Inverters vs wallwarts...

[Phil Hobbs]

jlarkin@highlandsniptechnology.com wrote:

On Wed, 29 Jun 2022 08:18:39 -0700 (PDT), Fred Bloggs
bloggs.fredbloggs.fred@gmail.com> wrote:

On Wednesday, June 29, 2022 at 10:20:07 AM UTC-4, jla...@highlandsniptechnology.com wrote:
On Wed, 29 Jun 2022 08:23:22 -0400, legg <le...@nospam.magma.ca> wrote:

On Tue, 28 Jun 2022 13:37:59 -0700 (PDT), Phil Allison
palli...@gmail.com> wrote:

legg wrote:
================

** Learn to trim - asshole.


Anybody measured the ripple?

Electrolytic life is rms current dependent, among other things (like
temperature).

If you stick one next to a vacuum tube, it\'s received radiated that
dominates part temperature, and voltage stress of the app that
dominates.

** You just made that mad crap up.

If you\'ve ever calculated mtbf under Mil Hdbk 217, or Belcore,
you\'d be aware of dominating life factors.
Belcore has no derating factor for bad design. That often dominates
MTBF.

Temp is the single determining factor on expected life.
Almost always the local ambient completely dominates.

Current forces self-rise due the part\'s ESR and limited
body surface area, (unless you stick it next to a hot radiator).
The actual relevant temperature is measured on the component\'s
body.
We are designing a fancy switching power supply and need a 20 uF cap
that can handle several amps RMS, a 250 KHz triangle from a
half-bridge and an inductor. We are thinking about using four 4.7 uF
radial-leaded film caps in parallel.

We have samples of several types on order. I plan to set up a test rig
and push amps of triangle into them and see how hot they get. May as
well snoop the waveform across each cap while I do that.

Film cap data sheets are typically not much help. They might spec a
few sinewave loss tangents and maybe allowable voltage vs frequency,
but rarely spec ESR or ESL or any thermals. We have to measure all
that.

These people don\'t figure current into life expectancy, only voltage and temperature.
https://www.cde.com/resources/technical-papers/filmAPPguide.pdf

That\'s interesting but, typically, qualitative and theoretical. I\'ll
have to test actual caps.

We will have a lot of air flow too, which will increase allowable RMS
current and needs to be quantified too. We should orient and space the
caps to take advantage of the air cooling. This ain\'t simple.

It\'s distressing, in electronics data sheets and literature, how
seldom you find actual numbers. I recently bought a book about
electronic cooling, but it\'s packed with equations and theory, with
not a single worked-out case of blowing air over a flat plate. There
is an equation, but it\'s a nightmare.

If I had a 6\" square of 0.062 thick aluminum, and blasted 200 f/m of
air along both sides, what would theta be? I\'ll have to measure that.

And what would the temp rise be of my 4.7u film caps, in degc/watt, in
still air and with air flow? Gotta measure that too.

Even in an inviscid, incompressible fluid, the equations aren\'t all that
simple. Forced-air cooling of macroscopic systems runs at some huge and
highly variable Reynolds number, depending on where you are.

There was a bit of a fad in the \'90s for people to publish various
semi-empirical papers on fan cooling, but that sort of died out.

You can compute the thermodynamic limit, obviously, because you know the
inlet temperature, the air mass, the maximum component temperature, and
the heat dissipated, but most heat sink systems don\'t get anywhere near it.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[legg]

On Sat, 25 Jun 2022 02:10:59 -0000 (UTC), bob prohaska
<bp@www.zefox.net> wrote:

I\'m setting up a UPS for my computer/comms equipment using an
inverter/charger and battery from Amazon. The equipment draw
is only about 40 watts measured with a Kill-A-Watt, but all
the associated wallwarts use capacitive-input switching power
supplies. That means they only draw current at line peaks.

My seat-of-the-pants guess is that the duty cycle is around 10%,
meaning that the average 40 watts is really 400 watts 10% of the time.
That\'s well within the continuous power rating of the inverter, which
is 800 watts, so it\'s likely the setup will work as it is.

The question is: Can the peak load be made closer to the average
load by putting an inductor in the AC line feeding the wallwarts?

If anybody\'s been through this exercise I\'d be grateful for guidance.

Thanks for reading,

bob prohaska

I did some passive correction for off-the-shelf 60Hz linears in
the 80s. Best effects achieved using a saturable choke and
quasi-resonant capacitor, over a limited range of power levels
for any specific installation.

The parts are generally impractical for a hobbyist to get
ahold of, though restacking laminations from unvarnished scrap
is possible. Requires good VP Impregnation to silence the
final iteration.

The actual current phase angle shifted from leading to lagging
over the useful range. Output voltage into the 60Hz capacitive
load was flat-topped, affecting the low-line voltage performance.
The choke/cap combination supported the output difference
during line current reversal.

