E1DA Cosmos ADC

[IVX]

That last thing is interesting, I had already suspected that ESS introduce deliberate DC offset to "fix something", so basically a hack. The ADC cannot seem to handle a true zero signal properly and that's sort of weird, I mean the offset introduced is so small it could easily be swamped by the driver's offset.

Anyway, this ADC, like any other D/S with switched capacitor input, wants to see very low source impedance and therefore be DC-coupled to the driver.
As for that driver, I've always wondered why feedback isn't used around the glitch capacitor (1nF in the ESS datasheet circuit), that way very low output impedance could be held throughout the whole audio range:
View attachment 386045
This -- or something very similar -- is what I'm going to build, in an ESS DAC+ADC combo. The filter frequencies are lowered as flat bandwidth doesn't need to be higher than 30kHz or so.
The application calls for DC coupled input and I hope it's not causing any troubles...

1611 vs 1632 10db voltage noise difference at 20Hz..
PS: and one more potential problem, I have a few cases when 220uF input alcaps were blown up due to overvoltage(typically because of DC bias of single rail class D amp). Users just replaced 220uF with brand new and the ADC recovered completely. If you have no such caps at all, the opamp will be under risk, and probably ES9822 as well.

 

[KSTR]

1611 vs 1632 10db voltage noise difference at 20Hz..

One of the reasons the 1633 came out ;-)

PS: and one more potential problem, I have a few cases when 220uF input alcaps were blown up due to overvoltage(typically because of DC bias of single rail class D amp). Users just replaced 220uF with brand new and the ADC recovered completely. If you have no such caps at all, the opamp will be under risk, and probably ES9822 as well.

Input protection / current limiting (also for power-off state) should be always present but to make it robust and low-noise is not trivial, and obviously a DC-blocker cap helps to reduce problems and it's a cost effective and simple solution which I find personally totally OK for the Cosmos ADC.

 

[IVX]

One of the reasons the 1633 came out ;-)

oh, it's much better, I'll try that next week I guess.
PS: 1633 has a quite low Aol, hence, not too good PSRR, >30 times worse than 1612 at DC, who knows what is there at 3-10k. I need to get the spice model.

 

[KSTR]

No Spice model for 1633 at the moment, sadly.
But the Aol is high (90dB) and probably constant through most of the audio range, thus PSRR might follow.
Looking at a very similar new chip from that family, the THS2630 (also 90dB Aol), both CMRR and PSRR are good (90dB) up to 1MHz.
It could well be that 1633's PSRR is actually slightly better than 1612 above 3kHz.

 

[IVX]

No Spice model for 1633 at the moment, sadly.
But the Aol is high (90dB) and probably constant through most of the audio range, thus PSRR might follow.
Looking at a very similar new chip from that family, the THS2630 (also 90dB Aol), both CMRR and PSRR are good (90dB) up to 1MHz.
It could well be that 1633's PSRR is actually slightly better than 1612 above 3kHz.

I have extracted .subckt OPA1633 from Tina's file but MC12 won't run that without total debugging.

 

[KSTR]

@IVX, thanks for the model. I'll give it a try in LTspice 24 and see if it works...

 

[IVX]

Finally, I deleted some redundant spacebars and MC12 stopped erroring, however, the model isn't simulating correctly at all.

 

[JensH]

Perhaps you should try Tina TI?
OPA1633 simulates fine there. I normally use Tine TI for simulation anyway.

It will be interesting to see the performance of the OPA1633 compared to the OPA1612.
From the datasheet it looks like the distortion is higher for the OPA1633, but perhaps that is down to different conditions? And perhaps the OPA1633 is still "good enough"?

 

[IVX]

yes, Aol 1633 the same 1612 at 1k, a bit higher at 3k and better at 10k etc, so no prob at all.

 

[KSTR]

Tried to fix the TINA-TI model to work with LTspice, but no luck. There were some illegal characters, netlist order was also different (to 1632) but there are syntax issues I could not resolve.

I'm reluctant to install yet another simulator....

 

[JensH]

I fully understand.
I have made some simulations in Tina TI and put them in the attached document.
In terms of frequency response and noise the OPA1633 looks promising. Hopefully the distortion is also competitive.

 

[KSTR]

I have made some simulations in Tina TI and put them in the attached document.
In terms of frequency response and noise the OPA1633 looks promising. Hopefully the distortion is also competitive.

Thank you for this.
One advantage of the true FDA 1633 is that it is easier to include the breakout resistors and the load capacitance in the DC and AF feedback.
With the simple opAmp FDA emulation, one needs two separate feedback paths, that is, a second one for HF to the inverter's input node.
The advantage is that you are not chained to a single-source specific chip.

 

[Heinzfi]

Just a heads up but the new TI’s ISOUSB211 USB isolator by JDS Labs might interest some non-ISO users.

Introducing JDS Labs Synapse – Next Gen USB Isolator – JDS Labs Blog

blog.jdslabs.com

 

[somebodyelse]

Just a heads up but the new TI’s ISOUSB211 USB isolator by JDS Labs might interest some non-ISO users.

Introducing JDS Labs Synapse – Next Gen USB Isolator – JDS Labs Blog

blog.jdslabs.com

100mA current limit might be an issue.
https://jdslabs.com/product/synapse/

 

[Sokel]

100mA current limit might be an issue.
https://jdslabs.com/product/synapse/

It probably rules out most hi sample rate USB powered DACs and even worst interfaces.
A universal one should at least meet the 500mA USB standard.

 

[somebodyelse]

It probably rules out most hi sample rate USB powered DACs and even worst interfaces.
A universal one should at least meet the 500mA USB standard.

Or allow for external power on the device side. Bonus points for supporting PD.

 

[KSTR]

100mA max is quite useless. External supply make the whole idea of proper isolation moot.
With the SN6505 they are using and the proper transformer one could have easily achieved 500mA with 1A peak capacity, together with USB-C controllers.

I don't see anything that suggests they handle USB-PD. The 211 chip sure does not and I don't see anything else crossing the isolation barrier that would handle the CC signals.

The 100mA limit may come from the USB2.0 max rating for the non-initialized state of a connection, 500mA is only allowed once the connection is established.
EDIT: And I haven't found any TPS7Axxx regulator in SOT23-5 that can handle more than 300mA and many are 200mA only, so maybe that's the bottleneck.

 

[IVX]

it is USB and you have to give 500mA at least.

 

[JensH]

oh, it's much better, I'll try that next week I guess.
PS: 1633 has a quite low Aol, hence, not too good PSRR, >30 times worse than 1612 at DC, who knows what is there at 3-10k. I need to get the spice model.

Did you get a chance to test the OPA1633?
And if so, what were the results?

 

[IVX]

Did you get a chance to test the OPA1633?
And if so, what were the results?

not yet.