Class D and Very Low Impedance loads

The following question landed in the thread about digital vs analogue class D. I'm starting a new thread here.
LineSource said:
Hi Bruno,

Would a digital amp be a "good" solution to drive about 25 watts into a 0.1 ohm resistive load like a ribbon speaker?

The large inductor filters on the digital amp outputs I have looked at would seem a limiting factor for such a low, almost pure resistance, 0.1 ohm load.

Welcome to the class D section :)
Your question pertains to whether class D amps (digital or analogue) are suited for driving super-low impedance loads.
An amplifier of any kind (linear, class D or tube) is designed to deliver power efficiently in a certain load impedance. You can start with any impedance in mind and design an amplifier for it. So, just like you may never have seen a class D amp designed to deliver 2.24V rms and 22.4A rms with good efficiency, I've never seen a linear amp that does this. So it's not just the coil that's the limiting factor. The MOSFETs too will have a current rating just upward of that of the coil.ç
Yet, any amplifier topology can be adapted for any load impedance, although there sometimes are practical limits.

As it happens I do think that class D is more suited for this work than linear amplifiers because of the availability of optimised power parts, thanks to PC technology. Because modern PC microprocessors often run at only 0.8 to 1.2V but can consume a whopping 100W, there's a whole family of MOSFETs, drivers and coils especially designed for this kind of work. SMD FETs capable of 100A/7V are not unusual.

Compared to normal class D amps the layout is somewhat more critical (which means something!) if efficiency is to be good.

So the answer to your question is yes, but you'll have to, erh, diy.
 

bcarso

diyAudio Member RIP
2005-05-03 4:09 am
Canoga Park, California
Good point Bruno regarding using technology associated with low voltage processors etc.

For a 0.1 ohm load I think it would be mandatory to have each amp live with/at the speaker. If you can do that you might be able to confine the magnetic field radiation component to as to operate the amps without output filters. There will be a bit of extra dissipation in the ribbons but they ought to be able to handle it. The E field components will be small since the voltages are low.
 
bcarso said:
Good point Bruno regarding using technology associated with low voltage processors etc.

For a 0.1 ohm load I think it would be mandatory to have each amp live with/at the speaker. If you can do that you might be able to confine the magnetic field radiation component to as to operate the amps without output filters. There will be a bit of extra dissipation in the ribbons but they ought to be able to handle it. The E field components will be small since the voltages are low.

Ouch... I always shudder when suggestions of abolishing the output coil are made. 0.1 ohm suggests that it's really just a flat strip of metal. I'm not sure if inductance of this metal would not be enough to get the idle dissipation low enough. @linesource, what are the dimensions of the actual ribbon? This way we can calculate the inductance and see if a filterless version would fly.

You're right about the E field. Besides, I suppose there will be some metal grille in front to keep fingers out. The M field might require a bit more care.
 
Bruno,

Thanks for stating this discussion. A high efficiency digital amp that could drive low impedance would be useful for many applications besides ribbons.

The ribbon is 3" wide and 90" long of 8 micron thick aluminum foil. In addition, there are one long ~126" and one short ~36" guage 10 copper interconnect cables to the amp.

My original thinking was to use multiple output transistors, in parallel each with a modest size inductor. This would be similar to a Class-A amp with multiple bipolar outputs, each with a 0.1 ohm current balancing resistor.
 

bcarso

diyAudio Member RIP
2005-05-03 4:09 am
Canoga Park, California
"Ouch...." Yes I know what you mean---but if a phase-modulation "filterless" design was done it might be ok. Unfortunately that's twice the big fat silicon and the attendant Ron losses. But with a few milliohms per FET in each side it would still be bearable.

Linesource, could you do an inductance measurement---just of the ribbon, or at least of what the minimum L would be with presumably the one return wire from one end or the other?

It would be interesting to see how that return wire could be routed so as to miss messing with the magnetic circuit for the ribbon while still minimizing the loop area for B field radiation.
 
bcarso said:
"Ouch...." Yes I know what you mean---but if a phase-modulation "filterless" design was done it might be ok. Unfortunately that's twice the big fat silicon and the attendant Ron losses. But with a few milliohms per FET in each side it would still be bearable.

Linesource, could you do an inductance measurement---just of the ribbon, or at least of what the minimum L would be with presumably the one return wire from one end or the other?

It would be interesting to see how that return wire could be routed so as to miss messing with the magnetic circuit for the ribbon while still minimizing the loop area for B field radiation.
The 3-level (class BD) method is quite effective in combating losses in the load when no filter is used. This still leaves EMI which can only be addressed by shielding the ribbon.

On the other hand, 3-level PWM is not a good choice in terms of distortion because there is no soft-switching in the power stages. Especially at low signal levels you'll have to contend with significant zero-crossing distortion.

To make a low-impedance connection to the ribbon one could use the metal frame (chassis) that is supposedly already there to hold the ribbon and the magnets in place, and connect the far end of the ribbon through the frame. This would incur much less connection resistance than a wire, however thick. Connecting the chassis to the amplifier rules out filterless 3-level operation of course (chassis would be "hot").
 

bcarso

diyAudio Member RIP
2005-05-03 4:09 am
Canoga Park, California
As I do some rough estimates on inductance the standard class D doesn't look as terrible as I thought. Even with only 1uH the ribbon dissipation at idle for a ~20W full power amp with plenty of headroom is less than 38mW, supposing a 500kHz switching frequency. It's a lot of current but the resistive component is so low that the dissipation is small.

