Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[John Roberts {JR}]

I feel a little guilty asking here instead of doing the legwork to research this myself but perhaps the expertise here can save me some time and effort.

I have a battery powered application driving two speakers where I would like to use class D for the obvious benefit in battery life. The audio quality does not need to be Hifi. I am driving a pair of loudspeakers with sinewaves to excite sympathetic vibrations in a drumhead.

The linearity (purity) needs to be good enough that I don't excite spurious resonances from too much THD but dynamic range/linearity  can be < 8 bit...

My simple question is can I trick up a simple PS regulator chip to make a crude class D sinewave source capable of driving say 4 ohms? I would probably use two in a bridge configuration to KISS for DC coupling and PS interactions.

I don't need very high frequency sinewaves, say 1-2 kHz max, so I'm thinking a cheap PS chip might be adequate?

Is this worth pursuing further or a dead end? I am looking for a cheap alternative to a better than I need music quality class D chip set.  

I suspect most regulators do not have voltage comparators capable of rail to rail detection, so I need to inject the AC sinwave into the - feedback port and use it inverting.  I don't expect these to swing outputs rail to rail, but if I can get good efficiency I can live with that.

Any other obvious gotcha's ?

Thanks in advance

JR

[bruno putzeys]

Class D is four quadrant (all combos of current and voltage sign can occur with reactive loads), regulator IC's are usually one-quadrant only. Hacking a PS IC for class D would make your life harder instead of easier.

The simplest no-think class D solution would be a full bridge with a driver like the good ol' HIP408x series with a comparator in front. The HIP4080A has the comparator built-in but it has some funny startup artefacts unless they fixed that by now (last time I used it was 1998). Around it you can then apply some positive f/b from the output stage for hysteresis combined with negative integrating feedback (ie. make a feedback network with one LHP pole and one RHP zero). This circuit is called a hysteresis controlled oscillator and it is the simplest self-oscillating circuit I can think of. THD specs are actually quite acceptable (<0.1% THD, >100dB SNR) and PSRR is good if you take care to construct the feedback network differentially with good resistor matching. No op amps are needed unless you want the pole to be at DC.
The output filter isn't controlled of course but I don't think that's a problem in your application.

[John Roberts {JR}]

Thank you... that's why I asked...

Good point, some of the chips I looked at with enough grunt to drive a speaker had low side switches too, but only 1x versue 2x devices for the high side.

Yes, the self oscillating class D is very hip, but if I bail on a PS chip I may as well generate the PWM from the micro I'm using, since I'm making very specific sinewaves, and I need to do a D/A somewhere in this process..

From a quick look at those H drivers I'm not sure they are the right choice for my low rail voltage app.. I'm using 9V now but thinking of going lower voltage in my next design.

I was hoping to avoid rolling my own boost circuits for high side drivers and managing the switch timing. Since I don't care (too much) about linearity I can just throw in some dead time and maybe cobble together some discrete FET switches.  
------
Have you ever looked at those couple watt class D amp chips (sorry don't recall the PN) ? I think they're probably made for computer speakers or such?

I'm trying to avoid reinventing the wheel, but I'm also trying to keep this cheap.

Thanks again your comments were helpful. I may be making this more complicated than it needs to be. I just need to buffer my PWM. KI

[zmix]

John Roberts  {JR} wrote on Tue, 24 February 2009 14:39

... I am driving a pair of loudspeakers with sinewaves to excite sympathetic vibrations in a drumhead. JR

John,

Fascinating, what is the project?

[John Roberts {JR}]

zmix wrote on Tue, 24 February 2009 20:35

John Roberts  {JR} wrote on Tue, 24 February 2009 14:39 ... I am driving a pair of loudspeakers with sinewaves to excite sympathetic vibrations in a drumhead. JR John, Fascinating, what is the project?

Actually I'm working on the second generation version of my electronic drum tuner...

I figured out how to precisely match the lug tension of a drum's lugs electronically. I want to get better battery life in my next version.

If I really wanted to get clever I'd replace the speakers with an electric motor and crankshaft driving a piston, since I'm just making pure tones. But I don't want to invent any more than I need to, to get the job done,

JR

[bruno putzeys]

This could be your ticket then: http://products.monolithicpower.com/Products/productLanding. do?productId=71

Works from 7.5V upwards.

[John Roberts {JR}]

 I was hoping for perhaps even lower (min) voltage, but that part looks like it could work on 9V battery from what I can see on app note descriptions. My next generation processor will be running from 3.3V, but I need more than that to wiggle the speakers adequately so 9V is not out of the question.

I signed up for a account with them... so I'll look at the data sheet when they let me...

Thanks again...

JR

PS.. how many passwords do it take to get through modern life?

