diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Class D (http://www.diyaudio.com/forums/class-d/)
-   -   Class-D Amp with IRS2092 (http://www.diyaudio.com/forums/class-d/116056-class-d-amp-irs2092.html)

mag 23rd January 2008 09:05 PM

Class-D Amp with IRS2092
 
2 Attachment(s)
Hi All,
I have quite finished the design of a 12V powered class-D amplifier using the new chip from International Rectifier called
IRS2092.

The amplifier can be used in BTL mode delivering up to 500W into a 4ohm load (I hope....).

By now only the power supply has been mounted showing good
performances. The amplifier section is not yet assembled, so I cannot tell you anything about it.

I know that the schematics is still full of little mistakes and some
components values has been changed.
Please take a look at it and tell me what do you think.

mag 23rd January 2008 10:34 PM

2 Attachment(s)
Here the second part of the schematics with the 12V PSU and
the input stage

mag 27th January 2008 01:12 PM

2 Attachment(s)
Hi all,
I have finished assembling my car amplifier. Everithing looks promising but I haven't done any power test yet.

The mosfets remain cool even without heatsink and the waveforms on the SMPS transformer and on the amplifier switching node look perfectly square without any appreciable ringing.

Now the power supply PWM frequency is set to 120kHz and the class-D frequency is set to 380kHz and the amplifier is powered
at +/- 50V

The power supply seems to work without problem but the
amplifier output inductance gets very hot even at idle after few
minutes.

I suspect that core losses are too high since the same behaviour appears even at idle.

The output filter is a 4th order bessel type with 35kHz corner frequency with 2ohms load. I have calculated the ripple current on the first inductor and it is around +/- 1A so I suppose that the losses are not due to the wire gauge.
I have tried to rewind the inductance using home made litz wire but basically nothing changes... it still remains very hot.

The inductor I used is the Bourns 2300-HT-220-RC rated for 19A of continuous current and 8mohms of DC resistance.
I have measured the inductance value and it has the right value (22uH). It has 17 windings so the AL value of the core must be 76nH/t.

Can you recommend me an inductance or even a ferrite core that can be used at those frequencies and at those power level?

Thank you

Workhorse 27th January 2008 01:19 PM

http://toroids.info/T106-2.php

enjoy!!!!!!!!

lumanauw 27th January 2008 04:16 PM

Hi, Mag,

Very nice indeed :D You do the SMD soldering by hand? How does it perform, sonically?

Eva 27th January 2008 04:44 PM

It seems that you are developing a commercial product.

Did you make any core loss calculations? The selection of iron-powder materials suitable for class D is quite limited. You can find useful resources at www.micrometals.com


Consider the following changes:

- Use a toroid transformer in the SMPS for reduced leakage flux.

- Remove the big electrolytics before the 12V input filter inductor or they will overheat and vent. This is because they will be subject to extreme ripple currents (alternator and vehicle PWM) when the amplifier is placed in a vehicle. Use film and small high ESR electrolytics only.

- Use the on voltage drop across the primary side SMPS MOSFETs to trigger the overcurrent condition. You don't need a separate sense resistor.

- Change the "remote" scheme to avoid drawing 1mA when the amplifier is off.

- Separate primary and secondary side grounds. Use optocouplers to transfer signals.

- Consider implementing the protection features with a small microcontroller. It saves plenty of parts.

BTW: You have an input preamplifier and filtering stage, don't you?

mag 27th January 2008 05:54 PM

Thank you Workhorse,
so you think also that is the core that have high losses
at this frequency.

I see on the specs of the T106-2 core that the AL value is only 13.5. To reach 22uH I need 40 turns on that core so I can not use
a very thick wire.

Which wire diameter do you suggest knowing that the output
load will be 2ohms. Is a litz wire necessary or I can use a solid wire?

To assembly the SMDs I use solder paste and an hot air blower.
The smallest components on my amp are 0603 but with this technique you can even mount smaller components such 0402 or even 0201. Basically is the same technique used in mass production, the only difference is that you need to manually put the solder paste on the components pads, manually position the components on the right place and finally heat everything up until all the soldering are done.

mag 27th January 2008 05:54 PM

Quote:

Originally posted by Eva
It seems that you are developing a commercial product.

Did you make any core loss calculations? The selection of iron-powder materials suitable for class D is quite limited. You can find useful resources at www.micrometals.com


Consider the following changes:

- Use a toroid transformer in the SMPS for reduced leakage flux.

- Remove the big electrolytics before the 12V input filter inductor or they will overheat and vent. This is because they will be subject to extreme ripple currents (alternator and vehicle PWM) when the amplifier is placed in a vehicle. Use film and small high ESR electrolytics only.

- Use the on voltage drop across the primary side SMPS MOSFETs to trigger the overcurrent condition. You don't need a separate sense resistor.

Thank you Eva,
no it not a commercial product at all... it is just for fun and it is my first class-D amplifier desing. I have made it with SMD just because I think that they are much more easy to handle rather than standard components.

The SMPS transformer core is Epcos RM14 with N87 material. It can handle around 1kW at 100kHz; beeing a closed core I don't think that the radiated EMI will be a concern. I used an RM core just because it was readily available and it is much more easy to mount on a PCB. In any case I will try with a toroid and see what happens.


I will remove the first electrolytic to avoid problems. It was only to improve filtering but of course the ripple current will be huge...

You mean using the Rds-on of the mosfet for current sense? Nice idea, I like it. In any case the current protection does not work now, There is some noise picked up from the trace connecting the
sense resistor to the SG3526 that makes the protection tripping even at idle...

Eva 27th January 2008 08:21 PM

Ouch... If you are not experienced you are going to have a hard time making this current limiting (and the entire project) work. Consider leading-edge blanking and low-pass filtering to cancel shunt inductance, and don't underestimate the EMI radiating cpabilities of such a push-pull transformer, it's leakage inductance what causes trouble... Foil windings and a flux band may help, but it's just easier to use a toroid. Remember that you will need slope compensation for current limiting with duty cycles above 50%.

Another recommendation is to use fast recovery rectifiers rated at the lowest possible voltage because they have lower Qrr resulting in a smaller and shorter turn-on current spike at the primary side (and less EMI radiated by the transformer).

mag 27th January 2008 08:43 PM

I have added a 100nF cap in parallel to the current sense shunt and I have replaced the capacitor in parallel with the SG3526 current sense input pin with a 1uF cap (it was 10nF). Now the current limit protection does not trip anymore at idle but I don't know if it trips when needed....
I think that a 1uF capacitor will slow down the overcurrent protection making it almost useless.

For the diodes I use the RURP3060 from Fairchild rated 30A 600V.
No indication about QRR, in the datasheet only trr is shown and it is 55ns.
There is also another version of this diode called RHRP3060, same voltage and current ratings but faster, 40ns.

I have found another issue with the SMPS: when the input voltage reaches 14.8V and the output is regulated to 48V it start
to work in discontinuous mode. If I open the feedback loop the dutycycle comes back to around 50% and the output voltage is 60V.

Can the SMPS work in discontinuous mode at idle or there are some drawbacks in doing it?

Probably removing one turn from the transformer secondary side will solve the issue, what do you think about?

I think that the most elegant way to solve this problem is to use syncronous rectification but it involves a complete redesign of the power supply and I want to avoid it.


All times are GMT. The time now is 04:47 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2