Comparing LME49810, 49830 and 49811

[bobodioulasso]

O.K., thank you.

This beeing put, i allow myself to insist:

What about my choice of devices? (they are under my hand)

- Five to six pairs of 2SC3264/2SA1295 Sankens as output devices

- 2sc5200/2sa1943 as drivers,

- 2sc5171 / 2sa1930 as pre drivers.

[panson_hk]

Originally posted by klewis



Panson,

Yes, I am refering to THD. I set the bias at 2.8v. THD gets better if it's set higher - say 4.5v, but I don't know if the 49811 will got that high.

Ken

Thank you Ken. What was the output level for THD simulation? If it was not high, the transistors were probably in Class A mode with higher bias levels. See Andrew's message above.

[panson_hk]

Originally posted by AndrewT
Hi,
whether the transistor is forced to pass a non zero current or allowed to switch off, does not define or otherwise if it's operating in ClassA.

With drivers conducting non-zero current all the time, we avoid driver stage switching distortion. Switching distortion of driver (medimum power) might be insignificant.

Any comment?

[AndrewT]

Hi,
driver switching may give rise to increased distortion.
Modified versions of ClassAB in the output stage seem to give some distortion benefits in the performance.

I cannot see why applying similar modified ClassAB for the driver should not also offer some benefit.
What is quite clear is that the handover between pairs of halfwaves is currently not as clean as pure ClassA. It seems to be something to do with the sidebands.

But any stage that follows the sinewave or music signal for only one half of the wave form and becomes inoperative during the other half of the waveform will always be ClassAB and will always be in need of improvement.
Some methods will give rise to these much needed improvements, just don't believe the designer when he tries to convince us he has achieved ClassA when clearly this is not the case.
Some day someone might come up with a modified ClassAB scheme that betters a true ClassA stage, let's wait till that comes.

[bobodioulasso]

Aren't the drivers working in class A in the following scheme?

http://www.audiolabor.de/index.php?o...d=34&Itemid=51

Schaltplan = schematic

[klewis]

Originally posted by panson_hk


Thank you Ken. What was the output level for THD simulation? If it was not high, the transistors were probably in Class A mode with higher bias levels. See Andrew's message above.

Panson,

vin was 34v

Schematic below

[klewis]

Here's the Thd at 1khz for the 2sd669

[klewis]

Panson,

So, I don't know what happened here... my prior sim showed the thd for the 15030 at >2% now it's better than 0.2% - I don't know why..., sorry.

Thd for the 15030/31 combo subsituted for the 2sd669 in the prior schematic.

Regards,

Ken

[panson_hk]

Originally posted by klewis
Panson,

So, I don't know what happened here... my prior sim showed the thd for the 15030 at >2% now it's better than 0.2% - I don't know why..., sorry.

Thd for the 15030/31 combo subsituted for the 2sd669 in the prior schematic.

Regards,

Ken

Hi Ken,

Thank you for your data. Do you have .option plotwinsize=0 in your simulations?

[RocketScientist]

Originally posted by jackinnj

I've decoupled the LM4702 from the main supply rail with a small resistor -- but my amps generally use lateral MOSFETs which turn on at a low voltage to begin with (so the voltage is a few tenths LOWER).

I was thinking of exactly what was contained in your post -- and regulating the LM4702 a la Borbely's MOSFET designs. Just enough higher ...

I built one of Borbely's designs many years ago and, to be honest, I think regulating the driver stage causes more problems than it solves. The feedback loops in the regulators introduce essentially another pole into the stability equation. Many authors and designers, such as Self, Slone, etc. have since mostly shunned driver regulation for stability reasons, and because it's very easy for the common mode rejection (CMRR) of a well designed input stage and VAS to reject the low frequency power supply variations created by the output stage.

Back when Borbely designed that amp, there were no high power driver IC's available (unless you count the LM391 but it was only good for 70 watts). It was more difficult to design high CMRR input and driver stages without the benefit of matched devices on a single piece of silicon and the other benefits of a monolithic design. Today, the LM4702/LME498xx make it easy to have extremely high CMRR.

My Borbely amp, with the regulators, would try to oscillate when clipped--especially into a real speaker load. With some creative compensation I got it to where it was stable on the bench with a resistive load only to have it later fry my tweeters in the listening room. A scope in the listening room showed huge bursts of HF into the speakers at clipping. The relatively sharp spikes on the rails at clipping apparently caused a high frequency transient response from the regulators, which in turn, apparently set up a positive feedback mechanism to the driver stage. Bad news! I removed the regulators and the amp was extremely stable with my tweeters having a long and healthy life from then on.

And even with lateral MOSFETs you can gain output power by having the driver stage run on a higher voltage. Power, of course, is the square of voltage. So even gaining a few volts on each rail makes a significant difference. With the much higher RDSon creating more Vds drop at full power, lateral MOSFETs need all the help they can get--especially into low impedance or reactive loads.

If the output devices can get closer to the rails, for the same power output, you can use lower rail voltages. That decreases the size of the heatsinks and also lets the amp run cooler at idle (especially with high bias lateral mosfets).

And, as an aside, for those who like to leave their amp on... In most areas, every single watt of 24/7 electrical consumption is at least $1 per year. A lateral MOSFET amp with 2 devices per side typically needs around 300ma bias. With 60 volt rails a 7 channel home theater amp would dissipate a toasty 252 watts at idle. With a higher voltage driver stage, and 55 volt output rails, it would produce about the same power and idle at 231 watts. Adding in power supply losses you might save $25/year in electricity.