Balanced input LM3886 design and PCB layout

Swapped few opamps today to see if there is a difference in sound.

Opamp is loaded with 1k and has voltage follower or buffer topology.

OPA2134 - my favorite - sounds warmer and connects all of the sound stage together. Clear sound.

LM4562 - my second favorite - very accurate and spacious sounding - every instrument or sound is there to the point where it seems that you can stretch your hand and touch it (or point on it). At moments it felt like the amp was "singing" - such clear mids.

NE5532 - did not like the sound - it was blended and little messy, though clear when not many instruments were playing.

TL072 - similar sound to 2134
 
Last edited:
While going through the OPA2134 datasheet I noticed that it's THD is dependent on the supply voltage. 18V dual supply gives the lowest THD.

I therefore proceeded to increase opamp voltage from 14.85V to 16.95V. Doing so improved the sound even more.

Running it at 18V or close is something I'm little afraid of as it is the max rating for the power supply voltage for this chip.
 
Update on the project:

There was a mistake in the layout for the LM337 negative voltage regulator. Pins IN and OUT were swapped. Had to fix that.

Also measured THD and THD+N using modded w OPA1612 Focusrite audio interface. My guess is that I actually need to get better power supply - possibly regulated PSU using LM338?...
 

Attachments

  • THD_mes1.PNG
    THD_mes1.PNG
    252.2 KB · Views: 90
  • THD_mes2.PNG
    THD_mes2.PNG
    446.5 KB · Views: 98
  • THD_R_ch_mes1.PNG
    THD_R_ch_mes1.PNG
    274 KB · Views: 94
  • THD_sound_card_mes1.PNG
    THD_sound_card_mes1.PNG
    466.1 KB · Views: 101
As long as the gain is more then 10 , LM3886 is stable. 100k/10k gives gain of 10. 10k/1k gives gain of 10 too.

Having Lm4562 as buffers allows to preserve CMRR of difference amp. You can read more in THAT12xx datasheet, pages 4 and 5 - https://www.mouser.com/datasheet/2/407/THAT_1200-Series_Datasheet-532288.pdf

I have discussion on similar topic here - https://www.diyaudio.com/community/...-is-better-to-choose-as-a-summing-amp.386600/
I would go 10k/1k as high resistor values can cause phase shift and instability.
 
Update on the project:

I decided to go with simplified design and move 15V regulators out of the board. Attached pictures of newly assembled amp modules. Yes, I had to resolver poor LM3886 about four time now each.

Opamps in soic package were used - OPA1652 FET input.

I am also removing linear power supply and replacing it with Connex SMPS300RE.
 

Attachments

  • BDDBF8F6-7511-4405-A852-937D12DF26CB.jpeg
    BDDBF8F6-7511-4405-A852-937D12DF26CB.jpeg
    406.3 KB · Views: 117
  • CAB9A498-3892-4592-BB8F-36F29B19DA0F.jpeg
    CAB9A498-3892-4592-BB8F-36F29B19DA0F.jpeg
    444.1 KB · Views: 117
  • 6EC40EA2-9A19-4F96-852E-1C26FC47AD1C.jpeg
    6EC40EA2-9A19-4F96-852E-1C26FC47AD1C.jpeg
    484.6 KB · Views: 108
  • 4FBDAFE6-D019-4103-95EE-6EDB11D145A8.jpeg
    4FBDAFE6-D019-4103-95EE-6EDB11D145A8.jpeg
    382.7 KB · Views: 114