While some of us are still waiting for the official BIII 8 Channel PRO, I decided to do open-heart surgery by transplanting another ES9038 chip into an old BIII board for an upgrade to an 8 channel implementation I did some time back.

I can confirm that Legatos still work superbly being driven by the ES9038 in an 8 channel topology. More importantly, the ES9038 is a MEGA UPGRADE to this topology - the old ES9018 sounded flat and lifeless in comparison to the punch and resolution of the ES9038 driving 8 discrete channels.

Better yet, the default settings of the ES9038 chip worked perfectly for this config using 8 channel I2S running up to 192KHz - and meant simply removing the old BIII firmware chip.

The case isn't the prettiest, but aesthetics were not the main concern as this unit is housed in an enclosed cupboard. The primary design criterion was sound quality, and to this end the build features independently powered AVCC and XO power supplies (thus the number of transformers and regulators). This makes a HUGE improvement to the sound quality over the default topology of running all tridents from the same filthy VD power source

Edited by user Wednesday, January 16, 2019 6:13:42 AM(UTC)  | Reason: Not specified

I do indeed have a couple of scopes (100Mhz and 200MHz), and have previously observed the hf gunk on the DAC outputs - which is why all my newer builds now incorporate hand-wound ferrite bead inductors on the DAC outputs (the new Mercury IV apparently has hf filters but the Legato has none).

I did this build originally for a friend sometime ago. It took several days to mod the DAC chip, install additional transformer and regulater, and change the SE gain of 4 legatos to match the ES9038 - so I didn't have any time ledft to make up 8 inductors. Maybe a mod I'll do later along with isolating the USB I2S inputs...

The newer regulators I use have Murata chokes on the AC in and these also appear to reduce DAC noise. In particular, I now use a dedicated regulator/transformer winding for the AVCC (and also clock) power supplies - which significantly reduces DAC noise on the analog outputs. This gives a clearly audible improvement The improvement is so substantial I now consider it to be mandatory (the second pic shows blue/white twisted wire providing the AVCC SR module with clean 5V - the Input pin that usually feeds the reg with filth from VD is disconnected). I highly recommend this mod. It provides a far bigger and obviously noticeable audible improvement over filtering on the DAC outputs alone as it stops a large amount of crap from entering the analog stage in the first place (the AVCC regs obviously cannot remove all the noise injected into the VD supply by the digital/control stages of the DAC, and its why, no doubt, the chip designers completely separated analog and digital gnds and power inputs).


If the build hadn't required SE outputs I would have cut the Legatos in half and stacked 4 halves each side for much shorter runs from the DAC - but that was not an option unfortunately. I hope TPA come up with a modular stackable single channel I/V design for these kind of builds (hint hint)

will measure and let you know...

Well, all I know is what I hear.

I am running some ES9028s in mono mode into a Legato and initially there was a pronounced digital 'whistling' noise audible if you put your ear right up to the driver.

The ferrites substantially reduced the 'whistling'. I heard no deleterious effects to either high, mid, or low outputs when ferrites were used. I only ever heard improvements.

I use DSP speaker driver correction, I noticed no change in re-sampling the associated drivers

Make of that what you will...

Agree - BTW I was using a pure sinewave generator as test input into the DAC and didn't notice any appreciable variation in amplitude as I swept thru the aubile frequency band whilst testing that the digital xover was working as expected, nor did I see any new spikes appearing when looking at a fourier xform on the scope (only spikes diminishing).

One interesting observation is that I originally noticed the increased noise when upgrading from a 9018 to a 9028 (mono mode) - the 9028 was noisier but it does not output more current I believe. I attributed the noise increase to the increased processing power of the newer chip.

The noise is not really audible listening thru headphones (mind you, the Legato headphone buffer circuit does contain an inductor) - its only audible thru the amp. The inductors reduced the noise down to the level usually expected from any amp. In this case, a bit of phased white noise you can just hear if sticking ear up to the tweeter. I've had this same noise from an amp even when the DAC is off. I attributed it to the input leads acting as aerials (as unplugging the leads made it all-but vanish).

So the ferrite inductors appear to be quite effective IMHO.

A large part of the DAC is essentially a computer performing real time signal processing. It's using more current because it has much more processing grunt than the 9018. More logic gates switching in the MHz band = more HF noise. Its not current per se, but current being switched at high speeds that generates the noise...

I bridged the 'In' on both sides with a wire underneath the AVCC module that is not visible in the photo

(both IN pins are disconnected from the main board of course)

I would also recommend this approach with the XO trident powering the clock as it makes a further audible improvement to sound quality.

Once you've done that you might as well power the main board with 3.3v and only use the 1.2v trident

Current draw will be more with shunt regs as they shunt current by their very nature (50ma?)

Double reg will theoretically be better depending upon your power source, so if you have the TP regs you might as well do this - otherwise I get stellar listening results from the DIYINHK LDO regs