Buffers for ADC design with AK5394A and other ADC's ?

Status
Not open for further replies.
Jump to Latest
#43
1audio said:
Here is some of the models. I have no experience with Tina so i don't know how to "port" the models but they are all pretty generic. The Supertex models came from Supertex. Your circuit looks ok. I used depletion mode mosfets from Supertex for the cascode which work well. The source resistors you are using see high. It gives a wider stability margin but higher noise and lower gain on the first stage. All tradeoffs to be managed.

Distortion measurements in Spice are not worth much once the circuit is linear.

I need to find time to breadboard some of these circuits.
Click to expand...

You have to open the models in Tina and then copy the text. Save it to a text file with what ever name you want with a sub extension then associate an LT symbol for the model.

The Tina libraries are stored in a compressed format that's why they need this treatment to use them in LT.
 
#44
Using an opamp for the reference voltage could limit the noise to that possible with an opamp. Getting a really low source impedance from DC to MHz is important in this application since it is both the reference and it get charges zapped from it with every sample. I would be looking at an LT1115 for this location (expensive but one of the best options).

The Muses volume control is quite good but has its limits. It also needs to be AC coupled (the biggest performance constraint in the NuForce P20 project). Relays and resistors are better but more involved to implement. For a measurement system I would use relays and a differential input. There are several examples I can pull from my collection for ideas.

What is the real noise floor of the ESS ADC? We need to translate that into nV/rtHz to optimize the rest of the system. And see if it is an improvement over the AKM.

I'll try to get into the ESS room at CES and find out what they can share.
 
#47
I tested the 20k pot with 2Volts offset at the input and also another test with 1,65V offset on the output. That is not a condition in any normal audio application.
You can hear a click when changing the volume settings, but the muses works.
So I decide to kick out the capacitors.
 
#49
It seems like the noise floor of the ESS ADC is lower than the AKM, but only by 1 dB according to the specifications.

What I think is more important is that the distortion specification is a lot lower. So it would definitely be interesting to check the distortion of both. The distortion of the ES9102 seems to be much lower than the distortion of other good ADC's, including the AK5394A.

For noise measurements with a low (or no) signal the noise floor of the ADC is not terribly important. As long as the input amplifier has a low noise floor and some gain, it is still possible to measure very low noise levels.
 
#50
Here is a distortion plot I just made with the AKD5394A board. I upgraded the opamps in this channel to LME49710's. When I tie the two channels together the distortion rises to the level of the JRC5534's in the other channel.

The distortion is really quite low (probably in the area where you can't be certain of its real). Important- must use 12.288 MHz for the master clock. The distortion goes up with a 24MHz master clock. The input circuit is really very simple.

For what its worth I was using one of these really crude supplies for this measurement: JE215: JAMECO KITPRO: Education & Hobby Kits . Its not very good with a fair amount of ripple but I still get -130 dB hum. Same with a switcher.
 

Attachments

  • AK5394A distortion.PNG
    AK5394A distortion.PNG
    23.8 KB · Views: 471
#51
Thanks for providing this very useful information. Good to know that I should use a 12.288 MHz clock.

Did you use the balanced or the single ended input?

When you tie the two channels together, how do you do that? Just the inputs?

Do you know how much of the distortion comes from the generator?
 
#52
Hi all ...

In case one/some of you might be interested in using one of the esstech converters at a higher sampling rate (384 kHz) there's this USB to I2S card available:

USB Audio - Rigisystems

On my computer it works quite stable just receiving data (win7 ULT, newer Intel MB + CPU, SSD HD), although I haven't yet come around to connect an ADC.

Regards,

Jesper
 
#53
I was measuring using the single ended input (adds an inverter to drive the negative input). I just bridged the two channels with a BNC TEE. I'll swap out the IC's on the other channel later this week when more samples arrive.

The Rigisystems stuff works (its used in some DAC's and ADC's) but its Linux support is not good. 384 KHz may not be useful if the supporting software doesn't exist. It takes more than just adding some options to get higher sample rate analysis to work. Verification that its not getting hammered by system issues, especially going through system services, is important.
 
#54
Verification that its not getting hammered by system issues, especially going through system services, is important.
Click to expand...
... Yups, although I haven't tried it with 384 kHz I have experienced (what to me was) tricky setup issues at 192 kHz.

I personally won't be using the USBPAL for an audio analyzer but mentioned it just in case someone here knew how to use 384 kHz analyzer-software-wise and would like 384 kHz sampling.

... and then something completely different ... I just noticed a Christmas atmosphere in town today - clearly, for the first time ... am inspired to pass on a Merry Christmas to you all 😉 ... and also a thanks for much, much help from so many of you with (all) the topics I don't know about, or don't (yet?) know well enough ... :yes: 🙂

Greetings,

Jesper
 
#55
gentlevoice said:
... it just in case someone here knew how to use 384 kHz analyzer-software-wise and would like 384 kHz sampling
Click to expand...

The limitation is not the SW it's HW dependent. I just let any custom sample rates up to 1MHz using the sound card interface but could easy go to the 32 bit value limitation.

Well, ARDA tech just have 384kHz ADC/DAC chips and the performance looks promising 🙂

Currently I have not seen any digital I/O using 384kHz... as in the old days left & right channels where used to double the speed. I have build this already in my SW...

HpW
 
#56
Did you mean ESS? ARDA still only lists the ADC and I don't know of anyone who has put it in a product. ESS has some solutions. Still no commercial products I know of with their ADC. Certainly nothing to offset the costs of the masks yet. Its possible to get 384 through the XMOS interface both ways, on paper. Can you get the Windows sound layer to work at those sample rates?

However 384 is old news. Version 2 of HDMI supports 1536 KHz sample rates and 32 channels. No one in the alliance has any idea why, however, but they hope the big numbers will win sales, somewhere.
 
#58
Its good to see that they got a design win. However Ayre would be hard pressed to use a full wafers worth of chips. I hope they get other adopters.

Here is an IM measurement I just made with the AKD5394A. I'm using two generators mixing at the output with the peak voltage at the effective full scale on the AKD5394A input (4.5V RMS equivalent).

The levels of each tone (19KHz and 20KHz) are approx -8 dB. the 1KHz IM product measures -134 dB. This is using the LME49710's. Using the JRC5534's I get -126 dB for the IM product. In either case that is performance close to what can be had with the best of analog equipment.

However the input circuit is not too flexible in this setup. now that there is a baseline its time to look at and test options.
 

Attachments

  • 5394A IM test.PNG
    5394A IM test.PNG
    67.3 KB · Views: 389
#59
Demian - are these numbers substantiated? I can measure IM to -130dB (spec'ed) with the Panasonic if you need a back-up confidence test.

If this ADC works out for the DIY'er as a reasonably priced T&M instrument, it will be fantastic.

-Richard
 
Status
Not open for further replies.
Top Bottom