Composite LM1875 MiniAmp and HeadphoneAmp

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi Folks :wave2:

Its been a long time I posted here, several years ago (except my chipamp photos few minutes ago :) ), but read more...

Now I'm thinking about a compact headphone amp with capability to driving speakers as well. It would use LM1875 from about +-10V supply voltage. I think its enough for most of the headphones, while chip can deliver 10W for 4ohm and 5W for 8ohm speakers (see on Youtube).
Low supply voltage has more benefits: low heat dissipasion (small heatsinks) and makes it possible to feed the unit from a 24V adapter (~1A is enough considering max. 1/5..1/10 Crest factor of the musics).
So avoid of handling main voltage (230V here) is solved, voltage stabilization is solved and current limit is solved by buying a cheap adapter! ;)

Some technical background
I solved the known clipping recovery problem of LM1875/3886 composite arrangements by appliying only 1/2 divider after the driver OPA (inside the outer loop) and feeding the OPA from lower supply voltage (below +-9V in the case of +-30V main supply). It means only 6dB loss from the total open loop gain (with higher OPA supply voltage at least 1/6 divider is needed for LM3886 to avoid high frequency oscillations at clipping which endangers tweeters).
I also use Miller feedback capacitor for the driver OPA creating a pole around 30kHz to make it stable (while gain of the power OPA is around 16x). 30kHz pole not means 30kHz bandwith, but ~100kHz (-3dB) according to simulation.
I know, there was/are extensive research here as well about composite LM1875/3886, still, I didn't find nice clipping measurements with 1/2 divider so forgive me not refferring to any post! :eek:

Passive VGND induced supply imbalance
I would use a passive virtual ground (VGND) to keep things simple with 220R/5W resistors and 4700uF caps. Why 5W? In case of max. supply rail imbalane ~20V will drop on one resistor: 20V^2/220R=1,8W. Multiplying it by >=2 for keep temperatures not so high... Why 220R? Lower is better because causing lower rail imbalance, but lower resistor needs applying higher wattage. Why 4700uF? @30Hz, its impedance is 1ohm, 1/4th of 4ohm (minimum speaker impadance). I think its enough (2dB loss).
If I use AC coupling at the input I can keep input DC offset below approx 3,5mV (worst case OPA1642's specs). With 10x gain, 4ohm speakers and 220R in the VGND it means 3,5mv*10*220/(4||4)=3,5V supply imbalance.
Considering dropout of supply filters of total 3V and droput of LM1875 of 2,5V/rail, it means (24V-3V)/2-2,5V-3,5V=4,5V for the lower supply rail. It means 2,5Wpeak at 4 ohms.
Using OPA2140 (DC offset is lower than 0,1mV) max. 0,2V supply imbalance @4ohm speakers can be achieved.

Thermal considerations
Max. dissipation of LM1875 from +-10,5V is around 8W or so @4ohm speakers (extrapolating the curves in the datasheet). Using a 11K/W small heatsink and considering 3K/W from junction to heatsink of TO220 (referring description in the datasheet) it means 8W*(11+3)K/W=112K (112°C) above Tambient in the housing. Max. alloved junction temp. is 145°C for LM1875, means max. 33°C in the housing. Considering Crest factor of the music it will be good even with higher temperaures inside (because lower dissipation).

Sound quality considerations
LM1875's idle current is 70mA, means its a class-A amp for headphone applications, and a very good class-AB with speakers. (For lower idle consumption, buy LM675!)
In a composite design, class-A bias can be easily applied for the distortion cancelling OPA using a single resistor.
For exceptional sound quality, extensive supply filtering have to be used considering high frequency noise (LC-filter), with low output impedance (Kmultiplier) and a separate own rail filter one for the driving OPA.
For avoiding DC offset, I would use coupling caps (NP0, nothing other!). I remember to my CKKIII, in which I can hear the difference between different DC servo OPAMPs => avoid DC servo in my understanding.

Possible extras
I'm thinking about using a relay for switch off speakers when listening headphones.
Also crossfeed can be used for headphones.
And - its not so extra - gain switch 1x-10x is mandatory for covering a wild range of usable headphones. It have to be applied as a resistor divider at the input, because outer loop gaing cannot be lower than 8 (gain of 16 of LM1875 /2 considering the divider) to ensure gain>=1 for the OPA.

So what do you think?
Last edited:
Also I believe you can make a nice headphone amplifier with an LM1875 and an "OPA". I only have a few comments:
You write that the LM1875 has a quiescent current of 70mA. True, but at a higher supply voltage.
You like to use +/-10V supply and suggest a 24V adapter with synthesized mid-point rather leaving +/-12V. Fine, avoid going too close to the lower supply limit of the LM1875 for performance reasons.
I cannot say how much hum the synthesized mid-point may cause. Synthesized midpoints work with OP-AMP level currents but you take the design into LM1875 level currents.
Good luck with the project.
Last edited:
You're right: idle current is only ~50mA at +-10V. Good, it means a total 4,5W idle consumption - will be warm a bit.

If I can keep DC imbalance low enough I belive VGND will not be a problem. I used it for several HP-amp designs, down to 16ohm loads. As its the common ground for inputs and output, it only causes common mode signal/noise. For the LM1875 at least: for the OPA, the Zener supply will be damping this actively.
I will post if I have results..
I made two designs (schematics): an inverting and a noninverting one.Noninverting produces 6dB less noise: 9uV up to 10kHz (and 3uV up to 1kHz). I think its OK for sensitive IEMs too.

Other benefit is that the input impedancies are lower by an order of magnitude; although one of these is changing with gain swich and volume pot position.

As in the case of using OPA1642 or OPA2140 neither the "noninverting distortion" nor DC issues play a role, I decided to build the noninverting one.
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.