Author Topic: decent but affordable 16 bit DAC  (Read 6221 times)

0 Members and 1 Guest are viewing this topic.

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: decent but affordable 16 bit DAC
« Reply #25 on: June 11, 2019, 08:38:04 pm »
The tricky implementations of gain switching compensate for or are invariant of the series resistance of the switch.

https://www.analog.com/media/en/training-seminars/tutorials/MT-072.pdf

That is why I recommended different fets. Higher Rdson means lower capacitance and better transients, etc. but can introduce some errors. Higher capacitance has the opposite effect, therefore it's a tradeoff. Anyhow if the resistors are chosen to be large enough, often the  Rds is orders of magnitude better than the resistor tolerance itself, making it a non issue  ;D

I am just point out that with a slightly different circuit configuration, FET resistance becomes irrelevant because only the operational amplfier's input bias current passes through it.  Now JFETs or integrated CMOS multiplexers can be used without consideration of series resistance.
 

Offline OM222O

  • Frequent Contributor
  • **
  • Posts: 768
  • Country: gb
Re: decent but affordable 16 bit DAC
« Reply #26 on: June 11, 2019, 09:33:59 pm »
Interesting ... Can you please provide a schematic for that configuration? So far I have only came across the example which I provided  ???
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: decent but affordable 16 bit DAC
« Reply #27 on: June 12, 2019, 09:01:48 pm »
Interesting ... Can you please provide a schematic for that configuration? So far I have only came across the example which I provided  ???

Figure 7 on PDF page 5 of the document I linked as shown below shows the basic idea:

https://www.analog.com/media/en/training-seminars/tutorials/MT-072.pdf

Note that separate dividers in parallel with the output could also be used instead of one stacked divider.  Input resistance matching to remove the contribution of input bias current to input offset voltage can be done by adding a compensating resistance in series with each switch if necessary; the AD797 shown uses input bias current cancellation so this was not shown here but it might be done with a 741 or 324.  A low input bias current part would not require this either.
 

Offline OM222O

  • Frequent Contributor
  • **
  • Posts: 768
  • Country: gb
Re: decent but affordable 16 bit DAC
« Reply #28 on: June 12, 2019, 09:43:12 pm »
Sorry but I can't see how this is different from the schematic that I provided. All the drains are commoned and will go to the inverting input (R+ net), the gates are the "TTL gain control" and each source goes to one resistor in the chain. My idea used a fixed resistor in the feedback path as well, but now that I've seen that diagram, I realized it was just an extra part and you could remove it alltogether  :-+ please correct me if I'm wrong and the diagram in AD app notes is different from my schematic.

Edit: in hindsight I only provided the schematic for the range switching and not the entire circuit as I didn't have time and that schematic was already abailable to me quickly. Maybe that's where the confusion comes from. If so, my bad and I will create full schematics for the specific implementation rather than using words to describe it from now on ;D
« Last Edit: June 12, 2019, 09:46:40 pm by OM222O »
 

Offline nemail2Topic starter

  • Regular Contributor
  • *
  • Posts: 206
  • Country: at
Re: decent but affordable 16 bit DAC
« Reply #29 on: June 13, 2019, 08:04:34 am »
Hi

sorry I'm on vacation currently with (gladly  :-DD) limited access to a PC and writing stuff on the phone is just... meeeeeeeeeeeeeh :)
The parts choices seem reasonable  :-+
But that layout though   :-DD no offense but it seems very willy nilly without any rhyme or reason. Usually the layout follows the schematic, i.e. you have sections that do specific things, rather than just running traces all over the shop. What software are you using for the pcb design? If it's easy eda I can help you clean it up. Otherwise I can upload some pictures as hints. It would be beet to seperate your analog and digital supplies btw. LDOs are so cheap these days that there isn't an excuse for not doing so. You also won't need the ferrite beeds if you have a clean supply to start with  :-/O
I appreciate your feedback and I knew that the layout is not ideal. I wasn't aware that it is THAT horrible, though :)

The lower part is digital stuff, display, MCU power supply, relais switching, pwm fan control, etc.
Then there are big fat power traces for the series pass elements and to the output connectors and in between from left to right) there is the current measurement circuitry, the CC/CV opamp control loop and the onboard voltage measurement as well as the external voltage sense resistors.
I do have quite many LDOs and Vregs on the board, especially in the analog part I didn't go tight and did rather put in another LDO than going all over the board with a trace.

when crossing traces, I tried to only do it in 90 degrees and I tried to separate digital and analog stuff as much as possible and I tried to keep traces as short as possible, especially for the critical (analog) parts.

