Half-bridge woes.

Thread Starter

HellTriX

Joined Apr 11, 2008
83
I setup a IRF3205 half-bridge driver driven by my single 555 PWM to the Lin and Hin pins. The output for the low works as intended. For the high output I calculated the boot strap cap close to where it should be even tried a few values up and down. I used a 4001 Diode and tried some 4183s (I think) that where in my junk box, (testing circuit with 5khz signal). The output waveform didn't even resemble a square wave. At this point I removed the mosfet from the driver and hooked ran the high output through a diode and resistor to charge a capacitor up and measure its voltage. I'm getting a 6.58v output. Ironically this is the exact same voltage that is going into the HIN and LIN pins from my 555. The boost strap effect is not working and I'm not sure why. The driver and 555 is being driven by a 12V regulated supply and I verified 12V to the inputs of each chip.

I'm probably going to be asked for a schematic? The 555 is trivial and proven with my other PWM circuit, and the half bridge driver was hooked up with the addition of a diode and cap for the boost strap and thats it.

I couldn't find a half bridge driver in Eagle or electronics workbench, so maybe you can give me suggestions on how to create a schematic if you need one?

tnx
 

Thread Starter

HellTriX

Joined Apr 11, 2008
83
Does this driver handle its own dead time? or do I need to make an inverting PWM signal and address my own dead time before entering the driver chip?

Seems like this defeats the purpose of the driver.
 

jpanhalt

Joined Jan 18, 2008
11,087
Does what driver?

Your description does not include a conventional gate driver, unless you are planning on using the 555 to do that. If that is the case, you need to supply a schematic, but I suspect a single 555 cannot drive top and bottom mosfets and provide a dead time too.

John
 

Thread Starter

HellTriX

Joined Apr 11, 2008
83
Here is a schematic.
The bottom one is a duplicate of the top except I tested without mosfets to try to ascertain the max voltage being generated by the bootstrap effect. This was measured at 11.6v of my ~12 supply.
D1 = 4148
C1 = Tried values from 10nf to 47mf. (nominally 170nf).
Qx = Are not IRF540s (Only their for illustration).

The problem is, I don't think the boosted voltage is working, Or I should be seeing over 20 volts at the HO when referenced to ground?
Any clarification would be helpful. ty.
 

jpanhalt

Joined Jan 18, 2008
11,087
I think VS needs a connection, such as the source for the top mosfet. VB is referenced to it. Also, it is clear you are trying to capture the voltage at HO with C3 so you can measure it with a standard DVM. Do you have a scope you could check it with instead.

See: http://focus.ti.com/lit/ml/slup169/slup169.pdf

Pay particular attention to pages 24-28 and Figures 26-29.

John
 

Thread Starter

HellTriX

Joined Apr 11, 2008
83
I think VS needs a connection, such as the source for the top mosfet. VB is referenced to it. Also, it is clear you are trying to capture the voltage at HO with C3 so you can measure it with a standard DVM. Do you have a scope you could check it with instead.
I do have a scope but its a cheap one. Not very clear what the voltage levels are, still working up to the point where I trust the results I'm getting. Will check out that pdf in a few minutes.

You'll see that VS should also be connected between the half-bridge MOSFETS.

This must be the missing piece, I completely overlooked that connection in the last few days of messing with this. I will adjust the circuit later today when I get back from then university.

Tnx guys.
 

SgtWookie

Joined Jul 17, 2007
22,230
Here's a caution for you:
You're driving dual MOSFETS on the low side and the high side using just one driver IC.
You need to determine whether the driver IC can supply/sink sufficient current to switch the gates on and off in time; otherwise you will wind up with the dreaded "shoot-through" condition.

The IR2105 can only source 130mA, and sink 270mA with a dead time of 520nS. Those IRF3205's have a heck of a gate charge requirement, and you're doubling it.

Be conservative. You're far better off to have too much "dead time" than ANY "shoot-through" time.
 

Thread Starter

HellTriX

Joined Apr 11, 2008
83
Here's a caution for you:
You're driving dual MOSFETS on the low side and the high side using just one driver IC.
The IR2105 can only source 130mA, and sink 270mA with a dead time of 520nS. Those IRF3205's have a heck of a gate charge requirement, and you're doubling it.
Forgot to mention I'm using an IRS2183 with a source of 1.5A and sink of 1.8A, just didn't have any exactly like this in eagle so I used that one instead for illustration.

