Introducing the Buffalo III-SE-Pro 9028/9038

[stew1234]

I re-flashed the Amanero with current firmware and have gotten it working, but am getting the roaming intermittent static some have reported playing DSD 64. I can't get anything higher to play without losing lock.

Separately, I have been comparing the 9038/mercury to a 9028/mercury or legato. The 9028, even with the same mercury output stage, has better bass while the 9038 sounds thin. I'm not sure what's going on but have been trying to get the 9038 to sound "right" for a while now. DSD playback does sound better, but that's not an option for a lot of streaming and I wonder if there's some playback issue I need to address with the 9038.

[francolargo]

The 9028, even with the same mercury output stage, has better bass while the 9038 sounds thin.

Same resistors in Mercury for both 9028 and 38? If so, it could be an effect of whatever control you use to equalize the volumes - the 38 being presumably 4X more attenuated.

I'm not sure what's going on but have been trying to get the 9038 to sound "right" for a while now

I've found that some uFl cables are not good after too many installations/removals. Pure speculation: perhaps switch to new ones and as short as possible. Or as a test direct solder very short wires into the B3... I've started using a little edge modification to the B3s that employs gold pins. Although they are soldered onto the B3, they can be disassembled and re-assembled to the I2S/DSD source more often than uFl wires with little ill effect.

Good luck!

[marcelooms]

Did you change the resistors when switching to the SE9038?
They should be, the 38 puts out 4 times as much current as the 28!

Originally Posted by stew1234

I re-flashed the Amanero with current firmware and have gotten it working, but am getting the roaming intermittent static some have reported playing DSD 64. I can't get anything higher to play without losing lock.

Separately, I have been comparing the 9038/mercury to a 9028/mercury or legato. The 9028, even with the same mercury output stage, has better bass while the 9038 sounds thin. I'm not sure what's going on but have been trying to get the 9038 to sound "right" for a while now. DSD playback does sound better, but that's not an option for a lot of streaming and I wonder if there's some playback issue I need to address with the 9038.

[stew1234]

I initially set the mercury up for use with the 9038, so when switching to the 9028 I left the resistor and the output level was lower but still plenty of volume with my amp/speakers.

I have had issues with UFl cables before and am planning to try direct soldered silver but that still wouldn't explain the improvement in sound when going from the 9038 to 9028.

I am using a shunt power supply on the DAC and thought it may have been limiting current, but I get the same issue when using a linear supply. I have a vague sense that something may not be being powered fully/properly with the 9038 but given that any issues go away with the exact same set-up and swapping in the 9028 I'm at a bit of a loss.

[francolargo]

what's the peak-peak voltage coming from the 9038-Mercury at 0dB with a sine wave test signal? Are you using the onboard firmware?

[marcelooms]

Agreed, the chips should not sound very different. Even the es9018 does not sound very different from the 38 imo. So power issues may indeed be the issue, the 3,3V analog can draw lots of current. Are you using the latest trident moidules?

Originally Posted by stew1234

I initially set the mercury up for use with the 9038, so when switching to the 9028 I left the resistor and the output level was lower but still plenty of volume with my amp/speakers.

I have had issues with UFl cables before and am planning to try direct soldered silver but that still wouldn't explain the improvement in sound when going from the 9038 to 9028.

I am using a shunt power supply on the DAC and thought it may have been limiting current, but I get the same issue when using a linear supply. I have a vague sense that something may not be being powered fully/properly with the 9038 but given that any issues go away with the exact same set-up and swapping in the 9028 I'm at a bit of a loss.

[stew1234]

What points would I measure to test peak to peak voltage? + and gnd on se or balanced out?

I was using them, but have been using Sparkos regulators providing 3.6v. They have plenty of current and are fed by shunts set at about 300ma.

Earlier this year the DAC was working great until I accidentally swapped the VDD trident and XO trident trying to get sync mode to work. After I realized I mixed them up, I swapped them again but wasn't getting lock. I measured 345mv between gnd and out of the VDD regulator insetad of 1.3v. I ordered a new 1.3v trident and got sound again.

I don't think this could have done anything if I am still getting sound, but that's the only obvious issue I can think of.

When I was swapping in new power supplies I was getting good bass, but somewhere along the way it started sounding thin. I thought that was just the sound, until I had the 9028 to compare it to again.

