diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Power Supplies (http://www.diyaudio.com/forums/power-supplies/)
-   -   FAST high-side drivers/circuitry? (http://www.diyaudio.com/forums/power-supplies/69662-fast-high-side-drivers-circuitry.html)

mzzj 13th December 2005 10:21 AM

FAST high-side drivers/circuitry?
 
does anyone know fast, high-current(6A or more) high-side drivers for 300-400v dc bus voltage?
IR2110 and relatives are piece of sh*t...

Currently i am using toroidal gate drive transformers in my 4kW subwoofer-inverter but GTDīs leakage inductance is limiting performance somewhat. With DIY toroid I can get sub-250nH leakage inductance witch is great compared to commercial e-core **** from coilcraft etc but i want faster.... :)

4.5kW beast...

Pierre 13th December 2005 10:32 AM

I got good results with a TC1427 driver chip driving a T0584 Pulse GDT at around 60-100KHz.
I also had bad experiences with IR2110 for my offline power supply, but good results in a Class-D amplifier (+/-65V rails).

phase_accurate 13th December 2005 11:09 AM

If you don't mind some addidional circuitry then you might also consider one of those IXYS low-side drivers combined with a magnetic coupler (which can be used like an optocoupler but is chepaer and faster).

Regards

Charles

Eva 13th December 2005 11:30 AM

Insert a buffer circuit between the pulse transformer and the gates and use the own pulse transformer both for powering the gate driver and for transmitting the drive signal (two independent rectifications are required). You may probably need to modify your circuit in order to drive the transformers in both directions (bipolar pulses).

I've done that kind of high current gate drivers in a cheap way with CD4050 buffers (for a sharp clear drive transient) and ZTX851/ZTX951 fast high-current bipolar transistors. However, you seem to have far more resources and money for experimentation than me, so you may be able to buy dedicated driver ICs and order thight SMD PCBs. You may even use the same TI gate drivers, but place them after the transformer, that's the trick :)

mzzj 13th December 2005 12:15 PM

Quote:

Originally posted by Eva
Insert a buffer circuit between the pulse transformer and the gates and use the own pulse transformer both for powering the gate driver and for transmitting the drive signal (two independent rectifications are required). You may probably need to modify your circuit in order to drive the transformers in both directions (bipolar pulses).

I've done that kind of high current gate drivers in a cheap way with CD4050 buffers (for a sharp clear drive transient) and ZTX851/ZTX951 fast high-current bipolar transistors. However, you seem to have far more resources and money for experimentation than me, so you may be able to buy dedicated driver ICs and order thight SMD PCBs. You may even use the same TI gate drivers, but place them after the transformer, that's the trick :)

Yeap, something like this i had already in my mind. I am just lazy winding GDT's so all-in smoking driver chip would be nice. :D

I found quite nice looking pdf with separate power and pulse transformers using small ferrite beads. And it didnt have any limitations on pulse width as usual implementations have.
.http://www.itee.uq.edu.au/~walkerg/p.../fetdrvr-t.pdf

More resources and MONEY? :D :D
as a student my financial resources are extremely limited at the moment...
this is near-zero budget project, only paid for 4 pcs. LM361 comparators. ;) Everything else is from scrapped welding machines and other second-hand sources exept sample drivers from TI

Eva 13th December 2005 12:54 PM

You don't need independent power and pulse transformers, a single one may do both things.

http://eva.eslamejor.com/old_gdrv.jpg

This is a picture from some old gate drive boards. The one on top is from a buck stage and operates at any duty cycle, the transformer has a center tapped secondary (not strictly required) and the primary was driven from a SG3525 IC. It was able to provide aprox. 80ns crossover time driving a IRFP460 (1 ohm gate resistor). The other is from a full bridge stage, it works much in the same way but only one winding is required to drive it since effective duty cycle is never above 50%.

PD. The CD4050 buffers are absolutely fakes, this is the kind of cr4p that you can buy in local stores in Spain (some units were defective and failed, taking their IRFP460 and ZTX851/ZTX951 away, fortunately others worked fine).

subwo1 14th December 2005 04:54 PM

If approximately 20ns rise and fall times are good enough, you may like to consider buffering the outputs of your IR2110 with IRF7343 N/P channel MOSFETs. That way, you get 20-30 ampere peak drive and the MOSFETs protect the driver IC from spikes. Those MOSFETs have threshold voltage ratings of about 1 volt, so you can connect them as source followers for ease and simplicity.

EDIT: It is a good idea to place something like a ten or twenty ohm resistor from gates to sources of the source followers to settle down the turnoff voltage quickly. I have also separated the gates and driven the N-channel one through a resistor to slow down the the rise time some.

mzzj 15th December 2005 03:00 PM

Couldnt get any better rise times driving mosfets directly with ucc37321 than what i get with pulse transformers. Less ringing on gate voltage at least.

Joseph Hynes 23rd December 2005 02:37 PM

mzzj:
Your GDT x-formers look neatly made, and you've got a pretty impressive L(leak), but there is a possibility you could do better than that.

These are 2 GDT I made for a 1KW SMPS, (I was lazy, I made one trafo for all switches) .
The first, has 12 strands twisted together, 4 filar primary with 2-filar each secondary. It has a leakage L of about 0.14uH.

http://www.dicksongage.com/3.jpg

As I understand it, for lowest leakage L, the idea is the make the primary and secondary "occupy same space", or as close to that as possible. (ok, this is heresy if you're looking for regulatory approval--creepage/clearance are gone) .

The second GDT was made from 12 turns of USB cable wound on a 26mm toroid, and had a L(leak) of under 0.10 uH. It has 4 vinyl insulated conductors ( the secondaries) inside a braided shield (the primary), and probably has better insulation characteristics.


http://www.dicksongage.com/1.jpg

I had intended to test the "winding inside a tube" model and didn't bother to make it neat. It works well enough that you could find a bare shield braid, put insulated conductors through it, and make a much neater job than I did.
Ultimately the creepage/clearance deal depends on your particular primary-secondary voltage difference, and your level of comfort with with your handywork.

Joseph

http://www.dicksongage.com/2.jpg

poobah 23rd December 2005 03:48 PM

mzzj,

Here is a link to a chip-set that use one trans for both power and signal... just some food for thought. I used these once, and I did not like them, very complicated, but the principle is worth learning. They are now obsolete, as they should be.

http://focus.ti.com/lit/ds/symlink/uc3727.pdf

My best experience has always been with one trans for floating power and an opto for the signal. I agree, the IR21XX are s**t.
I don't know your requirements for propagation delay and timing accuracy. Maybe too much jitter for you. I don't know of a chip that has the 4-5 Amp drive you want... I think you'll need discrete buffers for the high amperage drive.


All times are GMT. The time now is 06:20 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright Đ1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2