I want to drive, 10 parallel FDP61N20 from the high side and 10 parallel FDP61N20 on the low side. These have a 58nC total gate charge typical and the frequency of operation will be 15-20khz. TBH, I haven't fully calculated the switching time/current, but if I ball-parked it right they should turn on/off without shoot-through and still switch fast enough to keep thermal switching losses low. These will also be highly derated. Meaning the theoretical 610Amp capacity at 200V will be 90-120V at 150 Amp average, with less then 2 second 300 Amp peaks. After some testing I will determine if I can move my average up to 250 Amp, and 500 Amp peaks. Which would be the absolute max I would need to switch at 15khz - 20khz.

If anyone wants to help me with the driver to mosfet gate charge x10 at 15-20khz mathematics, to check my work, would be totally welcomed :D

I had found a half bridge driver that was 4 or 9 amp (one of those two) but I couldn't remember where I saw it at. I was considering this for the parallel drive of all these chips. If not, I was going to use the 1.5A/1.8A, and if it turned out that it couldn't supply enough drive current, I was going to add a cascade stage with a second bootstrap to provide the required current to the output parallel stage. Math still a work in progress (lots of university home work keeping me from my hobby project).
 

jpanhalt

Joined Jan 18, 2008
11,087
If it helps, I have used the LT1158 to drive 5 parallel 1010E's top and bottom, and it works well. Frequency is 20 KHz or a little more. The LT1158 has a bit more current capability and more feedback control than the 2105. Don't know the actual amperage. Probably in the 300A ballpark under full load, but not stalled. I have stalled the motor and not blown anything. It's only at 12V.

To drive 10 in parallel might push it. And of course, one of the limitations of the T0-220 type mosfet is the package. Have you considered an IGBT --one of the really big ones -- that can be obtained on ebay fairly cheaply?

John
 

Thread Starter

HellTriX

Joined Apr 11, 2008
83
If it helps, I have used the LT1158 to drive 5 parallel 1010E's top and bottom, and it works well.
Have you considered an IGBT --one of the really big ones -- that can be obtained on ebay fairly cheaply?
John
From what I can gather the IGBTs are much less efficient than the mosfets.
I'm also not ready to try dealing with a negative bias to get them to fully shut off. I guess if my current idea falls through then maybe I should get some experience with IGBTs.
 

jpanhalt

Joined Jan 18, 2008
11,087
From what I can gather the IGBTs are much less efficient than the mosfets.
From what I can gather, you are considering something on the order of 100V @ 250 A (25 KW).

These references, tend to give a different view. IGBT are more efficient at that power range. John

http://www.irf.com/technical-info/whitepaper/choosewisely.pdf

MOSFETs are preferred in:
· High frequency applications (>200kHz)
· Wide line or load variations
· Long duty cycles
· Low-voltage applications (<250V)
· < 500W output power

IGBTs have been the preferred device under these conditions:
· Low duty cycle
· Low frequency (<20kHz)
· Narrow or small line or load variations
· High-voltage applications (>1000V)
http://www.st.com/stonline/products/literature/an/3717.pdf

http://www.techlearner.com/Apps/IGBTsIntro.pdf


MOSFET dramatically reduces the forward voltage drop at
current levels above 12 amps. By reducing the forward drop,
the conduction loss of the device is decreased. The gradual
rising slope of the MOSFET in Figure 1a can be attributed to
the relationship of VDS to RDS(on). The IGBT curve has an
offset due to an internal forward biased p–n junction and a fast
rising slope typical of a minority carrier device.
It is possible to replace the MOSFET with an IGBT and
improve the efficiency and/or reduce the cost.
 

Thread Starter

HellTriX

Joined Apr 11, 2008
83
From what I can gather, you are considering something on the order of 100V @ 250 A (25 KW).
This will be variable from 100V @ 0Amps up to an estimated peak of around 200-300A. 50-70% duty cycle at 100-150A expected.

MOSFETs are preferred in:
No · High frequency applications (>200kHz)
Yes · Wide line or load variations
Yes · Long duty cycles
Yes · Low-voltage applications (<250V)
No · < 500W output power

IGBTs have been the preferred device under these conditions:
No · Low duty cycle
Yes · Low frequency (<20kHz)
No · Narrow or small line or load variations
No · High-voltage applications (>1000V)
Used that like a checklist. It would seem Mosfets out weigh IGBTs for my current application unless I miss interpreted. Also a side note, This was one of my reasons for nearly doubling the amount of mosfets required because they have a good efficiency when you run them at half their rated current, IE: you get much lower Rds On values for the same current handling.

