small inverter driver chip

[John Larkin]

Hi,

I have a nice 4-winding ISDN transformer I want to use to generate
some odd supply voltages from a +10 basic supply. I need a biggish
voltage to bias a PIN photodiode, and some -5 to run some analog
stuff. The transformer is 1:1:2:2, so there are possibilities.

I want to drive both ends of one untapped winding, so what would work
would be an oscillator, a flipflop, and a pair of rail-rail drivers,
like maybe a full-bridge thingie or a dual mosfet gate driver. Current
will be low, 100 mA maybe.

So, anybody know of a single, tiny chip that might do all this?
(Flyback wouldn't work well with this xfmr.)

John

 

[Joerg]

Hello John,

I have a nice 4-winding ISDN transformer I want to use to generate
some odd supply voltages from a +10 basic supply. I need a biggish
voltage to bias a PIN photodiode, and some -5 to run some analog
stuff. The transformer is 1:1:2:2, so there are possibilities.

I want to drive both ends of one untapped winding, so what would work
would be an oscillator, a flipflop, and a pair of rail-rail drivers,
like maybe a full-bridge thingie or a dual mosfet gate driver. Current
will be low, 100 mA maybe.

So, anybody know of a single, tiny chip that might do all this?
(Flyback wouldn't work well with this xfmr.)

I have done that with motor control chips a la LM18293. But many of
these have fallen from grace, going lalaland. More pricey, but also more
spiffy:

http://www.allegromicro.com/ic/motor.asp

 

[Roger]

Hi,

I have a nice 4-winding ISDN transformer I want to use to generate
some odd supply voltages from a +10 basic supply. I need a biggish
voltage to bias a PIN photodiode, and some -5 to run some analog
stuff. The transformer is 1:1:2:2, so there are possibilities.

I want to drive both ends of one untapped winding, so what would work
would be an oscillator, a flipflop, and a pair of rail-rail drivers,
like maybe a full-bridge thingie or a dual mosfet gate driver. Current
will be low, 100 mA maybe.

So, anybody know of a single, tiny chip that might do all this?
(Flyback wouldn't work well with this xfmr.)

John

Commercial circuits would at least be designed for centre tapped
transformers at these low powers.

However you could use 1 chip and an external bridge.

Make a H bridge circuit with PNP/NPN totem poles. Take a a TL494/594
which is a PWM controller with two uncomitted output transistors. Cross
connect the base's of the H bridge via a resistor and an output
transistor, that is the base of each PNP transistor is connected to a
resistor and then to the collector of one of the 494's output
transistors. The emitter of the output transistor is connected to the
base of the NPN transistor on the opposite arm of the bridge.

The TL494 does have deadtime control......you are going to need it as
this circuit is not fast and you will quicklt destroy small bipolars
with a minimum of shootthrough ;-)

BTW, it would appear that your problem is getting different voltages
rather than achieving high efficiency. Have you considered using one of
those little audio amplifier chips of the sort they use in e.g.
portable radios? Most of these can easily be made to self oscilate at
the top of the audio scale by means of a simple feedback network. Sine
wave solutions can achieve low noise without resorting to hi spec
capacitors.

 

[[email protected]]

Hi,

I have a nice 4-winding ISDN transformer I want to use to generate
some odd supply voltages from a +10 basic supply. I need a biggish
voltage to bias a PIN photodiode, and some -5 to run some analog
stuff. The transformer is 1:1:2:2, so there are possibilities.

I want to drive both ends of one untapped winding, so what would work
would be an oscillator, a flipflop, and a pair of rail-rail drivers,
like maybe a full-bridge thingie or a dual mosfet gate driver. Current
will be low, 100 mA maybe.

So, anybody know of a single, tiny chip that might do all this?
(Flyback wouldn't work well with this xfmr.)

You could try Baxandall's class-D oscillator - if you connect the 1:1
windings as centred-tapped pair, the 10V supply gives you 31.4V
peak-to-peak across each of the 2:2 windings, 62.8V if you put them
both in series. You'd need an additional inductor to put between the
cenetre-tap and the +10V supply, and two transistors (or MOS-FETs) to
make the system oscillate. No IC's at all.

Baxandall invented the class-D oscillator to drive high voltage
inverters - it sounds very much what you want.

I've referred to it here before - check Linear Technology's application
notes AN-45, AN-51, AN-61, AN-65. Jim Williams does describes the
inverter as a Royer inverter, but he's being a bit too inclusive.

