The setup is pretty simple:
What isn't shown on the amp schematic is the local 100nf MLCC decoupling caps directly (physically) across pin 4 and 8 (VCC - GND) as recommended by the NE5532 datasheet.
The tests are run with both channels at -1dB as recommended by the RMAA software.
What isn't shown on the amp schematic is the local 100nf MLCC decoupling caps directly (physically) across pin 4 and 8 (VCC - GND) as recommended by the NE5532 datasheet.
The tests are run with both channels at -1dB as recommended by the RMAA software.
Attachments
Is the laptop isolated (battery power only)?
Are you paralleling the two halves of each dual opamp as in that diagram with 10 ohm resistors? What are the tolerances of the 20k feedback resistors?
What is -1dB as rms V into the opamps?
I'm wondering if the 400Hz spur with the NE5532's is because they take much less supply current than the discretes, and the power supply is actually motor-boating on light load (is it a SMPS?)
Are you paralleling the two halves of each dual opamp as in that diagram with 10 ohm resistors? What are the tolerances of the 20k feedback resistors?
What is -1dB as rms V into the opamps?
I'm wondering if the 400Hz spur with the NE5532's is because they take much less supply current than the discretes, and the power supply is actually motor-boating on light load (is it a SMPS?)
I've been running tests both with laptop plugged in and battery only to measure any ground loop effects.
Yes, both halves are parallelled with the 10R. The feedback resistors are 5% but I've matched them with a fluke to below 1%.
1dB appears to be around 310mVRMS (According to the software, I still need to physically measure it).
So I had a look at the power supply filtering and decided to add an RC filter at the power input as follows:
Seems to have done the trick (Green is before, White after):
[NE5532 Noise]
[Burson Noise]
[Results]
Yes, both halves are parallelled with the 10R. The feedback resistors are 5% but I've matched them with a fluke to below 1%.
1dB appears to be around 310mVRMS (According to the software, I still need to physically measure it).
So I had a look at the power supply filtering and decided to add an RC filter at the power input as follows:
Seems to have done the trick (Green is before, White after):
[NE5532 Noise]
[Burson Noise]
[Results]
Attachments
Cool - probably a SMPS then, often they back-off to a sporadic mode at low current drain which is the worst thing possible for audio, still its good to see only 10 ohms needed to filter than out.
Am I right in thinking the performances are fairly well matched overall between NE5532/Burson (except for power consumption, size, price) ? I'd be interested in current-noise measurements (the 10k in series with the input should show that). The NE5532 is something like 0.6pA/√Hz, which turns into 6nV/√Hz with 10k source impedance, a high current noise can be a show-stopper for circuits with impedances around 10k or more (active filters at low signal levels like an RIAA amp for instance).
Normally the current noise is related to the bias current, 0.56pA/√Hz for 1µA, and scaling as √current. The only datasheet(s) I can find for Burson's suggest a bias current of 180µA, suggesting 7.6pA/√Hz which is huge - perhaps this is a misprint or been fixed, but it was an immediate red-flag for me.
[actually the IM is rather different and the voltage? noise difference sticks out a bit]
Personally I'd always quote noise as spectral densities - A-weighting is very hard to calculate with, and more useful when dealing with things with a non-flat noise floor (such as tape + vinyl, not electronics).
Am I right in thinking the performances are fairly well matched overall between NE5532/Burson (except for power consumption, size, price) ? I'd be interested in current-noise measurements (the 10k in series with the input should show that). The NE5532 is something like 0.6pA/√Hz, which turns into 6nV/√Hz with 10k source impedance, a high current noise can be a show-stopper for circuits with impedances around 10k or more (active filters at low signal levels like an RIAA amp for instance).
Normally the current noise is related to the bias current, 0.56pA/√Hz for 1µA, and scaling as √current. The only datasheet(s) I can find for Burson's suggest a bias current of 180µA, suggesting 7.6pA/√Hz which is huge - perhaps this is a misprint or been fixed, but it was an immediate red-flag for me.
[actually the IM is rather different and the voltage? noise difference sticks out a bit]
Personally I'd always quote noise as spectral densities - A-weighting is very hard to calculate with, and more useful when dealing with things with a non-flat noise floor (such as tape + vinyl, not electronics).
Last edited:
I made a little dongle to make current noise measurements (JST male and female bridged with 2x matched 10k resistors):
Here are the noise runs:
[NE5532 w/ 10k]
[Burson w/10k]
[Combined]
[Results List]
[NE5532 Noise Before]
[NE5532 Noise w/10k]
[Burson Noise Before]
[Burson Noise w/10k]
Here are the noise runs:
[NE5532 w/ 10k]
[Burson w/10k]
[Combined]
[Results List]
[NE5532 Noise Before]
[NE5532 Noise w/10k]
[Burson Noise Before]
[Burson Noise w/10k]
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.