A sort of amusing issue though for the transducer in general is frequency response, since with a microhenry (which seems small at first glance) the 3dB rolloff is at omega = R/L, which is only around 16kHz. But then I guess this is a midrange element anyway.
 
bcarso said:
As I do some rough estimates on inductance the standard class D doesn't look as terrible as I thought. Even with only 1uH the ribbon dissipation at idle for a ~20W full power amp with plenty of headroom is less than 38mW, supposing a 500kHz switching frequency. It's a lot of current but the resistive component is so low that the dissipation is small.

A sort of amusing issue though for the transducer in general is frequency response, since with a microhenry (which seems small at first glance) the 3dB rolloff is at omega = R/L, which is only around 16kHz. But then I guess this is a midrange element anyway.

OK but you will have to shield your transducer electrically if you want to run it without inductor. Again, since inductors made for this power/current level are common in the PC world you might consider making life easy from the point of driver construction (not having to take EMC measures on the driver side).
 
Is this a full range ribbon - or just Mid - Hi?

If full range (Bass), have you considered that you’re going to have to use a full bridge (2x RDSon losses), with a Half Bridge I can't image how you would prevent PSU pumping with such a Low Load Resistance....

Regulated PSU that can Sink as well as Source maybe, but then you might as well use a full bridge.

John
 

bcarso

diyAudio Member RIP
2005-05-03 4:09 am
Canoga Park, California
Bruno, I agree it might be more reasonable just to use an additional L C. It would seem that the magnetic shielding is more trouble than it is worth (although there is probably a clever way to produce some mostly cancelling fields by routing of the wires. Again, more trouble than it is worth. It was just entertaining to consider.

John, I agree pumping would be a huge problem. However, this whole exercise leads to such constraints on the power supply anyway that it might as well be considered part of the design to begin with, and as such have as you say some low-loss sinking/sourcing capability from the outset. Whether that functionality entails the same amount of silicon and magnetics as a full bridge I don't know.

Brad
 
JohnW said:
Is this a full range ribbon - or just Mid - Hi?

If full range (Bass), have you considered that you’re going to have to use a full bridge (2x RDSon losses), with a Half Bridge I can't image how you would prevent PSU pumping with such a Low Load Resistance....

Regulated PSU that can Sink as well as Source maybe, but then you might as well use a full bridge.

John

Even for the mid/high I would seriously look at full bridge.
 
I was thinking along the lines that the only thing that limits how much the PSU rails increasing with a Half bridge design (during pumping) is the amount of energy that can be stored into the bulk caps – so the larger the caps the more energy that can be stored – reducing the increase in rail voltage (with non sinking PSU).

As the PSU rails would be much lower for a ribbon design, the 12V rating of these cheap Car audio “Bake Bean Cans” Mega Farad Caps. would allow them to be used in this application – however by choice I would still go for a full bridge design….

John
 

bcarso

diyAudio Member RIP
2005-05-03 4:09 am
Canoga Park, California
Bruno said: "Obviously the stored energy is no different from what you need for a 25W amp at 8 ohms, only that electrolytic's energy density becomes worse at lower voltages (the oxide becomes a smaller fraction of the volume)."

I wonder if the "ultracapacitor" parts are approaching sufficiently low ESR now to be adequate for this application, if assisted by low-ESR smaller C's in parallel? Their typical voltage limitation for single caps of 2.5V would be less of a problem here. They are certainly appealingly compact.
 
bcarso said:
I wonder if the "ultracapacitor" parts are approaching sufficiently low ESR now to be adequate for this application, if assisted by low-ESR smaller C's in parallel? Their typical voltage limitation for single caps of 2.5V would be less of a problem here. They are certainly appealingly compact.

Let me run though a few numbers while I think about what to answer.

A typical 25W amplifier will have something like 1.5J worth of energy storage per rail, sufficient for rectification and freedom from significant pumping. Point is, for an amp running off +/-2.5V (delivering 25A) this means 480000uF per side (albeit at only 2.5V).

For standard electrolytics, 480000uF/2.5V would be quite a bit bigger than 4800uF/25V, for technological reasons.

The smallest Ultracap (www.epcos.com) is 4F, 2.3V. Sounds like enough. Unfortunately the ESR is specified as 380mOhm. We're not going to draw 25A from these.

Of course, you'd need to use a synchronous buck converter anyway to make +/-2.5V at this kind of current. The caps can go on the input side of this converter so you get smaller ones.
You'd have to use two cpu style buck regulators to do that, one with the control chip somewhat unusually arranged for the negative voltage.

The alternative proposal is to use a full bridge. OK, that's twice the ON losses but you are allowed to run it at up to at least 5V (the FETs are 7V anyway). The extra losses are gladly incurred because otherwise you'd lose extra power in the +/-2.5V supply. You can directly run this amp off a computer SMPS and no large caps are needed.

The latter argument is so persuasive that even if you don't expect pumping (mid or tweeter ribbon), a full bridge is still the obvious choice.
 
High current

Hey all, I have been reading this topic and am interested in how efficient this amplifier will be. Although very low Rds on MOSFETs are avaliable, their Rds on does increase with the current. So, while having a quick look at things, it may be hard to get the amplifier to be any more than 25% efficient if the Rds on at 25A will be much more than 0.1Ohms.
If this is not a problem, then sweet.



http://carlselectronics.webhop.net