smps with 250v output

[jean-marc MERCY]

I am starting to work on smps and my goal is to achieve an output
ranging 250~300 volts. Power would be 250w or less. Are there any
(non-obvious) pitfalls I should be aware of with such a voltage
level. One of my concerns is rectifiers, the other is insulation.
Would anyone out there have any expertise/advice ? General discussion
can be carried out here, private one to my email.

Thanks

 

[Fritz Schlunder]

I am starting to work on smps and my goal is to achieve an output
ranging 250~300 volts. Power would be 250w or less. Are there any
(non-obvious) pitfalls I should be aware of with such a voltage
level. One of my concerns is rectifiers, the other is insulation.
Would anyone out there have any expertise/advice ? General discussion
can be carried out here, private one to my email.

Thanks
--
Ce message a ete poste via la plateforme Web club-Internet.fr
This message has been posted by the Web platform club-Internet.fr

http://forums.club-internet.fr/

Without more details about topology and whatnot you plan to use, there are
two general hints I might suggest.

Absolutely positively make certain that you use very effectual fast (IE,
pulse by pulse, something that can respond in say less than 10us and keep
the duty cycles thereafter to around 1% or less thereafter) current limiting
scheme capable of providing full protection from indefinite output short
circuits. Include this current limiting right from the very beginning,
don't wait until you've blown up several iterations of prototypes before
doing this.

The other thing I would suggest is to not get carried away with seemingly
good diode ratings. Some device like the UF4007 (1A 1000V very fast
recovery) is NOT appropriate for your output rectifier even though your
output voltage and currents are well within the datasheet ratings of the
device.

As you increase the blocking voltage of the diode, the reverse recovery
charge (don't worry too much about the "reverse recovery time" figure,
mainly it is the reverse recovery charge figure that determines the losses)
goes up quite a bit. Ultrafast diodes rated 200V and below can have some
pretty good reverse recovery charge figures, but when you start playing with
higher voltage rated parts like beyond 300V, the reverse recovery charge
gets rather substantial. Complicating this fact further is the increased
voltage stress caused by the higher blocking required in your high voltage
output application.

If you can figure out the reverse recovery charge of your diodes under your
operating conditions (use figures of 125 deg. C, the 25C values are much
smaller than real world will likely be), which will be measured in coulombs,
then multiply by the frequency they will see (measured in inverse seconds)
and you will get a product with the units (coulombs/second). You will find
those units are the same units for current, so you effecively get a somewhat
equivalent rms reverse recovery current given your switching frequency.
Then multiply this rms reverse recovery current figure by the reverse
applied voltage (in your case probably say 30% greater than 300V or 600V
depending upon your circuit topology and input voltage variation margin) to
find a power loss figure due to rectifier reverse recovery. The power will
be dissipated in both the power MOSFET(s) in your circuit as well as the
rectifier itself.

So as you might be able to see from the above, when you try to make SMPS
devices with high output voltage you end up with a double whammy in terms of
diode losses getting dramatically worse when compared against lower output
voltage switch mode power supplies.

As a result you need to operate at a lower frequency than powersupplies with
5V or 12V outputs (using nice schottky diode technology) to obtain best
efficiciency (or perhaps in some cases just to stay within the power
dissipation ratings of the output rectifier). Given your conditions I
suggest the output rectifiers should see a frequency of say something in the
range of 50kHz to 100kHz (so in say a half bridge the main MOSFETs should be
switching at 25kHz to 50kHz). This is unfortunate in that it makes your
magnetics and perhaps capacitors bigger, but is necessary to avoid
overburdening your output rectifier. Use nice TO-220 or TO-247 packaged
diodes with adequate heatsinking. The biggest problem with little devices
like DO-41 and DO-201 packages is their lack of easy to use heatsink tab,
despite their seemingly good current, voltage, and recovery time ratings.

 

[Winfield Hill]

Fritz Schlunder wrote...

