I've been poking around these forms for a few months now, listening, learning, and understanding chip amps the best I can. I came across this design for a bridge/parallel design, apparently tested by National Semiconductor and has promising results BPA-200
I have a PCB design mocked up in ExpressPCB from one of the schematics show in that paper (last schematic in the paper, one of the last pages). I've been comparing them with designs people here have, and I noticed a difference. Most people here (more or less) run their input signal directly into the LM3886, while this design does the same, but it has an extra op amp in the feed back. I read the reason for this, makes sense, however, why isnt everyone using this? Does it's function provide so little, people opt not to incorporate it? Should I have it in my design? Is it not needed for single chip designs, but necessary for multi chip designs? Can anyone shine some light onto this design?
I have a PCB design mocked up in ExpressPCB from one of the schematics show in that paper (last schematic in the paper, one of the last pages). I've been comparing them with designs people here have, and I noticed a difference. Most people here (more or less) run their input signal directly into the LM3886, while this design does the same, but it has an extra op amp in the feed back. I read the reason for this, makes sense, however, why isnt everyone using this? Does it's function provide so little, people opt not to incorporate it? Should I have it in my design? Is it not needed for single chip designs, but necessary for multi chip designs? Can anyone shine some light onto this design?
AN-1192 BPA-200 has two opamps, one is the buffer you mentioned and the other is the dc servo and it is in the feedback loop. I assume MotoMan_Yz400 to be asking about the one in the feedback loop.
This circuit is a dc servo. It integrates the output signal (calculates the mean voltage) and then it substracts it to the input signal (not as easy but that's what it does).
This is a way to reduce dc offset that has some advantages:
· Does not use caps in the signal path
· Offers better precision
But it has many problems:
· It complicates the layout and forces (almost always) to add a second power supply to feed that little opamp.
· In fact it is a magnified capacitor and it may also have some of it's defects, severely attenuated by the fact that it operates at low frequency and it does not *theoretically* interact with the audio signal.
· It requires high precision parts to do it's job well. That does mean close tolerance, not audiophile grade.
I havn't experimented with it already so I can't give any clue about if its sound improovement is worth the time and money
Well one would think, if it's helping regulating the DC offset by eliminating it the best it can, then couldnt you conclude that the LM3886 would then run cooler? If theres always extra current running through the IC, then it's not aiding to the audio output, just heating things up. That sound right?
The difference would be 8 mV best case (gainclones usually have offsets arround 2 mV and this means 30V*(2mV/8Ohm) = 7.5 mW.
This isn't worth the extra price, and a DC servo circuit is more prone to fail that a single cap. This is for peple who fiercely belive that there shouldn't be any caps in the signal path.
This isn't worth the extra price, and a DC servo circuit is more prone to fail that a single cap. This is for peple who fiercely belive that there shouldn't be any caps in the signal path.
You are absolutely right, I was refering to AN-1490 where it states:
I have a whole folder full of application notes, sorry for the mistake.
I actually belive that the DC servo must also degrade the sound and probably in a much less benefical way than the coupling cap since it (partially) referenciates the feedback loop to an active voltage source whose low impedance is reached trough incredible ammounts of NFB and will vary severely with frequency.
I would be very pleased if anybody has done capacitor versus servo A/B comparison and can give a clue.
I have a whole folder full of application notes, sorry for the mistake.
I actually belive that the DC servo must also degrade the sound and probably in a much less benefical way than the coupling cap since it (partially) referenciates the feedback loop to an active voltage source whose low impedance is reached trough incredible ammounts of NFB and will vary severely with frequency.
I would be very pleased if anybody has done capacitor versus servo A/B comparison and can give a clue.
Well I have little to no control over any possible DC offset in the signal. I'll be using sources anywhere from a computer, CD player, or even a TV cable box. So understanding that those sources can be quite horrific with their signal, not sure if any offset will be present, just would like to eliminate it the best I can. This isnt going to be any kinda of award winning amp system, but I would like to "try" to make it the best I can with a reasonable budget in mind. If the cost is an extra $20 for the ICs, etc. thats fine with me. I'm just not trying to spend $400-$800 down at my local stereo store for a system with bells and whistles that I just dont need or want (ie. Dolby Digital
, Stadium reverberations
, and Theater EQs
)
, Stadium reverberations, and Theater EQs )
I really feel that the cap doesn't degrade the sound so much, so if you aren't going to use very expensive speakers with very expensive sources using the cap is probably the way to go.