I think my next project or thereafter, I will play with some IGBTs. So I can have an experienced voice when weighting their pros/cons.
 

Thread Starter

HellTriX

Joined Apr 11, 2008
83
On another note:
Got the circuit operating from 12V supply and getting 18-20 volts at the high side gate (PS might be sagging below 12v). Showing that my bootstrap is working.

I put a 1ohm resistor between high and low mosfet with a center tap going to the Vs to simulate a motor and protect against shoot through for the initial test.

Which brings me to the following question:
Whats the best way to measure if I may have shoot through? I only got a single channel oscilloscope so I can measure high/low side and compare them. I don't have any more bright ideas on what to do to check this.
 

Thread Starter

HellTriX

Joined Apr 11, 2008
83
Master of answering my own questions.
At no load I measured a current of 18 microamps of current between the two mosfets, I guess this rules out shoot through. I know the chance will go up as load goes up, but its at least currently safe to being load testing.
 

SgtWookie

Joined Jul 17, 2007
22,230
I believe I'd try to figure out what the dead time really is.

Try connecting a couple of resistors in a divider network across the bridge to the middle, so that when both upper and lower sides are off, the divider network will pull the bridge halfway between supply and ground. You'll see a "stairstep" effect of sorts on the middle of the bridge when it's running. High, middle, low, middle, etc. It won't be exact, but you'll have a ballpark idea. You'll likely see an RC time curve when it's returning to the middle, due to all the capacitance from the multitude of drains/sources you have connected up. That's OK, just measure from the leading edge of "returning to the middle" to where the next up or down kicks in.
 
Last edited:

SgtWookie

Joined Jul 17, 2007
22,230
Hmm, that's not a lot of dead time. I haven't looked at the specs for the driver you now admit to be using. (you CAN use the Value button to change the text of the IC's on your schematic - on the toolbar on the left, click the resistor with "10k" highlighted underneath it, then click the IC whos part number you want to edit)

Try testing it over temperature, see if the DT changes. Preheat your oven to it's lowest setting for a half-hour, and then try running your board inside the oven on a couple of sheets of cardboard. Try sticking it in the freezer for a few hours, and test it again the moment it comes out. It won't take but a few seconds to warm back up, so you'll need to be quick.

If I'm guessing correctly, your DT will increase as the components cool, and decrease as it heats up. This is because MOSFETS have a positive temperature coefficient; as they heat up, their resistance increases. I'm betting that your gate driver IC has MOSFET outputs, and as it warms up it'll take longer and longer to charge/discharge the gates of your MOSFETS.

Just how much it'll increase or decrease I can't tell you offhand. I suggest you should find out though.
 

Thread Starter

HellTriX

Joined Apr 11, 2008
83
Looking at the datasheet, its telling me 280ns min, 400ns typical, and 520ns max dead time. Which is a lot more then the on/off times they list. The on/off delays are 180 and 220ns, with a delay matching of up to 35ns.

Guess this chip has really close tolerances. I was looking for something that
was quick when I bought them. Hope they are not too fast.

*Also know how to change values on the schematic sheet now :) tnx
 

SgtWookie

Joined Jul 17, 2007
22,230
Actually in that case, 2.8uS is much more than I would expect from the specs you just mentioned.

Here's a neat trick for you.
Copper Pour using the Polygon tool
In the Board layout,
1) Click the Polygon icon (just below the Arc tool icon)
2) Select the layer you wish to create the polygon (copper fill) on.
3) If your board is rectangular, select one of the right angle wire bends.
4) Miter:0
5) Width: 0.01 or larger (don't select zero! I suggest using your smallest trace width.)
6) Solid crosshatch (the icon just to the right of the Width box)
7) Isolate: 0.012 to 0.024 (experiment with this)
8) Spacing: same as Isolate
Now, draw the polygon on the perimeter of your board:
1) Left-click on the lower left corner of your board.
2) Left-click on the upper right corner of your board.
3) Left-click on the lower left corner of your board.
The polygon should now be closed/completed - if it is not, right-click to close it.
Filling the polygon:
Click the Ratsnest icon (the big X)
Your board will be magically filled with a copper pour.

You can change the name of the signal using the NAME tool (the resistor with R2 highlighted) to GND (or whatever name you're using for your ground reference) and the polygon will "merge" with all of your ground traces, creating a large ground plane.

The copper pour can make it difficult to see what you're doing. I just close the board and reopen it, but you can use the Ripup tool and click on the edge of the polygon to remove the fill.
 
Top