HS gate drive transformer?

[Hammy]

I wound a toroid to use as a HS gate drive transformer for a 100W two
switch flyback. I'm using the typical DC restore technique AC couple
the transformer.

Here's my schematic;

http://i42.tinypic.com/9gvyow.png

I'm using a PIC and a FET driver to simulate the extremes of the PWM
controller duty cycle and drive level and it works well with the
exception of Burst Mode. Burst mode is when the controller modulates
the original gate signal at light load for anyone who doesn't know.

A picture is worth a thousand words.

Here's my gate waveform at 90kHz 70% duty. A little sloping but
acceptable.

http://i39.tinypic.com/25kieyq.jpg

Now here's where the fun begins I'm modulating the 90kHz to simulate
burst mode.

http://i44.tinypic.com/qy5l69.jpg

Here's a zoom of one of the burst.

http://i43.tinypic.com/2eumxbk.jpg

My primary and secondary are 1:1 ; inductance of 1.5mH.

Is there anyway to clean this up?

 

[Joerg]

I wound a toroid to use as a HS gate drive transformer for a 100W two
switch flyback. I'm using the typical DC restore technique AC couple
the transformer.

Here's my schematic;

http://i42.tinypic.com/9gvyow.png

I'm using a PIC and a FET driver to simulate the extremes of the PWM
controller duty cycle and drive level and it works well with the
exception of Burst Mode. Burst mode is when the controller modulates
the original gate signal at light load for anyone who doesn't know.

A picture is worth a thousand words.

Here's my gate waveform at 90kHz 70% duty. A little sloping but
acceptable.

http://i39.tinypic.com/25kieyq.jpg

Now here's where the fun begins I'm modulating the 90kHz to simulate
burst mode.

http://i44.tinypic.com/qy5l69.jpg

Ah, another guy with an Instek scope.

Here's a zoom of one of the burst.

http://i43.tinypic.com/2eumxbk.jpg

My primary and secondary are 1:1 ; inductance of 1.5mH.

Is there anyway to clean this up?

Looks like you haven't wound your toroid transformer in bifilar fashion.
That is essential for apps like this, leakage inductance must be
minimized, else this rings like crazy.

 

[Hammy]

What are the values of the parts? Something is resonating at 12 KHz.

The burst has some very low-frequency DC components. Dump burst mode!

John

I'm using 0.1uf caps and a 4.7k gate resistor the diode is just a
1n4148.

The burst mode frequency isn't specified in the data sheet so your
right I'm probably going to have to dump it. The controller is the
NCP1217D 100kHz.

The only reason I wanted to keep the feature is the supply may not
have a load on it all the time or a small load. The data sheet doesn't
say what the minimum duty is and I don't want the OVP kicking on and
latching the controller off when lightly loaded.

I guess I'll test the controller for minimum duty and ensure there is
a minimum load for it.

 

[Hammy]

Ah, another guy with an Instek scope.

Joerg do you know how to take screen shots that show the measurement
screen ? I'm assuming you can do it because they show pictures of
screen shots in the manual with the measurement window.

Looks like you haven't wound your toroid transformer in bifilar fashion.
That is essential for apps like this, leakage inductance must be
minimized, else this rings like crazy.

Yep its wound Bifilar. Its also 5 turns past the minimum to avoid
saturating for its worst case Duty and Drive level and I checked it
using a current transformer.

Its wound tight and all the way around the toroid.

 

[Hammy]

That is the one thing i hate about those stupid "controllers" that
supposedly are good for DC-DC conversion.
Why simulate?
Toss the verdammdt controller and use the PIC directly; then you can
have the program be as dirty or clean (or both) as you want...
..Namely slow non-giga-amp startup current spikes!

I have read through some decent DC/DC conversion app notes using PIC's
and for LVDC I might try one down the road. This is an offline flyback
for 90-140VAC in 100W out. It would get a little messy using one for
that.

The bridge rectifier is beefy and more then capable of handling the
turn on current surge the fuse and line filter help to reduce it. I
see your point though you can deal with a lot of things in software.
Adjusting the frequency (PFM) to suit the load demands, soft start and
less dissipative current sense etc.