Absolutely positively make certain that you use very effectual fast (IE,
pulse by pulse, something that can respond in say less than 10us and keep
the duty cycles thereafter to around 1% or less thereafter) current limiting
scheme capable of providing full protection from indefinite output short
circuits. Include this current limiting right from the very beginning,
don't wait until you've blown up several iterations of prototypes before
doing this.

The other thing I would suggest is to not get carried away with seemingly
good diode ratings. Some device like the UF4007 (1A 1000V very fast
recovery) is NOT appropriate for your output rectifier even though your
output voltage and currents are well within the datasheet ratings of the
device.

As you increase the blocking voltage of the diode, the reverse recovery
charge (don't worry too much about the "reverse recovery time" figure,
mainly it is the reverse recovery charge figure that determines the losses)
goes up quite a bit. Ultrafast diodes rated 200V and below can have some
pretty good reverse recovery charge figures, but when you start playing with
higher voltage rated parts like beyond 300V, the reverse recovery charge
gets rather substantial. Complicating this fact further is the increased
voltage stress caused by the higher blocking required in your high voltage
output application.

If you can figure out the reverse recovery charge of your diodes under your
operating conditions (use figures of 125 deg. C, the 25C values are much
smaller than real world will likely be), which will be measured in coulombs,
then multiply by the frequency they will see (measured in inverse seconds)
and you will get a product with the units (coulombs/second). You will find
those units are the same units for current, so you effecively get a somewhat
equivalent rms reverse recovery current given your switching frequency.
Then multiply this rms reverse recovery current figure by the reverse
applied voltage (in your case probably say 30% greater than 300V or 600V
depending upon your circuit topology and input voltage variation margin) to
find a power loss figure due to rectifier reverse recovery. The power will
be dissipated in both the power MOSFET(s) in your circuit as well as the
rectifier itself.

So as you might be able to see from the above, when you try to make SMPS
devices with high output voltage you end up with a double whammy in terms of
diode losses getting dramatically worse when compared against lower output
voltage switch mode power supplies.

As a result you need to operate at a lower frequency than powersupplies with
5V or 12V outputs (using nice schottky diode technology) to obtain best
efficiciency (or perhaps in some cases just to stay within the power
dissipation ratings of the output rectifier). Given your conditions I
suggest the output rectifiers should see a frequency of say something in the
range of 50kHz to 100kHz (so in say a half bridge the main MOSFETs should
be switching at 25kHz to 50kHz). This is unfortunate in that it makes
your magnetics and perhaps capacitors bigger, but is necessary to avoid
overburdening your output rectifier. Use nice TO-220 or TO-247 packaged
diodes with adequate heatsinking. The biggest problem with little devices
like DO-41 and DO-201 packages is their lack of easy to use heatsink tab,
despite their seemingly good current, voltage, and recovery time ratings.

Thanks for the interesting and useful info, Fritz. The issue of diode
reverse recovery charge is a bit messy, because this is generally not
given by the manufacturer. While one can check the reverse capacitance
curves when attempting to evaluate one diode vs another, the efficiency
calculation that you suggested requires a charge value that's painfully
determined, I suppose, by integrating the diode's nonlinear capacitance
vs voltage curve.

Do you have any low reverse-recovery charge diodes to suggest for us?
For example both the MUR160 (ON Semi) and UF4005 (Vishay) data sheets
show about 8pF of capacitance at 50V, but we know the capacitance of
high-voltage diodes is generally less than low-voltage parts in that
same series (e.g. see the Fairchild UF4007 data sheet - we'd really
like to see more detailed data).

(BTW, doesn't the latter observation mean that using higher-voltage
diodes than necessary can reduce rather than worsen reverse-recovery
charge, when evaluated at a fixed voltage? If one finds some charge
"specs" to the contrary, this could be because the higher-voltage
diodes are specified for higher voltage operation, duh, so Q = CV is
higher because even though C is slightly lower, V is much higher.)

