I've been looking and searching this forum for the right (best) sort of rectifier to use in a conventional psu (50-60 Hz). I haven't been able to find an answer so hopefully this thread will help me and others to choose the right type of rectifier.
For what i've seen there are 3 types available (if you don't include tube-rectifiers)
a- conventional slow diodes or bridges
b- ultrafast diodes
c- fast slow recovery
beside this classification you have the way the diodes are built, like: p-n diodes, schotky diodes, avalanche diodes. I'm not into the semiconductor physics so i don't have a clue what the advantages of each of these are. If someone could explain what the pro's and con's of each build-type are?
The main function of a rectifier is to rectify AC, while producing as little as possible harmonics and other unwanted frequencies in the output. These harmonics and unwanted frequencies, or from now on noise are created because a diode is not perfect linear device and by the reverse current?
How do these 3 types of diodes handle this task and wich one does the best job (lowest noise at output)? Let's asume for this discussion that voltage drop is not a real issue as long as it stays within reasonable range (say 4 volt)
I know from an article of Nelson Pass that he favors the slow conventional diodes, although in his latest XA amps, he uses fast slow recovery. I asked him about this and he said it was to meet EMI standards for various country's. Im a bit confused about this, if these fast-slow recovery produce less EMI and other types of interference then they should be superior to the conventional diodes?
For what i've seen there are 3 types available (if you don't include tube-rectifiers)
a- conventional slow diodes or bridges
b- ultrafast diodes
c- fast slow recovery
beside this classification you have the way the diodes are built, like: p-n diodes, schotky diodes, avalanche diodes. I'm not into the semiconductor physics so i don't have a clue what the advantages of each of these are. If someone could explain what the pro's and con's of each build-type are?
The main function of a rectifier is to rectify AC, while producing as little as possible harmonics and other unwanted frequencies in the output. These harmonics and unwanted frequencies, or from now on noise are created because a diode is not perfect linear device and by the reverse current?
How do these 3 types of diodes handle this task and wich one does the best job (lowest noise at output)? Let's asume for this discussion that voltage drop is not a real issue as long as it stays within reasonable range (say 4 volt)
I know from an article of Nelson Pass that he favors the slow conventional diodes, although in his latest XA amps, he uses fast slow recovery. I asked him about this and he said it was to meet EMI standards for various country's. Im a bit confused about this, if these fast-slow recovery produce less EMI and other types of interference then they should be superior to the conventional diodes?
Well, since you are turning a 50 Hz sine into a 100 Hz current spikes to load your caps, you'll get lots of harmonics no matter which diode you choose.
What choosing the right diodes does is mitigating the oscillations when the diode turns off.
When the diode is conducting, the circuit is L-Cs with L=transformer and Cs=supply caps. This one doesn't usually oscillate.
When the diode switches off, you have L-Cd-Cs with Cd the diode capacitance. All the nasties in the transformer can also come into play. So basically you have a LC resonant circuit which DOES exhibit damped oscillations, at a rather high frequency since the diode capacitance is small. These oscillations are undesirable.
As you probably know (see wikipedia), in a damped resonant RLC circuit :
- Frequency depends on L and C
- Damping depends on R
- The amount of energy that will have to be dissipated during the oscillations, is the amount of energy at in the circuit T=0 (when the diode switches), that is, 1/2 L i^2.
These effects are readily visible on any good scope, but you will need to make a probe to remove the huge 50 Hz wave (a RC highpass works well).
Just Spice it. Good luck finding a proper transformer model, though.
So basically,
- you use a superfast recovery diode in the hope that the i^2 part is reduced because the diode switches off as soon as it can, before reverse current has the time to develop
- you use a soft recovery diode in the hope that the softer recovery will dissipate some of the energy and provide less excitation to the LC circuit
- you use a snubber circuit to resistively dissipate the energy so the oscillation is well damped and not a problem anymore
- or perhaps the particular values of components in your circuit provide all the damping you need (like, the ESR of your caps, pure luck, or extraordinary competence, etc)
- or a mix of the above
(Note : this isn't advice, it's a list of the various solutions and the reasons behind them).
EMC directives are mostly concerned about the crap the gear sends back into the line, and the ambient medium.
What choosing the right diodes does is mitigating the oscillations when the diode turns off.
When the diode is conducting, the circuit is L-Cs with L=transformer and Cs=supply caps. This one doesn't usually oscillate.
