Development of a "reference" class D starting point

Status
Not open for further replies.
Jump to Latest
#362
Hi.
I would choose iron powder cores, as they can withstand much higher currents in the same size
Click to expand...
That is true, but powdered cores have extremly nonlinear L versus H characteristic. While they are good choice for SMPS inductors, I do not think you want another nonlinear element in class D design. See also Bruno's advice on that matter.

Best regards,

Jaka Racman
 
#363
Hi ssanmor!
I agree to Jaka that iron powder cores would be a LoFi solution due to their heavy nonlinearity.

Well the B/H characteristic of a ferrite itself may be better than iron powder. But it is still poor and nonlinear.
The key for low distorsion in this output chokes is the air gap in the choke design. In a ferrite design you need this air gap anyway otherwise you will run into saturation issues. The gap dominates the magnetic resistance of the magnetic path and by this it linearizes the overall behaviour.

Another promising idea would be some cut cores of a high saturation material. Some Materials would offer high saturation, low losses and an adjustable air gap as desired.
I.e Amorphous C from Magnetics or
Vitrovac&Vitroperm of the VAC.
Unfortunately such cut cores are not popular in small shapes
So this proposal is probably only suitful for high power Class D.



Bye
Markus
 
#366
Well, I fully agree in that ferrites are more linear, but for high power they need to be much larger. If you want a HiFi solution, you can use oversized iron powder cores, with current ratings so high that they will never approach saturation, for example 25-30A.
For powers above 200-300W, ferrite toroids are quite unusable, and the ones pulse and the other company makes are small, too. They seem to be gapped powder cores.

Best regards
 
#367
BTW, I've try to order EP17 N49 (15-20 turns, 1mm wires, .3-.5mm gap) for up to 200W@8ohm, but was offered only more then 100(!! and only ungaped EP20 N87🙂 cores to order..if i'll try order for iron powder cores, probably that situation would repeat. What wanna i say? While our pro's here and they still gives us free hints, we can save our money for more useful things, isn't it? :2c:
 
#368
Guys,

We use for the UCD180 an EP17 coil and for the UcD400 an EP20.

A toroidel is much more difficult to create a good electrolytic shielding. Thereby it's easier to design an iron powder coil as a ferrite coil, because you have to carefully calculate the air-gap.

And not to forget a good Class-D amp runs on 400-500kHz. You need really very good iron powder material to have the magnetic losses unther control, so I would go for ferrite 😉

Regards,

Jan-Peter

www.hypex.nl
 
#369
OK, it goes to ferrites.
Question no.2. What is the size of this final inductor, related to rated output?
In the sch it is only said like 300uH. But I can make 300uH using mini core or using 1cm2 Ae core.
I look at several commercial unit, they use 2 or 3 toroids, stacked together (maybe glued). Is this to increase power handling in this final choke?
 
#371
Well, seems no-one wants to speak for "everyone". What we have been doing in the latter part of this thread is to make an embodiment of Bruno's patented design, with a few minor additions (hopefully not detrimental ones 🙂). So it's right here...
It uses only discrete transistors for the comparator and mosfet drivers, which could be both a strength and a weakness. It's good from an educational point of view, but it also has constraints on how high you can go in terms of supply voltage and output power.
 
#372
Hi,

Few things I would like to mention:

-Output filter coil is 30uH in all shown designs, not 300uH!

-I don't think the discrete transistor versions are all that limiting, if you want more power just select suitable transistors, plus a few adjustments, coil, deadtime, etc.

- I doubt Bruno has a "shareable" design, and I don't think he will until UCD has become old news (when he's got something better and more saleable) so don't hold your breath on that one.

Here is my question:

Seems our posted versions of the UCD have the problem that they don't work into a low impedance load. At 4ohms it's OK, 2 ohms it is no longer working very well, but still modulates somewhat, at 1 ohm it's at resonant frequency with no modulation, as feedback can no longer track the reference. Would modifying the feedback network to a lead/lag help? Or can anyone think of a simpler way?
Throw an op amp in and crank the gain perhaps??

I think if anyone wants to bridge two into an 8ohm load or like myself, power a 4ohm speaker, it could pose a problem with dropping to resonance intermittently, whenever the dynamic load drops in value.

Thanks
Chris

EDIT: I made a mistake, it doesn't switch at "resonance" it switches at whatever the input frequency happens to be with a low impedance load. So, no modulation and a dangerously low frequency.
 
