50dB (-PSRR) is flawed enough.
In particular with 15.000uF effective rail capacitance in the case of the JR model 10.
(Major downside of bridging these amp chips, imo. Getting an additional 30-40dB would require a huge lytic size)
Where are we getting the -50dB number from? The stated distortion specs don't reflect this poor result.
PSRR graph of the 3886 datasheet, fig 47.
(in a bridged/parallel setup, reduces by a mere 10dB, thanks to +PSRR holding up)
(in a bridged/parallel setup, reduces by a mere 10dB, thanks to +PSRR holding up)
Noted, so how does the model 10 achieve the additional 30dB? The specs are on page 15 of the manual.
PSRR is customarily referred to the input. Since the min recommended gain is 20dB the -50dB figure becomes -30dB at the output.
One way around this is to use a single rail and refer the input to the -ve rail. This is how I'd rig it if I were designing a bridged LM3886 (hypothetically, since I'm no longer a big fan of this particular chip). With all signals referenced to the -ve rail the -ve PSRR becomes irrelevant. The +PSRR is very decent indeed for this chip. Bridging with only a single rail much more sense from the pov of optimizing the PSU - you use caps at twice the working voltage which store more energy than caps at lower voltage for the same volume.
The manual doesn't show PSRR, only CMRR. That would be a function of the input circuitry.
Yes, but the manual also shows THD+N of 0.008% which translates >80dB so he seems to have addressed the chip's PSRR inadequacies. What am I missing?
As any other power amplifier, with the ripple reducing smoothing caps.
(at Io=0, only determined by the total quiescent current level. c=a/v)
Weren't we just saying that its power supply capacitance was inadequate to achieve anything respectable?
😉
😉
Assuming the THD+N is across the whole band (the LM3886 DS shows about 0.008% distortion @ 20kHz) then the PSU ripple needs to be only 52dB below the output if the PSRR in practice is 30dB. Since the full output is 150W into 8R, this corresponds to 35VRMS and hence ripple needs to be below 88mV @ 20kHz. With a current draw of 4.3A we get a supply impedance around 20mohms which is quite doable. The main reservoir caps are going to be ESR/ESL limited at 20kHz since 20mohms corresponds to only 400uF.
That's my handwaving answer, anything I missed out?
That's my handwaving answer, anything I missed out?
(10.000uF-12.000uF lytics have a max impedance of ~10mΩ at 20kHz. Times 4, divided by 5, makes a SL of current)
just guessing here -- perhaps JR has notched it (ie:broskie style or R.F. style, since the frequency(ies) are known). i used to build R.F. CW-audio filters that way and they were very effective
Last edited:
although JR may have used a broskie or DSP solution, here's another thought: since the lousy negative side PSRR is still over 90db at 120hz then if the PSU is linear there's no issue. and if the PSU is SMPS then the ripple frequency is high and determinate and quite easy to filter out within the PSU or between PSU and chipamp. so with good design and layout the amp could meet it's THD+N spec without much fuss. am i missing something?
The Model 10 has a standard power supply with toroidal, bridge rectifier and electrolytics.
Ripple on the rails is determined by the mains frequency and the smoothing cap value, I = C * dV/dt.
Aka Vp = I / (2 * f * C)
(the THD+N spec of 0.008% is for 1W)
Ripple on the rails is determined by the mains frequency and the smoothing cap value, I = C * dV/dt.
Aka Vp = I / (2 * f * C)
(the THD+N spec of 0.008% is for 1W)
OK, so
bridge PSU rectifier ==> 120hz ripple freq ==> negative side PSRR > 90db per fig. 47
starting with any reasonable rails filter, reject what's left by 90db ==> dead quiet (unless the layout is screwed up)
Last edited:
hifi-advice.com
http://www3.hiendy.com/hififorum/forum.php?mod=viewthread&tid=48342
It does not seem to be so clear as to what type of PSU was used.
http://www3.hiendy.com/hififorum/forum.php?mod=viewthread&tid=48342
It does not seem to be so clear as to what type of PSU was used.
Last edited:
interesting quote from that site:
"Although I haven't heard the 112, I still have a pretty good idea of what it should sound like as my Model 10's switched power supply broke down and no matter what the importer tried, couldn't be fixed. In the end the amp went back to Jeff Rowland after which the man himself installed a linear power supply in place of the problematic switched one. The inporter first didn't tell me but I immediately noticed the difference in weight when I lifted the power supply. And it wasn't only the weight that had changed: the sound was also completely different. . . ."
"Although I haven't heard the 112, I still have a pretty good idea of what it should sound like as my Model 10's switched power supply broke down and no matter what the importer tried, couldn't be fixed. In the end the amp went back to Jeff Rowland after which the man himself installed a linear power supply in place of the problematic switched one. The inporter first didn't tell me but I immediately noticed the difference in weight when I lifted the power supply. And it wasn't only the weight that had changed: the sound was also completely different. . . ."
It's typical of Rowland amps, updates without adding something to the model description.
There's a thread on the Model 7 of the 1980s, turned out there were at least three different versions. Power device change from Motorolas to Japanese, or back, updated front-end boards.
The higher power Model 9 was manufactured with both bipolar power device output stages and bridged/parallel IC's (not 3886s though), with a traditional power supply and SMPS. There was also the option to order one or the other, he last updated model did get a TI add-on.
Afair, the dual case models started with a regular power supply, soon after updated by the switch mode version (within the first year or two).
Though I understood that they could also be converted back to a standard PS.
There's a thread on the Model 7 of the 1980s, turned out there were at least three different versions. Power device change from Motorolas to Japanese, or back, updated front-end boards.
The higher power Model 9 was manufactured with both bipolar power device output stages and bridged/parallel IC's (not 3886s though), with a traditional power supply and SMPS. There was also the option to order one or the other, he last updated model did get a TI add-on.
Afair, the dual case models started with a regular power supply, soon after updated by the switch mode version (within the first year or two).
Though I understood that they could also be converted back to a standard PS.
Last edited: