Tír Na HiFi

A line-level buffer's a bit of a different application to an I/V - I'm a bit of a novice with JFETs but what to me seems less than optimal with that schematic is it isn't going to give you an invariant power supply draw. Reason being the load current is in parallel with the CCS JFET rather than in parallel with the active JFET. In the latter case the JFET's acting like a shunt regulator across the load so you end up with no PSU ripple from the audio signal. My hypothesis for audio SQ is that its the noise on the supply rails (rather than the THD figures) which gives rise to colouration. For zero colouration you either need infinite PSRR (pretty difficult to do) or you need zero signal-correlated PSU noise (a bit easier to engineer).

What I suspect is determining the SQ with your schematic nige is the PSRR of the JFET as source follower. Not something I've ever modelled but it will depend on drain current and I would guess lower drain current will give better PSRR. But lower drain current also gives lower gm so the two are going to cancel to a first order. Could be why differences between FETs didn't much show up?

There's one feature of JFETs which MOSFETs can't manage and that's accurate DC offset tracking with matched parts. But then I don't mind using caps at all, with the impedance levels involved a 100nF C0G does the job nicely for coupling.

Here's what I'm doing in my latest TDA1387 incarnation - biassing up the I/V MOSFET to give lower input impedance. In this case the bias current is around 10mA so this gives about 250mS for 4ohms Zin. It needs a bit of tweaking in-circuit because there's no DC offset cancellation - the offset's given by the difference in currents in the two bipolar biassing sources.

ok cool

ill have to order some parts for that,
looks like the mosfets are only in mouser

I see them in Digikey too, just have to look in the correct section of FETs, not the 'RF' part.

Assume we are using single 1387 chip?
Not paralleled

Great, I have those parts & board mounted TDA1387s
I guess substituting a pot for the 10K R to gnd in the top CCS would allow for DC offset balancing?
As Nige asked - single 1387 - what PS voltage & how are you treating pin 7?

I'm using two paralleled in my most recent prototype. I was running without an LC filter, then added one to see what difference it made to the SQ and because the filter loses 6dB I added another DAC (tacked on underneath the PCB) to get back to the same output.

Supply voltage is 6V and that's coming from a single K339 MOSFET, fed with 7V (TL431) at its gate. The current sources are also coming from the same 6V (not 8V like it shows on the LTspice schematic). Running more current through the K339 gets its Zout down.

As for the 10k pot suggestion - it will only make a slight difference to the output current coz the output current is determined by the 56 (or 62) ohm Rs.

I see the absolute max voltage for 1387 from d/s is 6V - i'm thinking of 2 batteries to give 6.6V - have your tried this high a V?
So a pot substituted for 56 or 62ohm could adjust for V offset or what's the best approach?

I built a tda1387 at 6.6v
Works fine

nige2000 wrote: ↑Thu Jul 06, 2017 6:31 pm I built a tda1387 at 6.6v
Works fine

Right, thanks, Nige
Who's going down to Fran's this weekend for more assembly duties so I can give a headphone buffer to them to bring down?