| Pierre |
Hello.
I am planning to make a Class D output stage using big mosfets, TO247 devices with around 4000pF gate capacitance and 120nC gate charge. I plan to switch at around 350KHz, needing less than 100ns rise/fall times.
To swith such mosfets that fast, I need a mosfet driver after the IR2113 in order to upply much more current to the gates. My plan is to add a TC4420 (6 amps) or similar at the output of the high side and another one at the output of the low side. Has anyone tried this approach?
I have seen some schematics (Crest LT and others) using similar mosfets driven directly by the IR2113 with 39ohm resistors, but I guess that can't switch too fast at that frequency (perhaps they run at 100KHz or so?)
The supply of the lower one is straight forward, just -VSS + 12V.
The supply for the upper one is the bootstrap voltage, referenced to the switching point between mosfets. But... I suppose I need a large bootstrap cap.
What is a recommended starting value for the gate resistor (I use antiparallel diode as well)?
Thanks! |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
Hello.
I am planning to make a Class D output stage using big mosfets, TO247 devices with around 4000pF gate capacitance and 120nC gate charge. I plan to switch at around 350KHz, needing less than 100ns rise/fall times.
To swith such mosfets that fast, I need a mosfet driver after the IR2113 in order to upply much more current to the gates. My plan is to add a TC4420 (6 amps) or similar at the output of the high side and another one at the output of the low side. Has anyone tried this approach?
I have seen some schematics (Crest LT and others) using similar mosfets driven directly by the IR2113 with 39ohm resistors, but I guess that can't switch too fast at that frequency (perhaps they run at 100KHz or so?)
The supply of the lower one is straight forward, just -VSS + 12V.
The supply for the upper one is the bootstrap voltage, referenced to the switching point between mosfets. But... I suppose I need a large bootstrap cap.
What is a recommended starting value for the gate resistor (I use antiparallel diode as well)?
Thanks! |
Crest Audio LT1800 & CD3000 Runs at 250KHZ!
I have used 4 X IRFP460N with Rg=4.6Ohms,Series Schottky's MBR4045, HFA series FREDs, 2 X MIC4420 Gate Driver per side, Isolated +-15V for both hi-side and low side gate drivers....Switching Frequency= 250KHZ, 6N137 optocouple to drive the gate drivers...Rails were at +- 180VDC..idle, Rise/Fall time less than 75nS |
|
|
| Pierre |
Thanks.
I would like to keep the IR2113 for the level shift, shutdown, etc, and only buffer its outputs, using two TC- (or MIC-) 4420. |
|
|
| alfsch |
hmmm...
2A in 4nf gives about 20ns rise time, seems ok.
i would try 10 ohm + pnp for closing gate
my favorite pnp: zetex fmmt717 , 10A peak !!
that should close every mosfet :-) |
|
|
| Pierre |
Yes, but that will overheat the IR2113.
Now, with a IR2110 (SMD versio) and 3nF mosfets, it gets quite hot.
Mmmm. Have to try...
On my currently working-well circuit, I have tried the modification:
Added two TC4420's and 2.2ohm gate resistors, and the new mosfets, but there seems to be a lot of cross-conduction (the mosfets get quite hot even with no signal). Maybe two possible causes:
a) The bootstrap voltage is not enough and the upper mosfets gets into linear region. (I have used 100nF ceramic || 1 uF tantalum || 10uF tantalum from switching node to IR2110's VB (VCC of the upper TC4420). Not enough?
b) The dead time is not enough (before it was), or the mosfets are not turning off fast enough (I use an antiparallel diode, through it, the driver should turn it off strongly).
Any ideas? |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
Thanks.
I would like to keep the IR2113 for the level shift, shutdown, etc, and only buffer its outputs, using two TC- (or MIC-) 4420. |
Then your Deadtime must be at least 180nS or more, if you want to avoid destruction of mosfets and also the Turn-OFF would be a problem, since you only have the option of reducing the Vgs during TURN-OFF to around Zero...which doesnot TURN-OFF the mosfet much faster...but In my case The availability of -5V VGS during TURN-OFF makes the mosfet to switch-off much faster and also deadtime is just 35nS....
