Sure it's possible but John expressed the wish not to do that, for noise reasons.
You can feed an artificial rail from that as long as something also regularly "eats" from that rail.
That is the "pull the hand brake a little while on ice" method :-)
An inductor is better as it wastes much less energy.
-- Regards, Joerg http://www.analogconsultants.com/
It is impossible to divert current instantly from one stray inductance because of the other stray impeadance. That's the problem. Even as small as 10nH makes for only 100 A/us/V
This is more like an engine brake.
The goal is making maximum net profit; isn't it?
Vladimir Vassilevsky DSP and Mixed Signal Design Consultant
Not sure what you are concerned about here. I thought you meant the voltage spike. That can be handled with diodes and a "dump rail".
That's what you normally don't want because it costs efficiency and generates heat. On trucks people don't care (yet) because at around $4/gallon the price of Diesel still seems not to be not high enough to recapture the energy :-)
Sure. Net profit in many markets occurs when your system has less noise then that of the competitor. Done that for years, for Doppler in medical ultrasound. One of the lithmus tests for docs was to see who had the least amount of noise bands in the FFT. Without such inductors in the switchers it was next to impossible to play. OTOH, people not doing things like that brought me a lot of business. Ka-ching :-)
-- Regards, Joerg http://www.analogconsultants.com/
when engine braking the injectors are off, the engine is just a big air pump it doesn't cost any fuel
-Lasse
Problem is the stray impeadance of the dump path, plus the turn-on time of the diodes. All of that appears to be surprisingly non-zero.
Sure. And I've designed the KW class SMPS; with severe cost constraints.
Vladimir Vassilevsky DSP and Mixed Signal Design Consultant
The engine will heat up during that effort and all that energy is vented off, lost. This is why the buses around here now have regenerative braking (except when the driver has to brake hard). It gets recycled into a battery and then used again for acceleration.
-- Regards, Joerg http://www.analogconsultants.com/
Turn-on time? That is blazingly fast. Turn-off is usually the issue. Well, since Si-C has become affordable not so much anymore.
We should have a contest, maybe eat cherries and see who can spit the pits the farthest :-)
-- Regards, Joerg http://www.analogconsultants.com/
"Blazingly fast" is not an engineering term. As mentionned already, 10nH is just 100 A/us/V
Larkin ?
Maybe not.
Or,
Hey, back on topic (isolated converters), how's this for an SSR from hell:
.-----. >-----|cheap|--+----.|------->
| DC- | | || | DC | R1 ||->- | | | | .---| |--+---------+--->
| '-----' =3D=3D=3D
-- Cheers, James
That's what's so neat about adding the capacitor to ground in the half-bridge. When a fet turns off, the current goes into the cap, and you get a nice smooth transition to the opposite rail. Neither fet ever sees a high-dissipation zone; no diodes have high di/dt to mess with their holes and electrons. There's no big dV/dT or ringing anywhere and no shoot-through. It's cool and quiet.
A really smart fet driver would synchronize the gate drives with this capacitive trapezoid thing. That's patentable. I hereby donate it to the world. Well, every million chips, you have to buy me a beer.
John
That's cool. We're already using the LTM8023, 2 amps at about $9 each. I could festoon my boards with a mix of them. Great word, festoon.
The other thing that would be nice would be a module with a lot of outputs, 4 or 5 maybe.
I've done that! A $4 dc/dc SIP and a 50 amp fet is a cheap 50 amp SSR.
John
I think some already do that, by monitoring gate current flow, but they're not telling.
RL
John Larkin wrote:
John, I'm having trouble discovering what is so unusual about your circuit. I made a half bridge driving an inductive load. I added 4 diodes to shunt the back emf away from the mosfet body diode so I can measure the current. I used 250KHz because I misread a post saying you were using
150KHz. But I don't think that affects things much.When a mosfet turns off, the inductor forces the voltage to go to the opposite rail, where it turns on the bypass diode. The current through the inductor starts decreasing linearly. There is no ringing on the output voltage or inductor current, but there may be on an actual layout due to stray inductance.
A little while later, the mosfet fet on that side turns on. But nothing happens since the inductor is still forcing current through the bypass diode. So there is no shoot-through and no power dissipated in the mosfet at turnon.
When the mosfet turns off, the opposite happens. The inductor forces the voltage to the opposite rail and turns on the bypass diode. The inductor current decreases linearly, and a little while later, the opposite mosfet turns on. Again, there is no current through the mosfet, so there is no power dissipated at turnon.
When you add a 2nF cap to ground, basically very little changes. The voltage transient becomes a ramp, but it is still over before the opposing mosfet turns on, so nothing has changed.
The only difference seems to be a small hump in the mosfet voltage at turnoff. This may increase the power dissipated slightly.
I don't know which DRQ127 inductor you are using, so I had to guess. I tried many different values with little overall change in the operation. Since you mentioned the waveforms are trapezoidal, I ended up with 50uH, but I don't think it matters a great deal.
