After the great success and really amazing sound quality of my LM3886-based gainclone project, Ive decided I want to build a version that is better suited to driving low impedance loads, like my Yamaha NS1000Ms.
This has lead me to the LM4780 chip from TI, which is really just two matched LM3886 dies in the one package can be used as one stereo amp, or a bridged or paralleled mono amplifier. The datasheet is the best place to start when designing an amp or implementing a chip like this. Things to look for include the SOA or safe operating area graphs, which will help determine power supply rails and maximum power dissipation into various loads. The bridged configuration gives double the voltage swing into 8 ohms, but essentially the same performance into 4 ohms. I wanted the amp to be optimised for driving 4 ohms and that means connecting both amplifiers in parallel.
In this particular amp, I want to optimize certain design aspects and capitalise on some of the LM4780 inherent advantages. Design goals for the LM4780 gainclone project amplifier are as follows:
- Optimised for driving 4 ohm loads
- Therefore each LM4780 will be configured for PA100 operation as per the TI application notes
- Very low power supply impedance
- Therefore dual mono, including separate mains transformers
- Shortest wiring possible
- Localised power supply C, wherever possible
- Minimum number of parts to obtain desired characteristics
- Therefore no active regulation
- No input network
- No output network
- Careful matching of load sharing R
- Minimal feedback
- Complete amplifier to be a one box design
I am going to use Peter Daniel’s well-regarded LM4780 dual-mono kit as the basis for my design, and make whatever adjustments and modifications I choose to along the way.
I have decided that I am going to use an unregulated power supply, except for the regulation is provided by the mains transformers – so no active regulation. A few of the DIY guys are championing gainclones with regulated supplies, but my thinking on not going down this path is that the LM3886 and LM4780, by design, have extremely good rejection of PSU ripple and noise – so good in fact that many people choose to use just 1500uF of C per rail, right at the chip with no other ripple filtering. From the spec sheet, the rejection is 100 – 120dB per rail, and 110dB common mode rejection which is amazing really.
I’m going one better and keeping the 1500uF of C per rail at the chips, but backing that up with 22000uF per rail further back. All caps will be bypassed with 0.1uF film caps. Having caps so close to the chip should provide extremely low power supply source impedance, and the main reservoir is still sitting at a very low impedance just a little further back. This large energy storage should help to maximise impulse performance and the amplifiers ability to drive low impedance loads like my Yamaha NS1000s.
I see no point in regulation of the power supplies here as any noise and ripple will be rejected by the chip and there will not be much noise and ripple anyway. I didn’t want to raise power supply Z by introducing regulators and the transformers won’t be loaded up significantly, so the rail voltages should therefore never sag by much. Hopefully there will not be any/much modulation of rails by signals present at the output – this could be the limitation of the unregulated pathway.
I’m going to pay close attention to ground and signal returns as I’ve done with the first amp. I’m also not going to cannibalise the first amp as I thought I might – that way I keep it as a fully functioning unit and can do comparisons etc. I was going to re-use the PSU boards and filters, but that just leaves me with a bunch of non-functioning crap! This way I will have two complete amps.
Watch this space as I will be updating my progress here…