Capacitors across rectifiers;why?
- Thread starter johnheath
- Start date
Help Support GroupDIY Audio Forum:
moamps
Well-known member
ruffrecords said:
The di/dt is zero at diode's switching on and off (approximately) moment. There are two time areas of high di/dt, positive where diode's current supplies the load and charges an elco, and negative when the diode's current drops and the elco takes over the load.
The real cause of spikes isn't high di/dt per se but high remaining reverse charge Qrr of the diode when it is switched off as Tim already pointed out. This reverse charge is higher if the di/dt is higher at the moment when the diode should be switched off and if diode's transition time is higher. This Qrr is zero when diode starts conducting so there can't be spikes at diode's switching on moment (and dynamic resistance of the diode is very small, etc.).
So, if we like to have smaller spikes, we should have smaller di/dt or we should use diode with lower Qrr (schottky) in the first place. But, if we already use a diode which generates spikes, we can dump it more or less by using a RC network on diode or transformer's secondary where R is set for critical damping of ringing and C is just a pass capacitor.
(edit: ???, my 1k post 8) )
would like to see scope shots showing the effect of adding ferrite beads to the B+ and CT leads,
but if that helped suppress noise they would already be doing it so cancel that order and make it a double cheeseburger hold the bacon,
but if that helped suppress noise they would already be doing it so cancel that order and make it a double cheeseburger hold the bacon,
ruffrecords
Well-known member
moamps said:
I largely agree with the above except for a couple of points. This lecture from the Dubkin Institute of Technology makes interesting reading.:
http://eleceng.dit.ie/kgaughan/notes/DT021PE/Microsoft%20Word%20-%20Lecture%203%20-%20Diodes.pdf
In particular:
"When a normal silicon diode (PN junction) is forward biased the junction region
becomes flooded with minority charge carriers. If the voltage is reversed these carriers
must be removed from the junction to form a depletion layer in order for the diode to start
blocking. The diode will actually conduct in the reverse direction for a small time after
negative voltage is applied until the stored charge is removed. "
It seems to me this is independent of di/dt during the conduction period - once the junction is flooded it is flooded and will remain so until the diode becomes revers biased. Furthermore it states:
"In some diodes the reverse recovery current decays very suddenly (snaps off) this high
di/dt can cause very high voltage spikes in any circuit inductance. "
It seems to me this is the mechanism we are talking about.
From the above, the di/dt is due entirely to the diode properties, not the di/dt of the forward current. And:
"Schottky diodes use a metal semiconductor junction (schottky barrier) rather than
a PN junction. They have a forward voltage drop typically half that of an
equivalent PN diode. Also they do not suffer from charge storage. "
So Schottky diodes would seem to be a good idea except:
"Schottky’s suffer from high reverse leakage and are only suitable for use in low voltage
circuits (up to 100V). "
So they will be no good for HT supplies. Perhaps we should be looking at fast recovery types which are a:
" Fast or ultrafast recovery diode optimised for gentle trun off of reverse
current. Recommended for most high frequency switching applications to avoid problem
of high di/dt. "
Which is basically back to where we started. My question was why critical damping?? By definition it allows a significant high amplitude part of the spike to remain. Surely we require a severely over damped response to remove the spike almost completely - in other word R equals zero?
Cheers
Ian
moamps
Well-known member
ruffrecords said:
Qrr depends of dI/dt. Look the attached picture with an example.
For HT you should check out SiC Schottky diodes.
You cannot dump an oscillation without a resistive element. Period. By adding only a capacitor, oscillation is just moved to another frequency. If you add a small resistor, the pass capacitor should be large and the losses become very high.
Attachments
ruffrecords
Well-known member
Interesting. Where does that come from?moamps said:
Thanks for the tip. I will check them out.
Again we are going back to the beginning.
There is still resistance - the winding resistance.
Cheers
Ian
moamps
Well-known member
I can't find the source on web right now. It's a data for a FRED diode. But you can find this type of data in datasheets for any modern fast, SIC and similar diodes.ruffrecords said:
Yes, but the ringing-oscillating is still there because this resistance is very small in a high power linear or SMPS transformers. This dumping is maybe not needed for HT linear supplies where currents are small and secondary real part of impedance is significant. Anyway, some designers, for example, use resistors (20-100ohms) in series with bridge and some capacitors across the bridge in HT power supplies for that purpose (Menno van der Veen), or just use simple RC network across the secondary.