 

[Joerg]

Hello Roger,

BTW, it would appear that your problem is getting different voltages
rather than achieving high efficiency. Have you considered using one of
those little audio amplifier chips of the sort they use in e.g.
portable radios? Most of these can easily be made to self oscilate at
the top of the audio scale by means of a simple feedback network. Sine
wave solutions can achieve low noise without resorting to hi spec
capacitors.

I often design isolated power transfer circuitry and have tried audio
amps. Mostly disappointing. Many of these do not like to go close to the
rails, lose steam up there and occasionally become quite unorthodox
(foldback, pumping etc.). And they got rather hot. In most cases I ended
up using cheap motor drivers. Digikey has tons of them.

 

[Roger]

Hello Roger,

I often design isolated power transfer circuitry and have tried audio
amps. Mostly disappointing. Many of these do not like to go close to the
rails, lose steam up there and occasionally become quite unorthodox
(foldback, pumping etc.). And they got rather hot. In most cases I ended
up using cheap motor drivers. Digikey has tons of them.

The key here is efficiency, the OP appears to have quite low power
requirements and therefore would not be needing to drive them to thier
limits if efficiency is not a problem. He mentioned 10V/100mA, that's
1W total output power (or 100W super real cool music power), so even at
50% efficiency it aint going to get that hot unless it is really tiny
;-)

Having said all that, your solution of using motor drivers does make
more sense unless he wants low noise (negative supplies suggests that
he is doing precision analog stuff).

 

[Joerg]

Hello Roger,

The key here is efficiency, the OP appears to have quite low power
requirements and therefore would not be needing to drive them to thier
limits if efficiency is not a problem. He mentioned 10V/100mA, that's
1W total output power (or 100W super real cool music power), so even at
50% efficiency it aint going to get that hot unless it is really tiny
;-)

Having said all that, your solution of using motor drivers does make
more sense unless he wants low noise (negative supplies suggests that
he is doing precision analog stuff).

That's a good point. With an audio amp you could actually regulate the
input level into the transformer if you wanted to. If that loop makes
sure it stays far enough off the rails you'd be enjoying almost the
whole PSRR of that audio amp in noise isolation. There could be a
thunderstorm going on on that 10V rail and this supply would remain
blissfully unaware of it.

If it had to be rock bottom in cost I'd probably do this one as a
blocking oscillator, in discretes.

 

[Terry Given]

Hi,

I have a nice 4-winding ISDN transformer I want to use to generate
some odd supply voltages from a +10 basic supply. I need a biggish
voltage to bias a PIN photodiode, and some -5 to run some analog
stuff. The transformer is 1:1:2:2, so there are possibilities.

I want to drive both ends of one untapped winding, so what would work
would be an oscillator, a flipflop, and a pair of rail-rail drivers,
like maybe a full-bridge thingie or a dual mosfet gate driver. Current
will be low, 100 mA maybe.

So, anybody know of a single, tiny chip that might do all this?
(Flyback wouldn't work well with this xfmr.)

John

Why not use a pair of complementary FETs, in a diagonal half-bridge
arrangement. Drive the low-side N-FET as per usual, and the N-FET drain
can provide the gate drive for the high-side P-FET. I do that all the
time, its pretty darned cheap. All one needs to do is ensure the FETs
allowable gate voltage range is sufficient for the Vcc used

Cheers
Terry

 

[Mike Monett]

Why not use a pair of complementary FETs, in a diagonal half-bridge
arrangement. Drive the low-side N-FET as per usual, and the N-FET drain
can provide the gate drive for the high-side P-FET. I do that all the
time, its pretty darned cheap. All one needs to do is ensure the FETs
allowable gate voltage range is sufficient for the Vcc used

Cheers
Terry

How do you handle shoot-through?

Regards,

Mike Monett

 

[Terry Given]

How do you handle shoot-through?

Regards,

Mike Monett

its a diagonal half-bridge, so the primary inductance is in the way
(P-FET - top LHS switch, N-FET = bottom RHS switch, the opposite
diagonals are diode)

ergo no pr0oblems at all with shoot-thru, and the P-FET gatedrive can be
fairly slow and it still works fine.

I have made hundreds of thousands of these, ranging from 50kHz to 2MHz.

Cheers
Terry

 

[Clifford Heath]

its a diagonal half-bridge,

I think you're saying that you generate +ve and -ve supplies
out of one inductor just by deciding which FET to turn off
first... have I got that straight? Cute idea anyhow.