Using IR2110/2113 along with Bootstrap supplies for driving High Capacitance Large Die High Voltage Mosfet isnot a good way..you would eventually end up with nothing but smokes and flames of different colors..... |
|
|
| Pierre |
| Well, Crest LT does this way and they have fame of being very rudged and reliable... |
|
|
| IVX |
| Pierre, maybe choice of the new mosfets much better idea vs raping heavy old ones? How much volts & amperes you need? |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
Well, Crest LT does this way and they have fame of being very rudged and reliable... |
Who told you this....LT series was discontinued since past 3 years just because there were reliability problems associated with them, only 230 units were produced out of which 89 were returned to factory just because of "HIGH FIELD FAILURE RATES", They were replaced by new CD3000...which is much more rugged and used robust techniques such as series schottkys and high side cascoded gate drive and also their LT series uses low gate charge APT single pair Mosfets...with rails voltage +-120V thats why IR2110 survived on their test bench but failed on field operation...
| quote: | Originally posted by IVX
Pierre, maybe choice of the new mosfets much better idea vs raping heavy old ones? How much volts & amperes you need? |
IRFP460 is replaced by IRFP460N with improved features....less gate charge....but there are many more with much improved features from IXYS and Fairchild... |
|
|
| IVX |
| :bigeyes: IRFP460?? It's old hat from the other century, that's too far from current state of the semiconductors industry. I guess, that at moment Infineon is a leader at 400+ nomination, easyly to find the Infineon mosfet ~40nC with about ~100mOhm, i don't know if IXYS already can offer such one, but Fairchild can't (AFAIK). |
|
|
| Pierre |
| I was thinking about STW75N20's... |
|
|
| IVX |
| quote: | Originally posted by Pierre
I was thinking about STW75N20's... |
IRFB4227 looks better yet. |
|
|
| Pierre |
Mmmm, Ciss=4600pF!
Ok that gate charge is lower, but what dominates gate driver requirements? |
|
|
| IVX |
| I think Ciss=4600pF isn't so bad, more important, that miller charge lower (faster switching, lower spikes on the gate through Crss), Rds_on lower and more linear vs current, Trr much lower 100nS vs 222nS. |
|
|
| Workhorse |
Infineon's IPW50R140CP has Qg=48nC & Ciss=2540pF only and is rated at 550V, 23A, Pd=192W
I think its much more promising device... |
|
|
| IVX |
| quote: | Originally posted by Workhorse
Infineon's IPW50R140CP has Qg=48nC & Ciss=2540pF only and is rated at 550V, 23A, Pd=192W
I think its much more promising device... |
It's exactly what i've suggested, if you want to replace old irfp460, IPW50R140 the best. BTW, Ciss & Pd rather superfluous parametres for class D. |
|
|
| Bender.ru |
| Hi Ivan, i think IRFB4227 really excellent solution for midpower amp (0.5-1kW:) ) btw this fets recently come in to the market in Moscow :D. I hope it successfully replace my PHP28NQ15. |
|
|
| IVX |
| hmm, i've doubt regarding such replacement yet, for subwoofer maybe. |
|
|
| Bender.ru |
| Yes of cause, for subwamp with carrier ~100kHz. |
|
|
| Workhorse |
| quote: | Originally posted by IVX
It's exactly what i've suggested, if you want to replace old irfp460, IPW50R140 the best. |
IRFP460 older type Qg=240nC
IRFP460N newer version has much lower Qg=120nC |
|
|
| Pierre |
| I'd like to stay with TO-247 devices for rudgedness and thermal reliability... |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
I'd like to stay with TO-247 devices for rudgedness and thermal reliability... |
Some amplifiers like UCD, ZAPs, NCDs, ICE all rely on TO-220 FET packages............
While TO-247 is all rounder best.........47N60C3, IRFP460N |
|
|
| Pierre |
Hello.
Some of you mentioned that Trr (body diode reverse recovery time) should be low, but TO247 devices have Trr>200ns in most cases.
Can this be ammended by the addition of a ultrafast diode in parallel?
I have seen some designs, some as Tripath:
(http://www.tripath.com/downloads/RB-TA0105.pdf , see pag. 14)
...that use a "small" 4A ultrafast diode (MUR420) in parallel with the body diode, with much lower Trr (35ns or so). Is this in order to speed up recovery time?
What should be the current rating of the diode?
So, can I choose a mosfet with all parameters OK except for a slightly large Trr and then add two ultrafast diodes in parallel to improve situation?
Thanks! |
|
|
| luka |
Hi
IRFP4227 desn't have that big Trr=100ns I think |
|
|
| Pierre |
Yes, I know, but it is difficult to find.
