Low efficiency from SEPIC converter

[Liam]

Hi
My company has had a long standing design based on the LT1371 in a
SEPIC configuration. Its worked fine in that application but we had a
bit of shock when we found its efficiency is only 55%.
Our new application requires as high a figure as possible.

The circuit used is the one in the data sheet. One possible problem is
that our application normally draws 80mA ( 55 % efficient ) but
occasionally requires a pulse of 0.5A.

The components are as shown on the circuit except for the diode which
is a MBRS340T3 and C2 which is an AVX TPS series ( could'nt get hold
the specified one ).

Does anyone have any idea where we've gone wrong. The linear tech
switchercad software reckons we should get 85% efficiency.

Thanks in advance
Liam

 

[Pooh Bear]

Hi
My company has had a long standing design based on the LT1371 in a
SEPIC configuration. Its worked fine in that application but we had a
bit of shock when we found its efficiency is only 55%.
Our new application requires as high a figure as possible.

The circuit used is the one in the data sheet. One possible problem is
that our application normally draws 80mA ( 55 % efficient ) but
occasionally requires a pulse of 0.5A.

The components are as shown on the circuit except for the diode which
is a MBRS340T3 and C2 which is an AVX TPS series ( could'nt get hold
the specified one ).

Does anyone have any idea where we've gone wrong. The linear tech
switchercad software reckons we should get 85% efficiency.

I've had disappointing efficiency from an LT switcher device many yrs back
too.

I would suspect indifferent inductor performance as something worth
investigating initially.


Graham

 

[TP]

I've had disappointing efficiency from an LT switcher device many yrs back
too.

I would suspect indifferent inductor performance as something worth
investigating initially.


Graham

As well as trying an inductor with lower resistance, you might look at
the DC coupling capacitor and the output capacitor. Higher voltage
rating, higher uF, lower ESR will each contribute to better efficiency
if that is where the weakness lies.

TP

 

[Winfield Hill]

Pooh Bear wrote...

I've had disappointing efficiency from an LT switcher device many
yrs back too. I would suspect indifferent inductor performance as
something worth investigating initially.

Indeed. Many inductors are quite lossy at 50kHz, suffering from
copper losses (exacerbated by skin depth, and proximity effect) and
from ferrite core losses at high flux levels. Liam, does your poor
efficiency occur only under the light load, or at full load as well?
What's the converter's switching duty cycle at 80mA? What's your
input voltage? Have you examined the current waveforms?

 

[Winfield Hill]

Winfield Hill wrote...

Indeed. Many inductors are quite lossy at 50kHz, suffering from ...

Oops, make that 500kHz.

 

[Ken Smith]

[... snipped ..]
(I changed 50 to 500 per followup)

Indeed. Many inductors are quite lossy at 500kHz, suffering from
copper losses (exacerbated by skin depth, and proximity effect) and
from ferrite core losses at high flux levels. Liam, does your poor
efficiency occur only under the light load, or at full load as well?
What's the converter's switching duty cycle at 80mA? What's your
input voltage? Have you examined the current waveforms?

Also:

What is warming up?
Is the inductor getting warm?

If there is way too much stray inductance getting to the output rectifier
circuit, you will see a spike in the transistors waveform just as it turns
off. Chances are all of the energy in the spike ends up as heat.

Slow recovery in the output side rectifier will show up as a current spike
at the instant of the transistor switching on. This is bad because the
extra power that the transistor is having to eat is coming from the load
side of the system so all the other losses are applied to it then its
lost.

What does the gate drive of the MOSFET look like when measured right at
the package. If you have a lot of resistance or inductance in either the
source or gate leg, the drive as seen by the power device can be quite
different than the drive from the controller chip. An inductive current
sense resistor can be real trouble in this regard.

 

[Winfield Hill]

Ken Smith wrote...

Also:

What is warming up? Is the inductor getting warm?

Good suggestions, but a few caveats...

If there is way too much stray inductance getting to the output rectifier
circuit, you will see a spike in the transistors waveform just as it turns
off. Chances are all of the energy in the spike ends up as heat.

This is generally a potential problem when flyback transformer leakage
inductance is present, but Liam is using an inductor.

Slow recovery in the output side rectifier will show up as a current spike
at the instant of the transistor switching on. This is bad because the
extra power that the transistor is having to eat is coming from the load
side of the system so all the other losses are applied to it then its lost.

Reverse recovery delay losses are a problem with silicon rectifier diodes,
but he's using Schottky barrier diodes, with no reverse recovery loss.

What does the gate drive of the MOSFET look like when measured right at
the package. If you have a lot of resistance or inductance in either the
source or gate leg, the drive as seen by the power device can be quite
different than the drive from the controller chip.

A potential problem if there were a MOSFET, but the LT1371 has an internal
BJT switch (which drops 0.2V at 0.5A in, 0.57V at 2A in for 100C junction).

Question for Liam, what are your input and output voltages and currents
under the two operating regimes? Perhaps you need a serious external-
MOSFET stepup converter. This thread has a discussion about using one
low-power low-voltage converter to create a high voltage to operate a
conventional high-power converter. http://groups.google.com/groups?
hl=en&lr=&ie=UTF-8&threadm=c97hvb023gq%40drn.newsguy.com

An inductive current sense resistor can be real trouble in this regard.

Yes, but the LT1371 has an internal low-inductance 0.04-ohm resistor.
http://www.linear-tech.com/prod/datasheet.html?datasheet=76

 

[Ken Smith]

Ken Smith wrote...

Good suggestions, but a few caveats...


This is generally a potential problem when flyback transformer leakage
inductance is present, but Liam is using an inductor.

I wasn't looking at his schematic as I wrote this. There still can be too
much stray inductance if the layout is very silly so I don't totally rule
it out.


A few more:

Non-coupled inductor SEPIC converters need very good coupling capacitors.
Is the cap getting warm?

Is the control loop galloping and taking the inductor up to saturation on
every other cycle?

Are you sure of your measurements? I've had DVMs do funny things with
high frequency noise.

 

[Winfield Hill]

Ken Smith wrote...

A few more:

Non-coupled inductor SEPIC converters need very good coupling capacitors.
Is the cap getting warm?

Is the control loop galloping and taking the inductor up to saturation on
every other cycle?

Are you sure of your measurements? I've had DVMs do funny things with
high frequency noise.

Those are good ones.