If you decide to try using a dc-servo remember that it's failure may lead to very high offset at the output, so a proper speaker protection (relay) is a must, there are a lot of great speaker-protection schematics arround.
If you aren't using very expensive loudspeakers and the build quality of the gainclone is up to the task a slow-blow fuse on each rail will probably be enough (This will stop the amplifier if a constant current is drawn for a sufficient time and that time should be much lower than the needed to burn a speaker), altough adding extra protection will not hurt and its always a good practice.
If you decide to try using a dc-servo remember that it's failure may lead to very high offset at the output, so a proper speaker protection (relay) is a must, there are a lot of great speaker-protection schematics arround.
If you aren't using very expensive loudspeakers and the build quality of the gainclone is up to the task a slow-blow fuse on each rail will probably be enough (This will stop the amplifier if a constant current is drawn for a sufficient time and that time should be much lower than the needed to burn a speaker), altough adding extra protection will not hurt and its always a good practice.
Well I spoke to my father lastnight (BIG time AudioPhile, and scientist) and he brought up a good point. The servos most likely perform their jobs quite well, most likely better than an input capacitor. However the over all quality increase will probably be so insignificant that unless you have $10,000 drivers, in an anechoic room, with a VERY good ear, someone like myself running it in the living room of my apartment would never know the difference. So he suggested the old philosophy of K.I.S.S. (Keep-It-Simple-Stupid). So keeping the servo in, or removing it from my design is still unknown at this point. I just did know it's exact function, and thought it was there for a major reason. It turns out it's there for a minor one.
Input capacitors
I personally simply do not believe the author of AN-1490 (Mark Brasfield) in his statement about sound degradation by the input capacitor. If you look at the schematic, it shows (for example) Cin1a and Cin1b as paralleled .47uF and 22uF caps -- this being a giveaway that what is intended here is an electrolytic (!!!) paralleled with a smaller film cap at the amp input, which is completely squirrelly. This is the setup usually used with the Ci cap blocking DC from ground, and with power supply smoothing caps, but I've never seen it before on the amp input.
Moreover, with the 68k input impedance of the amp, the 22uF input cap is incredible overkill anyway: the pole of the input network is 0.11 Hz! Clearly the designer has misapplied the rule of using a monstrously oversized electrolytic to avoid distortion; but in this application a more reasonable value of film capacitor would do just fine.
Since the 200uF Ci capacitor in the feedback loop (again chosen huge to avoid electrolytic distortion, but amazingly not bypassed with film here, though the board has a slot for a film bypass) raises the amp's pole to about 0.3 Hz anyway, a single 8.2uF film cap at the input would be fine; for that matter, a 4.7uF Cin would only raise the pole to 0.5 Hz, low enough for all but the most fanatical DC-to-Light advocates.
So I have to believe there was some mistake in the listening tests; perhaps trying various replacements for Cin1b (the film) while leaving Cin1a (the NP electro), which actually would be the source of any nonlinearities, in place. Otherwise the results make no sense.
Use an input cap. Murphy is alive and well, even in audiophiles' listening rooms.
Craig
I personally simply do not believe the author of AN-1490 (Mark Brasfield) in his statement about sound degradation by the input capacitor. If you look at the schematic, it shows (for example) Cin1a and Cin1b as paralleled .47uF and 22uF caps -- this being a giveaway that what is intended here is an electrolytic (!!!) paralleled with a smaller film cap at the amp input, which is completely squirrelly. This is the setup usually used with the Ci cap blocking DC from ground, and with power supply smoothing caps, but I've never seen it before on the amp input.
Moreover, with the 68k input impedance of the amp, the 22uF input cap is incredible overkill anyway: the pole of the input network is 0.11 Hz! Clearly the designer has misapplied the rule of using a monstrously oversized electrolytic to avoid distortion; but in this application a more reasonable value of film capacitor would do just fine.
Since the 200uF Ci capacitor in the feedback loop (again chosen huge to avoid electrolytic distortion, but amazingly not bypassed with film here, though the board has a slot for a film bypass) raises the amp's pole to about 0.3 Hz anyway, a single 8.2uF film cap at the input would be fine; for that matter, a 4.7uF Cin would only raise the pole to 0.5 Hz, low enough for all but the most fanatical DC-to-Light advocates.
So I have to believe there was some mistake in the listening tests; perhaps trying various replacements for Cin1b (the film) while leaving Cin1a (the NP electro), which actually would be the source of any nonlinearities, in place. Otherwise the results make no sense.
Use an input cap. Murphy is alive and well, even in audiophiles' listening rooms.
Craig
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.