Anyway, I am now curious about that diodes in parallel. Can anyone tell us what are they for exactly and how to use them? |
|
|
| Workhorse |
Simply paralleling a Fast Recovery Diode with the mosfet isn't going to give you any benefit, because the Forward drop of Diode is greater than Forward Diode of Body Diode of mosfet....for a given amount of freewheeling current conduction..
Unless you use a Series Schottky to achieve Bulletproof solution as in Crest CD3000...
or a Centretapped inductor in some cases as in Crest LT1200....[but not that much robust] |
|
|
| Pierre |
Mmmm. a bit confused, then.
Why does Tripah use them in their TA0105 amplifier? |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
Mmmm. a bit confused, then.
Why does Tripah use them in their TA0105 amplifier? |
A so called low cost alternative which does nothing at all when it comes to real life testing.... |
|
|
| Pierre |
Mmmm. interesting...
What problems will a high Trr (240ns or so) cause? Shoot-through? And will it be noticeable with no signal? I mean, if I have shoot-through will I notice that the mosfets get warm with no signal, right?
And what about the V/ns specification, is it related to Trr?
Thanks and sorry for so many questions |
|
|
| luka |
HI
You will see shoot-through on current sense resistors, but if you have big fet, this woun't heat it up, even without heatsink |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
Mmmm. interesting...
What problems will a high Trr (240ns or so) cause? Shoot-through? And will it be noticeable with no signal? I mean, if I have shoot-through will I notice that the mosfets get warm with no signal, right?
And what about the V/ns specification, is it related to Trr?
Thanks and sorry for so many questions |
Aren't you the same guy who made postfilterfeedback class-d[with Charles help] with complementary LM319 comparators....now it seems that you donot understand the concept of Freewheeling currents in switching circuits....How this could be.........Now I am confused.....
With high Trr, the Freewheeling current doesnot have quick discharge interval thereby increases the shoothrough, which leds to the Mosfet breakdown due to avalanche condition..... |
|
|
| Pierre |
Yes, I am the same guy... :-)
But wanted to clarify some concepts before trying to modify my old prototype for higher power and use larger mosfets.
The Tripath schematics have confused me a bit, sorry. Would like to hear them justify the use of that two diodes, perhaps we're missing some detail... |
|
|
| Workhorse |
If the forward drop of body diode is greater than forward diode of anti parallel fast recovery diode than its good, but this never happens unless one uses a schottky instead of fast recovery diode.......but higher voltage schottkys have higher forward voltage drop...
In order to overcome this drawback, another solution is using schottky's in series with mosfets to stop the body diode conduction phenomena and than the anti-parallel connected fast recovery diode takes care of freewheeling current very easily and shootthrough current is reduced and mosfets are saved from premature failure...... |
|
|
| Pierre |
| Thanks, very interesting reading... |
|
|
| Workhorse |
| Got some enlightenment Pierre....:) |
|
|
| IVX |
| Come on Pierre, relax and take to-220, what is the class D amp at all, if each mosfets will dissipate >150W? :) |
|
|
| alfsch |
from the datasheet, i think the irfb4227pbf looks good...
anybody tried it for a class d amp?
good or bad...? |
|
|
| luka |
Hi
Don't know about that, but irfp 4227 sure is. Read here and here
I wanted to use them, but with price of over 7€/each is too much for me, and too good fet to waste :D |
|
|
| alfsch |
well, price is about 2,70 (at 50pcs), so thats no big problem...
what i really dont understand:
datasheet :
Tst : shoot through blocking time : 100ns
never seen this spec....? means? |
|
|
| luka |
Hi
Yea this seems to be new, but if you look at page 7 you will see what they mean.. |
|
|
| alfsch |
no... :-( i dont see...
there is a Ids pulse shown, when both gates are driven "off"...
