berning type output is easily adaptable to exotic tube output topologies, like partial cathode feedback, separate screen windings, or ultralinear, by simply partitioning the HV windings and using commutating diodes on each winding. And the # of turns required on the ferrite cores is no more than on a SMPS. Once you have one of these set up, its easy to try all sorts of tube output topolgies with minimal effort.
Perhaps enlarging the scope of the problem will ultimately provide an even better solution. Since you want the high voltage for a tube amplifier anyway, try the Berning output stage. Glass Audio No 1, 2000. Solves the tube output xfmr and the HV supply all in one shot. Saves a lot of weight too.
I would suggest modifying the Berning design to use slew rate limited switching to avoid noise. Even better, use a two phase setup instead of the original single phase switching. By using two ferrite xfmrs, two switch pairs, and two sets of commutating diodes for the tubes, each phase only need cover 90+ degrees instead of 180 degree switching. That way, you can really slew rate limit the switching (and without any efficiency penulty either) and the audio signal always has a solid path thru the network, no need to filter switching gaps out of the audio output then. Its an ideal solution. If you have qualms about the SS devices in the audio path, the two phase design is formally equivalent to a diode and an always turned on Mosfet in series with the tube plate, not much sonic effect from that.
By the way, the Berning type output is easily adaptable to exotic tube output topologies, like partial cathode feedback, separate screen windings, or ultralinear, by simply partitioning the HV windings and using commutating diodes on each winding. And the # of turns required on the ferrite cores is no more than on a SMPS. Once you have one of these set up, its easy to try all sorts of tube output topolgies with minimal effort.
Looking carefully with the scope to what happens in a traditional 50-60Hz mains supplied transformer-rectifier-capacitor power supply reveals some ugly things
- 50-60Hz transformer windings have very highg self capacitances and self inductances making them ring when diodes start or stop conducting or when any pulse or AC source excites these resonances. Ringing frecuencies depend on transformer size but its Q is markedly higher in toroidal transformers due to its higher capacitances between turns and windings and this also allows resonances in one winding to cross-talk easily to the others
- Mains waveform isn't a pure sine wave, actually it looks more like a hardly clipped sine wave or a square wave with limited slew-rate. Of course, this clipping effect ocurrs due to all the DC- equipment powered by rectifying mains directly or through transformers, making aprox 300% of their average current consumption only during 33% time and no current consumption during the rest of the waveform
- The amplitude and the waveform of mains suffer random variations over time and this causes random operating point variations on most circuits with unregulated supplies and random clipping points in unregulated amplifiers
- Mains carries lots of common mode noise to ground and some diferential mode noise between phase and neutral, including 50-60Hz harmonics up to 20Khz and RF of all flavors [AM, FM...]
- 50-60Hz transformers create stray magnetic fields, of much more intensity in the classical E-I cores, that induce audio-frequency voltages on all small signal circuits near them
- Loose wiring making loops and conducting AC creates magnetic fields that also induce voltages on small signal circuits near them and this includes all the common mode and diferential mode RF carried by mains
For all these reasons, the output of the traditional transformer-rectifier-capacitor power supply contains aperiodic ripple, random DC variations, ringing from the transformer, lots of common mode noise and diferential mode noise in all frecuencies
Even worse, it creates magnetic fields dependent on the load and up to RF, think that faraday-cage shielding only works for electric fields, magnetic fields suffer little atenuation
In comparison, a high quality well designed and shielded switching power supply has input and output common mode and diferential mode filtering introducing high mains and circuit-generated RF and ripple attenuation, very low capacitance from mains to output opposed to classic toroidal transformers, pure DC regulated output independent of input wavevorm/amplitude or load, output current limiting, undervoltage and overvoltage protections, soft start, low weight, etc... and uses about the same space as a toroidal transformer of the same power rating
For all these reasons, I seriously think that the traditional 50-60Hz transformer-rectifier-capacitor power supply :
- Actually has much dirtier output
- Affects more small signal circuits
- Is bulky
- Has poor power/space ratio
- Has very high weight
- Is outdated
- Component costs are similar than for same power SMPS
- Is not adequate for audio circuits
- GENERATES LOTS MORE CONDUCTED AND RADIATED EMI THAN AN OPTIMIZED SMPS, because noise is also present but there are no efforts in things like damp ringing, shielding, filtering, cancelling out magnetic fields, etc...
