A New Take on the Classic Pass Labs D1 with an ESS Dac

[buzzforb]

GB list:
Parts GB
buzzforb - 1
Magsy - 1
vitalica - 2
Reenberg - 1
Emphrygian - 1
qusp - 1
Iancanada - 1
crazikid - 1
bigpandahk - 1
Rolle - 1
Coolhead - 1
woresseo - 1
Tehrkn - 1

[qusp]

I'm as PSU obsessed as the next guy (those who have seen my dac know thats an understatement) and for parts of the circuit that readily show improvement without dominating the enclosure, I generally go OTT with shunts, batteries, flea type regs (shunts being the only one that will stretch to this voltage), but I was actually surprised at how little the improvement was.

yes there was an improvement, but in a dac like mine, especially one thats about to jump to 100W for 4 channels, it needs to be a very large improvement to justify the space/heat. caps and zfoils dont take up extra space, or create heat.

i'll try it again, especially when a custom enclosure with external heatsinking comes into being, but for now i'll save the crazy PSUs for AVCC regs, Clock regs, DVDD regs for the dac itself, USB boards, fifo etc.

I will only do it at the full +/-45v because the improvement in measurements and sound boosting the voltage and thus gm of the fets is larger than the marginal improvement offered by the lower noise regs, the regs are not the limiting factor.

every build has trade-offs, different for everyone I guess.

[opc]

Excellent discussion above about the PSU overall.

analog_sa:

I think when you go to design a proper shunt reg, you'll see the rather monumental task of achieving something that can deliver +/- 45VDC at 250mA per channel.

There has to be a point in the chain where you stop with shunt regs and live with either standard voltage regs, or no voltage regulation at all, which I would guess is the case with your power amplifiers. It's just not practical to use them under these conditions unless power output is exceedingly low.

The higher operating voltage and constant current draw characteristics of the circuit itself make it less prone to PSU related issues, but just the same, I would imagine a very quiet supply would produce a measurable difference.

If I were you, I would look closely at the superb linear voltage regs from TI. They make a +/- 1A pair in the TPS7A series that could probably be implemented with zener diodes to increase the voltage input and output range up to 45V. The noise is extremely low, and dynamic performance is superb, even though it's of little consequence. This setup would provide a tangible improvement over the 317/337 combo without driving dissipation through the roof like shunt regs would.

Cheers,
Owen

[analog_sa]

Excellent points Owen, thanks. I'll see if any of my Jung regs could possibly be rigged for the task. Shunt is indeed silly.

[qusp]

so you think we can squeeze the full -120dB THD? -113 vs -120 is tangible? wont stop us trying for the record hey Owen?

i'm already planning that for other builds, already started playing with the PCB for them, those regs are superb, ordered a cheap rework station for soldering them and still thinking about heatsink, probably just some bga types... and the PCB of course

can you be a little more explicit about the zener protection? you mean just to protect them while the caps are charging yes? just like with the LM series, bypassing Vin to Vout? with these the sense and Vset are separate, but the sense connection shouldnt care about the voltage should it? or perhaps you mean to elevate the gnd pin above ground?

[opc]

Originally Posted by qusp

or perhaps you mean to elevate the gnd pin above ground?

That would be the one! It's an old trick that works well with the 317/337, and I would imagine it would work with the TPS series as well, but it would have to be tested. The zener should be bypassed with a small 0.1uF and a larger 100uF to reduce the noise across the zener, which will show up across the reg output if you're not careful. A 24V zener should do the trick.

Alternatively, you could try and use a higher voltage pass transistor in a circuit similar to that of a 317/337 for HV use.

As for THD+N, the only improvement would obviously be the noise floor, which is already quite low. You might irk out another 1dB, but I get the feeling the third harmonic will still be the dominant contributor which might leave the THD+N numbers right where they were.

Cheers,
Owen

[qusp]

ahh yes sorry I forgot to type the all important N, I meant THD+N ie. the datasheet ESS THD+N spec thats possible with a great deal of futzing with expensive opamps. (ha futzing is a real word)

possibly 1db? awesome! OK .i'm on it

another (perhaps more important) bonus will be the low profile, it could probably even fit under the IV board with lots of thermal vias to bring the heat to the bottom copper and just stick it to the same heatsink.


OK so elevate the GND pin above ground to 24V the zener obviously has to be rated for the >1A current. then we set the Vout for ~20v and sensing works as normal?

[analog_sa]

Originally Posted by qusp

the zener obviously has to be rated for the >1A current.

Uhm, why? It's only the ground terminal current passing through the zener, which seems to be in the region of a few hundred uA for the 317 iirc. It's only for the internal reference.

The series resistor from the output would set sufficient current through the zener, but no need to exceed 20-30mA

[qusp]

we arent talking about the 317, we are talking about the tps7a3301/tps7a47, which actually have 'real' GND pins unlike the lm317/337. I presumed opc was meaning to actually elevate the whole reg above ground by 24v with the zener, effectively putting the reg 'in series' with a 24v zener, meaning it has to cope with the return currents and the reg is rated for 1A, of course we arent using it all so could be somewhat less.

maybe i'm not thinking this through properly here, its 3.3AM here, but its a distinctly different case when we are talking about the actual ground pin, not a 3 pin reg

[opc]

analog_sa is correct... the zener only needs to be rated for a few mA.

All the current is still returned via the true circuit GND, and the old GND reference is simply elevated by 24V but only passes the operating current of the TPS part which is less than 1mA.

You might have been thinking in terms of a shunt reg which would have to be capable of passing the entire operating current and the overhead to GND at any given time.

The zener will need to be forward biased somehow, and it might be best to do that with a CRC filter off the input rather than the output to ensure the input to GND voltage on the regulator is not exceeded on startup. It would be quieter to take it from the output, but you could end up exceeding the regulator input voltage if the output ramps up too slowly and doesn't forward bias the zener in time.

I might give some of these options a try for fun. I'll report back if I get around to it in the next little while. A good pass transistor implementation might be more robust overall, and would alleviate some of the dissipation concerns with the quirky TPS thermal implementation. Both might be worth a shot.

Cheers,
Owen