 

[Terry Given]

I think you're saying that you generate +ve and -ve supplies
out of one inductor just by deciding which FET to turn off
first... have I got that straight? Cute idea anyhow.

no.

its piss easy, I shouldnt need to draw a picture.

- make a single-ended forward converter.
- bung a p-fet in between the +Vdc supply and the winding
- connect its gate to the drain of the lower N-FET
- slap a diode from P-FET drain to 0V (anode to 0V)
- slap a diode from N-FET drain to +V (cathode to +Vdc)

voila, now you can throw away the reset winding and diode.

either both FETs are on, and current flows into the winding,

or both FETs are off, and the mag current commutates back to the supply
thru the 2 diodes.

limit duty cycle to 50%


Cheers
Terry

 

[Roger]

How do you handle shoot-through?

Low power MOSFET designs can get away with shoot-througth ;-) MOSFETS
can handle large peak currents without the secondary breakdown problems
that plagued bipolars. Add to that the fast switching speed and all you
actually get is a brief spike that is limited only by circuit
impedance. But as such crude design techniques are often associated
with lo grade caps on the supply, then the current in the spike will
not blow the MOSFET.

BTW, I am not recommending the technique, just pointing out why you can
do it

 

[Mike Monett]

BTW, I am not recommending the technique, just pointing out why you can
do it

I'm glad you are not promoting it. The current spike would be noisy and
hard to filter. It would be uncontrolled and would change with temperature,
supply voltage, circuit timing, and normal part variations.

Maybe ok for cheap toy products, but it's usually a good idea to avoid
generating spikes like this in instrumentation designs. Or in designs you
intend to show in sed

Regards,

Mike Monett

 

[Mike Monett]

its a diagonal half-bridge, so the primary inductance is in the way
(P-FET - top LHS switch, N-FET = bottom RHS switch, the opposite
diagonals are diode)

ergo no pr0oblems at all with shoot-thru, and the P-FET gatedrive can be
fairly slow and it still works fine.

I have made hundreds of thousands of these, ranging from 50kHz to 2MHz.

Cheers
Terry

Hi Terry,

Thanks for the reply. John had asked for a full bridge due to the turns
ratio on his ISDN transformer:

I want to drive both ends of one untapped winding, so what would
work would be an oscillator, a flipflop, and a pair of rail-rail
drivers, like maybe a full-bridge thingie or a dual mosfet gate
driver. Current will be low, 100 mA maybe.

So I assumed you intended to use this technique in a full bridge. This
would definitely cause problems with shoot-through.

I played with your circuit in SPICE. Here are some observations:

1. Current through the load is always in the same direction. This
may cause problems with core saturation in John's application.

2. Multiple diodes reduce the voltage delivered to inductive loads.

3. Turnoff is not well controlled. A bit problematic at 2MHz.

Otherwise it may be useful in some applications.

Regards,

Mike Monett

 

[Spehro Pefhany]

its a diagonal half-bridge, so the primary inductance is in the way
(P-FET - top LHS switch, N-FET = bottom RHS switch, the opposite
diagonals are diode)

ergo no pr0oblems at all with shoot-thru, and the P-FET gatedrive can be
fairly slow and it still works fine.

I have made hundreds of thousands of these, ranging from 50kHz to 2MHz.

Cheers
Terry

The two-switch topology works nicely at high power and voltage (with
different gate drive, obviously) because the stress on the transistors
is low. See Billings Chapters 10/11. Pressman calls it "double-ended
forward converter" topology, also called the "two transistor (or
switch) forward converter" or "asymmetric forward converter".

The turns ratios may not match with the voltages John requires
(although starting with a 10V supply they might, depending on what
'biggish' translates to, and if it could be series with the 10V input,
which could get it into the 20V+ range using a 1 & 2 windings in
series and series with the input, and still leaving a 2 winding for -5
or maybe -7V with max duty cycle ~0.4). Or perhaps -3 will be enough
for the negative supply and both 2 windings can be used for the bias
voltage.

Since deadtime isn't an issue, maybe an SOT-23 PIC10F, an SMT dual
complementary MOSFET and a dual Shottky (BAT54?)


Best regards,
Spehro Pefhany

 

[Mike Monett]

The two-switch topology works nicely at high power and voltage (with
different gate drive, obviously) because the stress on the transistors
is low. See Billings Chapters 10/11. Pressman calls it "double-ended
forward converter" topology, also called the "two transistor (or
switch) forward converter" or "asymmetric forward converter".

If that's the same thing Terry proposed, why even bother switching the top
end of the load? Why not simply tie it to VCC along with a diode clamp?