The only thing is PFM usually isn't recommended for line powered
converters because it's difficult to design a line filter when the
converter can operate over a wide frequency range. Soft start can be
implemented on the secondary side with a couple of components I've
done it several times before you have to tweak it but it works.

I like PIC's and use them a lot but they have their places the
controller in a line powered converter isn't one of them.

 

[Hammy]

Use optos, and the transformer to generate a floating voltage
for a driver directly connected to the MOSFET?

It's an option I have some fast Avago gate drive opto's.

 

[Tim Williams]

I'm using 0.1uf caps and a 4.7k gate resistor the diode is just a
1n4148.

0.1uF resonates with 1.5mH transformer inductance as 13.0kHz. Sounds like a
'duh' factor to me.

Tim

 

[Hammy]

0.1uF resonates with 1.5mH transformer inductance as 13.0kHz. Sounds like a
'duh' factor to me.

Tim

Yea I thought of that later. I have been known for a few brain farts
though;-).

The problem is the burst frequency isn't specified and it can vary.
They don't even give a frequency range just some scope shots at
various loads showing different frequencies. On semis usually pretty
good about data sheets but this ones pretty sparse on some important
stuff.

It is a nice feature though if your supply is going to be powered with
a small or no load at least it wont pump up past the nominal voltage.

 

[Joerg]

Joerg do you know how to take screen shots that show the measurement
screen ? I'm assuming you can do it because they show pictures of
screen shots in the manual with the measurement window.

Ahm, didn't you just do that? Not sure what you mean here. Mostly I use
their Freewave program but on the road also a USB stick. You can turn
the menu off with the little gray round button. If you want a totally
clean screen with just the plot, or an Excel-capable file you can also
do that via Freewave. Also from the bare scope, if you hit "save all" it
stores both formats on the USB stick.

Yep its wound Bifilar. Its also 5 turns past the minimum to avoid
saturating for its worst case Duty and Drive level and I checked it
using a current transformer.

Its wound tight and all the way around the toroid.

But something is ringing. How are you measuring this? Are the caps sized
right? Your schematic is only partial so I can't see what happens south
of the FET's source. If the source is flying up and down that's a
challenge. This scope isn't that great for differential measurements at
high sample rates.

 

[Joerg]

That is the one thing i hate about those stupid "controllers" that
supposedly are good for DC-DC conversion.
Why simulate?
Toss the verdammdt controller and use the PIC directly; then you can
have the program be as dirty or clean (or both) as you want...
..Namely slow non-giga-amp startup current spikes!

True PWM is next to impossible with a regular uC. That is because it has
a finite granularity of no better than one master clock cycle. At
contemporary frequencies that's not very good. For example, my last
design runs above half a megahertz. Try that with a PIC

 

[Joerg]

It's an option I have some fast Avago gate drive opto's.

Optocouplers aren't very well suited to drive FETs. They can't muscle a
gate capacitance around within tens of nanoseconds and at a couple of
amps. Transformers are the better choice.

Size the caps right, and maybe optimize the transformer coupling some
more (twisting the wires before winding etc. unless you did that
already). Burp mode in a converter isn't that great with transformers
but there aren't many alternatives. In ultrasound we do this a lot,
transmit a sequence of fast power pulses across a transformer, followed
by a major pause (to wait for echoes).

 

[Hammy]

Ahm, didn't you just do that? Not sure what you mean here. Mostly I use
their Freewave program but on the road also a USB stick. You can turn
the menu off with the little gray round button. If you want a totally
clean screen with just the plot, or an Excel-capable file you can also
do that via Freewave. Also from the bare scope, if you hit "save all" it
stores both formats on the USB stick.

I meant the side screen which shows measurements automatically like
rise/fall, PP,AVG,RMS.....

Probably the only way to do it is with Freewave. The software doesn't
work on my XP machine but it works on my W2K pro machine?

But something is ringing. How are you measuring this? Are the caps sized
right? Your schematic is only partial so I can't see what happens south
of the FET's source. If the source is flying up and down that's a
challenge. This scope isn't that great for differential measurements at
high sample rates.