I also tried different gate resistors, and larger values definitely affect the performance. Again, I had to use whatever mosfets were in the LTspice inventory, so that definitely would affect things if the ones you use have much lower input capacity.
Since everything seems to be governed by the inductor forcing current through the bypass diode until it goes through zero and the mosfet turns on, there does not seem to be any need to synchronize the gate drive.
About the only thing I can see is the cap slows down the transition, as you would have to expect. If it helps reduce the noise injected into the ground plane is not clear, since it is connected to ground and all the current through the cap goes into the ground plane.
So all things considered, I don't see much difference between having a bypass cap to ground, except it slows down the transitions. But it's not clear if that helps the noise injected into the ground plane.
I am including the LTspice ASC and PLT files. If you could, would you like to see if the analysis is correct?
Thanks,
Mike
SHEET 1 2108 800 WIRE 1072 -416 1056 -416 WIRE 1280 -416 1072 -416 WIRE 1424 -416 1280 -416 WIRE 1424 -400 1424 -416 WIRE 1056 -384 1056 -416 WIRE 1280 -304 1280 -416 WIRE 1424 -304 1424 -320 WIRE 1056 -288 1056 -320 WIRE 848 -208 816 -208 WIRE 992 -208 928 -208 WIRE 1008 -208 992 -208 WIRE 816 -192 816 -208 WIRE 1280 -192 1280 -240 WIRE 1344 -192 1280 -192 WIRE 1584 -192 1424 -192 WIRE 816 -96 816 -112 WIRE 1056 -96 1056 -192 WIRE 1056 -96 816 -96 WIRE 1168 -96 1056 -96 WIRE 1280 -96 1280 -192 WIRE 1280 -96 1168 -96 WIRE 1312 -96 1280 -96 WIRE 1344 -96 1312 -96 WIRE 1456 -96 1424 -96 WIRE 1488 -96 1456 -96 WIRE 1584 -96 1584 -192 WIRE 1584 -96 1568 -96 WIRE 1168 -80 1168 -96 WIRE 1056 -64 1056 -96 WIRE 1584 -48 1584 -96 WIRE 1168 0 1168 -16 WIRE 1280 0 1280 -96 WIRE 1056 48 1056 0 WIRE 1584 48 1584 32 WIRE 848 128 816 128 WIRE 976 128 928 128 WIRE 1008 128 976 128 WIRE 1280 128 1280 64 WIRE 816 144 816 128 WIRE 1056 160 1056 144 WIRE 816 240 816 224 FLAG 976 128 M2G FLAG 1056 160 0 FLAG 816 240 0 FLAG 1424 -304 0 FLAG 1584 48 0 FLAG 1312 -96 Vout FLAG 1072 -416 VCC FLAG 816 128 Vin FLAG 992 -208 M1G FLAG 1280 128 0 FLAG 1168 0 0 FLAG 1456 -96 L1R1 FLAG 816 -208 Vin2 SYMBOL voltage 816 128 R0 WINDOW 123 24 134 Left 2 WINDOW 3 25 95 Left 2 SYMATTR Value PULSE(0 12 500n 50n 50n 1.5u 4u) SYMATTR SpiceLine Rser=1 SYMATTR InstName V1 SYMBOL Voltage 1424 -416 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V3 SYMATTR Value 24v SYMBOL Voltage 1584 -64 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V4 SYMATTR Value 12v SYMBOL res 1584 -112 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R1 SYMATTR Value 1 SYMBOL diode 1040 -64 R0 SYMATTR InstName D2 SYMATTR Value ES1D SYMBOL res 944 112 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R3 SYMATTR Value 10 SYMBOL ind 1328 -80 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L1 SYMATTR Value 50µ SYMATTR SpiceLine Rser=1u Rpar=1e9 Cpar=1pf SYMBOL diode 1264 -240 M180 WINDOW 0 24 72 Left 2 WINDOW 3 24 0 Left 2 SYMATTR InstName D3 SYMATTR Value ES1D SYMBOL diode 1264 64 M180 WINDOW 0 24 72 Left 2 WINDOW 3 24 0 Left 2 SYMATTR InstName D4 SYMATTR Value ES1D SYMBOL diode 1072 -384 M0 SYMATTR InstName D1 SYMATTR Value ES1D SYMBOL cap 1152 -80 R0 SYMATTR InstName C1 SYMATTR Value 1p SYMBOL nmos 1008 -288 R0 SYMATTR InstName M1 SYMATTR Value STP8NM60 SYMBOL nmos 1008 48 R0 SYMATTR InstName M2 SYMATTR Value STP8NM60 SYMBOL voltage 816 -208 R0 WINDOW 123 24 134 Left 2 WINDOW 3 -60 136 Left 2 SYMATTR Value PULSE(12 0 50n 50n 50n 2.4u 4u) SYMATTR SpiceLine Rser=1 SYMATTR InstName V2 SYMBOL res 944 -224 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R2 SYMATTR Value 10 SYMBOL res 1440 -208 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R4 SYMATTR Value 1k TEXT 1064 -536 Left 2 ;'Larkin's MOSFET driver TEXT 1056 -504 Left 2 !.tran 0 120u 110u 100ns
