full bridge pwm controller

kubeek

Joined Sep 20, 2005
5,657
Hi,
i am looking for recommendations for a good pwm controller that could be used for push-pull or bridge converter. So far I found voltage-mode SG3525, and current mode UC3846 (1846 or 2846 are the same thing).
I would like to use current mode, but cannot find any working model of the 3846 for ltpsice. (found some very old pspice model of SG1846, but it would be PITA to convert it)

Do you have any suggestions for a controller that is not obsolete and has a model that would work in LTspice?
p.s. I am trying to stay out of phase-shift for now, as it is my first high power converter and would like to keep it within what I know and undestand.

tsan

Joined Sep 6, 2014
97
LTspice includes many controllers from Linear Technology. Look first on linear website for a current mode controller and then check if LTspice has a model for it. For example there is a model for LT1241 (and LT1619 and several others)
http://www.linear.com/product/LT1241

kubeek

Joined Sep 20, 2005
5,657
LTspice includes many controllers from Linear Technology. Look first on linear website for a current mode controller and then check if LTspice has a model for it. For example there is a model for LT1241 (and LT1619 and several others)
http://www.linear.com/product/LT1241
Thank you, but both of those are single-ended. I need to find something that does push-pull, half bridge of full bridge (which are equivalent topology)
So far I found LTC3721-1, so I am going through the datasheet.

kubeek

Joined Sep 20, 2005
5,657
God dammit, that stupid LT chip can only do up to 300ns of dead time, which is totally not enough. Any more suggestions?

ebp

Joined Feb 8, 2018
2,332
Why do you want longer dead time?

kubeek

Joined Sep 20, 2005
5,657
The IGBTs are kinda slow and I would rather lose some max duty cycle than destroy at least 20 bucks worth of parts. Better to creep up to that limit slowly than have the whole thing explode on the first turn on.
I am not completely sure that current mode control through a current transformer will be fast enough to prevent high current shootthrough and ensuing pandemonium. 4x680uF at 350V in my book says about 150J stored in the main capacitors, that would be one hell of a bang.

ebp

Joined Feb 8, 2018
2,332
I can certainly appreciate your concern with IGBTs. Many of the newer controllers do have short dead time to allow maximizing duty cycle. Clamping the input to the PWM in some may be possible, but it isn't likely particularly easy. I have a vague and possibly imagined memory of a controller with adjustable deadtime, but if it exists I don't recall which.

Linear Tech has focused more on low-voltage switcher products, some of which are quite marvelous. Unitrode (now part of TI) was more into parts for line-powered supplies. TI itself dabbled. Motorola dabbled. ST has some OK parts but is something of an also-ran.

It is very common to have to put some filtering on the current sense input to reject assorted transient currents associated with switching. By the time the current sense learns about the shoot-through the switches have already become light emitting FETs or IGBTs. Exciting, but not useful. Some of the newer controllers have active leading-edge blanking of the current sense input to eliminate the need for passive filtering. I'm not sure "dead time" isn't a deliberate double entendre.

"4x680uF at 350V in my book says about 150J stored in the main capacitors, that would be one hell of a bang."
Yes. Yet it is. Been there. Done that. Didn't like it. I had four 1000 µF caps (2s2p for a half-bridge with active power factor correction, 1.2 kW) at about 385 volts. Now I've never actually fired a shotgun indoors for accurate comparison ... You climb down off the ceiling, sit quietly for a bit to regain your composure then once more unto the breach. I was never sure exactly what happened since there were dead FETs too. I suspect a FET shorted due to a previous overstress, shorting half the series connected caps thus putting 385 volts on 250 V capacitors - somewhat beyond their short-term overvoltage rating. One place I consulted actually had custom-made acrylic blast shields for bench work. I often used 3/4" plywood. Right this moment the remains of a TO-247 FET sit on top of my microwave oven (don't ask, I don't know just how it found its way there, it is years old) - the plastic front of the package is blown away and there is a small crater in the middle of the drain paddle where the die used to be.

eetech00

Joined Jun 8, 2013
1,785
Hi,
i am looking for recommendations for a good pwm controller that could be used for push-pull or bridge converter. So far I found voltage-mode SG3525, and current mode UC3846 (1846 or 2846 are the same thing).
I would like to use current mode, but cannot find any working model of the 3846 for ltpsice. (found some very old pspice model of SG1846, but it would be PITA to convert it)

Do you have any suggestions for a controller that is not obsolete and has a model that would work in LTspice?
p.s. I am trying to stay out of phase-shift for now, as it is my first high power converter and would like to keep it within what I know and undestand.
Hi

I made an LTspice model for an SG1525 a few weeks ago as a basis for an SG3525. if you want to test it, let me know.

Or post the file for the SG1846 and Ill try to convert it to LTspice for you..

eT

Last edited:

kubeek

Joined Sep 20, 2005
5,657
...Or post the file for the SG1846 and Ill try to convert it to LTspice for you..
Thank you that would be very much welcome.
the model is here http://ltwiki.org/files/LTspiceIV/Mixed Part List/Spice-Models-collection/swit_reg.lib as sg1846, and the missing DIGIFPWR that you are going to need is here http://ltwiki.org/files/LTspiceIV/Mixed Part List/Spice-Models-collection/dig_io.lib That is as far as I got, I just didn´t have t he nerve to go through all the definitions of logic gates, which seem to be the main culprit, as I don´t know how ltspice treats those and their connection back to analog.

eetech00

Joined Jun 8, 2013
1,785
Thank you that would be very much welcome.
the model is here http://ltwiki.org/files/LTspiceIV/Mixed Part List/Spice-Models-collection/swit_reg.lib as sg1846, and the missing DIGIFPWR that you are going to need is here http://ltwiki.org/files/LTspiceIV/Mixed Part List/Spice-Models-collection/dig_io.lib That is as far as I got, I just didn´t have t he nerve to go through all the definitions of logic gates, which seem to be the main culprit, as I don´t know how ltspice treats those and their connection back to analog.
OK....might take a few days..but I'll give it shot.....