I have 0.1uf caps on the input and output. Tim's right the 1.5mH and
0.1uf caps are resonating with the burst frequency.

I'm not taking a differential measurement my probe tip is right at the
gate of the FET and the ground clip is right at the source.

I don't think its possible to do this without a floating supply and
using an opto to send the HS control signal to a driver. Even if I
disable the burst mode the controller could still go into burst mode
under fault condition like over current. To many chances something
could go wrong.

I suppose I could use an old UC384X it has none of these fancy
features and a 0.5 duty cycle clamp. Other then the unpredictability
of the burst mode it does work well.

I was also planning on using synchronous rectification a shottky
dissipates 4W but this FET would have less then 1W conduction losses.
I thought I would try something different;-)

http://www.nxp.com/documents/data_sheet/PSMN8R2-80YS.pdf

They are coming out with some nice FETS.

 

[Joerg]

There are some fast optocoupled mosfet gate drivers.

The home-brew way to do this is to use a cheap (ca $4) potted DC-DC
converter to make a supply that floats on the fet source. Then use a
fast logic-type optocoupler and a real gate driver chip up there.

That's a lot of stuff for a 50W-100W converter

The advantage of optos is DC coupling without duty-cycle problems. And
not having to wind transformers.

Another option is to use a chip with proper bootstrap circuitry.

So far transformers have been good to me. The only problem is that
off-the-shelf gate transformers are IMHO way overpriced and the
reasonably priced ones from China are custom runs and have MOQs. Which
for a product design doesn't really matter. It seems the usual COTS
vendors don't really realize how much of their lunch is being eaten.
Maybe time to invest in Rinminbi, if I just knew where ;-)

 

[Joerg]

I meant the side screen which shows measurements automatically like
rise/fall, PP,AVG,RMS.....

If you don't turn off the menu then the side screen will be stored in
the image file.

Probably the only way to do it is with Freewave. The software doesn't
work on my XP machine but it works on my W2K pro machine?

It does work on XP, it runs on a laptop on the lab bench here and that's
an XP machine. However, MS screwed up backwards compatibility in .NET
and you must load its version 1.1 environment. Freewave did not install
correctly with version 2.0 but .NET 1.1 can coexist peacefully with 2.0.

It's really nice, I take measurements for clients and then send the
results over the LAN into the client directory on the office server.
Then head to the big PC to compile the reports. Ok, it doesn't transfer
my cup of coffee along

I have 0.1uf caps on the input and output. Tim's right the 1.5mH and
0.1uf caps are resonating with the burst frequency.

Oops ...

I'm not taking a differential measurement my probe tip is right at the
gate of the FET and the ground clip is right at the source.

With a flying source that's a problem because the Instek connects the
probe shield to PE. So you get ground loops and tons of ringing. Plus
you neighbors listening to AM radio won't be very happy about that. You
can get battery packs for many Instek models but it's very expensive. I
never checked whether one can hack into this and connect one's own
battery. Even then you'd at least need to wrap the probe through a nice
big ferrite a few times.

Maybe at your frequency range you could try a diff measurement with two
channels.

I don't think its possible to do this without a floating supply and
using an opto to send the HS control signal to a driver. Even if I
disable the burst mode the controller could still go into burst mode
under fault condition like over current. To many chances something
could go wrong.

With properly size caps it should be possible.

I suppose I could use an old UC384X it has none of these fancy
features and a 0.5 duty cycle clamp. Other then the unpredictability
of the burst mode it does work well.

Question: Since this is a flyback why is the source flopping about?
What's it doing? Maybe post the whole schematic if it isn't confidential.

I was also planning on using synchronous rectification a shottky
dissipates 4W but this FET would have less then 1W conduction losses.
I thought I would try something different;-)

http://www.nxp.com/documents/data_sheet/PSMN8R2-80YS.pdf

They are coming out with some nice FETS.

But you have to drive over 3500pF of gate capacitance. Make sure the
Rdson versus gate capacitance deal is somewhat optimized for the chosen
frequency and power level. Ideally the losses due to both effects would
be about on par.