It\'s no good guessing what the current waveshape is; you\'ve
got to measure it / scope it. Line current transformers are
pretty cheap these days, often included in <$10 wattmeters
from off-shore sources. A scope is useful, but more expensive.

A lot of modern wall-warts are actually PFC compliant, through
the use of dedicated low power integrated controllers. These
employ valley-fill or critical-conduction (FM) off-line switchers
economically, at power levels as low as 5W.

Don\'t guess. Measure. Read specs of devices involved.

Don\'t go overboard. Your UPS output may be more tolerant of peak
loads than you assume, and your loads may be less peaky, simply
due to industry commodity trends and available parts.

RL

 

[Anthony William Sloman]

On Saturday, June 25, 2022 at 5:04:19 AM UTC+2, Don Y wrote:

On 6/24/2022 7:10 PM, bob prohaska wrote:
I\'m setting up a UPS for my computer/comms equipment using an
inverter/charger and battery from Amazon. The equipment draw
is only about 40 watts measured with a Kill-A-Watt, but all
the associated wallwarts use capacitive-input switching power
supplies. That means they only draw current at line peaks.

My seat-of-the-pants guess is that the duty cycle is around 10%,
meaning that the average 40 watts is really 400 watts 10% of the time.
That\'s well within the continuous power rating of the inverter, which
is 800 watts, so it\'s likely the setup will work as it is.

The question is: Can the peak load be made closer to the average
load by putting an inductor in the AC line feeding the wallwarts?

If anybody\'s been through this exercise I\'d be grateful for guidance.

You can actually make some very strange inverters

https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=1062238

from 1983 talks about doing some interesting stuff. Buying two transformers rather than just one back then made the whole approach too expensive to be interesting. It might be worth looking again while thinking about printed windings.

--
Bill Sloman, Sydney

 

[whit3rd]

On Saturday, June 25, 2022 at 5:31:43 AM UTC-7, bill....@ieee.org wrote:

On 6/24/2022 7:10 PM, bob prohaska wrote:

The question is: Can the peak load be made closer to the average
load by putting an inductor in the AC line feeding the wallwarts?

If anybody\'s been through this exercise I\'d be grateful for guidance.

You can actually make some very strange inverters

https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=1062238

from 1983 talks about doing some interesting stuff. Buying two transformers rather than just one back then made the whole approach too expensive to be interesting. It might be worth looking again while thinking about printed windings.

Oh, the time for that HAS come, this is one of the results:

<https://www.ti.com/lit/ds/symlink/ucc12050.pdf>

Good for a half watt, and kinda... tiny. The actual info on the circuit design is sketchy, because
there\'s not a lot of good symbols available to describe an integrated magnetcs circuit.

 

[whit3rd]

On Friday, June 24, 2022 at 9:31:00 PM UTC-7, Jasen Betts wrote:

On 2022-06-25, bob prohaska <b...@www.zefox.net> wrote:

The question is: Can the peak load be made closer to the average
load by putting an inductor in the AC line feeding the wallwarts?

Yes but only a little, the current phase angle is typically only
about 20 degrees leading so a line reactor will not help much. Most
of the power factor comes from crest factor rather than cos(phi).

It will work better if you put a bridge rectifier before the inductor.
(because now you can use a larger inductor), but now you\'ll have to
figure out which wall warts actually need AC, and only connect the
DC-capable ones.

The inductor coming after the rectifier, though, means it has to handle the DC
current without saturating; that can be a problem, and it does add weight
and cost. Doing such a design is a tricky task, not easy to simulate or
calculate, so is usually a cut-and-try exercise.

 

[legg]

On Wed, 29 Jun 2022 07:19:46 -0700, jlarkin@highlandsniptechnology.com
wrote:

On Wed, 29 Jun 2022 08:23:22 -0400, legg <legg@nospam.magma.ca> wrote:

On Tue, 28 Jun 2022 13:37:59 -0700 (PDT), Phil Allison
pallison49@gmail.com> wrote:

legg wrote:
================

** Learn to trim - asshole.


Anybody measured the ripple?

Electrolytic life is rms current dependent, among other things (like
temperature).

If you stick one next to a vacuum tube, it\'s received radiated that
dominates part temperature, and voltage stress of the app that
dominates.

** You just made that mad crap up.

If you\'ve ever calculated mtbf under Mil Hdbk 217, or Belcore,
you\'d be aware of dominating life factors.

Belcore has no derating factor for bad design. That often dominates
MTBF.

The stress factors for measured voltage current and temperature are
a fair indication of design integrity. Designs can be \'bad\' for
other reasons entirely.

Temp is the single determining factor on expected life.
Almost always the local ambient completely dominates.

