• State Resolved
  • Locked Locked
  • Replies 8 replies
  • Subscribers 49 subscribers
  • Views 580 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

THS4561: Updated 2020 model does not appear to be modeling input noise correctly

Michael Steffes
Guru 47535 points
Part Number: THS4561
Other Parts Discussed in Thread: OPA1637, OPA1611

Well I was trying to start a fresh thread on the audio forum with Iwata-san, but apparently the THS4561 forces it to the amplifier forum

Anyway, Iwata and I have been iterating on the relative merits of the OPA1637 vs the THS451 as an Audio DAC output transimpedance stage. He had noticed the higher 1/f on the current noise on the THS4561 really impacted the low frequency spot noise, and it does. 

He had used the 2018 original RTM model, I found there is a 2020 update on the product folder and dropped it into his file, now it appears much lower (erroneously) than the OPA1637. Frankly, afraid to check the Zol and Aol in this THS4561 model update, with the noise so far off, I will have to stick with the 2018 model for this next AX article and not waste time on this update, 

here is that file where you have to reverse the supplies to use the new THS4561 model, 

Audio-FDA_comparizon Zt with 2020 THS4561 model.TSC

Here is the input voltage noise sim for just the THS4561 2020 model, the 1/f corner is correct, but the flatband is way low from the PDS 4nV

And then the input current noise, this is way low on both flatband and 1/f if the PDS plot is correct, 

  • 0
    TI__Genius 15555 points

    Hey Michael,

    Thanks for looking into this, please use the updated model attached below for fixes on these issues. This updated version should be on the web within the next week or so.

    Best, 

    Hasan Babiker

    THS4561.lib

  • 0 in reply to Hasan Babiker31
    Guru 47535 points

    Well Hasan, thanks for moving on this so quickly - in the meantime, I have tested the 2018 model in a simple circuit upgrade to (Figure 35, OPA1611 PDS) and find the original model produces about 21deg phase margin (before I fix that). That warns me of danger, while the 2020 model dropped into the same circuit produces about 32deg. gonna have to stick with the original, here is that 2018 test case, (I did find the input voltage noise line, and changed this one to 4nV, from the 5nV that was in there?) 

  • 0 in reply to Hasan Babiker31
    Guru 47535 points

    Continuing on this Hasan, 

    I needed to get the output spot noise simulating well for this audio dac article - was having a lot of trouble with it jumping around in a discontinuous fashion. I went back to a known working file that Iwata had sent on an audio forum post comparing 3 different FDA solutions on output noise. he was using a different PCM audio dac source approach, not sure if that is what fixed the sim problem, but it worked and got the plot I needed for the article - coming back to the THS4561 model accuracy issue, Here I took the set up for that first diff. Zt stage and ran it for noise using the 2018 model where I had modified the eni from 5nV to 4nV and then the file you sent on the 2020 model with updated voltage noise - I had noticed the voltage noise looked right on your update, but the current noise is way off still - that shows up here at low F, which was Iwata's point about the THS4561 being perhaps less useful than the OPA1637 due to the currrent noise 1/f corner difference - he is right below about 100Hz, but the HD benefits of the THS4561 make it more attactive in my opinion. Make sure you fix the current noise if you update the 2020 model online, 

    And this file, note this Zt schematic is very low phase margin and I have fixed it in the article, but that is happening out around 100MHz, so not in this noise sim. 

    THS4561 Zt stage using Iwata source different model noise testing.TSC

  • 0 in reply to Michael Steffes
    TI__Genius 15555 points

    Hello Michael,

    Current noise seems to be modeled correctly in newest update from what I can tell. I've attached my test circuit below.

    08 Current Noise.TSC

    Also I seem to be running into errors when trying to open your .TSC file. 

    Best,

    Hasan Babiker

  • 0 in reply to Hasan Babiker31
    Guru 47535 points

    That is kind of an odd way to do it, you understand the outputs are saturated against the rail and inputs have found some non-zero voltage delta

    I had inserted a V9 TINA file, here is a V7

    6518.THS4561 Zt stage using Iwata source different model noise testing.TSC

  • 0 in reply to Michael Steffes
    Guru 47535 points

    Morning Hasan,

    Going back to this input current noise question - I think you can plot that out in a closed loop sim by simply putting your output sense element as a current meter in one of the inputs and run an output noise sim, that yields this in that diff Zt sim I am using to compare models, you can see they are quite different - to get the flatband numbers I put the markers on the curves at far left then slid them over, the old model is correct at 0.4pA the 2020 update is too low at 0.26pA

    The way this was done back in 2018 was using a macro that I think Art Kay did, here that is in the 2018 model. The 1st term is a low frequency number to enter that noise at, 1Hz here, then the noise at that frequency, then the higher F flatband value in pA, you can see it is trying to produce a 40pA at 1Hz, just like the plot above, that very last line and number is where I changed the voltage noise from what was Z=5 to Z=4. 

    The 2020 model, I could not find this 

    However, even for the 2018 model, it looks like I need to decrease  the 1Hz number to about 8pA as per the PDS plot (right scale) - assuming that is correct and the model needs to match it - the point here is that Iwata has correctly pointed out in the audio transimpedance app, this current noise really increases the low F output noise from this stage - changing the model to this lower 1/f on the current noise will help that on the THS4561 implementation. 

    So, as long as I have this set up, let me test the model input noise for the other part we were looking at, the new OPA1637 FDA's, Yea, the model has basically no 1/f corner in it for the current noise - that will of course simulate better at low F, the OPA1637 datasheet shows about an 8Hz 1/f corner so this model is not correct either on the input current noise profile, Looks right for the flatband. 

  • 0 in reply to Michael Steffes
    TI__Genius 15555 points

    Thanks Michael,

    It does look like the method you provided gives a more accurate representation of the current noise. I was getting the same result when biasing the circuit at DC by using inductors, after digging into it further it looks like the differential capacitor what was causing the discrepancy in results. Please use the model attached below for a fix on the current noise.

    2500.THS4561.lib

    Best,

    Hasan Babiker

  • 0 in reply to Hasan Babiker31
    Guru 47535 points

    Yes Hasan, that looks right on the current noise now, here I dropped in your new netlist into the comparison sim I have been using over on the audio forum (the OPA1637 is also wrong on its current noise 1/f corner, not as much).