Maker Pro
Maker Pro

Speaking of high frequency transformer stuff...

T

Tim Williams

Jan 1, 1970
0
How would you tackle the problem(?) of 1MHz, 100V common mode on
transformers or whatever coupling you're using? Think high side drive. Is
an ordinary gate drive transformer suitable, does it need special
construction or drive? Or would it be better to use a big fat CM choke to
bring things down to earth?

I don't think optical coupling would be very useful. I've already had poor
results with 6N136's at 20kHz. I know there are fancier parts available,
too, some with built in gate drivers. I've seen it before where fast edges
will couple through the pri-sec capacitance of a GDT, even when driven hard.

Tim
 
T

Tim Williams

Jan 1, 1970
0
Yzordderrex said:
Is the 1MHZ on the primary side - and getting into the secondary?
Sorry, I work better with drawings.

Ok. Picture the average half bridge MOSFET circuit:
http://webpages.charter.net/dawill/tmoranwms/Elec_IndBridge.gif
At 1MHz, the FETs need local gate drivers (e.g. TC4420), so plant some
inbetween the GDT and FETs. Add a DC-DC converter to power the drivers.
Now, the GDT can be a higher impedance, since it only has to drive the
TC4420's, not gates directly, which helps.

But now the problem is the pri-sec capacitance on the GDT and DC-DC conv.
secondaries. The current flowing through that capacitance has to go
somewhere. If it goes through the windings, you get coupling, which can
lead to feedback and self destruction (I've seen the feedback before). The
windings can be shielded, but that makes more stuff in the transformer =
more leakage inductance. A tradeoff as usual.

Note that leakage in the DC-DC converter isn't a big deal -- I could filter
and regulate that locally. But the GDT has to be fast and square (<50ns t_p
and t_r), with no bouncing to upset the driver.

The big question I'm wondering is, is it feasible to shield the transformers
and sink away the dV/dt, without sacrificing GDT bandwidth, or is an
optocoupler suitable (I'll have to check if there are any fast enough), or
is the whacky idea of a three-wire common mode choke feasible?

One thing in favor of transformers is the high frequency. I'm not
interested in anything under 200kHz, so the maximum pulse width is 0.25us,
and a 5V pulse would carry a maximum flux of 1.25 uWb*t. A medium or low-mu
ferrite would work okay (like mix #43, not too lossy at 1MHz), and a typical
1" toroid will easily handle that in just one turn. And let's say I put on
three turns of coax for this transformer: how much capacitance is there in
4" of RG174? 30pF/ft, or 10pF for the whole thing. I'm expecting edges of
200V and maybe 20ns, or dV/dt = 10V/ns, or peak currents of only 100mA.
That's a lot for a signal level, but I could drive this transformer with a
FET driver (TC4427 or whatever), which would happily sink a current like
that. So it's not looking too bad.
For gate drive transformer for
inverters you normally want very good coupling and low capacitance.
These contradict each other so that the design is a compromise. I've
used some of the HP-Agilent-Avago opto-driver parts and I think
they're rated at near 10kv/us.

That number is meaningless in at least one case. For the 6N136, they spec
10V/ns, by applying a 10V, 1ns ramp. Sure, at 10V/ns, the peak current
might be high, but with only ten volts difference, there simply aren't
enough coulombs delivered to have any effect. So it's a lie of a rating.
One optocoupled gate drive I've made suffered from unreliable behavior with
320V/0.1us edges. The datasheet doesn't say anything about external
shielding, but I had to put some on myself, which helped a little bit (it
still chatters if you bring your finger near it!).

Tim
 
T

Tim Williams

Jan 1, 1970
0
John Larkin said:
One nice combo is a dc/dc converter to make 10 volts or whatever, and
a fast logic coupler to get the gate drive signal up there, and then a
proper gate driver chip floating with everything else.

Yup, that's the plan. But I'm not sure about transformer (and what design
to use) or opto anything.
Several people also make nice gate drivers with built-in isolators.
Avago, IR, like that.

The HCPL-9000 does look pretty nice (and it's specified for 15kV/us at *1kV*
common mode). The datasheet is suspicously quiet on what a "symmetric
magnetic coupling barrier" is, but later on it states supply current is
drawn in 2.5ns gulps. I wonder if the output state is stored in a
flip-flop, and I wonder what its initial state is.
I've made my own transmission-line gate-drive transformers for
short-pulse stuff, like 100 ns pulse widths/400 volt swing/5 ns
risetimes. Longer pulses and high duty cycles become problematic.

Extreme case: use fiber optics. Megavolt isolation!

Perfectly acceptible If I were building a "valve house". ;-)

Tim
 
T

Tim Williams

Jan 1, 1970
0
John Larkin said:
Several people also make nice gate drivers with built-in isolators.
Avago, IR, like that.

WTF... Si8233 would do quite perfectly, except there's no data? I must've
learned something from Jeorg, cuz that doesn't make me too comfortable!

Oh well, I suppose one of those HCPL9000's with a TC4420 isn't so bad. Hmm,
needs 5V though... zener will be good enough.

Tim
 
F

Fred Abse

Jan 1, 1970
0
One nice combo is a dc/dc converter to make 10 volts or whatever, and
a fast logic coupler to get the gate drive signal up there, and then a
proper gate driver chip floating with everything else.

I'd go along with that. There are optocouplers characterized for high
common mode noise immunity (TLP759, HCPL4504) that might help.
 
J

Joerg

Jan 1, 1970
0
Tim said:
How would you tackle the problem(?) of 1MHz, 100V common mode on
transformers or whatever coupling you're using? Think high side drive. Is
an ordinary gate drive transformer suitable, does it need special
construction or drive? Or would it be better to use a big fat CM choke to
bring things down to earth?

