Was thinking of making yet another chip amp, and maybe will be trying the minimalist approach, a gainclone I guess. Am wondering about the benefits and risks of omitting various components. Will be using either a LM3886 or LM4780, probably 2x parallel.
Input cap. Seems unnecesary if you can guarantee no DC on the input.
Cap from NFB net to ground. Not quite sure of the issues of omitting this.
Zobel/Boucherot. Obviously to deal with the load from your speakers being reactive rather than wholly resistive. I have no way of measuring my speakers' reactance. What might be the downside to omitting it?
Nothing but a few uF of smoothing caps on PSU. People do this and I have no idea why.
Your help is appreciated.
Input cap. Seems unnecesary if you can guarantee no DC on the input.
Cap from NFB net to ground. Not quite sure of the issues of omitting this.
Zobel/Boucherot. Obviously to deal with the load from your speakers being reactive rather than wholly resistive. I have no way of measuring my speakers' reactance. What might be the downside to omitting it?
Nothing but a few uF of smoothing caps on PSU. People do this and I have no idea why.
Your help is appreciated.
Input cap - yes you can omit if you are sure you will never use a source with any DC.
NFB cap - careful selection of feedback resistors and input impedance resistor is required to minimise DC offset. May drift with temperature too. You should aim for less than 50mV.
Boucherot - not a good idea to omit in my opinion. Amp will oscillate either all the time or with certain signals, May fry tweeters or the amp itself will overheat and shutdown or sound bad.
Low smoothing capacitance - done by people who like the sound of increased distortion. Not my cup of tea.
If you aim to parallel chips using the above approaches you might run into problems. Start with a single chip first. Paralleling chips requires very well matched components and performance from each amp.
Remember that unless you have identical unmodified amps to a/b with or means to measure the differences, the excersise is likely to be futile!
NFB cap - careful selection of feedback resistors and input impedance resistor is required to minimise DC offset. May drift with temperature too. You should aim for less than 50mV.
Boucherot - not a good idea to omit in my opinion. Amp will oscillate either all the time or with certain signals, May fry tweeters or the amp itself will overheat and shutdown or sound bad.
Low smoothing capacitance - done by people who like the sound of increased distortion. Not my cup of tea.
If you aim to parallel chips using the above approaches you might run into problems. Start with a single chip first. Paralleling chips requires very well matched components and performance from each amp.
Remember that unless you have identical unmodified amps to a/b with or means to measure the differences, the excersise is likely to be futile!
Hi Robert.
Omitting the NFB cap will likely increase the DC offset on the output of the amp. (having a cap here ensures that there is NO gain at DC, so what comes in is what comes out.
If it needs the zobel and is not there, then the likely result will be oscillation.
I initially made my p2p LM3886 gainclone with 1000uF / rail and it performed fine, adding 4700uF extra at a later date reduced the amount of hum on output (it wasn't that high to start with).
The other effect of having too low an amount for smoothing caps will be that the DC voltage will be lower, and the ripple will be much larger.
Hmmm I went away for about a half an hour in the middle of writing this post, seems like some others beat me too it
Tony.
Omitting the NFB cap will likely increase the DC offset on the output of the amp. (having a cap here ensures that there is NO gain at DC, so what comes in is what comes out.
If it needs the zobel and is not there, then the likely result will be oscillation.
I initially made my p2p LM3886 gainclone with 1000uF / rail and it performed fine, adding 4700uF extra at a later date reduced the amount of hum on output (it wasn't that high to start with).
The other effect of having too low an amount for smoothing caps will be that the DC voltage will be lower, and the ripple will be much larger.
Hmmm I went away for about a half an hour in the middle of writing this post, seems like some others beat me too it
Tony.
OK, Zobel stays, big PSU caps stay, input cap goes. Points remaining
Thanks again.
I imagine the problem may be the impedance seen by +in will vary according to your source and especially your volume control's position, so matching with the impedance seen by -in might be difficult.
I would parallel the amps on a LM4780s, so on a single die, with <0.1% resistors. I thought one benefit of omitting caps is that caps are harder to match (with my not too accurate cap metre).
Thanks again.
Last edited:
Smaller supply filtering needlessly throws away output power headroom as the tops of the output waveform hit the ripple. The graph shows the results of a test I performed awhile back using a TDA2040, 9-0-9 2A transformer and various filter cap values. The Y axis is the maximum nonclipped RMS voltage into a non inductive 4 Ohm load. 2,200uf seemed to be the point where going higher would result in diminishing returns. Less cap value, the output voltage (thus power) drops off significantly. Keep in mind that this is only one channel.
As a side note, the amplifier had virtually no hum until I got down to the 100uf caps, but is was still very weak.
An externally hosted image should be here but it was not working when we last tested it.
As a side note, the amplifier had virtually no hum until I got down to the 100uf caps, but is was still very weak.
Last edited:
That is exactly right. The DC offset will vary with volume position. With careful value selection you should be able to achieve an offset that stays within -50mV to 50mV.
I'm not condoning this design approach, just giving you some information.
OK. This is good to know. Some kits seem to ignore it though, for example AudioSector's DIY Chip Amplifier Kits, PCB's, Components and Information.http://www.audiosector.com/lm4780.shtml
Exactly the opposite. Tie the +INs together (with 1k stopper resistors) and feed them from a common blocking cap and bias resistor, this takes current offset out of the equation. Each amp needs its own complete and perfectly matched feedback network with oversized/matched GND leg caps for unity gain at DC so we only have to deal with voltage offset and with no gain applied (20mV max). See TI/NS "AN-1192 Overture™ Series High Power Solutions" application note.
If you want a DC-coupled parallel amp (N sections) you need to servo N-1 sections to zero differential output voltage against the unservoed "master" chip. Any servos become as critical as coupling caps and feedback caps in terms of perfectly matched time constants.
Last edited:
It would be very convenient if you are right. I hope you are right: That it is better to tie the +INs together rather than to DC block each individually.
With individual DC blockers you would add two more variables, the corner frequencies of the roll-off and offset current mismatch.
A slight DC offset isn't dramatic, if you had, say, 200mA of cross current idle dissipation will rise but basically that's it. Actually, a moderate cross current could be considered a good thing as it widens the class-A operation region.
A slight DC offset isn't dramatic, if you had, say, 200mA of cross current idle dissipation will rise but basically that's it. Actually, a moderate cross current could be considered a good thing as it widens the class-A operation region.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.