So.. I thought it was halfway decent but appearently not  :-DD so I'll gladly take any advice on how to optimize the layout. I don't really know what to optimize there, apart from relayouting everything. However if I'd do that, it might end up worse than it is now. Or quite the same :-DD
Boron rhymes with moron
 

Offline OM222O

  • Frequent Contributor
  • **
  • Posts: 768
  • Country: gb
Re: decent but affordable 16 bit DAC
« Reply #30 on: June 13, 2019, 09:53:28 am »
when crossing traces, I tried to only do it in 90 degrees and I tried to separate digital and analog stuff as much as possible and I tried to keep traces as short as possible, especially for the critical (analog) parts.

So.. I thought it was halfway decent but appearently not  :-DD so I'll gladly take any advice on how to optimize the layout. I don't really know what to optimize there, apart from relayouting everything. However if I'd do that, it might end up worse than it is now. Or quite the same :-DD

I can see your point about the traces being 90 degrees to each other, but that's not how it works  :-DD it's never that simple in electronics  :scared:
It has more to do with loop area, rather than overlap of the traces themselves, although this will reduce the coupled noise, it's not as effective as you think  :-//

I will do some layout examples on how to properly segment the board into blocks and how to separate the signals / grounds properly.
« Last Edit: June 13, 2019, 10:12:33 am by OM222O »
 

Offline OM222O

  • Frequent Contributor
  • **
  • Posts: 768
  • Country: gb
Re: decent but affordable 16 bit DAC
« Reply #31 on: June 13, 2019, 01:08:14 pm »
here is a very basic schematic and layout, just to get the point across (didn't spend too much time on cleanliness, but it's still not too shabby):
https://easyeda.com/theepicn008/layout-example

Thinks to note:
  • I didn't use a proper MCU breakout board. just 2 4-pin headers as an approximation. use the proper teensy breakout pins
  • I went with a linear regulator since it was easier to draw than the switch mode one you have. again this isn't a major point here
  • VCC is a thick track running at the bottom. this is the voltage input to everything

as you can see, the digital and analog traces are completely seperate: digital stuff on the left and analog stuff on the right. if you draw a line in the middle of the board, you can see the DAC and ADC both have the digital IO on one half, and the analog IO on the other half. this way you can have a solid ground plane, but it acts like 2 ground planes (analog and digital) joined only under the chips that are both analog and digital. if you want to be extra sure,separate AVDD and DVDD on the ADS1219 (i.e. connect 3.3V to reset and DVDD lines rather than 5V to both).

Also you can see the "blocks". top right corner is the linear regulator section, to its left are the mixed signal chips (DAC and ADC), under those there is the reference voltage (layout copied from T.I datasheet with minimal routing changes), the two LDOs and their caps, and the digital stuff on the very left. I hope this helped you clean up your board a bit. I used too large of a package (1206) which makes things look less neat, but I think yours were 0603 so you should be fine in that regard. just keep the layout like the schematic (i.e. in blocks) and be careful of the mixed signal chips. that should be it. just make sure you have the switching regulator on the digital side, not the analog  :-DD
 

Offline nemail2Topic starter

  • Regular Contributor
  • *
  • Posts: 206
  • Country: at
Re: decent but affordable 16 bit DAC
« Reply #32 on: June 15, 2019, 11:21:29 pm »
Hi

Thanks for drawing that example. However I'm still not sure how I should do this with my design.
I mean, I have this already:
- digital and analog traces are separated, only the SPI and I2C lines necessary for the DAC and ADC go to the analog part of the board (lower part is digital, upper part is analog)
- current return paths should be reasonably short everywhere
- ground planes are split and star grounded, teensy 3.5 has a ferrite bead which connects the AGND pin (which is connected to the analog part of my board) to the GND pin (which is connected to the digital part of my board)
- all traces as short as possible or at least reasonably short, at least for the analog part
- all power supply traces are big ass
- separate LDOs and VREGs for digital and analog circuitry
- no switching regulators on the analog side

i admit, what's not ideal in my opinion or rather say what i really think to understand whats not ideal:
- I2C traces are quite long
- SPI are as well, as I need SPI for the DACs and for the Display as well, which is on the opposite side of the teensy board on the digital side and the DAC is obviously on the analog side
- power supply traces (27V and 15V) are quite long because i just need them in totally different sections of the board
- NTC trace (TH0) goes from the TIP3055 footprint all the way down to an analog input of the teensy 3.5 -> quite long trace as well

for none of these issues I have a solution. I have simply no idea on how to solve those issues - the board is quite large and a bit complex as well (at least for a beginner like me) and many things are already tightly packed and I simply do not know how to make the layout better.
even if I'd relayout everything, I'm pretty sure I'd end up in a similar layout (if not worse) like I have now.

Of course I don't expect you to do my project for me but I'd be really curious how you would layout this design. I just can't see how your clean but simple example layout could scale to a design like mine and still stay as clean as yours is. At least not if I'd do the layout.

I have done some more cleanup and a bit of relayouting to reduce the size of the board and make some traces shorter and I have updated the github repository with the lastest version:
https://github.com/mamama1/LabPSU_Darlington/raw/master/preview.PNG
Boron rhymes with moron
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf