From: CIC on
On Thu, 5 Aug 2010 17:33:56 -0700 (PDT), sarah kent
<sarah6541(a)gmail.com> wrote:

>On Aug 5, 7:09�pm, "Tim Williams" <tmoran...(a)charter.net> wrote:
>> "sarah kent" <sarah6...(a)gmail.com> wrote in message
>>
>> news:afad00db-3e31-493f-bd1e-7650e7abae39(a)k19g2000yqc.googlegroups.com...
>>
>> > Hello All,
>>
>> > I am tring to generate a square wave of 100KHz and 200 volts peak to
>> > peak across a resistor via H- bridge. The current could go upto 30A.
>>
>> AC only, no DC? �Variable duty cycle? �If so, what kind (balanced ala
>> TL494 or differential ala class D audio)?
>>
>> Especially if the supply is line-operated (maybe this won't be an option
>> for your load, I don't have a clue), I like to drive AC-only, balanced-PWM
>> loads with a gate drive transformer. �This can be driven directly from a
>> TL598 (or UC3525, or...), or your choice of gate driver (as long as it's
>> not a complementary emitter follower type).
>>
>> > I am looking into two following ways to drive the NMOS ( four NMOS
>> > transistors ) H - bridge.
>>
>> > 1.
>>
>> > a. Using H bridge Drivers HIP4081A. I tried to look for the
>> > replacement of HIP4081A but only found A4940 from Allegro microsystems
>> > which a little slower than the HIP part regarding rise and fall time.
>>
>> Do you have a spec for how fast you need to run? �100ns is fine at 100kHz,
>> you'll most likely burn more power in conduction than switching. �Anything
>> faster is a bonus, but not particularly useful (and could be dangerous,
>> due to higher dI/dt forcing improved layout!).
>>
>> > According to my understanding the Driver chip has to charge the input
>> > capacitance of the transistors. I am planning to use NMOS transistor
>> > (IRF540).
>>
>> Whaaa? �Did you read the Vds(max) on that!? �200Vp-p needs a 100V power
>> supply. �That means >150V transistors!
>>
>> > So, according to HIP 4081A data sheet trise = 10nsec at C= 1000pF and
>> > if power supply = 50Volts (60% of 50V = 30volts) �, So, I calculated
>> > (1000pF x 30V) / 10nsec = 3 Amp.
>>
>> > The gate to source charging capacitance of IRF540 is 14nC; So, the
>> > charging time would be
>>
>> > (14nC x 30 V) / 3 = 140nsecs. Are my calculations accurate and in
>> > right direction?
>>
>> No, C == A.s, so 14nAs * 30V = 420nVAs == nJ (energy). �Divide by amps
>> gets you Vs, which is flux, not time.
>>
>> Since C == A.s, 14nC / 3A = 4.67ns. �This is less than the rated t_r so it
>> will be chip-limited, not capacitance-limited.
>>
>> > Which proves that the A4940 is a slower part. But how will it effect
>> > the voltage waveform across the resistive load. I meant to say what
>> > would be the difference between the two waveforms generated by two
>> > different driver chips.
>>
>> Likely the HIP is faster. �You'd use it if you needed to drive
>> particularly large FETs particularly fast. �If you don't need the speed,
>> don't put down the money for it. �A regular IR2112 is enough for mere
>> IRF540s at 100kHz. �Well, a little sluggish for 100k, more like 50k for
>> best results, but still not bad.
>>
>> > 2. The second method is to use four NPN transitors to drive two PMOS
>> > transistors (High side) and two NMOS (Low Side) transistors.
>>
>> No. �Check Vgs(max). �Complementary MOS is impossible over 20V supply
>> without extra circuitry, which adds enough complexity that you are better
>> off with a driver chip and NMOS (which work better anyway).
>>
>> > Microcontroller will provide the PWM pulses to control the 2N2222
>> > transistors. You can find the diagram at the following link.
>>
>> >http://a.imageshack.us/img39/3520/201115.jpg
>>
>> > I am thinking of using 2N2222A to drive two (IRF540N) but unable to
>> > find the PMOS equivalent of IRF540 N. Any suggestions!! Do I need them
>> > of the same ratings?
>>
>> > Please also suggest that H - bridge driver approach is the better
>> > approach or the NPN transistors driving the PMOS and NMOS transistors
>> > is the good approach. I heard that the NPN driving the PMOS and NMOS
>> > is the most efficient way, if yes than how?
>>
>> > Would 2N222 be able to charge the MOSFETS quickly as HIP4081A.
>>
>> Nope:
>>
>> 2N222
>> Low Power, General Purpose, Ge, PNP, 70mW, 15V, 12V, 70mA, 85>C, 400KHz,
>> 70, 20T, RCA, TO1
>>
>> Neato, it's germanium, 400kHz fT. �:^)
>>
>> Typoes aside, forget 2N2222 anyway. �Like 2N3055 and LM741, it is ancient
>> and best left forgotten, thrown to the gutter, on the road of progress.
>> 2N4401/03 are quite reasonable substitutes, and there are numerous others
>> with even better performance if you insist on rolling your own gate drive.
>> ZTX651 comes to mind, it's even smaller than 2N4401 yet drives 3A peak
>> quite easily. �TC4420 and similar are quite popular for big jobs (6A peak,
>> good for monster IGBTs).
>>
>> Tim
>>
>> --
>> Deep Friar: a very philosophical monk.
>> Website:http://webpages.charter.net/dawill/tmoranwms
>
>Hi,
>
>Thanks for the reply.
>
>I have following questions
>
>1. Why complementary MOSFET scheme can not work with power supply
>greater than 20 Volts?
>2. Is there a way that I can generate the sine wave of 200 volts peak
>to peak at 100KHz with out using H bridge driver or four MOSFETS?
>
>
>Sarah

Do you need a sine or a square at 100KHz? Earlier you mentioned a
square...
From: markp on

"markp" <map.nospam(a)f2s.com> wrote in message
news:8c2ld0F923U1(a)mid.individual.net...
>
> "markp" <map.nospam(a)f2s.com> wrote in message
> news:8c2l73F80aU1(a)mid.individual.net...
>>
>> "sarah kent" <sarah6541(a)gmail.com> wrote in message
>> news:afad00db-3e31-493f-bd1e-7650e7abae39(a)k19g2000yqc.googlegroups.com...
>>> Hello All,
>>>
>>> I am tring to generate a square wave of 100KHz and 200 volts peak to
>>> peak across a resistor via H- bridge. The current could go upto 30A.
>>>
>>>
>>> I am looking into two following ways to drive the NMOS ( four NMOS
>>> transistors ) H - bridge.
>>>
>>> 1.
>>>
>>> a. Using H bridge Drivers HIP4081A. I tried to look for the
>>> replacement of HIP4081A but only found A4940 from Allegro microsystems
>>> which a little slower than the HIP part regarding rise and fall time.
>>> According to my understanding the Driver chip has to charge the input
>>> capacitance of the transistors. I am planning to use NMOS transistor
>>> (IRF540).
>>>
>>> So, according to HIP 4081A data sheet trise = 10nsec at C= 1000pF and
>>> if power supply = 50Volts (60% of 50V = 30volts) , So, I calculated
>>> (1000pF x 30V) / 10nsec = 3 Amp.
>>>
>>> The gate to source charging capacitance of IRF540 is 14nC; So, the
>>> charging time would be
>>>
>>> (14nC x 30 V) / 3 = 140nsecs. Are my calculations accurate and in
>>> right direction?
>>>
>>> b. I calculated the same things using A4940 and found the following
>>> results
>>>
>>> (1000pF x 30V) / 35nsec = 857 mA
>>>
>>> (14nC x 30V) / (857mA) = 490nsec.
>>>
>>> Which proves that the A4940 is a slower part. But how will it effect
>>> the voltage waveform across the resistive load. I meant to say what
>>> would be the difference between the two waveforms generated by two
>>> different driver chips.
>>>
>>> Is there true replacement part available for HIP4081A. The reason for
>>> finding the replacement for HIP4081A are
>>>
>>> a. What if Intersil discontinue this part in the future.
>>> b. May be there is a cheapaer part available than HIP4081A.
>>>
>>> 2. The second method is to use four NPN transitors to drive two PMOS
>>> transistors (High side) and two NMOS (Low Side) transistors.
>>> Microcontroller will provide the PWM pulses to control the 2N2222
>>> transistors. You can find the diagram at the following link.
>>>
>>> http://a.imageshack.us/img39/3520/201115.jpg
>>>
>>> I am thinking of using 2N2222A to drive two (IRF540N) but unable to
>>> find the PMOS equivalent of IRF540 N. Any suggestions!! Do I need them
>>> of the same ratings?
>>>
>>> Please also suggest that H - bridge driver approach is the better
>>> approach or the NPN transistors driving the PMOS and NMOS transistors
>>> is the good approach. I heard that the NPN driving the PMOS and NMOS
>>> is the most efficient way, if yes than how?
>>>
>>> Would 2N222 be able to charge the MOSFETS quickly as HIP4081A. What
>>> parameters should I consider before I choose the microcontroller.
>>>
>>> This is not a home work. :)
>>>
>>> Thanks
>>> Sarah
>>>
>>
>> Have you considered using the Baxandall class D resonant oscillator for
>> this? The benefits are no high frequency PWMing (so good for EMI), no H
>> bridges or high voltages on the driving side, and it is a very simple
>> circuit. All depends on how accurate you want the frequency and the size
>> of the load etc, but you can simulate with LTspice as long as you have a
>> good model of your transformer. Bill Sloman has some good stuff on this:
>> http://home.planet.nl/~sloma000/Baxandall%20parallel-resonant%20Class-D%20oscillator1.htm
>>
>> Mark.
>>
>
> Oops, just spotted you want a square wave. Sorry!
>
> Mark.

Actually, you did ask if there was a way of generating a sine wave in
another post, so maybe the Baxandall class D resonant oscillator might be
worth looking at.

Mark.


From: sarah kent on
On Aug 5, 7:09 pm, "Tim Williams" <tmoran...(a)charter.net> wrote:
> "sarah kent" <sarah6...(a)gmail.com> wrote in message
>
> news:afad00db-3e31-493f-bd1e-7650e7abae39(a)k19g2000yqc.googlegroups.com...
>
> > Hello All,
>
> > I am tring to generate a square wave of 100KHz and 200 volts peak to
> > peak across a resistor via H- bridge. The current could go upto 30A.
>
> AC only, no DC?  Variable duty cycle?  If so, what kind (balanced ala
> TL494 or differential ala class D audio)?
>
> Especially if the supply is line-operated (maybe this won't be an option
> for your load, I don't have a clue), I like to drive AC-only, balanced-PWM
> loads with a gate drive transformer.  This can be driven directly from a
> TL598 (or UC3525, or...), or your choice of gate driver (as long as it's
> not a complementary emitter follower type).
>
> > I am looking into two following ways to drive the NMOS ( four NMOS
> > transistors ) H - bridge.
>
> > 1.
>
> > a. Using H bridge Drivers HIP4081A. I tried to look for the
> > replacement of HIP4081A but only found A4940 from Allegro microsystems
> > which a little slower than the HIP part regarding rise and fall time.
>
> Do you have a spec for how fast you need to run?  100ns is fine at 100kHz,
> you'll most likely burn more power in conduction than switching.  Anything
> faster is a bonus, but not particularly useful (and could be dangerous,
> due to higher dI/dt forcing improved layout!).
>
> > According to my understanding the Driver chip has to charge the input
> > capacitance of the transistors. I am planning to use NMOS transistor
> > (IRF540).
>
> Whaaa?  Did you read the Vds(max) on that!?  200Vp-p needs a 100V power
> supply.  That means >150V transistors!
>
> > So, according to HIP 4081A data sheet trise = 10nsec at C= 1000pF and
> > if power supply = 50Volts (60% of 50V = 30volts)  , So, I calculated
> > (1000pF x 30V) / 10nsec = 3 Amp.
>
> > The gate to source charging capacitance of IRF540 is 14nC; So, the
> > charging time would be
>
> > (14nC x 30 V) / 3 = 140nsecs. Are my calculations accurate and in
> > right direction?
>
> No, C == A.s, so 14nAs * 30V = 420nVAs == nJ (energy).  Divide by amps
> gets you Vs, which is flux, not time.
>
> Since C == A.s, 14nC / 3A = 4.67ns.  This is less than the rated t_r so it
> will be chip-limited, not capacitance-limited.
>
> > Which proves that the A4940 is a slower part. But how will it effect
> > the voltage waveform across the resistive load. I meant to say what
> > would be the difference between the two waveforms generated by two
> > different driver chips.
>
> Likely the HIP is faster.  You'd use it if you needed to drive
> particularly large FETs particularly fast.  If you don't need the speed,
> don't put down the money for it.  A regular IR2112 is enough for mere
> IRF540s at 100kHz.  Well, a little sluggish for 100k, more like 50k for
> best results, but still not bad.
>
> > 2. The second method is to use four NPN transitors to drive two PMOS
> > transistors (High side) and two NMOS (Low Side) transistors.
>
> No.  Check Vgs(max).  Complementary MOS is impossible over 20V supply
> without extra circuitry, which adds enough complexity that you are better
> off with a driver chip and NMOS (which work better anyway).
>
> > Microcontroller will provide the PWM pulses to control the 2N2222
> > transistors. You can find the diagram at the following link.
>
> >http://a.imageshack.us/img39/3520/201115.jpg
>
> > I am thinking of using 2N2222A to drive two (IRF540N) but unable to
> > find the PMOS equivalent of IRF540 N. Any suggestions!! Do I need them
> > of the same ratings?
>
> > Please also suggest that H - bridge driver approach is the better
> > approach or the NPN transistors driving the PMOS and NMOS transistors
> > is the good approach. I heard that the NPN driving the PMOS and NMOS
> > is the most efficient way, if yes than how?
>
> > Would 2N222 be able to charge the MOSFETS quickly as HIP4081A.
>
> Nope:
>
> 2N222
> Low Power, General Purpose, Ge, PNP, 70mW, 15V, 12V, 70mA, 85>C, 400KHz,
> 70, 20T, RCA, TO1
>
> Neato, it's germanium, 400kHz fT.  :^)
>
> Typoes aside, forget 2N2222 anyway.  Like 2N3055 and LM741, it is ancient
> and best left forgotten, thrown to the gutter, on the road of progress.
> 2N4401/03 are quite reasonable substitutes, and there are numerous others
> with even better performance if you insist on rolling your own gate drive..
> ZTX651 comes to mind, it's even smaller than 2N4401 yet drives 3A peak
> quite easily.  TC4420 and similar are quite popular for big jobs (6A peak,
> good for monster IGBTs).
>
> Tim
>
> --
> Deep Friar: a very philosophical monk.
> Website:http://webpages.charter.net/dawill/tmoranwms

