ESS THD ‘Hump’ Investigation

[syn08]

shame on ESS for their non-communication/non-information policy on this blatant issue... now AKM also doesn't have everything covered in their datasheets but still much better than ESS.

As an amateur in audio, I share every bit of frustration with ESS policies regarding information. From a business perspective though, I can understand ESS considering support as a potential revenue stream. We do not know their policies on support, perhaps they are asking their customers for X dollars per month, or year, or per chip, etc... in support fees, something that scales with the chip sales? That wouldn't be unheard of, and it only comes against the business/support model, as we know it, of the big semi companies like TI and AD. Supporting their products with public information would certainly kill this revenue stream, and given the minuscule market they are targeting, probably they are penny pinching.

What I find inexcusable is the fact they don't let everybody know about these policies. Their distributor, at least on this side of the pond (don't know about Ismosys in Europe) behave like aholes, shutting down any communication channel, as soon as they find out you are not going to buy the chips in quantities, essentially telling you a "fuck off, your money is of no good here", instead of sending what could be for example a boiler plate message, outlining the sales/support policies they have to follow.

 

[mansr]

Cirrus sure, but i haven't seen much high performing stuff using TI chips, if any. Would be interesting to see someone like topping try their hand at a TI based design, like they did on the D30Pro with Cirrus.

A PCM1972A or PCM1794A based design can easily exceed 100 dB SINAD. If that's not adequate, you must be Rob Watts.

 

[pma]

Nice connectors, @syn08

 

[Billy Budapest]

Cirrus sure, but i haven't seen much high performing stuff using TI chips, if any. Would be interesting to see someone like topping try their hand at a TI based design, like they did on the D30Pro with Cirrus.

The TI stuff is sold under their trade name Burr Brown.

 

[syn08]

Nice connectors, @syn08

Chinese stuff, pennies a pop. I wouldn't bet on them for anything over 1GHz, but perfectly adequate for audio.

 

[syn08]

A PCM1972A or PCM1794A based design can easily exceed 100 dB SINAD. If that's not adequate, you must be Rob Watts.

Adequate for audio reproduction, no doubt about that. Not so much for instrumentation.

 

[chris719]

Cirrus sure, but i haven't seen much high performing stuff using TI chips, if any. Would be interesting to see someone like topping try their hand at a TI based design, like they did on the D30Pro with Cirrus.

PCM1792A/4A was popular for a while and arguably the best performing DAC chip until ES9008 came out. It’s just old, that’s all. The current outputs make it a bit more costly than an AK4490 or 93 to implement.

 

[mansr]

Adequate for audio reproduction, no doubt about that. Not so much for instrumentation.

Take a look at the name of this website. If I'm not mistaken, it includes the word audio.

 

[Veri]

PCM1792A/4A was popular for a while and arguably the best performing DAC chip until ES9008 came out. It’s just old, that’s all. The current outputs make it a bit more costly than an AK4490 or 93 to implement.

I'm pretty sure it's just a bit more of a pain to implement, the PCM1794A is. Still a good part though, but also not exactly cheap. Especially when you want to use them in dual mono mode to maximise performance.

 

[chris719]

I'm pretty sure it's just a bit more of a pain to implement, the PCM1794A is. Still a good part though, but also not exactly cheap. Especially when you want to use them in dual mono mode to maximise performance.

Yes, the I/V converter is extra components. They are more distortion limited, I think, considering you can reach 127 dB DNR with only one (on paper). Still pretty good.

 

[chris719]

Take a look at the name of this website. If I'm not mistaken, it includes the word audio.

Well, measurements are a hot topic here generally, are they not?

 

[mansr]

I'm pretty sure it's just a bit more of a pain to implement, the PCM1794A is.

Compared to what? Most of the competitors can't be used without a microcontroller.

Still a good part though, but also not exactly cheap.

It's $7. AK4497EQ is $48. Pricing for ESS isn't available.

 

[Veri]

Compared to what? Most of the competitors can't be used without a microcontroller.

Compared to what's "popular" nowadays. We're not seeing that many recent devices with burr brown DAC chip parts do we? Then again maybe ASR paints a skewed picture, I'm not really sure.

 

[syn08]

It's $7. AK4497EQ is $48. Pricing for ESS isn't available.

ES9038PRO is $76, available in stock @Mouser. ES9028PRO is $60, ES9018S (aka PRO) is $50, all in singles.

I've read several times the name of this website, and while it says "audio" it is not "Cheap Audio Science Review".

 

[pma]

In fact it very often is.

 

[chris719]

Compared to what? Most of the competitors can't be used without a microcontroller.

I think we've moved well past the days where audio designers were unable to proceed if the product needed I2C or SPI. Bottom line, the PCM1792A doesn't compete with ES9038PRO or AK4499, it's closer to the voltage out AK449x and CS4398 in performance and requires more external components than those do, which is why it never saw adoption in the pro audio industry. It's easier to screw up and not reach the datasheet specs.

It also only officially supports 192 kHz sample rate which is marketing suicide these days.

 

[chris719]

ES9038PRO is $76, available in stock @Mouser. ES9028PRO is $60, ES9018S (aka PRO) is $50, all in singles.

I've read several times the name of this website, and while it says "audio" it is not "Cheap Audio Science Review".

My biggest surprise is that there are no industrial DACs that can match this performance at any price. Without a ton of work on a custom deglitcher, anyway. I suppose that speaks to non-audio application requirements more than anything else.

 

[syn08]

Attached is the MOD IMD distortions plot. As expected, since harmonic distortions and IMD are mathematically related, no suprise: hump is gone. Signal is the standard 60Hz+7KHz, 4:1.

BTW, I've noted that for Vcm under 0.6V, the ES9038PRO distortions increase significantly (overall, not as a "hump"). Since I don't like to be that close to this lower limit with Vcm=0.68V, I've settled for Vcm=0.82V. Not much of a difference in distortions (perhaps 0.5dB worse) and still no trace of hump. Coincidentally, 0.82V is almost exactly AVCC/4.

 

[syn08]

PCM1792A doesn't compete with ES9038PRO or AK4499, it's closer to the voltage out AK449x and CS4398 in performance

Now I got the itch to compare the ES9038PRO with the AK4499. Unfortunately, the scratch is for 2022, if ever... I have the AKM demo board, but it would not be an apple to apple comparison, too many things are implemented differently.

 

[chris719]

I recall a very old post @diya from Dustin Foreman, who I believe was the design lead on the original Sabre, if not others. The implication was that the current output was almost an afterthought; the design was a voltage source and they merely observed cancellation of some distortion components when operating into a virtual ground.