Testing...

C

Chaos Master

Guest
Hey.

Testing some stuff I done in my newsreader (Gravity), and I don't want to
subscribe to alt.test.

Just to keep this on topic:

Does someone here knows where to find info about "auto-ranging" (I don't know
the exact term for this) SMPS power supplies? E.g. the type used in TV's that
automatically "detect" if the line voltage is 120 or 240v.

[]s
--
Chaos MasterŽ | "I'm going under,
Posting from Brazil! | drowning in you
ICQ: 126375906 | I'm falling forever,
ask for e-mail/MSN | I've got to break through"
---------------------. -- Evanescence, "Going Under"
 
"Chaos Master" <ig@ig.com.br> wrote in message
news:MPG.1b136b7ab41341f9989996@130.133.1.4...
Hey.

Testing some stuff I done in my newsreader (Gravity), and I don't want to
subscribe to alt.test.

Just to keep this on topic:

Does someone here knows where to find info about "auto-ranging" (I don't
know
the exact term for this) SMPS power supplies? E.g. the type used in TV's
that
automatically "detect" if the line voltage is 120 or 240v.

[]s
--
Chaos MasterŽ | "I'm going under,
Posting from Brazil! | drowning in you
ICQ: 126375906 | I'm falling forever,
ask for e-mail/MSN | I've got to break through"
---------------------. -- Evanescence, "Going Under"
Click to expand...
normally they dont bother with "autoranging" and just accept a wide
variation - 90Vac - 265Vac = 125Vdc - 375Vdc = 3:1 variation in line
voltage. This usually turns into a 3:1 variation in duty cycle - it actually
depends on the converter. For a peak current mode controlled flyback
converter in DCM, input duty cycle D varies in direct inverse proportion to
Vin (as dIp/dt = Vin/Lmag goes up, for Ipeak = constant D goes down). D also
varies in proportion to the square root of output power. [NOTE: output duty
cycle D1 does not vary; "dead" duty cycle D2 soaks up the variation in D,
maintaining the equality D+D1+D2 = 1]

Say you had D=0.45 at 125Vdc. Thus D = 0.15 at 375Vdc. (say Lp = 600uH. do
the maths)

If T = 10us (100kHz Fsw) Then Ton = 4.5us at 125Vdc, 1.5us at 375Vdc

If Tonmin = 0.5us (UC3842 will give about this, due to current sense filter,
comparator & flip-flop delays, and time gate driver takes to discharge below
Vth of FET)


At Vin = 375Vdc:
then 1.5us/0.5us = 3 = Dmaxload/Dminload = sqrt(Pmax/Pmin)

so Pmax/Pmin = 9 ie this supply will work down to 11.1% load, at which time
squegging (pulse skipping) will occur - this is because the minimum on time
causes a finite quanta of energy to be delivered to the load, which (below
11.1% load) is more than required so Vout rises. The lower the load, the
longer it takes for Vout to discharge to its setpoint, ie the more cycles
are skipped.


Note that at 125Vdc:
Dmaxload/Dminload = 4.5us/0.5us = 9, so Pmax/Pmin = 81, ie will regulate
down to 1.2% load (which is bloody low, and often good enough - some
supplies just burn a watt or two in minimum-load resistors....).

for a DCM flyback, a simple trick is to use a fast SMPS chip, like the
UCC3801 - this will respond in about 70ns + your gatedrive. including
current sense blanking, it can halve the minimum on time, thereby
quadrupling the load range.

Another good trick is to use a somewhat lower frequency - 50kHz doubles Ton,
immediately quadrupling the load range. So the aforementioned supply, at
50kHz can regulate down to 2.8%, and that would drop to about 0.7% using a
UCC3801 and a decent gate drive. (at Vin = 375Vdc; results are 0.3% & 0.08%
at 125Vdc)

other tricks invlove stuff like making F = kVin (marty browns on-semi app
note (google wide input range converter) does this).

For all the OEM supplies I have done, I know what my minimum loads are. I
have designed smps (still in production) that run from 90Vdc to 1000Vdc at
Pout = 10 - 100W. Now that eats up a lot of duty cycle.

in theory you could use a "higher order" topology ie one with a transfer
function like (D/(1-D))^2 (you can create an arbitrarily large number of
arbitrarily complex converters if you can be bothered) but I have never seen
a real paper (ie one with an actual physical design, complete with
measurements) on one, let alone a real product.

cheers
Terry
 
mv /var/posts/Terry Given/the_domes@xtra.co.nz /dev/null:

[snip: Auto-ranging SMPS info.]

other tricks invlove stuff like making F = kVin (marty browns on-semi app
note (google wide input range converter) does this).

For all the OEM supplies I have done, I know what my minimum loads are. I
have designed smps (still in production) that run from 90Vdc to 1000Vdc at
Pout = 10 - 100W. Now that eats up a lot of duty cycle.
Click to expand...
1000VDC? wow!!!

in theory you could use a "higher order" topology ie one with a transfer
function like (D/(1-D))^2 (you can create an arbitrarily large number of
arbitrarily complex converters if you can be bothered) but I have never seen
a real paper (ie one with an actual physical design, complete with
measurements) on one, let alone a real product.
Click to expand...
Interesting reading... :)

--
Chaos MasterŽ | "I'm going under,
Posting from Brazil! | drowning in you
Win 98 + BB4Win | I'm falling forever,
Slackware Linux | I've got to break through"
---------------------. -- Evanescence, "Going Under"
 
"Chaos Master" <chaos_master@ig.com.br> wrote in message
news:MPG.1b1db0b3d501c3969899f2@130.133.1.4...
mv /var/posts/Terry Given/the_domes@xtra.co.nz /dev/null:

[snip: Auto-ranging SMPS info.]

other tricks invlove stuff like making F = kVin (marty browns on-semi
app
note (google wide input range converter) does this).

For all the OEM supplies I have done, I know what my minimum loads are.
I
have designed smps (still in production) that run from 90Vdc to 1000Vdc
at
Pout = 10 - 100W. Now that eats up a lot of duty cycle.

1000VDC? wow!!!
Click to expand...
AC motor controller that would run from 208Vac - 480Vac. With regeneration
the DC bus would get as high as 850Vdc (small spikes up to 900Vdc) so I
chose 1000Vdc to give me a bit of margin. Allowing it to run down to 90Vdc
gave us enough headroom that it would happily keep the smps running during a
brownout.

I sketched out a design for one that would run at 1500Vdc (690Vac operation)
at slightly higher power levels, but changed jobs before I finished it.


in theory you could use a "higher order" topology ie one with a transfer
function like (D/(1-D))^2 (you can create an arbitrarily large number of
arbitrarily complex converters if you can be bothered) but I have never
seen
a real paper (ie one with an actual physical design, complete with
measurements) on one, let alone a real product.

Interesting reading... :)

--
Chaos MasterŽ | "I'm going under,
Posting from Brazil! | drowning in you
Win 98 + BB4Win | I'm falling forever,
Slackware Linux | I've got to break through"
---------------------. -- Evanescence, "Going Under"
Click to expand...
The girl from evanescence is gorgeous!

Cheers
Terry
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top