It's pretty brazen that they advertize this as a 130W device. I wonder
what it's lifetime will be under that sort of load (in milliseconds,
from turn-on to a sharp popping noise).

 

[Hammy]

That's a lot of stuff for a 50W-100W converter



Another option is to use a chip with proper bootstrap circuitry.

Yes but I think you might have problems as well with burst mode.

So far transformers have been good to me. The only problem is that
off-the-shelf gate transformers are IMHO way overpriced and the
reasonably priced ones from China are custom runs and have MOQs. Which
for a product design doesn't really matter. It seems the usual COTS
vendors don't really realize how much of their lunch is being eaten.
Maybe time to invest in Rinminbi, if I just knew where ;-)

My input is North American only it's my supply; my use only. I've made
small floating supplies with a cheap PIC10F controlling a small pulse
transformer and floated a driver on it for an SSR. I used the PIC for
zero cross detect and to drive an opto for controlling the floating
driver. The supply just ran open loop the output was clamped with a
zener so it never exceeded the Vgs of the FET's. The footprint was
small and I could vertically mount it I guess.

The whole thing only consumed a couple of mA, it wasn't switching
though just on/off.

Switching at 100kHz with a PIC and fet driver I'm getting about 6mA
quiescent current 3mA or so is probably due to the 78l05 for the PIC.

Avago does make some opto's with totem outputs sink/source couple of
amps at least. Depending on what you want to spend 1uS propagation
delay's down to the low nS.

 

[Joerg]

Yes but I think you might have problems as well with burst mode.

Not really, just make sure that the bootstrap voltage remains 10V or so
above the source at all times.

My input is North American only it's my supply; my use only. I've made
small floating supplies with a cheap PIC10F controlling a small pulse
transformer and floated a driver on it for an SSR. I used the PIC for
zero cross detect and to drive an opto for controlling the floating
driver. The supply just ran open loop the output was clamped with a
zener so it never exceeded the Vgs of the FET's. The footprint was
small and I could vertically mount it I guess.

The whole thing only consumed a couple of mA, it wasn't switching
though just on/off.

Switching at 100kHz with a PIC and fet driver I'm getting about 6mA
quiescent current 3mA or so is probably due to the 78l05 for the PIC.

Sound about normal.

Avago does make some opto's with totem outputs sink/source couple of
amps at least. Depending on what you want to spend 1uS propagation
delay's down to the low nS.

If you want to go opto you'd probably be better off with John's
suggestion. Get a fast but run-of-the-mills weak optocoupler, then
follow it with a staunch MIC4421 or some other driver for a little over
a buck.

The fancy chips will always be expensive because us guys never use them
in product designs. We only would if they'd be cheap and 2nd-sourced,
sort of the chicken and egg problem.

 

[Hammy]

With a flying source that's a problem because the Instek connects the
probe shield to PE. So you get ground loops and tons of ringing. Plus
you neighbors listening to AM radio won't be very happy about that. You
can get battery packs for many Instek models but it's very expensive. I
never checked whether one can hack into this and connect one's own
battery. Even then you'd at least need to wrap the probe through a nice
big ferrite a few times.

Yep they defiantly aren't giving the battery option away.

Maybe at your frequency range you could try a diff measurement with two
channels.



With properly size caps it should be possible.

I haven't given up yet I'll play around with what I have part wise and
see if I cant get something reliable.

Question: Since this is a flyback why is the source flopping about?
What's it doing? Maybe post the whole schematic if it isn't confidential.

No not confidential I haven't made up a full schematic yet.

Here's my actual test schematic. I'm not showing Bypass caps and the
lm317 for the driver and the 78l05 for the PIC both regulators get
their input from the 20 V supply V1.

http://i39.tinypic.com/1zf4gpl.png

That is just a test set-up to test the high side drive for a active
clamp flyback shown below.

http://i39.tinypic.com/10s7ybl.png

I don't have a schematic for the flyback because I just started
considering it yesterday. I'm just working out how to drive the
High-side FET.

