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

[opc]

Hi Guys,

After having messed around for quite a while with several variations of Nelson's D1 with various modern DACs, I've finally come up with one that I think is worthy of being posted for everyone to use.

I've combined the excellent transconductance of a power mosfet, with the low distortion of a JFET buffer to get to the circuit below. It's basically a low impedance D1, with a different mosfet, and a B1 buffer in place of the old mosfet based buffer. You could probably call it a D1B1.

The circuit below provides about 2 VRMS at 0dBFS output when used with a ES9006 run in stereo mode. Performance is superb, and all the measurements are attached. THD+N is 0.000515% (-105.5dB) at 1kHz -8dBFS. This is without matched fets, and a seriously kludged home-made PCB, which means it can be made better with a decent implementation. Even the way it is, I think it sounds excellent.

This circuit could easily be adapted to work with any DAC, and provide almost any gain needed. The gain was set so low to minimize voltage swing at the DAC output.

The source of the mosfet sits at 1.65V which is what the DAC wants to see, and the drain sits at a little over 9 volts which allows direct coupling of the B1 input to the drain. Gain can easily be increased by increasing the value of R1, R2, R3, and R4 while maintaining the same ratio.

I'll be starting a PCB layout with this circuit and the ES9018 DAC which should provide some pretty incredible performance.

Anyone feel like helping with the digital side of the DAC?

Cheers,
Owen

[needsp]

Hi Owen

Now that hits the spot- many thanks indeed! Now all we need is for someone to produce a pcb. Though it looks so promising, I may try lashing it up on stripboard

Have you listened to it in comparison to Twisted Pear's IVY implementation on their Buffalo 32S?

Paul

Paul

[opc]

Hi Paul,

I've compared it to my Resolution Audio CD50, to the stock ESS reference board for the ES9006, as well as a reference AK4399 and a reference CS4398. I find it sounds better than all of them, and I also find the ESS DAC to be better than the AK or the CS parts. I don't have any of the Twisted Pear stuff here, so I can't compare it to that. Personally though, I don't like the use of op-amps where simple discrete circuitry can provide the same functionality with better performance. Just my opinion though.

I'm already halfway through a PCB layout, but I'm not sure how to handle the PSU options. Discrete PSU's for each channel will be used, but people seem fussy about what regulators they like. I personally would only insist that they be on the PCB itself, and not wired from separate boards. The circuit shown uses simple adjustable linear regs (LM337/317) to get the +/-18V, which if implemented correctly, work quite well. Any suggestions on a favorite regulator would be appreciated.

My PCB will also have the DAC and all PSU's onboard, but I can always spin a separate board with just the output stage if there's interest.

To answer your question from the other thread:

What's the best way in your opinion to use the circuit with single ended-input amps?

For SE DAC output applications, just half the circuit should work perfectly. You'll end up with a phase inversion, and higher 2nd order distortion, but the former is easy to deal with on the DAC, and the latter is par for the course with SE.

If you mean SE output from a balanced DAC, then you might need to try a few different things. You could send the balanced output to an op-amp/filter and get SE that way, or you could drive a transformer directly. The low impedance voltage output of this IV stage would be much more suitable for driving a 1:1 transformer (compared to directly off the DAC). If you had a 1:1:1:1 transformer you could get away with only one cap between the two series primary windings, and eliminate the two coupling caps at the outputs of the buffers.

There's also the possibility of just taking one of the two outputs of the balanced connection to ground, but I haven't tried it yet, and distortion might be high. I'll give it a shot and I'll let you know.

There are several other discrete options, and I'll take a look into it since it seems to be a popular request with other DACs.

Cheers,
Owen

[NicMac]

Originally Posted by opc

My PCB will also have the DAC and all PSU's onboard, but I can always spin a separate board with just the output stage if there's interest.

You certainly have my interest for a separate board with the output stage!
Nic

[needsp]

Hi Owen

All good stuff. I certainly wouldn't normally use op amps in my system- it's only just recently I accepted FETS as useful assistants to triodes! But I wanted to try the Sabre chip, and the Twisted Pear Buffalo 32S is how I "chose" to do it (actually Nic forced me into a corner with the logic of his argument.... :-) )
I may try the circuit with DIYParadises' Monica NOS DAC (which aleady uses a discrete I/V design). But, yes, I was mainly asking about getting a SE output from the circuit used wit he the Sabre 9018, and so will be very interested in your findings when you try taking one balanced output to ground (I had hoped that the way you'd drawn your circuit, with each current output half referenced to ground, meant you were confident of this approach working!)

I'm a great fan of Paul Hynes' regs BTW

Keep up the good work, and put me down for any pcb too

THanks

Paul

[jkeny]

Owen, I have an idea to make a modular design between your pcb & ACKO's Sabre DAC board - here's my post from the Buffalo tweakers thread in response to qusp wishing that the tweakers board had something to plug into it by which I assumed he meant an output stage:

I have an idea about that - basically to allow output boards to be plugged in from the top into the tweakers board using standard (low impedance) pin connectors i.e a 3D design. I reckon this would give the shortest path to the output stage (probably shorter than the Buff with on-board o/p stage?) & would be stable with 8-10 pins on each side PLUS would make it modular - just plug in another IV stage or whatever as long as it has the correct standard socket I'll be floating this one with Acko & Owen

What do you think?

[opc]

Hi jkeny,

That would probably be the best way to go if people want to be able to swap various stages. We would just need to come up with a standard pinout for the connector, and I think that should be optimized based on the DAC outputs. With good layout, you could probably keep those connections to less than half an inch which would be about as good as having it on the same board.

I'm guessing the standard 0.1" pitch inline connectors would be the best bet, but I'm open to suggestions. I have seem much more expensive and finer pitch connectors for video applications, but that might just complicate things for people.

Maybe another two connectors one either side of the IV stage could be used for swapping regs for the +/- 18V. I'll look into it.

I should have the layout mostly finished by the end of this weekend so I'll post a preliminary version for comments.

Cheers,
Owen

[ichiban]

If I may throw my 2 cents into this. I'm no EE
so maybe this isn't possible with the D1 half of
your design. However, would it be possible to
cobble the D1 with the "symmetric psu B1"
thus eliminating coupling caps? Also, builders
of this shunt regulated B1 seem to be unanimous
in their praise of it!

Building a symmetrical psu B1 buffer

[relder]

Originally Posted by ichiban

If I may throw my 2 cents into this. I'm no EE
so maybe this isn't possible with the D1 half of
your design. However, would it be possible to
cobble the D1 with the "symmetric psu B1"
thus eliminating coupling caps?

Given the DAC output is around ground potential you'll have DC above the MOSFET, you have to get rid of it, which is most easily done with a cap, sorry.

One of these days I plan on doing something similar (B1 output), but I'll likely attempt a lovoltech/qspeed JFET instead of a MOSFET - but need to finish other projects 1st. Nice work BTW

[relder]

Looking at this circuit again has me wondering: why are the filter caps, C7/C8 here, tied to the positive rail instead of ground? Seems like an opportunity for PS noise to creep in. I suppose something to do with the fact that the output signal is created across the resistor and therefore the filter cap should therefore be across the same nodes?

Any particular reason you loaded the output so much with 1K output resistors?