LME49810 - a new cousin for LM4702

[glennb]

Originally posted by Russ White
The 10uf cap sees a resistance of around 7K to GND. A 10uf cap should be sufficient for that.

Yep, agree. There is no reason why RS=6.81K can't be increased to 22K. C7=10u can then be reduced to 0.47u - 1u, which is a more reasonable size in a film cap. That will give f-3 of 14 Hz or 7 Hz.

But the Power GND and Signal GND does seem to be disconnected.

They do need to be connected. The AGND should be a 'star' network and the PE (PGND) should be 'star' network, and the two star points should be linked by one thin conductor. This prevents high currents in the PE ground circulating into the AGND connections, and degrading THD.

RIN=243 Ohm is not actually required for the amp to work. However, it can be used to limit the HF bandwidth to prevent RF getting into the amp, provided a small cap is added across RS, say about 10nF. f-3 with then be 65 KHz. It can be scaled, ie. 2430 Ohm and 1nF to get it in the range of a polystyrene or mica cap.

[BWRX]

Quite right Russ. I looked at his schematic and thought I saw 10uF in both places, but the schematic clearly shows 220uF in series with the feedback resistor. My mistake.

As glenn also mentions, I also think it would be a good idea to increase the resistance values so smaller value caps may be used.

[gmikol]

Thanks for all the replies everyone. My initial goal was to lay things out according to the datasheet, and to provide the option to short the input and feedback caps. I think there's a typo in the National datasheet that describes the input pole for Rin/Cin. I agree on increasing the resistance.

Re: The grounding. I know that PE and AGND are separate on the board. I was planning on having the star on the PSU board. Any reason not to?

Aside from the grounding issues, it sounds like the layout is generally sound. Power amp layout is a lot different than submicron SRAM layout.

--Greg

[Triodas]

I don't think that increasing the resistance values would be a good idea. I have got experince with LM3875: lower resistance values - better sound. But I don't know how to explain this phenomena.
Pls excuse my english.

[gmikol]

Originally posted by Gcollier
I'm going to be pushing 60 volts on each rail so I will have to make some adjustments, mostly to the output stage

Yeah...some big adjustments. I've been trying to learn about SOA and thermal calculations, and by the spreadsheet I put together, it looks like you'd need 4 pairs and a pretty beefy heatsink (< 1 C/W). Of course, there's no guarantee that my spreadsheet is right.

Good luck...

--Greg

[BWRX]

We mean use slightly higher resistance values, not outrageously higher values. For example, increase the 243ohm resistor to 2k and the 6.81k to 56k. Now a 10uF cap will give you a -3dB frequency of 8Hz in the feedback loop, and a 1uF cap will give you a -3dB frequency of 3Hz at the input. Now you can use very good quality caps in the signal and feedback paths. The tradeoff is of course slightly higher noise because of the higher value resistors.

The next step up is to get rid of the caps and DC couple the signal.

[Avenida]

there will be a demo some time next month in London to show the design of using these high voltage driver

[AndrewT]

Originally posted by BWRX
........ Now a 10uF cap will give you a -3dB frequency of 8Hz in the feedback loop, and a 1uF cap will give you a -3dB frequency of 3Hz at the input. .........

I recommend that the filter frequencies be reversed so that the input filter provides the effective bandwidth limit for the amplifier.
I think that using the NFB roll-off as the bandwidth limiter is not a good way for good sound.

I would go even further.
Keep the F-3db=3Hz at the input and change the NFB F-3db to about 2Hz (needs an RC about 4times higher than the BWRX proposal).

[glennb]

Originally posted by gmikol
You hardly need the driver for this configuration. Since 50mA * 75 hFE(Min) = 3.75A, for MJL3281, good for 30V rails into 8 Ohms. This is about as hard as you'd want to push a single device anyway. But I laid it out as a demonstrator for doing larger output stages. If and when I get around to building any of this, I may try a driver-less single device output stage just for the heck of it.
--Greg

Greg as you probably know, hFEs on the spec sheet are meant to be a guide only. A real-life batch of transistors from a particular manufacturer tested at a load current of a few Amps on a heatsink might show hFE anywhere from 50 - 120.

Also, your example was good for an 8 Ohm resistive load, but for a speaker system load you should allow for impedance dips and V-I phase angle changes around the speaker resonances. The output transistors' peak current and average power dissipation should be calculated as if a 3-4 Ohm load existed, for an 8 Ohm nominal speaker system. There is a fair bit of leeway here because the low impedance loads only exist at some frequencies and the average long term power in music at these frequencies may not be significant.

Even with a 50mA drive capability from the '48910, its still not much good to directly drive a typical output transistor to achieve over about 25 Wrms output into 8 Ohms.

MOSFETs and Darlingtons are a different situation...

[monkey29]

The LME49810 as the core, one to three pairs of Toshiba output stages.

heatsink size: 230mm x 110 mm
PCB size: 210mm x 60mm

Further experiments and then published the results .
Display samples.