#374
I'm not an expert, but I had problem with such a "Charge Pump" upper mosfet driver (when I experimenting with HIP4080). It cannot drive too many mosfets, while if we design for subwoofer Class D it almost need parrareled mosfet (like 3 or 4)
You can parrareled mosfets, but very picky. The result is a hard to find mosfet.
I think floating 12V for upper drive is more stable.
Throw an op amp in and crank the gain perhaps?
Click to expand...
You just answered your own question. Make this : the differential (left and right), put the base of right differential to ground. And the base of left differential to the output of an opamp.
Input and feedback are put in the opamp, not in the differential.
It makes the differential and the rest of the cct is acting like comparator.
 
#375
Hi Charles,

You nailed that one, way to take the fun out of it, again, thanks! 😀

That's exactly the sort of solution I was hoping for, simple!

It now keeps full modulation down to at least .8ohm in spice, so I don't think there's a need to worry about using a 4ohm speaker with it. Values I tried are 15uH and 1uF. just some numbers I plugged in to test with, no idea what the Q is or anything but I can mess with it later. Fs is up as expected but not much, it's right at 420khz. Just right.

Thanks again.

Chris
 
#376
lumanauw said:
I'm not an expert, but I had problem with such a "Charge Pump" upper mosfet driver (when I experimenting with HIP4080). It cannot drive too many mosfets, while if we design for subwoofer Class D it almost need parrareled mosfet (like 3 or 4)
You can parrareled mosfets, but very picky. The result is a hard to find mosfet.
I think floating 12V for upper drive is more stable.

You just answered your own question. Make this : the differential (left and right), put the base of right differential to ground. And the base of left differential to the output of an opamp.
Input and feedback are put in the opamp, not in the differential.
It makes the differential and the rest of the cct is acting like comparator.
Click to expand...

Hello,

I'm not an expert either, not by far.

I think the HIP4080 and the like, use a combination of bootstrap supply + charge pump. The bootstrap supplies the initial charge to pump the gate capacitances to enchance it, and the charge pump just maintains the charge on the bootstrap capacitor to keep the gate from dropping low. I would have thought a solution of that type should be able to drive at least a few fets with ease? What mosfets were you testing?

I agree with you a floating 12V high side supply would likely be far superior. We welcome all circuit contributions if you feel up to it 🙂 Can model that with a VDC source and it works well, but to actually implement it requires a bit more work. I guess a few extra windings around the secondary and a zener?

A MOS based driver would be superior as well, I plan on taking a crack at that someday soon.

You're op amp plan is workable but it messes with the rest of my circuit as it stands now, and I much prefer the simple change of a few values to get the same result, also changing the filter values should fix the situation for everyones circuit versions, so I think I'll go with that. I had plans for adding an op amp in the feedback loop but, maybe later, or someone else can do it.

As it stands now I plan on posting a schematic for a fully differential version, slightly changed from before, along with it's full bridge con/version, sometime this week unless my computer should explode first. (DOOM 3 is out!!)


Regards,
Chris
 
#377
You're op amp plan is workable but it messes with the rest of my circuit as it stands now, and I much prefer the simple change of a few values to get the same result, also changing the filter values should fix the situation for everyones circuit versions, so I think I'll go with that. I had plans for adding an op amp in the feedback loop but, maybe later, or someone else can do it.
Click to expand...
Again, I'm not an expert. The way I look self oscilating classD, it have to got very big gain, so that the differential and the rest will work like comparator, giving output hi-lo only.
Putting this opamp infront helps alot in making this. It's kind of differ from the previous idea, but I think it will work better.
Maybe someone with SIM can look the advantage of having this opamp?
 
#378
Hi,

I played around with it a bit more yesterday and noticed that while a smaller inductor does improve things when driving a lower impedance load, the benefits aren't there once you start increasing output power.

So it seems an op amp would be a requirement. Another Idea I had to simply make it more robust, is to capacitively couple the output..

This more than does the job, but we don't really like that idea, do we? Gotta run, had a small kitchen fire and I need some air.

Chris
 
#380
Hi,

I did. In transiant analysis cap value seems to make very little difference. I dont' have a working model for AC analysis so that's the best I can tell.

The gain is there, but the rate of change is too slow to enable feedback to track the input at all. So it ends up with one or two modulations as their slopes intersect, and then nothing, producing a nice triangle on the output.

Compensating for that in the feedback loop is beyond me at this point. It seems easier to add the op amp. I'd probably just opt for putting it in the feedback loop and not part of the input stage. I suppose there's also little reason why it couldnt' be an integrating op amp either? Would it be best to place it before or after the current network? Anyway, I'm open to ideas as to what would be best.

Thanks,
Chris
 
Status
Not open for further replies.
Top Bottom