diyAudio (
-   Headphone Systems (
-   -   LM3875 as a headphone amp (another try) (

rjm 19th August 2005 04:58 PM

LM3875 as a headphone amp (another try)
2 Attachment(s)
Although there is no evidence to support it, I want to believe that the LM3875 makes a decent headphone amplifier. It almost does as a standard gainclone, but the unregulated supply is just a wee bit noisy.

The case for: plenty of current reserves, designed to drive even lower loads than headphones with minimal distortion. Stable, easy to work with.

So, I figure, when not treat the LM3875 as you would any other op-amp in a headphone circuit: add in regulated supplies, and run it off 12 or 15V rails. (datasheet minimum is 10V)

This is a standard non-invering circuit, with no input coupling cap (might add it?). Simple as can be. PCB is 2"x1.4"

No clue if this is a good idea or a bad one, but I thought I'd share it.


Peter Daniel 19th August 2005 05:07 PM

I tried a regular amp's version as a headphone amp and it works pretty well, better than Benchmark used in the same application. It is especially good with AKG K1000, as you can read here:

Vil 19th August 2005 05:50 PM

but why use LM3875 ? LM1875 should be much more than enough for that .

rjm 20th August 2005 03:34 AM

The K-1000s aren't really a good example of a "typical" headphone... not too surprising a LM3875 based amp works well with them. I wonder if the LM3875 would still be a good choice when the load impedance and sensitivity start climbing into Sennheiser HD600 territory. My concern is how much noisier the chip is compared to typical audio op-amps. Let's try to figure that one out:

From the datasheet, you can guestimate that the THD+N at 10mW into 8ohms (26dB gain test circuit, 80kHz bandwidth) is 0.1%. This, one can safely assume, is almost entirely noise... the N part of THD+N.

The output voltage at 10mW/8ohms is: 283 mV.
The input voltage at stated gain 20x is: 14 mV.
0.1% of this is 14 uV.
Divided by sqrt of 80kHz, the e_n is 50 nV/sqrtHz

The input referred voltage noise (e_n) of the LM3875 is 50nV/sqrtHz about 10 times (20dB) higher than a typical audio op-amp. It comes down to whether a 20dB increase in noise is a price worth paying for the (presumably) lower distortion.

The output voltage for headphones is about 10 or 20 dB lower than loudpeakers. So the amplifier noise will be that much more more "apparent".


evo 20th August 2005 11:15 AM


I wonder if the LM3875 would still be a good choice when the load impedance and sensitivity start climbing into Sennheiser HD600 territory
How would i find this out? Build the circuit i suppose- could anyone suggest cap/resistor values.
BTW thanks Richard for starting this.

tschrama 20th August 2005 01:03 PM

I have thought about using the LM1875 ,LM3875, LM3886 etc to use as a headphone amp too. I think I actually tried it once but I cann't teally remember. My first motivation was that the those power-chip-amp would probbably stay well the in Class A region when used as a headphone amp.

Concerniong the noise: I really don't think that is an issue. If the dynamic range is good enough for speaker, it probbably will be good enough for headphones.

50nV/sqrt(Hz) gives 109 dB S/N when the max input signal is 2Volt. Offcourse then you need to build a normal highes voltage version with a attenuator resistor network at the output to acomplish that.


To me the voltage regulator really aren't nessecery; my speakers are completely hum free. What's good enough for speakers, will be good enough for headphones provided a attenuator netwerk is used out the output.. say.. 680R-out-33R--GND

Peter Daniel 20th August 2005 02:27 PM

I also tried AMP-1 with both Senns HD650 and Grado RS-1 and it worked fine. Hard to say how good, as I don't have any real reference headphone amp (but will have soon for comparison).

rjm 21st August 2005 01:43 AM

If you are actually building a n.i. Gainclone-type circuit from scratch just for headphones, I suggest lowering the gain from 30x down to 10x, by reducing the feedback resistor from 22k to 6.8k. Everything else can be the same, though an output series resistance of 120 ohms or so (see next paragraph) can be added.

An attenuator (resistor) at the output will indeed improve the signal to noise. However, I'm not too sure that a values above 120ohms are such a good idea. The load impedance of headphones is relatively flat, but still... on the other hand it could be just the ticket - what do I know? The correct value can be easily found by trial and error.

If you want to listen to headphones with a standard chip amp (Gainclone or otherwise), then the series resistance can be easily added as part of a wiring circuit connecting the output to the headphone jack. Add another pole to the mute switch, and you have a speaker/mute/phone switch. Very professional!

It may well be that the internal amplifier noise is greater than the output noise caused by the power supply, even when unregulated. If thats the case adding voltage regulators wont make any difference to the output noise.

Has anyone compared the noise of a Gainclone powered by regulated vs. unregulated supplies, or batteries vs. unregulated supplies?


rjm 21st August 2005 08:31 AM

Test results
Tried the Grado SR60 and Sennheiser HD600 with my gainclone, thorugh 100 ohm resistors. From the computer CD-ROM, I tested the line out directly into headphones, line out into the gainclone, and digital optical out through a NonOS TDA1543 DAC to the gainclone.

With the inputs left open, volume turned to max, and the gainclone case grounded, there was absolutely no audible noise with either of the headphones tested. As in "is the power on?".

The Grados cause the computer output to distort. They dont sound terrible even then, but for clean sound a separate amplifier is needed. Unfortunately, with the Gainclone, the sound was a little dull, as if rolled off in the treble. My guess is 100 ohms series resistance is too high a value for these 32 ohm cans.

The Sennheisers work quite well from the computer output directly, and to be honest putting the gainclone in the chain didn't make much difference. In either case the limitations of the source were obvious. The HD600s made clear the benefits of using the external DAC, but at the same time showed up the limitations of that component, too, as the DACs noise floor was considerably higher than the gainclones.

So I can confirm Peter Daniel's original assessment that the Gainclone works quite well both high and low impedance headphones. No need to add regulated supplies, or even change the gain. I doubt that even the output series resistance is needed.

The gainclone as a headphone amplifier seems to sound pretty much the same as the gainclone as a loudspeaker amplifier. Mine sounds a little dark, a little closed in. Always has done. Overall though it neither adds, nor subtracts, much from the source.


tschrama 22nd August 2005 02:10 PM

Mind you that I did not suggest a series resistor, but a attenuator network! This way you can create any output resistance you want. I would suggest a 600R-33R-GND network, and take the output from the top of the 33R resistor to GND. This give about 30R output resistance and -26dB attenuation. So you do not need to change the gain of the amp, and you get about the same S/N.

good luck,

All times are GMT. The time now is 12:00 PM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 18.75%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Copyright ©1999-2018 diyAudio