[barrows]

Originally Posted by stew1234

I re-flashed the Amanero with current firmware and have gotten it working, but am getting the roaming intermittent static some have reported playing DSD 64. I can't get anything higher to play without losing lock.

Separately, I have been comparing the 9038/mercury to a 9028/mercury or legato. The 9028, even with the same mercury output stage, has better bass while the 9038 sounds thin. I'm not sure what's going on but have been trying to get the 9038 to sound "right" for a while now. DSD playback does sound better, but that's not an option for a lot of streaming and I wonder if there's some playback issue I need to address with the 9038.

My build works perfectly with DSD 256 via Amanero/Cronus-

Buffalo 9038 PRO-Mercury

Amanero/Cronus with 2006be 11 firmware, with Native DSD 256 served up from a Sonore Renderer running its proprietary, linux based OS, Cronus with a 45.1584, NDK SDA aeries XO, with .01µF bypass cap (C0G) on the Rhea board

I sync clock the 9038 (master clock from the Cronus, with clock regulator removed on the Buffalos' DAC board to de-power the onboard oscillator).

The 9038 DAC requires more current than 9028/9018, so make sure the power supplies are stout. Also make sure you are using the newer Trident versions, especially the 1.3 volt version (not the older 1.2 V model).

I have no problems using u.fl, just keep them to 10 cm or less, and do not reuse many times.

This set up sounds fantastic for me, pretty much as good as one can get with any ESS based DAC. I do have a few parts mods on the Mercury, but doubt that those make any huge differences in sound quality. (Z-foils in the feedback resistors, Amtrans film/foils for the filter caps).

I only send DSD 256 to the DAC, via oversampling either with Roon or HQPlayer, and much prefer this to any PCM rate.

[stew1234]

Originally Posted by Paul M

Reviving a dormant thread:

I'm wondering if it would be possible to supply both AVCC on the 9028 with a completely separate supply. I'm thinking either a 3.3V LiFePO4 cell for both, or 1 cell for each AVCC L and R.

Has anyone tried this? Or tried powering the AVCC pins of the 9028/9038 independently from the rest of the BIII board?

Looking through old posts here, yes I have been doing this. Possum has posted a few times on the TPA website about doing this also. I've tried feeding them 3.3v directly and also feeding the tridents with their own external power. Feeding the AVCC tridents with external power was a big improvement. I'm currently feeding two Sparkos regulators that output 3.6v with two separate ultrabibs.

I plan to ultimately have one toroid power the left side of the i/v stage and AVCC L and another to power the right side.

[francolargo]

Originally Posted by stew1234

I'm currently feeding two Sparkos regulators that output 3.6v with two separate ultrabibs.

I plan to ultimately have one toroid power the left side of the i/v stage and AVCC L and another to power the right side.

I'm curious about how the level of AVCC supply 'separation' affects the sonic benefit. IIRC @Barrows added a dedicated regulator (or two?) on a shared transformer, which is obviously easier than adding another AC source. How does that measure?

In the rigs I'm tweaking now, the 'clean' linear supply at the Buff3 terminals (from a TPA Centaur) is pretty quiet except for about +/- 10 millivolts at 100 megahertz. ...and that's with the Buff3 clock un-powered. If I take that 5v and power an LT3045 set up for 3.3v, run it through a twisted pair over to the B3 clock trident socket, then finally run it through a mini passive filter with ferrite (pictured), *that power source* is quieter than what otherwise comes from the 3.3v trident output. (My old scope isn't good enough to say exactly how much quieter.). What I would love is to accomplish some of any benefit of a fully separate power circuit to the AVCC only with whatever might fit on a small active 'pre-filter'. If the Sparkos discrete regulator isn't really optimal, then the idea may not be worth pursuing. I think I read somewhere that general current and impedance considerations for amp supplies (in addition to noise) also apply to the AVCC supply. For me, adding a separate transformer is probably never gonna happen but ironing out all but about +/- 1uv of ripple over most of the audio band seems quite feasible.

Frank

PS - the reason for the external supply to the B3 on-board clock is that it is switched to remotely enable all DAC modes for various input sources. SPDIF uses the onboard clock, other sources an external MCLK.