High output current audio driver

I thought you said before 10 Vpp. Are you counting on a same-as-load-impedance series resistor?

If so and if that is tolerable, I consider it no worse to have a scheme with a combination of current and voltage feedbacks (maybe negative voltage feedback and positive current feedback) to generate an output impedance same as that of such a series resistor - that can reduce your amplifier voltage requirement by half.

If you have a severe supply voltage requirement, consider having two amplifiers, with their inputs driven inverted from each other. Maybe or maybe-not so-good if this device has to feed an "unbalanced line", such as 50 or 52 ohm coaxial cable whose shield is a ground that your supply rails must have low AC voltage/potential with-respect-to.

- Don Klipstein ( snipped-for-privacy@misty.com)

Thanks for the reply. I am not concerned about the power in the board as a whole. I am concerned about the board space of having to use large power resistors to handle unspec'ed error conditions such as a shorted output. That is why I am asking about alternate circuits.

I will be asking my customer for more info on the need for driving 50 ohm loads. But he doesn't have a lot more info that I have given here. I expect the end load is high impedance and they are just using

50 ohm cable because it is handy. Or they are using a 50 ohm terminated load and don't need 10 Vpp. But similar equipment does provide 10 Vpp output and the end customer is using 50 ohm cable and my customer decided I should drive it as a resistive terminated load. So until I get a change of direction, that is what I need to provide.

I can make the circuit work just fine. I can't get all the parts on the small board I have to work with. I have figured out that I don't need to size the input terminations as large as the output series resistors. That will give me some relief. But I would like to simplify the design more if I can.

Actually, I was hoping someone would be able to point me to a driver chip that is a bit more complete/in a smaller package/is not a high freq part needing compensation/fits the single ended codec to differential I/O, much like the TI DRV135, but can drive 10 Vpp with

12 volt supplies. That part is almost perfect, but the input is 10kOhms, so I need an amp in front so I can add high and low pass filters.

BTW, I am looking at using an AKM AK4556 sigma-delta codec. I understand how they operate with a very high rate single bit conversion that is then filtered into a lower sample rate with a wider width word. They include filtering as part of the device so that the only external filtering required is at much higher freqs of the I/O sample rate. But is this filter required at all? It appeared that some of the codecs I looked at included some analog filtering. Does that mean an external anti-alias filter is not needed?

Thanks for the insight. Actually, I am designing more than one device with the same circuit board, but different populations are allowed. The requirement for driving the 50 ohm line unfortunately is in the same configuration that requires 10 Vpp output. It does not, however, require wide bandwidth, transmitting only a 1 kHz AM signal modulated at 100 baud. The other two modes can change values of filter components, gain settings and the termination components. So if the load has a complex impedance the positive feedback can be removed and the output can be driven with the initial 50 ohm or higher resistance.

Is that any more clear? I know this sounds like an unusual application, but mine is to do or die... however I will be asking questions about this later this week. Right now I want to be able to say I have covered the bases and considered every technical approach.

Yes, for the 50 ohm load configuration, that is exactly what I am considering.

Yes, that is what I was thinking about. I have read a couple of articles about positive feedback to multiply the output resistor. This reduces the power spent in the output resistor, although not in the total circuit unless you can lower the supply voltages. But the total power is not my immediate concern. The dissipation in the resistors is since they get very large with power. The amps can be effectively heatsinked to help remove the heat, so they are better equipped to dissipate the Watt or so required.

Yes, I have been thinking about this and unless you are referring to a differential arrangement, I don't understand what you mean about the two amplifiers. I don't think I can use a differential output for the

10 Vpp, 50 ohm configuration.

"rickman" ...

High resolution? I hope it is not also high accuracy. A long cable with a 50 Ohm load will have a gain certainly less than one. And temperature dependent. How are you (is your) customer dealing with that? You are not doing remote sensing, are you?

Arie.

That is a very good question. Of the parts I have looked at, the answer is no... mainly because they don't operate at the required voltages, or they use too many components.

Thanks for the suggestion though.

Yes, I understand that. I am using the best heat sinking I can in a limited design such as this. The parts have a thermal pad and I am using vias to connect to one of three power planes underneath (four counting the copper pad on the opposite side). If push comes to shove, I can layout the board to put the active parts on the top and use actual heat sinks as will fit in the height provided. But it will really be up to the customer to provide adequate cooling air. Besides, the heat generated on board in normal operation is not so bad at about 0.6 watts total in normal operation (per channel). I think that is workable and higher heat with a shorted output should be survivable; I calculate a max junction temp rise of 20C over ambient which is entirely acceptable. Of course, that depends on the heat getting off of the board at some point which is the customer's design problem.

OK...

I have not thought this through... but how about a LM317 or similar configured as a 100mA constant current source, supplying an unprotected rail-to-rail output opamp. E.g. AD8531. Opamp is single supply, feed the LM317 from 12V. If there is a short circuit, the current limits and the opamp supply is reduced. The LM317 can heat up and eventually will go into thermal shutdown depending on heat sinking.