The BYT79 14A 500V fast-recovery diode datasheet has a good example of Qrr reduction.
http://trigger.ru/content/Catalogue/pdf/BYT79.pdf
Wrt to snubber part values, shunting an AC winding with more and more capacitance starts to have negative consequences such as larger ac current, and larger physical device and cost - especially for high Vac, when it should probably be X-rated, and possibly faster transient current spike within winding, and hence more easily able to transfer to other windings.
There may even be a benefit from using series ss diodes - perhaps by taking longer to transition through the on-to-off voltage (as it is doubled/trebled ...), and assuming the same reverse current/time has to flow, as it will consume the Q in each device at the same time.
http://trigger.ru/content/Catalogue/pdf/BYT79.pdf
Wrt to snubber part values, shunting an AC winding with more and more capacitance starts to have negative consequences such as larger ac current, and larger physical device and cost - especially for high Vac, when it should probably be X-rated, and possibly faster transient current spike within winding, and hence more easily able to transfer to other windings.
There may even be a benefit from using series ss diodes - perhaps by taking longer to transition through the on-to-off voltage (as it is doubled/trebled ...), and assuming the same reverse current/time has to flow, as it will consume the Q in each device at the same time.
ruffrecords
Well-known member
Thank you. I will check that out.trobbins said:
[/quote]
Wrt to snubber part values, shunting an AC winding with more and more capacitance starts to have negative consequences such as larger ac current, and larger physical device and cost - especially for high Vac, when it should probably be X-rated, and possibly faster transient current spike within winding, and hence more easily able to transfer to other windings.
[/quote]
I am not suggesting more and more capacitance. Just the capacitance you would normally have in series with the snubber resistance. I use an X-rated 100nF capacitor across the bridge input in my HT supply designs and this completely kills the spikes. They are not expensive parts.
I agree. I am not sure fast turn off devices have any benefit because they will just increase di/dt at turn off.
Cheers
Ian
moamps
Well-known member
ruffrecords said:
Are you referring to spikes on picture you posted in post 34?
ruffrecords
Well-known member
moamps said:
Yes.
Cheers
Ian
ruffrecords said:
The key advantage of fast turn off diodes is their low Qrr. As Merlin's plots show, the fast diode almost achieves a simple step in voltage. Another batch of test results from Mark Johnson also show similar outcome, where low Qrr devices cause effectively the same step in voltage (due to pre-existing V=Ls.dI/dt, stepping to V=0 when dI/dt becomes zero). Standard diodes with significant Qrr allow a noticeable negative current to flow, and the recovering dI/dt to get current back to 0A adds an extra transient increment to the voltage step.
ruffrecords
Well-known member
trobbins said:
OK, that makes sense; fast diodes are fast because of their low Qrr.
Cheers
Ian
5v333
Well-known member
does anybody treat heater windings with caps/snubbers aswell?
keeping rf crap away from filaments and leaking through to cathodes etc...
keeping rf crap away from filaments and leaking through to cathodes etc...
In the context of noise from a rectified B+ supply coupling over in to a heater winding, then I'd suggest there is no benefit in adding a snubber across the heater winding, as the coupling mechanism is likely to be capacitive coupling between windings on the power transformer.5v333 said:
If the heater winding was itself being rectified for a DC heater supply, then yes a local snubber across the heater winding would be well worth considering.
An article discussing rectification relation noise is:
https://www.dalmura.com.au/projects/Power supply issues for tube amps.pdf
ruffrecords
Well-known member
5v333 said:
I am pretty certain I have seen some rf schematics with series chokes an parallel C to the heaters of crucial tubes - possibly in old VHF TV turret tuners - do you remember them - amazing pieces of workmanship.
http://www.modip.ac.uk/sites/modip/images/large/BXL974.jpg
Cheers
Ian
Gene Pink
Well-known member
Awww man, I certainly do remember. Memories, the arguments at the shop about which is the best brand of contact cleaner to use on them, the revelation that you can adjust the two slugs of the selected channel by simply pulling off the knob and using the provided hole to align, with one of the plastic hex drivers in the set....ruffrecords said:
The UHF tuners had nuvistors, saved a bunch, wish I thought to save the sockets, too. Condenser mic amplifiers?
Gene
ruffrecords
Well-known member
Gene Pink said:
Yes, there was some very well thought out engineering in those beauties. Nuvistors are still available on eBay, some types quite cheaply. I am corresponding with a guy who has a bunch of them and wants to build one of my tube mice pres using them instead of the regular tubes. Nuvistors are incredibly sturdy, they should have low microphonis so they could be good for a condenser mic.
Cheers
Ian
