When I showed the measurements of the my old Arcam Back Box 3 DAC @restorer-john noticed the burned out state of the diodes in the 5V digital power supply. @SIY recommended UF4007 which I could order (the original BYD33G were not available to me). And while I replaced the diodes I also undid all changes I had done to "improve" SQ without really knowing what I did back then
. Apart from the exchanged diodes the unit is now back in its original state and I thought I should repeat the measurements.
Before the fix the voltage before the regulator had 390 mVpp ripple and some noise (not shown), and after the regulator 150 mVpp noise and no ripple:
Here is an image of the fixed power supply (4 brand new UF4007 diodes):
The voltage before the regulator still has 400 mVpp ripple:
But after the regulator it's now reduced to 60 mVpp noise and no ripple:
So let's see how the unit measures now. For all measurements I fed the Toslink input of the BB3 with 44/16 sinus by an old Edirol UA-25 or an RME ADI-2 PRO fs, and measured the analog output with either the RME and REW, or a SIGLENT SDS 1202X-E (2 channel 200 MHz digital scope).
Output level of the main output:
The left channel (yellow) delivers 2.44 Vrms, the right channel 2.29 Vrms, a difference of 0.5 dB (as with the MK1 version).
Output level of the aux output:
The left channel (yellow) delivers 1.22 Vrms, the right channel 1.15 Vrms, a difference of 0.5 dB (as above).
THD:
THD and noise are both the same as the MK1 version. The OPA134 I had rolled in was clearly no better choice than the original LM627.
IMD (CCIF = 19.000 and 20.000 kHz):
The actual IMD (at 1 kHz) is not that bad, but there is a lot of spurious stuff. However the reason why it's now much better than the MK1 version is that I forgot to enable dither in REW back then (not here). If I disable dither here as well (not shown) the result looks as bad as the MK1 version.
Looking closer the spurious signals are spikes with 100 Hz distance in between. I had hoped that with the fixed 5V power supply these 100 Hz spikes would disappear but this is not the case. However while playing around I found out that these spikes are not related to the power supply at all. Changing the lower frequency of the 2 signals just slightly changes the distance between the spikes and their amplitudes as well
.
With 19.050 kHz and 20 kHz and zoomed in the distance is 50 Hz:
With 19.025 kHz and 20 kHz and zoomed in the distance is 25 Hz:
With 19.020 kHz and 20 kHz and zoomed in the distance is 20 Hz:
With 19.010 kHz and 20 kHz and more zoomed in the distance is 10 Hz:
The only reason I can think of is inproper upsampling and/or digital filtering in the DAC chip. Anyone a better idea?
Multitone (44.1 kHz samplerate, not 192 kHz which the BB3 cannot handle):
Not that bad, and no ultrasonic noise as well despite this being a 1 bit DAC.
Finally the frequency response measured with white noise, to see the characteristic of the reconstruction filter. First full bandwith up to 96 kHz:
In my view this is not a bad performance for a 16 Bit DAC.
Zoomed in around the Nyquist frequency:
I would expect more damping at 22.5 kHz. It seems as if this filter characteristic we also see at many modern DACs (sufficent damping at 24 kHz and above) is not so modern at all.
Thanks again to @restorer-john and @SIY for the help. Since I don't need the BB3 any longer and must clear my work room it is for sale now.
. Apart from the exchanged diodes the unit is now back in its original state and I thought I should repeat the measurements.Before the fix the voltage before the regulator had 390 mVpp ripple and some noise (not shown), and after the regulator 150 mVpp noise and no ripple:
Here is an image of the fixed power supply (4 brand new UF4007 diodes):
The voltage before the regulator still has 400 mVpp ripple:
But after the regulator it's now reduced to 60 mVpp noise and no ripple:
So let's see how the unit measures now. For all measurements I fed the Toslink input of the BB3 with 44/16 sinus by an old Edirol UA-25 or an RME ADI-2 PRO fs, and measured the analog output with either the RME and REW, or a SIGLENT SDS 1202X-E (2 channel 200 MHz digital scope).
Output level of the main output:
The left channel (yellow) delivers 2.44 Vrms, the right channel 2.29 Vrms, a difference of 0.5 dB (as with the MK1 version).
Output level of the aux output:
The left channel (yellow) delivers 1.22 Vrms, the right channel 1.15 Vrms, a difference of 0.5 dB (as above).
THD:
THD and noise are both the same as the MK1 version. The OPA134 I had rolled in was clearly no better choice than the original LM627.
IMD (CCIF = 19.000 and 20.000 kHz):
The actual IMD (at 1 kHz) is not that bad, but there is a lot of spurious stuff. However the reason why it's now much better than the MK1 version is that I forgot to enable dither in REW back then (not here). If I disable dither here as well (not shown) the result looks as bad as the MK1 version.
Looking closer the spurious signals are spikes with 100 Hz distance in between. I had hoped that with the fixed 5V power supply these 100 Hz spikes would disappear but this is not the case. However while playing around I found out that these spikes are not related to the power supply at all. Changing the lower frequency of the 2 signals just slightly changes the distance between the spikes and their amplitudes as well
. With 19.050 kHz and 20 kHz and zoomed in the distance is 50 Hz:
With 19.025 kHz and 20 kHz and zoomed in the distance is 25 Hz:
With 19.020 kHz and 20 kHz and zoomed in the distance is 20 Hz:
With 19.010 kHz and 20 kHz and more zoomed in the distance is 10 Hz:
The only reason I can think of is inproper upsampling and/or digital filtering in the DAC chip. Anyone a better idea?
Multitone (44.1 kHz samplerate, not 192 kHz which the BB3 cannot handle):
Not that bad, and no ultrasonic noise as well despite this being a 1 bit DAC.
Finally the frequency response measured with white noise, to see the characteristic of the reconstruction filter. First full bandwith up to 96 kHz:
In my view this is not a bad performance for a 16 Bit DAC.
Zoomed in around the Nyquist frequency:
I would expect more damping at 22.5 kHz. It seems as if this filter characteristic we also see at many modern DACs (sufficent damping at 24 kHz and above) is not so modern at all.
Thanks again to @restorer-john and @SIY for the help. Since I don't need the BB3 any longer and must clear my work room it is for sale now.