When the diode switches off, you have L-Cd-Cs with Cd the diode capacitance. All the nasties in the transformer can also come into play. So basically you have a LC resonant circuit which DOES exhibit damped oscillations, at a rather high frequency since the diode capacitance is small. These oscillations are undesirable.
As you probably know (see wikipedia), in a damped resonant RLC circuit :
- Frequency depends on L and C
- Damping depends on R
- The amount of energy that will have to be dissipated during the oscillations, is the amount of energy at in the circuit T=0 (when the diode switches), that is, 1/2 L i^2.
These effects are readily visible on any good scope, but you will need to make a probe to remove the huge 50 Hz wave (a RC highpass works well).
Just Spice it. Good luck finding a proper transformer model, though.
So basically,
- you use a superfast recovery diode in the hope that the i^2 part is reduced because the diode switches off as soon as it can, before reverse current has the time to develop
- you use a soft recovery diode in the hope that the softer recovery will dissipate some of the energy and provide less excitation to the LC circuit
- you use a snubber circuit to resistively dissipate the energy so the oscillation is well damped and not a problem anymore
- or perhaps the particular values of components in your circuit provide all the damping you need (like, the ESR of your caps, pure luck, or extraordinary competence, etc)
- or a mix of the above
(Note : this isn't advice, it's a list of the various solutions and the reasons behind them).
EMC directives are mostly concerned about the crap the gear sends back into the line, and the ambient medium.
Factors to consider, PIV or peak inverse voltage rating. The max allowable surge current, if you are feeding say 10000mfd capacitor from a high power transformer this is very important. Diodes as PSU rectifiers exhibit a form of commutation effect whereby you can get a burst of R.F. energy at the point of conduction. This is easily cured, add a 0.01 to 0.1 mfd capacitor of suitable voltage rating across each diode. Several major manufacturers have fallen into this trap, their products producing a large unidentifiable harsh buzz on the audio. Even touching rectifier with finger stops it. Volt drop across diodes in mains PSU's is not an issue, it will be under 1volt even at the highest currents.
Regards Karl
Regards Karl
How does a active (or pasive) PFC work? Is it also aplicable to a normal PSU, i thought it only made sense for SMPS?
I'll do some pspice on various diodes and see what it gives. Unfortunatly i don't have a good scope, only a 10mhz tektronix from ancient times (it's a very old and heavy one with interchangeble bases and a lot of little lights inside
)
Wouldn't a 50hz notchfilter be better then a lowpas?
I've been wondering about the snubber caps aswell, do you place them over the diode, or over the input/output of your rectifier? I've read somewere (builder of extreme dac, french i believe) that it would be better to place snubbers over the input output of your bridge. I can't find the page anymore Does anyone have a idea about this?
I'll do some pspice on various diodes and see what it gives. Unfortunatly i don't have a good scope, only a 10mhz tektronix from ancient times (it's a very old and heavy one with interchangeble bases and a lot of little lights inside
)Wouldn't a 50hz notchfilter be better then a lowpas?
I've been wondering about the snubber caps aswell, do you place them over the diode, or over the input/output of your rectifier? I've read somewere (builder of extreme dac, french i believe) that it would be better to place snubbers over the input output of your bridge. I can't find the page anymore
Does anyone have a idea about this?
But your cap(s) across the diode are simply increasing the total capacitance but both are still in series with the transformer windings? What they actually do is lower the ringing frequency, which may is actually a good thing since lower frequency means less coupling into adjacent circuits and can also easier be filtered by the main PS filter capacitors. If you want to dampen the ringing you would still need some resistance. Cf. the various papers on snubbers, particularly the one from Jim Hagerman (on his WEB site I think).
Hi Martin, The caps I add across the bridge (or single diode in the case of half wave rectifier) are to supress the burst of RF energy that the diode can ( and I say can, not always) produce as it comes in and out of conduction. This was a very real problem in some products, and a certain Leak tuner suffered from this effect. I once owned an Hitachi mains/batt radio and the buzz on audio running on mains was unbelievable. You only had to bring your hand near the PCB and the buzz altered, and it was this same problem, "commutation" noise generated by the bridge. A finger on the bridge rectifier stopped it totally. Adding any cap from 1000pf to 1mfd across the input to the bridge was a permanent cure. As you say, if you have a problem with ringing then yes, a C/R network may be the way to go but, as you hint at, the components will have to be calculated for each individual case.
Regards Karl
Regards Karl
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.