A New EC-Composite LM3886 Amp

PCB arrived.
 

Attachments

  • Fig204 Min 6xLM3886 PCB 1.jpg
    Fig204 Min 6xLM3886 PCB 1.jpg
    28.7 KB · Views: 54
  • Fig205 Min 6xLM3886 PCB 2.jpg
    Fig205 Min 6xLM3886 PCB 2.jpg
    19.9 KB · Views: 53
Power on. No smoke. But both speaker output terminals had +13VDC offset.
Checking all DC points found out -15V of the full differential OP amp is open. After power off and soldering, everything back to normal.

+SP has an DC offset -2.8mV, -SP -2.6mV, total DC offset at speaker terminals is -0.2mV.
 

Attachments

  • Fig209 Min 6xLM3886 PCB power on.jpg
    Fig209 Min 6xLM3886 PCB power on.jpg
    35.9 KB · Views: 45
  • Fig210 Min 6xLM3886 pos offset.jpg
    Fig210 Min 6xLM3886 pos offset.jpg
    48.3 KB · Views: 44
  • Fig211 Min 6xLM3886 neg offset.jpg
    Fig211 Min 6xLM3886 neg offset.jpg
    48 KB · Views: 37
  • Fig212 Min 6xLM3886 speaker offset.jpg
    Fig212 Min 6xLM3886 speaker offset.jpg
    44.7 KB · Views: 45
Few tests about the amp DC offsets.

1. No ground in the middle of R18 and R8,C46 connnected,Speaker +/- terminals has 15mV offset to GND,but the SP +/- has only 0.2mV offset
2. No ground in the middle of R18 and R8,no C46, Pin 2 short to GND, Speaker +/- terminals has 0.3mV offset to GND, the SP +/- has 0.2mV offset
2. ground in the middle of R18 and R8,roughly same DC offset as 1. But for whole differential operation, 2 is used.
 

Attachments

  • Fig213 few tests about DC.jpg
    Fig213 few tests about DC.jpg
    73 KB · Views: 32
Testing gears:
Power supply: 2xindustrial SMPS 24-29VDC/5A
Signal source: E70 DAC
ADC: E1DA Cosmos ADCiso
Input XLR for P<10W
Aux Input for P >10W
Load: 8 Ohm Resistor
Condition: Open on a work bench
Measurement of E70: XLR output, XLR cable open bare wire end with shield ,
E1DA Cosmos ADCiso: XLR input 10V, XLR cable with alligator clips on the output bare wire ends.
 

Attachments

  • Fig214 E70 XLR output 2Vrms.jpg
    Fig214 E70 XLR output 2Vrms.jpg
    56.7 KB · Views: 24
  • Fig215 E70 XLR output 3Vrms.jpg
    Fig215 E70 XLR output 3Vrms.jpg
    54.3 KB · Views: 24
  • Fig216 E70 XLR output 4Vrms.jpg
    Fig216 E70 XLR output 4Vrms.jpg
    55.5 KB · Views: 24
Figure compares the THD+N with a straight XLR cable from E70 output to E1DA Cosmos ADCiso XLR input 10V. The number is better.

With the comparison, I'd like to show that:
Cutting a straight XLR cable into two and connecting it back by 3 alligator clips can introduce about 10db noise. THD stays almost same.
 

Attachments

  • Fig217 E70 ADCiso straight XLR cable 4V 10V in.jpg
    Fig217 E70 ADCiso straight XLR cable 4V 10V in.jpg
    61.2 KB · Views: 19
Last edited:
The noise performance of the Amp
FFT length1M 8 averge 20-20khz with REW, 8 ohm resistor load.
Fig 218: ADCiso XLR input short noise level
Fig 219: E70 XLR output noise level (direct connect to ADCiso)
Fig 220: Power off noise level of the amp with OPA1642 as the input buffer
Fig 221: Power on noise level of the amp with OPA1642 as the input buffer (input shorted)
Fig 220: Power off noise level of the amp with OPA2891 as the input buffer
Fig 221: Power on noise level of the amp with OPA2891 as the input buffer (input shorted)

From the comparison, it can be seen:
The measuring system has background noise of 0.1uV.
E70 has a noise level of 1uV
The amps have a noise level 5X higher tha E70. A hum of 60Hz(close 10uV) and its harmony can be seen
 

Attachments

  • Fig218 ADCiso XLR input short noise level.jpg
    Fig218 ADCiso XLR input short noise level.jpg
    49 KB · Views: 15
  • Fig219 E70 ADCiso straight XLR noise level.jpg
    Fig219 E70 ADCiso straight XLR noise level.jpg
    52.1 KB · Views: 13
  • Fig220 noise 6xbtl amp power off .jpg
    Fig220 noise 6xbtl amp power off .jpg
    49.4 KB · Views: 15
  • Fig221 noise 6xbtl amp power on .jpg
    Fig221 noise 6xbtl amp power on .jpg
    55.4 KB · Views: 15
  • Fig223 noise 6xbtl amp power on 2891 .jpg
    Fig223 noise 6xbtl amp power on 2891 .jpg
    49.6 KB · Views: 16
  • Fig222 noise 6xbtl amp power off 2891 .jpg
    Fig222 noise 6xbtl amp power off 2891 .jpg
    49.4 KB · Views: 16
THD+N Performance
The performance exceeded my expectations.


The amplifier gain is set to 10x (20 dB). Clipping occurs at 38.5 Vrms into 8 ohms and at 32 Vrms into 4 ohms. During testing, the power supply (620 VA, 140,000 µF) voltage dropped from ±34.2 VDC at idle to ±28.7 VDC under a 240 W load into 4 ohms (one channel), and to ±30.9 VDC under a 180 W load into 8 ohms (one channel).


  • Maximum output power into 8 ohms: 175 W with THD+N of 0.00023%
  • Maximum output power into 4 ohms: 240 W with THD+N of 0.00042%
 

Attachments

  • Fig227 THD+N 8 ohms from 1W to 175W.jpg
    Fig227 THD+N 8 ohms from 1W to 175W.jpg
    73.9 KB · Views: 10
  • Fig228 THD+N 4 ohms from 1W to 240W.jpg
    Fig228 THD+N 4 ohms from 1W to 240W.jpg
    75.6 KB · Views: 10
Last edited:
I tested both the OPA1642 (FET input) and OPA2891 (BJT input) op-amps as input buffers. While the overall THD+N is virtually identical for both, the harmonic distortion profiles differ noticeably.

At typical listening levels (1W–10W), the OPA1642 exhibits a dominance of even-order harmonic distortion, resulting in a warmer tonal character. The amplifier, as a result, also produces more even harmonics in this range.

At higher output levels (above 100W), the differences between the two op-amps become negligible, and both perform similarly in terms of harmonic structure.
 

Attachments

  • Fig229 THD+N opa2891 8 ohms from 1W.jpg
    Fig229 THD+N opa2891 8 ohms from 1W.jpg
    35.7 KB · Views: 9
  • Fig230 THD+N opa2891 4 ohms from 1W.jpg
    Fig230 THD+N opa2891 4 ohms from 1W.jpg
    36.7 KB · Views: 10
  • Fig232 THD+N opa1642 4 ohms from 1W.jpg
    Fig232 THD+N opa1642 4 ohms from 1W.jpg
    31.8 KB · Views: 10
  • Fig231 THD+N opa1642 8 ohms from 1W.jpg
    Fig231 THD+N opa1642 8 ohms from 1W.jpg
    31.2 KB · Views: 10
  • Like
Reactions: twitchie