[zmix]

John Roberts  {JR} wrote on Wed, 25 February 2009 00:42

Actually I'm working on the second generation version of my electronic drum tuner... I figured out how to precisely match the lug tension of a drum's lugs electronically. I want to get better battery life in my next version. If I really wanted to get clever I'd replace the speakers with an electric motor and crankshaft driving a piston, since I'm just making pure tones. But I don't want to invent any more than I need to, to get the job done, JR

Of course! I forgot that you make those...  How about an automatic drum tuner ala Gibson's 'robot guitar'?

[John Roberts {JR}]

zmix wrote on Wed, 25 February 2009 13:29

Of course! I forgot that you make those...  How about an automatic drum tuner ala Gibson's 'robot guitar'?

A friend of mine who used to make drums wanted me to make one for him. The fly in that ointment is that my tuner is subtly altering some of the acoustic relationships across the drumhead to isolate one lug from the others for measurement and tuning. Any built in automatic tuner would have to be kluge of a robot moving around above and below the drum or a similarly complex mechanism inside the drum.  

I could servo a glorified power driver from the existing up/down output from my tuning arrows, but that is of only minor utility when I already have go/no-go LEDs.

JR

[John Roberts {JR}]

Found some nice parts there, even one small class D amp, but none a slam dunk. I suspect I will just roll my own from dual or quad mosfets.

PIA that I'll need to come up with my own drive boost supply and sundry clamps. IIRC some of those switching FETs have body diodes that might cover some of that.

Thanks...

JR

[johnR]

John Roberts  {JR} wrote on Fri, 27 February 2009 16:22

IIRC some of those switching FETs have body diodes that might cover some of that.

The body diodes have a pretty slow reverse recovery. In the past I've had to put fast recovery diodes in series with bridge connected switching MOSFETs (to block the body diodes), in order to keep the smoke in. Fast recovery clamping diodes then have to be added externally.

An alternative would be to limit the drive signal slew rate during switch-on to prevent shoot through currents, if you can put up with the extra heat dissipation.

[bruno putzeys]

Things have changed dramatically on the FET front in the last ten years. Modern FETs have very good internal diodes. I've experimentally gone up to 2kW in full bridge with no external diodes and no reliability or EMI problem. So the 10W amp that JR wants to build can be built along the simplest lines possible.

If a FET does have slow recovery, adding an external diode hardly works, because that diode won't catch much of the current. There's a FET with the channel in parallel with the internal diode and through some 20nH of wiring inductance there's your external diode.

Imagine the following scenario:
*hard (non ZVS) transition with 10A reverse current.
*Schottky external diode (say Vf=0.5V at 10A)

During conduction the channel takes the current. At some point the driver turns the channel off before turning the opposing device on. The wiring inductance will initially keep the current flowing through the chip i.e. force it through the body diode (say Vf=0.8V at 10A). You have a voltage differential of about 0.4V to ramp up the current through the Schottky from 0A to 10A and correspondingly to ramp down the current through the body diode to 0. dI/dt works out as 0.4V/20nH=0.02A/ns. Transferring 10A takes a whopping 500ns. Directly parallelling a fast diode only helps when Ron is highish and then only to a minor extent. That still doesn't mean the body diode has recovered. So if you really can't afford body diode recovery you have to add another diode (a low voltage schottky) in series with the FET before adding an external catch diode.

But again, I'm only using this for high power amps (the UcD700 and up).

Slew limiting the gate drive signal is always a very good idea. In fact it does not appreciably affect efficiency (unless you take it too far), as the diode losses go down well before conduction losses start going up. Making the dead time as short as you can (without increasing idle losses) is also a good idea because it minimises the time during which the body diode can actually accumulate charge. Such effects are subtle in modern, fast devices but when you're using really large devices this makes quite a difference in full power efficiency.

[John Roberts {JR}]

Thanks again..

I am clearly outside my area of expertise talking about switchers (assuming I have one somewhere), but I have actually used the body diode once before in a switching PS I inherited. The original design had external catch diodes that were expensive, not to mention PCB real estate and placement cost. I just removed the diodes and from inspection the circuit worked fine. This was several years ago, so no hidden gotchas at least on that design.

For this discussion my output power is probably more like 1 watt?  Currently 2x 8 ohm speakers in series driven class B from 9V rail. This is questionable to even use class D if it wasn't battery powered and for my desire to use to lower rail voltage. With Class D I can drive the speakers in parallel from lower rails and get similar power with better battery utilization (I am planning to bridge the output for 2x battery swing).

My performance criteria is low enough that I could almost use a slow (bipolar) motor driver chip or such, while I am still leaning toward just  throwing a few dual mosfets at it.  A simple boost circuit for the high side drivers shouldn't be very heavy lifting.  

Thanks for everybody's comments, this is helping me refine the problem.  Maybe I should just parallel up a bunch of CMOS CD4007s  

JR

[bruno putzeys]

Have you looked at the analogue-input class D chips from TI? They've got some that'll work between 2.7V and 5.5V. That'd be good for say 3 AA cells.

[Dan Kennedy]

I've been using the SSM-2211 analog in part in an intercom
application off 5v with good success.