Last edited:

Jony130

Joined Feb 17, 2009
5,127
I made an LTspice model for an SG1525 a few weeks ago as a basis for an SG3525. if you want to test it, let me know.
Can you share the LTspice files?

kubeek

Joined Sep 20, 2005
5,657
These two appnotes seem they could help me a lot with understanding the error amp and comparator in the 1846, so I might be able to roll my own version along with SG3525 and be able to simulate it properly.
And here is the model of SG3525 that I use.

Attachments

• 402.8 KB Views: 6
• 178.9 KB Views: 8
• 1.2 KB Views: 2
• 2.6 KB Views: 4

kubeek

Joined Sep 20, 2005
5,657
Isn´t SG1525 just a tighter tolerance and wider temperature range sG3525?

eetech00

Joined Jun 8, 2013
1,785
Isn´t SG1525 just a tighter tolerance and wider temperature range sG3525?
Yes..from looking at the data...it seems to be.
I usually don't model temperatue anyway...so it would be easy to tweek a 1525 model for a 3525.

eT

eetech00

Joined Jun 8, 2013
1,785
These two appnotes seem they could help me a lot with understanding the error amp and comparator in the 1846, so I might be able to roll my own version along with SG3525 and be able to simulate it properly.
And here is the model of SG3525 that I use.
I'm curious...
Where'd you get the SG3525 model?

eT

kubeek

Joined Sep 20, 2005
5,657
Here is my simulation, it seems that the SS/shutdown functionality of tha SG3525a is as described, but I had before some issuses getting the pulse by pulse limiting to work correctly.
Do you guys see any possbile room for improvement? I don´t like the small oscillations on startup, but they are not too horrible. The circuit is going to be a TIG welder, powered by single phase 230V 16A socket. I think the biggest challange will be getting low noise signal from the shunt resistor, frankly I have little idea how much noise will the arc generate, requiring much heavier filtering and leading to much slower feedback and transient response.

Attachments

• 14.8 KB Views: 7

ebp

Joined Feb 8, 2018
2,332
A current sense transformer can be a good alternative to a sense resistor.

If I had the choice between spending $10000 on a welder and building one myself for a small fraction of the cost for materials, I'd buy one (I probably have most of the power components on hand - now just more junk I have to decide what do do with - off to the dump or try to find a home for them). I've designed a lot of switchers. I hate them. They are horrible unspeakable disgusting things. Here's a little something to mess up your slope compensation: Unless you use gapped ferrite for the inductor or are willing to use a very large expensive core, you will have to accept something in the range of 30% or greater reduction in initial permeability due to DC bias ("swing"). This isn't necessarily a big problem, but if you intend to model the circuit that must be part of the model. Gapped ferrite has several advantages over powder cores but it saturates very abruptly - not a disaster with fast cycle-by-cycle current limiting, but still something to be avoided. Unitrode/TI makes at least one average current mode controller suitable for bridge type converters. I'll try to dig up the part number. " Isn´t SG1525 just a tighter tolerance and wider temperature range sG3525?" Yes, it dates from the days when three temperature versions of each basic part were common. Thread Starter kubeek Joined Sep 20, 2005 5,657 A current sense transformer can be a good alternative to a sense resistor. I don´t think I can use a CT at the output, as there is almost pure DC current. Would have to use two CTs that can handle 200A and still would have the issues of getting a true image of the actual average output current. I was trying to infer this fom the primary CT, but didn´t have much luck with any simple function of the primary current corresponding to the average output current, especially at low duty cycle. If I had the choice between spending$10000 on a welder and building one myself for a small fraction of the cost for materials, I'd buy one (I probably have most of the power components on hand - now just more junk I have to decide what do do with - off to the dump or try to find a home for them). I've designed a lot of switchers. I hate them. They are horrible unspeakable disgusting things.
With the cost i think I will be about even with a simple machine (should have a working DC only TIG inverter at roughly 1000 USD), but I hope to get lots of experience building this thing.
Here's a little something to mess up your slope compensation:
Unless you use gapped ferrite for the inductor or are willing to use a very large expensive core, you will have to accept something in the range of 30% or greater reduction in initial permeability due to DC bias ("swing"). This isn't necessarily a big problem, but if you intend to model the circuit that must be part of the model. Gapped ferrite has several advantages over powder cores but it saturates very abruptly - not a disaster with fast cycle-by-cycle current limiting, but still something to be avoided.
I was planning to use an air-cored output inductor, because I couldn´t find anything suitable for the core. This should also allow me to (hopefully) couple HF start ignition voltage into it, when I later improve this to HF start and AC welding machine.
Already have all the main parts for the power board, so hopefully in a few weeks I will finalize the power and control boards and send it to a pcb manufacturer.
Unitrode/TI makes at least one average current mode controller suitable for bridge type converters. I'll try to dig up the part number.
Thank you that would be great.

Last edited:

ebp

Joined Feb 8, 2018
2,332
It looks like TI no longer makes the Unitrode ACM part.

Air core certainly won't saturate, but they surely do slop magnetic flux all over the landscape.