Hi,

I am unable to understand that why we do consider Gate to Source
Charge (14nC) to calculate the time?

Thanks
Sarah
From: Tim Williams on
"sarah kent" <sarah6541(a)gmail.com> wrote in message
news:e7ddfee9-2223-4f54-872c-ea3c042d51ba(a)e15g2000yqo.googlegroups.com...
> > > (14nC x 30 V) / 3 = 140nsecs. Are my calculations accurate and in
> > > right direction?
> >
> > No, C == A.s, so 14nAs * 30V = 420nVAs == nJ (energy). Divide by amps
> > gets you Vs, which is flux, not time.
> >
> > Since C == A.s, 14nC / 3A = 4.67ns. This is less than the rated t_r so
> > it will be chip-limited, not capacitance-limited.


> I am unable to understand that why we do consider Gate to Source
> Charge (14nC) to calculate the time?

Quoted above.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms


From: Jon Elson on
sarah kent wrote:
> Hello All,
>
> I am tring to generate a square wave of 100KHz and 200 volts peak to
> peak across a resistor via H- bridge. The current could go upto 30A.
>
>
> I am looking into two following ways to drive the NMOS ( four NMOS
> transistors ) H - bridge.
>
> 1.
>
> a. Using H bridge Drivers HIP4081A.
The HIP4080 series is only good to about 50 V, despite glowing claims by
Harris/Intersil. I made the switch about 13 years ago and haven't
looked back!

Try the International Rectifier IS2113 and related parts.
I tried to look for the
> replacement of HIP4081A but only found A4940 from Allegro microsystems
> which a little slower than the HIP part regarding rise and fall time.
> According to my understanding the Driver chip has to charge the input
> capacitance of the transistors. I am planning to use NMOS transistor
> (IRF540).
>
You won't get 30 A out of an IRF540. You need at least an IRF260 or
other newer-generation transistors.
Also, you have to carefully tune the dead-time between one transistor
turns off and the other turns on to avoid "shoot through" where current
flows through both transistors from + supply to ground. These currents
can build to hundreds of amps in tens of ns and cause great heating. I
use an R-C-diode circuit at the input of the IR driver chip to control
the dead-time.

The problem is the IR2113 chip can only source/sink 2 Amps, and the gate
charge+Miller charge goes way up as the transistors get larger. Miller
charge is the effect of drain to gate capacitance as the drain swings
over the output voltage range. The effect of D-G capacitance and a 200
V drain swing is GREATER than the gate charge alone! You can dream of
10 ns switching, but it is a fantasy without about 20 Amps of gate drive
current. Then, the lead inductance of the transistor start to become
significant, and you can't maintain safe G-S voltages with these kinds
of di/dt into the gate.

Maybe you'd better explain what you REALLY need.

Jon