Question for gainclone amp owners....

[lbstyling]

Hi all,
How do you find the noise floor on your amps? That is to say hiss and or hum when music isn't playing.

I'm particularly interested if you have very high efficiency speakers 110db@1w+. How close can you get to your speakers before you detect any noise with the gain turned up and nothing playing? Please post your speaker efficiency and clone type if nessacary when answering.

Thanks chaps!

[GloBug]

My LM3875 is extremely quiet. (Peter Daniels Classic Kit).
I am not sure what my speakers are, but they are efficient. I have about 5 speakers hooked up per channel.
They are Jensen P12, P10 etc. AlNico type speakers that are used with tube amps, so they should be fairly efficient.

If I get my ear around 8" or 9" I can hear a faint noise. But the furnace has to be off, and the room completely silent to hear anything that close.

I am not running a volume control so it is "full out"

It is a very quiet amp, for my purpose I consider it silent.

[lbstyling]

That's a good start then! I'll look up the efficiency of those speakers.

*update* 97db.

[GloBug]

Thanks, now I know!

[AndrewT]

if you have 110dB/W @ 1m speaker and you think you cannot hear 20dB SPL @ 1m
then the amplifier must have a noise output = -90dB ref that 1W level.
-90dB ref 1W (2.828Vac into 8r0) is ~90uVac.
You will hear a silent chipamp amp through 110dB speaker when at 1m. At 2.5m listening distance you may think it is silent.

It is very easy to build any amplifier, including chipamp, that has 3000uVac of noise at the output.

[GloBug]

The LM3875 has around ~98dB Signal to Noise ratio at 1 watt using +/-35V rails 8Ohm resistive load. Measured at 1 kHz.
114dB StR @40w
122dB StR @100w peak.

The LM3886 has a little more noise, ~92.5 Signal to Noise ratio at 1 watt using +/- 28V rails into a 4Ohm resistive load. Measured at 1kHz. I don't know if it gets better into a 8OhmL or not.

[lbstyling]

98db snr sounds good, the problem is my drivers are actually 115db/1w and the listening distance is 2-3m, so ideally I want something with 90db+ snr at less than 0.1w! The pass f5 has been suggested, but I idealy really want a cheaper solution.

I'm wondering if adding a parallel load to the driver (resistor) would help as it would bring the level above the noise floor, but use up the unnessasery headroom on the amp for this application.
Would a regular 5w 10ohm resistor be sufficient you think?

[AndrewT]

I would guess that an amp for 115dB/W speaker must be designed very specifically for lowest noise. ICs generally cannot achieve the lowest noise targets.
The National chipamps are stable at gains >10times and perform well around gains of 24times to 30times.

This is going to leave you with an amp that is constantly turned down at the input. You will probably do better with some type of discrete buffer as an amplifier, i.e. 1times gain.

[pinkmouse]

Following on from what Andrew said, you could have a look at the OPA548/9. Unity gain stable, and about the same power handling as the LM chips. They do have slightly higher distortion, but running them in inverting mode helps and they sound very pleasant.

[lbstyling]

I'll have a look thanks.
Out of interest, how do the above amps compare with the tk2050 tripath? I only mention this as I have a sure 2x100 and find the noise floor too high for this application.