My DIY UCD

Hi folks,

Well, I've mentioned my projects from time to time, fully intended on doing this P2P to keep costs as low as possible, and things have taken a turn that's forcing me to abandon my project for the time being, though I fully intend on getting back to it as soon as I can, maybe even sooner than I think right now.

In the meantime, I've decided to post what I consider to be a fairly final version of what I was going to wire up, have at it as you will.

I still have a very high powered full bridge design in the works that I started long before this, however I've learnt more during this half bridge project and know things need to be changed on it, so you won't be seeing that anytime soon.

I strongly recommend if you want an excellent amplifier, buy it from Hypex, but this is a good DIY project and much can be learned by it from many, I believe.

I'm not at all claiming it's 100%, but I think it's OK and should give a reasonable start to a working circuit.

This was made for ~65W RMS, but should handle much more, gain as shown is 27dB.

About the only thing not shown is the current source (it's ideal) and the delay circuit to switch it on via said current source.

However, this has zero protection circuitry whatsoever.

This simulates in transient very respectably, but since that's meaningless.. keep in mind alot of these values will need tweaking in a real world circuit.

Please excuse the untidyness of the circuit, I didn't have time to clean things up. I think you can all manage to figure it out though.

For a 1V input @ 20kHz:

TOTAL POWER DISSIPATION 1.04E+00 WATTS
DC COMPONENT = 7.670111E-03
TOTAL HARMONIC DISTORTION = 5.350052E-02 PERCENT

Regards,
Chris

Output waves,

Enjoy.

Hi there,

Thanks Subwo1. You're most welcome to make use of it in your project should you so desire, or a variation thereof.. or not

I appreciate the comment anyway.

Hi Sander, it was actually a total coincedence! Hmmmmmm, or, was it?

I thank you for your offer and your interest. If you'd like to prototype it or anything.... I only intended on doing it P2P for cost reasons but a nice PCB would be something too.

It of course has to purely be done for the love of it for the obvious reasons.

If you're curious as to what my new full bridge looks like, its' very much like this half bridge, just double (mirror) the cascoded outputs, increase the current to the LTP as the quad outputs divide it, then adjust the mirror balance resistor accordingly , and also mirror the output section.

I'm not yet satisfied with it enough to post it as a few of the original component selections aren't as good as I'd originally thought, it did simulate extremely well though, but I think it can be alot better. I'd rated the parts to handle 2kW continuously, I figured the drivers would hit a ceiling long before that though, but you know, doesn't cost anything at all to do on spice.

I'll help out anyway I'm able to if you decide to take it on, and you're welcome to email me or post here, I should be getting them, just won't have the time to tinker for awhile is all.

Take care,
Chris

Hi,

I already had a few more screens from simulations of this same circuit so I thought I'd post them as well.

This first one is of the same circuit prior to a few changes, where it had only ~20dB of gain, ideal power supplies for the mosfet drivers, aaaand.....yeah, only 2 ohm series gate resistors.

Next screen will be of the gate drive signals only. I know it seems like the gate signals intersect rather high ~2.5 to 3V, but that seems to be the sweet spot that keeps shoot through in check and power dissipation low, at least for the simulator.

Regards,
Chris

Gate signals + switching instance of half bridge.

That's it for now.

Regards,
Chris

Hi Ivan,

Simple. It is only my standard mosfet for simulating with, IRF540 advanced analysis model. It is the one component that has always been a constant, so other changes are better evaluated with things I've tried before.

I've also not found a better working model, and the parameters are fairly realistic.

The device I use currently in the actual circuit however is Fairchild's FDP3682, free samples. I've also tried FDP3672, they run a little warmer, bigger input capacitance, and I can only do so much as far as optimizing goes with no scope at hand, which will soon change. I'd say they're good mosfets though.

An IRF540Z might be fun to try though. Use whatever works anyway, it's all part of the fun of it.

Thanks,
Chris

PS:

Fairchild does have models of the mosfets I've mentioned using, however I've found that by the time you've tweaked settings enough for them to converge, it cant' simulate with any accuracy at all and still fails, so their newer spice models aren't worth trying.

Hi Sub,

No need to thank me for the thread it didn't cost me a dime

Those gate drive waveforms show the timing shift between the upper and lower drives due to capacitive interaction.

Click to expand...

That it does, and I found that minimized with these values, it can be much worse.

With these values I can actually push the power up a good amount before shoot through becomes apparent, I'd bet the same could be done in a real circuit, as JohnW said the sweet spot has to be found. There needs to be a certain amount of overlap I believe, which should be fine as long as it is below threshold, and have intersection occur between 2.5 and 4V. The goal is of course to strike equilibrium between overlap and deadtime. This works best in simulation anyway.

However at some level certain measures need to be taken to beef up the driver and also parallel schottky diodes with the body diodes like I have in the full bridge version.

Neither one is of any concern at only 25V (or 35V) rails.

Regards

Hi,

Consider this a continuation of my previous post, the intent is to demonstrate the importance and usefulness of striking the previously mentioned "sweet spot" with respect to the gate drive signals.

It has taken a long time for me to figure this much out and I figure if anyone wants to attempt this circuit, or a similar one, it is need to know information.

With this circuit such performance is most easily attained by setting the comparators output current to a reasonable level, 3mA works great for me, then play with the ratio of turn on/off bias resistors, you can keep the series gate resistors rather low while doing this, 2 to 10 ohms let's say. You have some room to play with it before it affects that sweet spot you've found, but if you need to increase it beyond that and it messes things up, a slight increase in the turn on bias resistor should do the trick.

What you want to be looking at while doing this is the intersection point (aim for between 2.5 and 4V, 3 works good for these models) and slew of the gate drive signals (twice as fast for turn off), as well as the current through each mosfet. Once the current looks reasonable (no kilo amp spikes) then you're getting into the ballpark and power dissipation should be reasonable automagically.

You've already seen how nice the current though the output devices looks in my simulation at low power with 25V rails, now I prove my point with a demonstration at 50V rails and nearly max modulation. The increase of rail voltage is the only change made to the circuit I've posted in this thread.

I do hope this is found to be useful, and saves people the _many_ hours I've spent on it to get it this far. It was actually the biggest obstacle I've had to overcome with it, without doubt.

Yes, I'm aware Vth declines with increased temperature, so it stands to reason in a real circuit you'd want to optimize this at operating temperature... you'd have a hard time not doing that anyway

The 50V, incedently, is worst case for the IRF540 model, at 35V rails the max output will look very much like it does at 25V rails, with no current spikes at all.

I hope this makes some sense I've typed it up rather quickly.

First screen is all waves including power, next screen will better serve the intent of this post.

I hope this is helpful.

Regards,
Chris

.... you can see in the above screenshot that power dissipation has increased a fair bit from the simulation with lower rails, I think it shows ~0.5W dissipated in one of the prior screens.

The reason it has increased in a way related to the obvious increase in rail voltage, but more directly related to the mild current spiking. If the drivers were re optimized to remove that at this rail voltage and output level, dissipation would drop back down.

You see now how critical it is to find the sweet spot.

If this is off by just a little bit, shoot through and dissipation will be up into the Kilo range, resulting in a rapid smoke show, and maybe some fireworks too.

Regards,
Chris

Hi Charles,

Actually Digikey has them in stock and they can be bought in single or small quantities, I was thinking of trying a few on my next project (which.. is this thread's circuit) to see how they compared to my fairchild devices.

Regards