Current forces self-rise due the part\'s ESR and limited
body surface area, (unless you stick it next to a hot radiator).
The actual relevant temperature is measured on the component\'s
body.

We are designing a fancy switching power supply and need a 20 uF cap
that can handle several amps RMS, a 250 KHz triangle from a
half-bridge and an inductor. We are thinking about using four 4.7 uF
radial-leaded film caps in parallel.

We have samples of several types on order. I plan to set up a test rig
and push amps of triangle into them and see how hot they get. May as
well snoop the waveform across each cap while I do that.

Film cap data sheets are typically not much help. They might spec a
few sinewave loss tangents and maybe allowable voltage vs frequency,
but rarely spec ESR or ESL or any thermals. We have to measure all
that.

It\'s always been difficult determining film cap ratings.
The dielectric losses usually have a positive temp co
above 50C and the parts themselves can have some of the
lowest part body temperature limits in the deck.

Part construction and materials quality, though largely the
product of automation, can vary. Not many other parts are
dependent on a coat of paint for env integrity.

Keep in mind that self-healing construction can mask
a considerable defect level. You\'ve probably pulled simple
film decoupling caps out who\'s capacitance measured almost
nothing compared to their original mfred values, due to
repeated intervening self-healing processes.

Philips published good information on polycarbonate, polyester
and polypropylene parts, when they were still in the business
of using them. Check their \'Components and Materials Part 15\'
prior to 1990. I don\'t think there\'s been much improvement on
that. Siemens tended to be more spotty. CDE did some mil stuff
that must have been backed up by something more than body
temperature rise. Polycarbonate development has lost some steam
due to issues with flammability.

There were film caps in most early fluorescent bulb replacements,
besides the usual HV electrolytic. At that temperature, it was a
toss-up which failed first.

Pulse-rated products get more attention to their reliability
and ratings. If your part doesn\'t provide sufficient data,
chances are it\'s a misapplication waiting to happen. Pulse
steepness dV/dT (~peak current) is dependent on both the
process characteristic voltage AND the part body size, so that
the amps/uF can actually reduce with an increase in leadspacing.

PPK and polyphenylene sulphide are potential rivals in some
applications, with slightly higher temperature limits, but I
don\'t recall anyone providing tabular or graphical data.

Can\'t imagine where my notes could be on that stuff. Today, even
the binder labels are illegible. Probably need a different kind
of \'specs\'.

RL

 

[legg]

Anyways, wade through it and you may come up with
something like this:

http://ve3ute.ca/query/Polycarbonate_Current_100KHz_Philips.pdf

It will be different for each dielectric, frequency of operation,
temperature etc.

RL

 

[bob prohaska]

Jan Panteltje <pNaonStpealmtje@yahoo.com> wrote:

On a sunny day (Sat, 25 Jun 2022 02:10:59 -0000 (UTC)) it happened bob
prohaska <bp@www.zefox.net> wrote in <t95qrj$r99$1@dont-email.me>:

I\'m setting up a UPS for my computer/comms equipment using an
inverter/charger and battery from Amazon. The equipment draw
is only about 40 watts measured with a Kill-A-Watt, but all
the associated wallwarts use capacitive-input switching power
supplies. That means they only draw current at line peaks.

[snip]
The question is: Can the peak load be made closer to the average
load by putting an inductor in the AC line feeding the wallwarts?


Interesting question, my cheap UPS seems to put out a square wave
I wondered if the flat tops are actually not better for the wall warts
as the charging part is longer than with a sine wave top...

AIUI, current into a capacitor is dv/dt x C, so the square wave should
be much worse in terms of peak current than my sine-wave case. Charging
is completed before the flat top of the cycle can commence. Seems likely
the leakage inductance of the inverter\'s transformer is dominant.

Been working now fine for a year or so with this thing,
comes in almost every day these days with mains company fiddling,..
flashing light bulbs sometimes here too.
To backup for longer times I have a pure sine wave 2 kW converter and a 250 Ah lifepo4 battery..
So I can keep watching sat TV or even cook food.
More than 10 wallwarts on that UPS now, some Raspberry Pi, some USB hubs, some cameras., also
security recorder, monitors... .. audio amp... 4 TB harddisks...
I would personally not bother with a a series inductor...

Your use case sounds very close to mine. If you aren\'t having trouble it
seems doubtful I will.

Thanks for writing,

bob prohaska

 

[Jasen Betts]

On 2022-06-25, whit3rd <whit3rd@gmail.com> wrote:

On Friday, June 24, 2022 at 9:31:00 PM UTC-7, Jasen Betts wrote:
On 2022-06-25, bob prohaska <b...@www.zefox.net> wrote:

The question is: Can the peak load be made closer to the average
load by putting an inductor in the AC line feeding the wallwarts?