Thanks,
- Win

 

[Fritz Schlunder]

Thanks for the interesting and useful info, Fritz. The issue of diode

reverse recovery charge is a bit messy, because this is generally not
given by the manufacturer. While one can check the reverse capacitance
curves when attempting to evaluate one diode vs another, the efficiency
calculation that you suggested requires a charge value that's painfully
determined, I suppose, by integrating the diode's nonlinear capacitance
vs voltage curve.

Indeed the issue is extremely messy. Most diode manufactures don't ever
bother to include reverse recovery charge data at all, and those that do
usually provide extremely skimpy data considering how important the
information is.

To their credit though I guess it is a difficult device parameter to
characterize. The reverse recovery charge is a strong function of numerous
variables: applied reverse voltage, forward current waveshape, foward
conducting current magnitude, temperature, and the rate of change of current
during reverse recovery if I'm not forgetting anything. While it is
relatively easy to hook a device up to a curve tracer and figure out its
characteristics as a function of one variable, five variables is a whole
nother beast. Similarly it would also be difficult to display the results
since a normal graph usually only has two axis (axises?).

The diode's nonlinear capacitance curves are easier to characterize, but
nevertheless not all manufacturers even provide consistent data here either.
Unfortunately the device's capacitance isn't an especially important figure
for determining the switching loss of most diodes used in many high
frequency switch mode power supply applications. Usually the diode loss due
to device capacitance in a high output voltage SMPS is much smaller than
either the forward conduction loss or the reverse recovery charge loss
(sometimes as much as an order of magnitude smaller).

Sometimes manufacturers will provide a value for the Irrm (peak reverse
recovery current) under some single static condition. This information
along with the device's reverse recovery time specification can sometimes be
useful as well as shown by this Maxim document:

http://www.maxim-ic.com/appnotes.cfm/appnote_number/849/ln/en

Do you have any low reverse-recovery charge diodes to suggest for us?
For example both the MUR160 (ON Semi) and UF4005 (Vishay) data sheets
show about 8pF of capacitance at 50V, but we know the capacitance of
high-voltage diodes is generally less than low-voltage parts in that
same series (e.g. see the Fairchild UF4007 data sheet - we'd really
like to see more detailed data).

The best high voltage diodes for use in high frequency high voltage SMPS
type applications that I can find are the International Rectifier "hyperfast
recovery" and "hexfred" type devices. Not all of the devices in these
product lines appear to be equally well performing, so it might be wise to
be careful.

Devices like the HFA04TB60:

http://www.irf.com/product-info/datasheets/data/hfa04tb60.pdf

Or perhaps the 8ETH06:

http://www.irf.com/product-info/datasheets/data/8eth06.pdf

And maybe the MUR1620:

http://www.irf.com/product-info/datasheets/data/mur1620ct.pdf

Seem to offer about the best performance I can find. Perhaps someday when
high voltage schottky diode technology becomes cheaper and more available
better performance can be practically obtained.

(BTW, doesn't the latter observation mean that using higher-voltage
diodes than necessary can reduce rather than worsen reverse-recovery
charge, when evaluated at a fixed voltage? If one finds some charge
"specs" to the contrary, this could be because the higher-voltage
diodes are specified for higher voltage operation, duh, so Q = CV is
higher because even though C is slightly lower, V is much higher.)

I'm not sure I fully understand what you are saying here, but I don't think
so. Generally lower voltage rated diodes have lower reverse recovery charge
figures than higher voltage versions even when everything else is held
contstant. In addition to reduced recovery charge, they usually also offer
lower forward conduction voltage. Therefore for best efficiency it is
usually best to use the lowest possible voltage rated diode for a high
frequency SMPS (even if it has a higher capacitance associated with it).
Generally the diode's capacitance and Q=CV formulas won't be very applicable
to finding the total diode switching power loss since the reverse recovery
charge (not a capacitive charge element, a different phenomena) dominates.