I don't think optical coupling would be very useful. I've already had poor
results with 6N136's at 20kHz. I know there are fancier parts available,
too, some with built in gate drivers. I've seen it before where fast edges
will couple through the pri-sec capacitance of a GDT, even when driven hard.

Not sure what you are concerned about but if it's the secondary winding
capacitance to whatever you could make that winding out of coax and hang
the shield onto the source of the upper FET. That's how we also get rid
of E-field stuff leaking into RF transformers.

It doesn't have to be any fancy coax. Thin audio line might do if it can
take the voltages and temperatures.
 
T

Tim Williams

Jan 1, 1970
0
So, here's a possible optoisolated solution:
http://webpages.charter.net/dawill/Images/1MHz_Induction_Heater.gif

Notice the feedback circuit has time limited positive feedback, so
shoot-through is impossible, high and low side on-times are equal (within
component variation), and if it stops oscillating, it just stops
(accordingly, something needs to be provided to kick it moving in the first
place, which isn't shown here).

Not sure if everything is quite fast enough. There's a good ~100ns of delay
in the loop, which is wholly 1/5th of the half cycle time. I'd rather not
spend more bucks on an LT1016 or something like that. Incidentially, Mouser
seems to have dropped LT from their catalog??

The timer is controlled for phase lock on the resonant tank, and either
phase or supply voltage controls output power. (Someone mentioned PFC SEPIC
recently, something like that could be handy here for generating 0-200V from
a 90-265VAC input. No need for postregulation in an app like this, 120Hz
ripple lets you know it's running.)

Regarding this thread, I've kept the application general, since 1MHz power
transformers work just as well for induction heating as for generic
switchers. Specifically, I am making another induction heater, but lessons
learned here will do just as well anywhere else (like D from BC's 1MHz
switcher with too much transistor-heatsink capacitance).

Tim
 
W

Winfield Hill

Jan 1, 1970
0
Tim Williams wrote...
How would you tackle the problem(?) of 1MHz, 100V common mode on
transformers or whatever coupling you're using? Think high side
drive. Is an ordinary gate drive transformer suitable ...

My favorite Intersil HIP4080 series of high-side/
low-side driver ICs easily goes to 1MHz. Check 'em
out. No transformers, no optical couplers, nada.
Ahem, HIP4081A in stock at DigiKey.

Rated at 80V (plus 15V swinging gate-drive), but hey,
go ahead and push it to 100V if necessary, or adjust
your turns ratio.
 
W

Winfield Hill

Jan 1, 1970
0
Jon Elson wrote...
I built a servo amp with the 4080 over a decade ago. The chip is rated
for 80 V, but the applications manager eventually admitted to me "Oh,
you're doing REALLY good to get them to run at 59 V, none of our other
customers ever got them to last above 54 V!" Oh, that was a REAL NICE
admission! So, I eventually redesigned the whole thing to use the
IR2113 half-bridge driver chip. They have been VERY reliable, way above
the voltages I normally run at. The one gotcha is that the common point
between the two transistors cannot be allowed to go negative, so I had
to put an ULTRA-fast diode across the low-side transistor. (The body
diodes in the FETS are incredibly slow to turn on.)

Unless the HIP4081A is on some improved process from the 4080 (I doubt
it) you'll never get close to 80 V, even. Winfield, have you ever run
the 4081A up above 40 V or so?

Whoa, I've been a fan of the HIP4081A for more than 15 years.
The "A" variant, I must admit. In that time I never heard
anyone 'dis them. My own designs use the parts (purchased
repeatedly over a 7-year period) up to and over the maximum
voltage spec rating, 80, 90 and 100V, and to and over 1MHz.

In my bench prototype testing I've had some ugly setups,
creating substantial V = L dI/dt voltages, that get out of
control in a real hurry (try 50nH*4A/10ns = 20V, be careful),
yet I only had one failure, during testing, which quickly
reverted to reliable operation after I tightened up critical
wiring to reduce the current-loop area.

I have decades (OK, 1.5 decades) of successful 24-7 operation
of my designs (the guys leave everything running continuously)
using these parts at or near their limits. Recommended.

BTW, I'm also a fan of the IR half-bridge drivers, but they
are relatively slow parts, meant for another kind of design.
Their slow speeds mean you won't so quickly run into wiring
issues, because for them L dI/dt has a much slower dt term.
I use these to make elegant 400V pulse generators (posted
on s.e.d.), but definitely not to run at MHz frequencies.
 
J

Joerg

Jan 1, 1970
0
Winfield said:
Jon Elson wrote...

Whoa, I've been a fan of the HIP4081A for more than 15 years.
The "A" variant, I must admit. In that time I never heard
anyone 'dis them. My own designs use the parts (purchased
repeatedly over a 7-year period) up to and over the maximum
voltage spec rating, 80, 90 and 100V, and to and over 1MHz.

In my bench prototype testing I've had some ugly setups,
creating substantial V = L dI/dt voltages, that get out of
control in a real hurry (try 50nH*4A/10ns = 20V, be careful),
yet I only had one failure, during testing, which quickly
reverted to reliable operation after I tightened up critical
wiring to reduce the current-loop area.

I have decades (OK, 1.5 decades) of successful 24-7 operation
of my designs (the guys leave everything running continuously)
using these parts at or near their limits. Recommended.

BTW, I'm also a fan of the IR half-bridge drivers, but they
are relatively slow parts, meant for another kind of design.
Their slow speeds mean you won't so quickly run into wiring
issues, because for them L dI/dt has a much slower dt term.
I use these to make elegant 400V pulse generators (posted
on s.e.d.), but definitely not to run at MHz frequencies.

Just my 2ct: I haven't designed in Intersil parts since a couple years
before I got my degree. And there's a reason ...
 
Top