Buffered LM3886 + overthetop regulated power supply

Status
Not open for further replies.
Jump to Latest
#21
sasmit said:
You're going to end up with a larger heat sink with a regulated power supply, the power dissipation that the manufacturer of the IC has accounted for was with a unregulated psu. The sag in the rails which you would have had with a normal psu would have reduced the overall power dissipation. If you're using a regulated psu you're rails are going to stay pretty much the same even during heavy loading ( assuming your transformer is adequately sized ).

This will be in addition to the heat sinks which you would use for your
regulators. you would have to use insulating pads for either the LM3886 or the pass transistor if you plan to mount them on a single heat sink.

Designing the pcb with both the regualtor and the power amp together taking care of the heatsink mountings , is fairly challenging even for experienced pcb designers.

IMHO you should design separate pcb's for your regulator and power amp. If you're motivation for designing a single pcb was to reduce pcb cost, then I suggest put the art work in a single pcb in such a way that you can cut across the sections in the pcb's with a hacksaw or design perforations into the pcb with a panelling script ( for eagle ).
Click to expand...

Thanks for the comments. My experience level with electronics and PCB design is... moderate. First chip-amp project but not first project. So I'm aware of the amount of thermal management involved and have a suitable selection of whacking great lumps of finned aluminium... though I've some old solid copper CPU heatsinks that would make a nice base for a component-to-component soldered gainclone. Will have to think about that. Momentary thoughts of working out how to clamp 2 LM chip-amps to one the base of one of these but doing that would put anyone on slippery slope ending in them spending time in a rubber lined room with no windows.
 
#22
nightanole said:
I really never did like that power supply design. Hey lets make a poor lm317 regulate till it dies and then have an external part shunt the current. Why Not just get a 5amp rated lm338

http://diyaudioprojects.com/Technic...338-5-Amp-Adjustable-Regulator-Data-Sheet.pdf
Click to expand...

Um, the lm317 isn't "regulated until it dies", it will see very little current thanks to R5 (& R6): the current is almost entirely handled by the power transistor. It's a very standard design and, IMHO, quite elegant and efficient. I, frankly, don't see why you wouldn't like it! 🙂 The LM338 is very expensive and has a absolute max of 5A (nominally plenty, granted) where the power transistor (eg a TIP35C for the + line) can cope with transients upwards of 40A (and do so VERY quickly) so it will NEVER, in this case, drop out... with a HUGE safety margin.

Sorry for CAPITALS! Should have used ITALICS. Can't be BOTHERED.

Ahem. Better to over-engineer than not. Better to have a circuit 10 times overspec than only 1.2 times... 🙂 The huge amp handling capacity of the transistor means it will run fairly cool and allow the LM to run quite efficiently too.

I'm driving some efficient Warfedale 8ohm speakers at moderate volume levels in a small room so it's unlikely either option would ever meet any serious stress. Given the circuit I have costs much less than half a 338 based solution and has better overhead, I'm sticking with reg+transistor design.
 
#23
Ok, thanks so much to the people who have given advice, corrections and comments. I appreciate your time hugely. I clearly made several misjudgements in my 1st version. The op-amp + insane PSU is gone! I will buy a good quality stepped log volume control instead (same cost, when it comes down to it, and much better signal path).

I'm fairly confident in proceeding with the design now shown and will post info and pix on the result. And, for what it's worth, the final PCB design. There's already quite a few PCB designs out there for the LM3xxx but I want to have the cap bank (IMHO an inexpensive array of medium uF 105degree units will outlast, provide better interim reliability, have better ripple rejection and have better ESR and leakage characteristics than popular 8000uF+ chassis mount caps) + the regulator circuits on the same board mounted as closely as possible to the LM chipamps.

If anyone spots any errors or possible improvements in the chipamp stages of the circuit, let me know!

Roman Black's suggested regulator is staying. I'm fundamentally attached to insane over-engineering of 40ampss of regulator capacity for a barely 4amp amplifier chip even if some people say they prefer an unregulated PSU - quite a few people say the regulated supply provides better bass and midrange detail (and quite a few think it's impossible for there to be a difference to start with). Besides, if you can't provide a decent scientific double blind test between a test design and a conservative 'control' design, accounts as to what different designs (designs, cables, wires, capacitors, etc etc) 'sound' like are far more damaging than not saying anything at all: it just lends false credence to worthless subjective opinion. And yes, sorry, all subjective opinion without theory and supporting empiric data >>>IS<<< worthless. Oops, sorry, wrong forum for epistemology! *dismounts soapbox*
 
#25
#26
#28
Reading the LM3886 data sheet for the 5000th time, I realise I shouldn't remove the feedback cap without removing the compensating cap across the input too.

Remove both or keep them? High frequency oscillation seems unlikely but EMI suppression sounds good.

Also, there won't be much current in the feedback look will there? Decebel Dungeon's first design says "The feedback resistor (or in my case two as I used two in parallel to double the power rating) was then soldered between pins 3 and 8." I may just use a single 1W resistor (surely 1/4W is enough?).