The turns ratios may not match with the voltages John requires
(although starting with a 10V supply they might, depending on what
'biggish' translates to, and if it could be series with the 10V input,
which could get it into the 20V+ range using a 1 & 2 windings in
series and series with the input, and still leaving a 2 winding for -5
or maybe -7V with max duty cycle ~0.4). Or perhaps -3 will be enough
for the negative supply and both 2 windings can be used for the bias
voltage.

There still may be a problem with a dc component. The ISDN transformer may
not handle it very well when asked to deliver power.

Since deadtime isn't an issue, maybe an SOT-23 PIC10F, an SMT dual
complementary MOSFET and a dual Shottky (BAT54?)

Gawd, now I've heard everything. A programmable timer Will have to study
that some more.

My approach would be a 74AC74 oscillator, divide-by-two and deadtime
generator, 74AC00 output driver and gate, two npn's and two pnp's as the H-
bridge, and 5 diodes, 6 resistors and two caps (plus bypass as needed).
Lots more parts, but it would work from 5V and give the full H bridge as
John asked.

Best regards,
Spehro Pefhany

Regards,

Mike Monett

 

[John Larkin]

If that's the same thing Terry proposed, why even bother switching the top
end of the load? Why not simply tie it to VCC along with a diode clamp?


There still may be a problem with a dc component. The ISDN transformer may
not handle it very well when asked to deliver power.


Gawd, now I've heard everything. A programmable timer Will have to study
that some more.

My approach would be a 74AC74 oscillator, divide-by-two and deadtime
generator, 74AC00 output driver and gate, two npn's and two pnp's as the H-
bridge, and 5 diodes, 6 resistors and two caps (plus bypass as needed).
Lots more parts, but it would work from 5V and give the full H bridge as
John asked.

74AC74 oscillator? I made an oscillator from a flipflop once, but it
took a lot of parts and broke some rules.

I can do it (and probably will) with a sot-23 schmitt/rc oscillator,
an sc-70 flipflop, and an so-8 dual mosfet gate driver (a little one,
rated 4 amps per side), total 5 parts and about 1 cm square, plus a
couple of bypasses somewhere. Oh, there's +5 available for other
reasons.

If I got brave, I'd make the osc with just the schmitt gate and a
resistor, no cap.

Maxim makes an oscillator-flipflop-driver, but the outputs are
open-drain, and that won't work with my transformer.

What I want is to derive some regulated -5 for opamps and something
like +20 to +30 to bias the photodiode. The pin photodiodes get faster
as the backbias increases, partly because the capacitance goes down
but mainly because the increased field sweeps carriers out faster.

So, 10 volt diff drive into a "1" winding. The other "1" winding gets
bridge rectified to -8, then lin reg'd to -5. The two "2" windings are
seriesed and diode rectified to +40ish, then zenered down to +28 for
the pin bias, something like that.

I have quantities of the IDSN transformer, so I'd like to use it here.
It's cute, about a cm cube, surface mount, under a dollar each. What I
don't have is a lot of space.

John

 

[Harry Dellamano]

its piss easy, I shouldnt need to draw a picture.

- make a single-ended forward converter.
- bung a p-fet in between the +Vdc supply and the winding
- connect its gate to the drain of the lower N-FET
- slap a diode from P-FET drain to 0V (anode to 0V)
- slap a diode from N-FET drain to +V (cathode to +Vdc)

voila, now you can throw away the reset winding and diode.

either both FETs are on, and current flows into the winding,

or both FETs are off, and the mag current commutates back to the supply
thru the 2 diodes.

limit duty cycle to 50%


Cheers
Terry

Nice, ...your 50% limit is because of the diode from N-FET drain to +Vdc.
Remove cathode from +Vdc and tie to a storage cap to ground. This output
forms a boost circuit. Set duty to 75% and get OPs 30V boosted output.
Windings on inductor/transformer will generate the other voltages needed.
POC!
Harry

 

[Harry Dellamano]

Nice, ...your 50% limit is because of the diode from N-FET drain to +Vdc.
Remove cathode from +Vdc and tie to a storage cap to ground. This output
forms a boost circuit. Set duty to 75% and get OPs 30V boosted output.
Windings on inductor/transformer will generate the other voltages needed.
POC!
Harry

Err.. But a simple boost configuration from 10V to 30V beats all. Set duty
to 67%. Use the two N=2 winding in parallel for the primary boost inductor.
Use the two N=1 winding in parallel, half wave rectify for -5VDC.
Send bottle of Tequila to Harry for this design effort.