Tír Na HiFi

With the headbuffer, there is very little or no discernible difference between running it from battery or one of those LM2596 DC-DC switchers

Yea that's pretty cool
Not to mention cheaper and easier ;)

Relative power supply grunge immunity seems to be a strong point of no global NFB circuits. The line stage I'm trying to perfect has GNFB and sounded rough until I introduced some draconian HF filtering on its rails...

Watching what happens in circuits that rely on NFB to correct poor HF behaviour, in something like LTspice, shows why this is the case, at least in some instances. The output stage starts to distort, the FB loop then drives the input and intermediate gain stages hard, very hard, trying to correct the error - and some parts of the circuit are driven into exhibiting very nasty current and voltages glitches - the transient response of the parts are not up to the job. Momentarily, the whole circuit is in an unstable state, and it requires time to restabilise ... and this is even before adding in the full parasitics that a real circuit would exhibit.

Thanks Frank - that's good background detail of the weaknesses inherent in feedback circuitry.
The transient response shortcomings are a possible explanation for why local feedback loops are considered more benign than global feedback loops - assuming the local feedback loop having better transient response characteristics.

Bandwidth limiting the input signal would seem to be one possible answer to this issue?

What's the plan for pin 7 then

jkeny wrote: ↑Sun Jul 09, 2017 12:42 pm Thanks Frank - that's good background detail of the weaknesses inherent in feedback circuitry.
The transient response shortcomings are a possible explanation for why local feedback loops are considered more benign than global feedback loops - assuming the local feedback loop having better transient response characteristics.

Bandwidth limiting the input signal would seem to be one possible answer to this issue?

Wouldn't work - the FB loop is totally in control in this situation, it's going to try and force the circuit to respond as fast as necessary to correct the error - which can easily be way higher in effective frequency than what's in the input. Bandwidth limiting the FB loop "solves the problem" - but then the FB is not doing its job, properly!

A good solution is to make sure the whole circuit exhibits competent behaviour well beyond the audio frequencies - Spectral is a highly regarded manufacturer that follows this philosophy. I've investigated a couple of DIY designs that really push the capability well beyond audio bandwidth, and the benefits are very obvious, when you see how well behaved the circuit is at the limits, just handling audio input. A rule of thumb I've developed is that I want to see the circuit be able to handle 200kHz signals - that is, the output waveform is visually in good shape handling such, at decent peak to peak voltages - many designs just produce a mess at these frequencies.

Thanks Frank - good info!

@Nige The variable voltage divider on pin 7 seems the way to go - volume control is a very useful function.
How did you implement this - a stepped attenuator with capacitor in parallel on one leg of voltage divider?

jkeny wrote: ↑Mon Jul 10, 2017 9:29 am Thanks Frank - good info!

@Nige The variable voltage divider on pin 7 seems the way to go - volume control is a very useful function.
How did you implement this - a stepped attenuator with capacitor in parallel on one leg of voltage divider?

my memory it a little vague think it was probably 10k stepped attenuator but it was on lifepo4 cells with 1uf film capacitor, think i might of had to put a resistor in parallel

i remember the volume adjustment had a very linear feel to it, i tried a 10000uf cap on pin 7 and gnd which made it very unresponsive as in it took too long to charge and discharge

is this a runner for a power amp?

stole from here
http://www.diyaudio.com/forums/pass-lab ... input.html