Looks like pin 7 should go to signal ground not power ground as I have it at the moment.

Nearly there...
 
#30
Schematic updated.

2x 10uF in parallel as as metalised poly preferable to polarised or bipolar or electros or tantalums in the signal path. Probably.

2x 120R in series instead of 240R because 240R 5W resistors nearly impossible to find locally.

Parallel resistors in the Zobel network as 3W+ non-wirewound resistors impossible to find locally.

I think I can leave this thread to die now unless anyone has any further comments or corrections.

Thanks a lot for the corrections, suggestions, comments and help - hugely appreciated.🙂🙂🙂
 
#31
AndrewT said:
Hi,
the 3886 does not need a buffer at it's input.
Your inadequate source may need a buffer at it's output.
Click to expand...
AndrewT is right.
One buffer before the potentiometer can be nice, though.

If you still use a buffer at the LM3886 +input pin,
you can remove DC-block: C31,C32,C34,C35, if you wish.

Then you may use R13, R14 ~1K (these will be driven by Op-amp output)
so that they matches negativ input R17, R21 and R18, R22
Paralleling, 1K || 22K is about 956 Ohm
So ideally R13/14 should be 956 Ohm.
But 1 kOhm is close enough.
 
#33
lineup said:
AndrewT is right.
One buffer before the potentiometer can be nice, though.

If you still use a buffer at the LM3886 +input pin,
you can remove DC-block: C31,C32,C34,C35, if you wish.

Then you may use R13, R14 ~1K (these will be driven by Op-amp output)
so that they matches negativ input R17, R21 and R18, R22
Paralleling, 1K || 22K is about 956 Ohm
So ideally R13/14 should be 956 Ohm.
But 1 kOhm is close enough.
Click to expand...

I'll maybe have a look at a buffer after building the PSU and chipamps.

The title of this thread is now misleading as I've shelved any pre-amp stages for now.

Thanks a lot for the info here.
 
#34
Couple of things before you let this die.
The lm3886 pair can suck 240 watts continous. Your transformer is only 150va, and regulated, so its 2-3x smaller then what it needs to be.

The lm3886 pair can suck 240watts continous, and its regulated. Your rails are 70vdc. So those .47ohm resistors after the rectifiers, are going to be dissipating 6 watts of heat with just the opamps, at thats not counting the regulator either.
 
#35
nightanole said:
Couple of things before you let this die.
The lm3886 pair can suck 240 watts continous. Your transformer is only 150va, and regulated, so its 2-3x smaller then what it needs to be.

The lm3886 pair can suck 240watts continous, and its regulated. Your rails are 70vdc. So those .47ohm resistors after the rectifiers, are going to be dissipating 6 watts of heat with just the opamps, at thats not counting the regulator either.
Click to expand...

Thanks for drawing my attention to this area.

Interestingly, this calculation http://www.national.com/appinfo/audi...gn_Guide13.xls (documents http://www.national.com/appinfo/audio/files/Using_Overture_Design_Guide.pdf) suggest maximum practical power draw is more like 60W per channel bringing the maximum draw to 120W... more or less.

For my application I don't think more than 20W output power into 8 ohms (per channel) is ever likely. That's one third power. Considerably less in normal use. Small room, efficient speakers.

This leaves me with the dissatisfying feeling of not doing a good job of engineering vs trying not to spend a fortune on a new 300VA torroid.

You have given me something to ponder.
 
#36
WAT?

70volt rails will give you 60 watts at 8 ohms. 70volt rails will give you a blown chip at 4 ohms. You need to pick your rails based on you load so you dont blow up the chip. The more va you have the more stable the power supply, but your going regulated so that might go out the window (no voltage droop for bass kick).

Also you do realize this is a class AB amp thats only 50% efficient at full tilt and 35%eff at 50% or less volume right? So if its putting out 60 watts of music, its also putting out 60 watts of heat, and i assume your using them in stereo so that doubles it. And your going regulated so the regulator has to eat the power of the extra voltage.
 
Last edited:
#37
nightanole said:
WAT?

70volt rails will give you 60 watts at 8 ohms. 70volt rails will give you a blown chip at 4 ohms. You need to pick your rails based on you load so you dont blow up the chip.

Also you do realize this is a class AB amp thats only 50% efficient at full tilt and 35%eff at 50% or less volume right? So if its putting out 60 watts of music, its also putting out 60 watts of heat, and i assume your using them in stereo so that doubles it. And your going regulated so the regulator has to eat the power of the extra voltage.
Click to expand...

As said, 8ohms. Doubt I'll ever use it above 20W.

VA Rating - Class-AB
The absolute minimum VA rating suggested is equal to the amplifier power. A 50W amp therefore needs a 50VA transformer, or 100VA for stereo 50W amps. Larger transformers (up to double the amp power rating) will provide a "stiffer" power supply, and this may be beneficial. For continuous operation at full power (never actually needed in audio), the transformer should have a VA rating of up to 4 times the amplifier power. Elliott Sound Products - Linear Power Supply Design
 
Status
Not open for further replies.
Top Bottom