But you have to drive over 3500pF of gate capacitance. Make sure the
Rdson versus gate capacitance deal is somewhat optimized for the chosen
frequency and power level. Ideally the losses due to both effects would
be about on par.

A two amp driver should take care of that. I've just been using a
spreadsheet from On semi for rough estimates for now.

It's pretty brazen that they advertize this as a 130W device. I wonder
what it's lifetime will be under that sort of load (in milliseconds,
from turn-on to a sharp popping noise).

They all do that it seems worse now because of the high 100 plus amps
they also quote for the part.

 

[Joerg]

Yep they defiantly aren't giving the battery option away.

Another old trick is to use a car battery and inverter. Mine just
croaked. Couldn't believe it, half a year ago it was showing slight
signs of weakness but would still crank a 2.6l gasoline engine alright.
So I replaced the one in the car. When I needed it for isolation
purposes a month ago ... nada, zip, zilch. It was totally hi-Z, charger
went to 17V and shut off, and the battery wouldn't even light a 1W bulb.
Puzzling how it could have gone so fast.

I haven't given up yet I'll play around with what I have part wise and
see if I cant get something reliable.

Try to find some smaller ones, maybe from EMC filters

No not confidential I haven't made up a full schematic yet.

Here's my actual test schematic. I'm not showing Bypass caps and the
lm317 for the driver and the 78l05 for the PIC both regulators get
their input from the 20 V supply V1.

http://i39.tinypic.com/1zf4gpl.png

That is just a test set-up to test the high side drive for a active
clamp flyback shown below.

http://i39.tinypic.com/10s7ybl.png

Since it can't bang past either rail you cold probably bootstrap that.
But transformers are a good option here, just not with 0.1uF in series.

I don't have a schematic for the flyback because I just started
considering it yesterday. I'm just working out how to drive the
High-side FET.

No chance just to do a half-bridge and be done with it? It would also be
a bit easier EMI-wise. And no air gap.

A two amp driver should take care of that. I've just been using a
spreadsheet from On semi for rough estimates for now.

Yes, 2A would be plenty.

They all do that it seems worse now because of the high 100 plus amps
they also quote for the part.

The topper is what I saw in some discount store aisle. An audio
amplifier in a blister pack, very ritzy packaging. It claimed 1000 (!)
watts PMPO and the power supply was <gasp> a wall wart. Yeah, right ...

 

[Hammy]

Try to find some smaller ones, maybe from EMC filters

Yea I'm digging through my parts. I tried a 10 nf and the oscillations
weren't as bad but my Vgs decreased by about a volt. I'll try some
other values. I don't think I can go below 10nf without adjusting the
inductance though.

No chance just to do a half-bridge and be done with it? It would also be
a bit easier EMI-wise. And no air gap.

That's my next PSU project;-). Coil Craft was kind enough to give me
some samples of flyback transformers. I used up all the small 10W ones
on projects but I still have some of the bigger ones left.

The topper is what I saw in some discount store aisle. An audio
amplifier in a blister pack, very ritzy packaging. It claimed 1000 (!)
watts PMPO and the power supply was <gasp> a wall wart. Yeah, right ...

Haven't you heard the Chinese can stick a small reactor in a wall wart
enclosure. That's not the really amazing part though ,they can do it
for $1.00 or less depending on quantities. ;-)

 

[Joerg]

Sounds like electrolyte dry out. See if you can find a way to "juice"
the cells again.

Strange thing is, the fluid level inside is ok. It's one of those
supposedly maintenance-free ones but one can pry off the lid assembly to
peek inside.

Why is it that this brand name battery is dead after 5-6 years, the gel
cell in my StatPower emergency kit also died after 5-6 years, while the
gel cell in my over 10 year old big $10 Chinese flashlamp is still fine?

[...]

Oh yes, I remember Peak Meaningless Pretend Output from the 1960s. A
play value created by inventive misapplication of theory that cannot be
duplicated in the real world. And look at the thin cord that carries it
from the wall wart to the device, better than room temperature
superconductors it must be.

They probably rely on the fact that the majority of shoppers does not
have an engineering degree