The point is to do the current limit externally, which means you don't have use a higher supply voltage than needed to drive the load - although perhaps there is a better way to do it.

--

John Devereux

(Sorry, not AD8531, you need a 12V capable part. I expect you already found some.)

--

John Devereux

On a sunny day (Mon, 21 Jan 2008 08:05:05 -0800 (PST)) it happened rickman wrote in :

Most, if not all, audio power amps use foldback current limiting, in the output stage. This limits dissipation, and current in case of short or overload. In case of the push pull complementary emittor follower there are 2 sense resistors in the emittor, the voltage is sensed by 2 transistors that clamp down the base. There is a resistor bridge to reference ground, in such a way that for lower output volatges the current limit is lower. So in case of a shorted output the output voltage is zero, and the current very limited. I have done this discrete, but I am almost sure there are audio power amps of lower wattage with this feature that you could use.

Interesting idea, but I don't think I would have the board space. The rails are + and - 12 volts, so I would have to have a current limiter for each voltage on each amp which would be eight limiters!

You have obviously never looked at real lines. Telephones use twisted pairs, which turn out to be very close to 100 ohm lines, and quite flat for frequencies in the megacycle range. Down at audio (say 2000 hz) they are quite close to 600 ohm lines. The shape of the attenuation vs frequency curve will vary seriously with the line quality. Also delay.

You can avoid those problems by doing such games as modulating 10 Mhz signals with the audio, and demodulating at the other end. Such games also allow you to pack many signals onto the one line.

--
 [mail]: Chuck F (cbfalconer at maineline dot net) 
 [page]: 
            Try the download section.

Late at night, by candle light, snipped-for-privacy@manx.misty.com (Don Klipstein) penned this immortal opus:

A TDA2002 or whatever is the modern replacement should do quite nicely. As other have stated, at audio the line impedance is of little consequence.

- YD.

--
Remove HAT if replying by mail.

concern.

resistors

base.

Unfortunately, I have not found an audio power amp that will fit all of my needs. Mostly if they are capable of driving the current and the voltage I need on the output, they are too large to be useful on my board.

I may have found a pretty good match to my requirements. Fairchild FHP3230 will drive 100 mA (typ) and to within 0.35 volts of the rails with a 150 ohm load. Looking at the graph for output swing for different loads (at Vcc = 3V) they get to within 0.2 V of the rails for a 75 ohm load. So I expect that (typ) I can drive a 50 ohm load to 10 Vpp using a 12 volt supply. It is a very small part and the short circuit output current is just 120 mA (typ), very close to the linear output current of 100 mA (typ). I would feel better if they had max and mins on some of these values and even better if they had a spice model. Using the lower Vcc, the total power dissipation will be much lower. I will need to use a small output resistor to improve stability of the amp, but that shouldn't be a real issue.

I am getting tired of looking and I am running out of vendors. I have searched TI, ADI, National, Fairchild, Maxim, Microchip and need to try LT again. I looked at Philips and ST, but their web sites suck so badly that I gave up trying to figure out if they had anything I could use.

Any other vendors I should check?

duals or a

will deliver

Remember that board space issue? It's a killer for sure! That is the whole reason that I can't use the current approach. The amps are small, but not microscopic and the passives take up more room that the amps! If I had to double or quadruple the number of amps I would need to double the size of the board.

concern.

output

resistors

base.

very

Tightening the current limit seems to be the best approach, so the Fairchild device looks a good solution.

Because this is only fault protection, you could also add your own current limiting solution (but that will add area)

Also if your Silicon has thermal shutdoiwn, do not be afraid of using the resistors to deliberately heat the chip (as so hasten shutdown, and lower the resistor sizes). It sounds like a protection tick-box, so a chip-recovery will likely be OK.

-jg

and now to deflate your balloon: I cannot see thermal shutdown in that fairchild device, so you'll need to find a device that does offer that.

Still, the idea was good :)

-jg

ple of duals or a

his will deliver

How much height do you have? Have you thought about using horizontal piggyback boards to give you effectively twice the surface area? Either one big secondary board, or multiple boards (e.g. one per channel). Short height vertical riser boards can often work a treat in this sort of scenario too. There is more than one way to skin a cat when it comes to PCB envelope design.

Dave.

"High resolution" is the customer's term for better than voice grade. I am using a standard 16+ bit codec which can sample at much higher rates than what is needed for audio. That is what he essentially asked for. I have been given no details about how the voice grade mode or the "high resolution" mode will be used. The voice grade mode I can figure out easily as I have the info on their current design and voice is pretty standard. It is the 50 ohm mode that I am actually building for him. But he would like this new card to replace the old one and pick up the "high resolution" mode as well... if possible. The issues of driving cables and varying loads aside, this card will do that. The rest is up to him since I was not given info to work with.

My only real concern with the Fairchild part meeting the spec for all of these modes (which can have different build options, btw, so resistors and such can change) is driving a capacitive load. They say to use an output resistor to isolate the capacitance. But if I use a very large one, it will reduce the voltage level. That is why I would like to simulate it, there are just too many variables to try to get it right the first time without simulation.