Comparing LME49810, 49830 and 49811

[AndrewT]

a1294, a1295, MJL1302, NJW1302, MJL4302 all do over 1A @ 90Vce and all are faster than the 1943.
The NJW0302 850mA @ 90Vce may be your best bet.

[RocketScientist]

Originally posted by panson_hk

I think you are actually referring to PSRR instead of CMRR.

I wonder that feedback loop introducing poles in amp's loop equation is a proper statement. The amp will act as a reactive load to the regulator. The regulator loop characteristic will be affected by the load condition. It should be designed to operate under all expected conditions.

Have you tried another regulator to supply the Borbely amp's driver stage?

I agree that designing an amp with sufficient PSRR should be attemped before considering a regulated supply (for input and VAS).

Yes PSRR is a better term. I was referring to common mode signals on the power supply rails so, in this case, they are the same thing.

I said the regulator "in effect" introduces another pole into the amplifier. It obviously doesn't directly introduce a real pole, but the net effect, in my experience, is similar.

The output impedance of the regulator rises at higher frequencies (just like any amplified feedback loop). And the open loop bandwidth of the Borbely amp is quite high. So those two things can interact to create high frequency stability problems well above the audio range. Amps with Mosfet outputs and high open loop gain and bandwidth can be very sensitive to even small things creating serious stability problems. National in their own App Notes have examples of this.

I agree, in theory, it should be possible to design the regulator and amp such that it should work OK. Obviously Borbely thought he did just that. But with the stability of Mosfet output amplifiers often being as much "magic" (some would say "art") as science, I think the regulators just complicate things and are not likely to offer any better performance when used with the LM4702 or LME498xx.

I long ago retired the Borbely amp (although I do still have it). It uses the original Hitachi metal TO3 output devices as used in the Hafler DH-200/220/500 amps. It's possible the particular grounding scheme, etc. of my amp was making it more sensitive to the regulators. But it was clear it was unstable with the regulators and very stable without them.

[bobodioulasso]

Hi Andrew,

Datasheets say:

fT = 30 mHz for all of these devices (same as A1943)

except

fT = 35mHz for A1295 and mjl4281.

But you say

"NJW0302 850mA @ 90Vce may be your best bet"

For what reason?

Is there a drawback using the Sankens as drivers. (i have them)

And as pre-drivers?

[AndrewT]

Originally posted by bobodioulasso
Datasheets say:

fT = 30 mHz for all of these devices (same as A1943)

except

fT = 35mHz for A1295 and mjl4281.

and look at the typical graphs.
If the manufacturers are telling the truth, then they all have a slightly different fT at different currents. The highest fT and lowest capacitance is from that 150W ONsemi

[AndrewT]

Originally posted by bobodioulasso
Is there a drawback using the Sankens as drivers. (i have them)

And as pre-drivers?

the output gain will fall to somewhere in the range 10 to 20 at very high currents.
If the driver is feeding a 2pair then it will see currents of the order of 10% to 20% of the peak transient output current. The driver must be able to pass that in the short term when high Vce is present.

The driver gain will also have fallen at these very high output currents.
Maybe in the range 20 to 30. That in turn determines the peak transient current that the pre-driver must supply. guess about 0.2% to 1% of the peak output transient current.

These requirements demand different devices.

A small output device (100W to 150W) could double up as a driver, but certainly not as a pre-driver.

[bobodioulasso]

"And as pre-drivers" was miswriten.

I was meanning: What do you suggest as pre-drivers?

[AndrewT]

it depends on the design. all the devices are interlinked.
You need to go through the amp output stage going from the worst case speaker current back through each stage checking the graphs of gain vs current and finding the base current, Then use that to determine the previous stage all the way back to the VAS.

Once you have candidates, enter the data into an SOAR spreadsheet with operating temperature and find what DC currents are permissible, what long term AC currents and short term AC currents for each device.

[bobodioulasso]

2sc 5174 as pre ?

http://www.datasheetcatalog.com/data.../2SC5174.shtml

2sc3263 as drivers ?

http://pdf1.alldatasheet.com/datashe...N/2SC3263.html

these driving four pairs of 3264 in the RB1090 Rotel (96v rails)

http://mesnotices.fr/manuel-notice-m...L/RB-1090-3-_E

[Tony]

Originally posted by AndrewT
that's where you and others fail to understand what the definition of ClassA mean by "for the full 360degrees of the cycle".

Andrew, can you post links that can further elucidate your point? thanks..

[RocketScientist]

Bobodioulasso, AndrewT brings up some very valid points about SOA, drive currents, etc. In a very high power amplifier such as you're proposing things get more difficult on many levels.

As your output transistors can only handle 75 volt rails, have you considered MOSFETs and switching to the LME49830? I suspect your transformers (for 90 volt rails) are worth more than your Sanken output transistors? Or, put another way, it might be cheaper to buy new output transistors and the LME49830's than new 1.2 KVA transformers?

MOSFETs tend to have a much more usable SOA at high voltages, and driving them is just a matter of having enough drive to get the gate capacitance related slew rate to an acceptable level (and I'm not sure I personally would even worry about a full power bandwidth to or beyond 20 khz with the power levels you're talking about).

Although I would still be cautious about 90 volt rails with the LME498xx parts. A 15% AC voltage surge (which are not that uncommon) might wipe out the IC and take output devices and possibly your speakers with it. It might be possible to decouple the rails to the LME498xx with resistors and use some sort of over voltage protection on those rails.

And the most simple definition of Class A is simply NONE of the transistors ever turn off in normal operation. In a typical push-pull Class A stage if you try to draw enough current it will fall back to Class B. But if you can design for the maximum real world current, it will always stay in Class A in normal operation. For driver stages feeding bipolar transistors, calculating the actual real-world current required can be challenging as there are many variables to consider. MOSFETs make things much more simple