so what? irfb4227 draws current after "off" at gate or before "on"(looks into future??) ? found no explanation.. |
|
|
| luka |
Hi
Looking at picture...First A is ON, charges L then it is OFF. Now is time that both are off, L produces negative voltage, when Id2 should be negative... Don't know for sure...it is all confusing :dead: Don't even know wich Id it is, but shoot through is only when both fets are conducting (not in this case) or body diode is too slow, so maybe this is happening..:apathic: |
|
|
| alfsch |
| ok, but if bodydiode : the current shoot would happen, when other fet switches on, not before! thats what is totally puzzling for me... |
|
|
| Eva |
| The datasheet is not clear at all, but I think that they mean "maximum-allowed-cross-conduction-time-just-at-turn-on-due-to-reverse-recovery" :D |
|
|
| luka |
Hi
I don't know... But fets, IRFP, were used in ClassD and that Tst isn't something bad, amp works 100%
PS: LOOL look at IRFP, now I realy don't know, since setup there is diffrent for same thing:confused: :bigeyes: :confused: :bigeyes: |
|
|
| alfsch |
| ok, so i order some irfb4227 and do my very best... |
|
|
| Eva |
It's Qrr what matters, as it's the integral of the reverse recovery current over the full recovery process (it's dependent on Id, di/dt and temperature, though). You may get an estimation of reverse recovery losses in watts with the following expression: (|Vss|+|Vdd|)*Fswitch*Qrr. You may find interesting Qrr plots in Infineon "CFD" datasheets (SPW20N60CFD and the like).
For 100V check IRF540Z. For 150V check IRFB4019pbf. For higher voltages there is not much stuff that I like in hard switched operation. Note that you don't need a TO-247 package at all, dissipation is going to be quite low in class D (particularly with music), and big MOSFET dies connected to widely spaced leads have their own sets of disadvantages. |
|
|
| luka |
Hi
Is IRFB4019pbf enough for +/-90 and 4ohme load? |
|
|
| Eva |
Download the datasheet and read. IRFB4019pbf is a 150V device so maximum rails are +/-70V or so. Next time do the math yourself :D:D
Assuming +/-70V and 4 ohms:
70^2/(2*4)=612Wrms (at full power sine)
70V/4=17.5A peak (at full power sine)
17.5A*.707=12.37A rms (at full power sine)
17.5A*36=6.3A rms with music (with 9dB crest factor)
12.37^2*.08=12.24W max. cond. loss. (full power sine)
12.24/2=6.12W max. cond. loss. per device
6.3^2*.08=3.17W cond. loss. (music 9dB crest factor)
3.17/2=1.6W cond. loss. per device
So conduction losses are ridiculously small!! Even smaller than predicted because the body diode will clamp Rds-on during reverse conduction. Switching losses due to diode recovery and parasitistic capacitances dominate.
Now some math on IRFB4019pbf switching losses...
Lets assume an average Qrr of 160nC. That's just a quick guess (and the value shown in the datasheet). It actually varies widely depending on the input signal. There is no recovery at all during resonant operation at low signal levels, when inductor current fluctuates around 0A changing direction before each switching period ends. Then Qrr rises progressively as the system enters continuous conduction mode and body diode current becomes higher. Also, Qrr is bigger than the figures stated in the datasheet because it increases for higher di/dt and 100A/us is quite slow for hard switching. 500A/us is a more practical figure and this may make Qrr two or three times bigger.
|Vdd|-|Vss|=140V
Fswitch=250Khz
140V*250Khz*160nC=5.6W of losses due to reverse recovery alone...
Note that this is not the full story since the transistors have to pass not only the reverse recovery current but the full inductor current while sustainging the whole 140V during the full Trr of each hard switching event. Real switching losses at full power may become as high as 30W (half in each half-bridge transistor).
Well, lets make another estimation, lets assume an average Trr of 60ns and an average |inductor-current| of 6.3A.
60ns*140V*6.3A*250000Hz=13.2W additional recovery losses...
And finally, lets account for crossover losses assuming an average crossover time of 28ns (just the typical 5V/ns body diode maximum allowed slope).
28ns*140V/2*6.3A=250000Hz=3W additional crossover losses...
In other words: Go for low Qrr rather than low Rds-on.
BTW: I know that those calculations are not precise at all, but they yield quite sensible results.
|
|
|
| luka |
hi
Ahh I allway forget that for +/-90 you need 200v fets, my mistake. Since this is half bridge, yea +/-70v is max with this fets.. as if this is not a lot... And yes low Qrr is allway better, but 4227's have quite low Qrr, compared ty fets that I have now
But if Qrr is so important then IRFB4610 would be even better or FDP3652 |
|
|
| odnaizutra |
hello Pierre,
have you got good results with 4227? |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
Yes, I am the same guy... :-)
|
Yeah we all know it by now.......
:D :D :D :D :D ;) |
|
|
|