Having some SMPS understanding, it´s relatively easy to modify for your needs an old AT or ATX computer supply, that costs almost nothing
A model with complete mains filtering should be selected as some models lack common mode or diferential mode filters because in theory can pass EMI tests without them
Originally these power supplies have a few mV of ripple and ringing on their outputs but this is due to cost cutting in output filters and poor PCB layout to be able to put all the components in so little space
Usable things from these supplies include input filtering, rectification, half bridge driving, switching transistors, driver transformer for them [made from a saturable reactor in AT units to make them self oscillate to be able to start as control circuit is powered from output side], control circuit usually based on a TL494 or equivalent, heat sinks, fan and little more
Letting the primary-driving intact you can get about 250W of output and wiht some modifications maybe 500W, but in the units I modified It was enough with 250W
ATX units are a bit trickier because you have to mantain the standby auxiliar supply to power the control circuit
Some links of interest :
Typical AT PS schematic :
Typical ATX PS schematic :
Same Proportional-base-drive bipolar transistor half bridge design with self-starting but at higher power and info :
Datasheet TL494 : http://www-s.ti.com/sc/ds/tl494.pdf
Additional info TL494 : http://www-s.ti.com/sc/psheets/slva001a/slva001a.pdf
PD: I suggest discarding the all-discrete control circuit alternative as you would need about 25-50 active components to get decent performance and reliability
- 50-60Hz transformer windings have very highg self capacitances and self inductances making them ring when diodes start or stop conducting or when any pulse or AC source excites these resonances. Ringing frecuencies depend on transformer size but its Q is markedly higher in toroidal transformers due to its higher capacitances between turns and windings and this also allows resonances in one winding to cross-talk easily to the others
- Mains waveform isn't a pure sine wave, actually it looks more like a hardly clipped sine wave or a square wave with limited slew-rate. Of course, this clipping effect ocurrs due to all the DC- equipment powered by rectifying mains directly or through transformers, making aprox 300% of their average current consumption only during 33% time and no current consumption during the rest of the waveform
- The amplitude and the waveform of mains suffer random variations over time and this causes random operating point variations on most circuits with unregulated supplies and random clipping points in unregulated amplifiers
- Mains carries lots of common mode noise to ground and some diferential mode noise between phase and neutral, including 50-60Hz harmonics up to 20Khz and RF of all flavors [AM, FM...]
- 50-60Hz transformers create stray magnetic fields, of much more intensity in the classical E-I cores, that induce audio-frequency voltages on all small signal circuits near them
- Loose wiring making loops and conducting AC creates magnetic fields that also induce voltages on small signal circuits near them and this includes all the common mode and diferential mode RF carried by mains
For all these reasons, the output of the traditional transformer-rectifier-capacitor power supply contains aperiodic ripple, random DC variations, ringing from the transformer, lots of common mode noise and diferential mode noise in all frecuencies
Even worse, it creates magnetic fields dependent on the load and up to RF, think that faraday-cage shielding only works for electric fields, magnetic fields suffer little atenuation
In comparison, a high quality well designed and shielded switching power supply has input and output common mode and diferential mode filtering introducing high mains and circuit-generated RF and ripple attenuation, very low capacitance from mains to output opposed to classic toroidal transformers, pure DC regulated output independent of input wavevorm/amplitude or load, output current limiting, undervoltage and overvoltage protections, soft start, low weight, etc... and uses about the same space as a toroidal transformer of the same power rating
For all these reasons, I seriously think that the traditional 50-60Hz transformer-rectifier-capacitor power supply :
- Actually has much dirtier output
- Affects more small signal circuits
- Is bulky
- Has poor power/space ratio
- Has very high weight
- Is outdated
- Component costs are similar than for same power SMPS
- Is not adequate for audio circuits
- GENERATES LOTS MORE CONDUCTED AND RADIATED EMI THAN AN OPTIMIZED SMPS, because noise is also present but there are no efforts in things like damp ringing, shielding, filtering, cancelling out magnetic fields, etc...
Having some SMPS understanding, it´s relatively easy to modify for your needs an old AT or ATX computer supply, that costs almost nothing
A model with complete mains filtering should be selected as some models lack common mode or diferential mode filters because in theory can pass EMI tests without them
Originally these power supplies have a few mV of ripple and ringing on their outputs but this is due to cost cutting in output filters and poor PCB layout to be able to put all the components in so little space
Usable things from these supplies include input filtering, rectification, half bridge driving, switching transistors, driver transformer for them [made from a saturable reactor in AT units to make them self oscillate to be able to start as control circuit is powered from output side], control circuit usually based on a TL494 or equivalent, heat sinks, fan and little more
Letting the primary-driving intact you can get about 250W of output and wiht some modifications maybe 500W, but in the units I modified It was enough with 250W
ATX units are a bit trickier because you have to mantain the standby auxiliar supply to power the control circuit
Some links of interest :
Typical AT PS schematic :
An externally hosted image should be here but it was not working when we last tested it.
Typical ATX PS schematic :
Same Proportional-base-drive bipolar transistor half bridge design with self-starting but at higher power and info :
An externally hosted image should be here but it was not working when we last tested it.
Datasheet TL494 : http://www-s.ti.com/sc/ds/tl494.pdf
Additional info TL494 : http://www-s.ti.com/sc/psheets/slva001a/slva001a.pdf
PD: I suggest discarding the all-discrete control circuit alternative as you would need about 25-50 active components to get decent performance and reliability
Sorry for missing the following link and for my bad english
Same Proportional-base-drive bipolar transistor half bridge design with self-starting but at higher power and info :
For more info, google shows lots of pages on these topics
Same Proportional-base-drive bipolar transistor half bridge design with self-starting but at higher power and info :
An externally hosted image should be here but it was not working when we last tested it.
For more info, google shows lots of pages on these topics
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.