Yes but only a little, the current phase angle is typically only
about 20 degrees leading so a line reactor will not help much. Most
of the power factor comes from crest factor rather than cos(phi).

It will work better if you put a bridge rectifier before the inductor.
(because now you can use a larger inductor), but now you\'ll have to
figure out which wall warts actually need AC, and only connect the
DC-capable ones.

The inductor coming after the rectifier, though, means it has to handle the DC
current without saturating; that can be a problem, and it does add weight and cost.

The current magnitude is the same as the ac current magnitude,
so not much of a problem. What bad thing happens if it saturates?

Doing such a design is a tricky task, not easy to simulate or
calculate, so is usually a cut-and-try exercise.

Last time I looked at it in detail, it seemed pretty easy to hit the
ballpark.

The aim is not a resonant filter, it\'s just making a current reservoir
so that (more) continuous current is drawn from the supply instead of spikes
on the peaks of the sine wave (or the edges of a modified square wave
inverter output)

--
Jasen.

 

[Clifford Heath]

On 30/6/22 05:38, Phil Hobbs wrote:

Even in an inviscid, incompressible fluid, the equations aren\'t all that
simple.  Forced-air cooling of macroscopic systems runs at some huge and
highly variable Reynolds number, depending on where you are.

Pretty sure that Reynolds number cannot be defined for an inviscid
liquid. But yeah, turbulent flows are difficult however you look at them.

 

[Phil Allison]

bob prohaska wrote:
======================

AIUI, current into a capacitor is dv/dt x C, so the square wave should
be much worse in terms of peak current than my sine-wave case.

** Nope - a rectified square wave is pure DC.
SMPSs used in pro-audio power amps are nearly all square wave inverters.

Battery to AC inverters are different.


...... Phil

 

[server]

On Wed, 29 Jun 2022 18:14:22 -0400, legg <legg@nospam.magma.ca> wrote:

Anyways, wade through it and you may come up with
something like this:

http://ve3ute.ca/query/Polycarbonate_Current_100KHz_Philips.pdf

It will be different for each dielectric, frequency of operation,
temperature etc.

RL

Gosh, real numbers. Thanks.

I just got this. It\'s a high voltage half-bridge test board for frying
inductors and film caps. People were doing another proto board so I
hung this on the end as a v-score breakaway.

https://www.dropbox.com/s/2x8z6yn29ab57xf/Z524_Wing_1.jpg?raw=1

 

[whit3rd]

On Saturday, June 25, 2022 at 3:10:38 PM UTC-7, palli...@gmail.com wrote:

bob prohaska wrote:

AIUI, current into a capacitor is dv/dt x C, so the square wave should
be much worse in terms of peak current than my sine-wave case.

** Nope - a rectified square wave is pure DC.
SMPSs used in pro-audio power amps are nearly all square wave inverters.

If the load on an AC source were a capacitor, you\'d get
current= dv/dt x C, and if the load were a rectifier into a (capacitor in parallel with a resistor)
you\'d get something more benign. In a power brick, the load is a rectifier and (resistor-loaded?)
regulator, like a capacitor in parallel with a current sink.

The ramp up when the rectifier diodes turn on will be abrupt, lots of peak current,
which was what the inductor was intended to moderate.

If the hypothetical inductor saturates, it\'ll make a buzzing sound; that\'s annoying.

 

[Phil Hobbs]

Clifford Heath wrote:

On 30/6/22 05:38, Phil Hobbs wrote:
Even in an inviscid, incompressible fluid, the equations aren\'t all
that simple.  Forced-air cooling of macroscopic systems runs at some
huge and highly variable Reynolds number, depending on where you are.

Pretty sure that Reynolds number cannot be defined for an inviscid
liquid. But yeah, turbulent flows are difficult however you look at them.

My point exactly.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[legg]

On Wed, 29 Jun 2022 19:36:21 -0700, jlarkin@highlandsniptechnology.com
wrote:

On Wed, 29 Jun 2022 18:14:22 -0400, legg <legg@nospam.magma.ca> wrote:

Anyways, wade through it and you may come up with
something like this:

http://ve3ute.ca/query/Polycarbonate_Current_100KHz_Philips.pdf

It will be different for each dielectric, frequency of operation,
temperature etc.

RL

Gosh, real numbers. Thanks.

I just got this. It\'s a high voltage half-bridge test board for frying
inductors and film caps. People were doing another proto board so I
hung this on the end as a v-score breakaway.

https://www.dropbox.com/s/2x8z6yn29ab57xf/Z524_Wing_1.jpg?raw=1

One good way of frying film caps is to thermally couple them
to hot chokes or power resistors.

Like a lot of other parts, they depend on thermal conduction
to the PCB.

RL