[Transient Analysis] { Npanes: 3 Active Pane: 2 { traces: 2 {589826,0,"V(vout)"} {34603014,1,"I(L1)"} X: ('µ',0,0,1e-006,1e-005) Y[0]: (' ',0,-3,3,27) Y[1]: ('m',0,-0.25,0.05,0.25) Volts: (' ',0,0,0,-3,3,27) Amps: ('m',0,0,0,-0.25,0.05,0.25) Log: 0 0 0 GridStyle: 1 }, { traces: 4 {589826,0,"V(m1g)-V(vout)"} {589827,0,"V(m2g)"} {34668551,1,"I(D4)"} {34668556,1,"I(D3)"} X: ('µ',0,0,1e-006,1e-005) Y[0]: (' ',0,-1,1,13) Y[1]: ('m',0,-0.02,0.02,0.26) Volts: (' ',0,0,0,-1,1,13) Amps: ('m',0,0,0,-0.02,0.02,0.26) Log: 0 0 0 GridStyle: 1 }, { traces: 3 {34668548,0,"Id(M1)"} {34668547,0,"Id(M2)"} {34603013,0,"I(V3)"} X: ('µ',0,0,1e-006,1e-005) Y[0]: ('m',0,-0.25,0.05,0.25) Y[1]: ('m',0,1e+308,0.04,-1e+308) Amps: ('m',0,0,0,-0.25,0.05,0.25) Log: 0 0 0 GridStyle: 1 } }Actually, if you were going to add a slowdown cap, I'd put it directly across the inductor. That way it slows down the dv/dt across the mosfets, but the current is dumped back into the inductor instead of into ground.
That will be 2 beer please:)
Mike
Never mind. There is still a current pulse dumped through the other end of the inductor to ground. Fast edges, substantial current. Very noisy.
The only thing that seems to improve the noise is to reduce the cap to around 100pf or so. There should be an optimum value that minimizes the noise from dv/dt across the mosfets, and the current spike into ground.
Mike
The=20
=20
Hardly an appropriate comparison. Cell phones vs instruments, and acceptable levels of disturbance for the purpose.
Indeed. It is probably covered already in a bunch of patents.
Hold on. You may end up having to donate to patent owners.
VLV
Consider the lower fet: while it's turning off, its drain voltage is going up but the inductor/load is still forcing current into it. That product of volts and amps, during the transition, is a lot of power dissipation.
Now you have two choices: start turning on the upper fet (and get a huge shoot-through current) or have a deadband with both fets off (and get a bunch of ringing). Either way there's dissipation in one fet or both. Switching faster helps, but that's noisy... I've seen hundreds of amps of shoot-through current in not many nanoseconds... nasty stuff.
With a cap from the switch node to ground, the lower fet turns off gently at zero volts and the inductive current shifts out of the fet and into the cap. The cap voltage then ramps to the opposite rail, and the upper fet turns on at zero voltage drop.
The switch node has to get from one rail to the other. It's better to have that happen through a non-dissipative component than through a dissipative one.
Our situations are a little different. I'm driving a transformer which has a bridge rectifier and a load on its output. I'm seeing both the magnetizing inductance of the transformer and its leakage inductance. To get a little closer to my case, make L=3u and R1=5. That rings like hell. Now the cap, maybe 10nF, looks useful.
Note that I'm using an IRS2153 driver that has a fixed non-overlap time of about a microsecond. So I don't have arbitrary timing available to, for example, close that gap to prevent ringing.
John
The current through
snip
theres an ir2157 with quite a few more features including deadtime set with a resistor
a bit in this too:
-Lasse
John,
Wow! That's a lousy transformer. Are you sure those values are realistic?
With no capacitance, the inductor rings as you stated. But that's not the main problem.
The problem is when a mosfet is turned on, current flows in the inductor. This builds up a magnetic field that stores energy.
When you turn off the mosfet, the energy has to go somewhere, just like the spark ignition in your car.
The result is a fast edge into the diode on the opposite rail. The risetime depends on the mosfet and the bandwidth of the transformer. With
3u, it can be pretty fast.The current spike is dumped into ground or VCC, depending on which way it is switching. You also get a current spike from the other end of the inductor into ground, which means you have numerous injection points with fast edges.
These fast edges will cause problems in nearby circuits, like a 250MHz ADC.
Later, when the inductor starts ringing, it dumps more current into the rails. That is not good either.
Now change C1 to 10nF and look at the current in C1.
You get a fast 1.8A spike into either rail, depending on which way it is switching. So all you have done is change the location of the current spike.
I don't think this method is quite optimum for low noise.
I looked at the Colpitts someone posted, but that configuration has severe problems with output swing.
I'd email Joerg a couple of beers and get him to do an online tutorial on resonant power transfer.
Mike
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.