ADN8831 Output Voltage Different from Datasheet - Outputs diverse before linear side saturates.

I am having a problem with the high-side (linear-side) FET overheating.  Some heating is normal in low current when the linear FET is not saturated and the voltage on that side of the TEC is not 5V or 0V.  My TEC is Laird SH14-15-06.  Its V-I curve is similar to a 0.25 ohm resistor. So 1V results in 4A. 

Here is the expected voltage output vs VOUT2 according Page 17 of the ADN8831 datasheet.

I calculated my FETs' expected temperature rise vs TEC Current

The problem is the low side (switching side) does not rise fast enough with respect to VOUT2, so the linear side transistor stays in linear (non-saturated) mode for higher TEC voltages.  Here is the measured TEC Voltage vs VOUT2:

Here it is in chart format:

OUT2 V[High] (Linear) V[Low] (Swiching) I[TEC] V[TEC] Transistor Pwr
1.255 1.38 1.31 0.22 0.07 0.79
1.245 1.83 1.69 0.35 0.14 1.09
1.241 2.04 1.87 0.48 0.17 1.40
1.236 2.31 2.11 0.7 0.2 1.85
1.225 2.86 2.5 1.5 0.36 3.14
1.217 3.3 2.82 1.5 0.48 2.48
1.208 3.78 3.14 1.88 0.64 2.20
1.199 4.11 3.28 2.22 0.83 1.86
1.189 4.63 3.65 2.69 0.98 0.86
1.176 4.94 3.82 3.02 1.12 0.03

The worst case is when OUT2 = 1.225V.  The transistor dissipates 3.14W.

Part of my issue is the TEC's low V/I ratio ("resistance"), but this would not be a problem if the TEC controller provided the voltage outputs on Page 17 of the datasheet.  

Might a larger inductor help?  Can someone knowledgeable about the workings of ADN8831 suggest why the voltage outputs are diverging so far before the linear side ("high-side") is saturated?  

  • I found a similar question about linear side heating from says it's related to R[DS,on] and other issues.  But R[DS,on] doesn't come into play in linear mode.  The part monitors the LFB pin and drives the linear-side transistor's gate to make LFB's voltage equal to the equations on page 17 of the datasheet.  It's purposely making the transistor operate the linear mode.  When TEC voltage is high enough (i.e. V[OUT2] is far enough from V[Ref]/2), ADN8831 is supposed to put the linear transistor in saturation mode.  Then TEC voltage is controlled by the switching side, minimizing losses.  It's all DC, no oscillation.  

    I wonder if may have been seeing something related to this five years ago when geno had a linear-side FET burn up.

  • Is this little discrepancy in the datasheet just something we have to live with?  Can I trust the numbers I took to be consistent from part-to-part?  If so I could size a larger transistor for the linear side.  Has anyone heard of this problem before?

  • 0
    •  Analog Employees 
    on Nov 5, 2019 11:03 AM

    Hi ,

     

    Thank you for posting your questions here in engineerzone. I am an applications engineer supporting the ADN883x products.

    With regards to your inquiry, here below are response:

    in linear (non-saturated) mode for higher TEC voltages.  Here is the measured TEC Voltage vs VOUT2:

    Please know that this curve is valid only during open-loop setting, without load connected between LFB and SFB.

    With a load connected across the LFB and SFB is a different scenario and the graph showing the SFB output is now not valid.

    With a TEC connected, the SFB out will now depend on the following factors:

    1. TEC resistance.
    2. Level of the current limit set (ILIM).

    Let’s consider the following condition:

    In your case, let us say OUT2 is around 1.15V; ILIM is set to 1.25A.

    With this condition, computing for the outputs as per page 17 will be:

    LFB = 5V (clamped to rail)

    SFB = 4.5V ; giving TEC voltage of 0.5V.

    Having a TEC resistance of 0.25ohms, and TEC voltage of 0.5V will ideally give ITEC (current across the TEC) = 2A.

    Given that ILIM is set to 1.25A thus not allowing 2A to flow across the TEC, the SFB will adjust accordingly.

    The LFB output will be the priority, then the SFB will adjust to comply. You can notice it on the formula for computing the LFB and SFB output.

    SFB is dependent on LFB, while LFB is dependent on VB which is a defined value.

    This can also answer your observation on why the switching side does not rise fast enough with respect to VOUT2.

    To clear this out, we need to confirm on the following:

    1. What is the current limit setting of your circuit?
    2. If it is ok with you, can you share to us your schematic so we can help you?

    On this part there are things that we would like to clarify:

    Here it is in chart format:

    OUT2 V[High] (Linear) V[Low] (Swiching) I[TEC] V[TEC] Transistor Pwr
    1.255 1.38 1.31 0.22 0.07 0.79
    1.245 1.83 1.69 0.35 0.14 1.09
    1.241 2.04 1.87 0.48 0.17 1.40
    1.236 2.31 2.11 0.7 0.2 1.85
    1.225 2.86 2.5 1.5 0.36 3.14
    1.217 3.3 2.82 1.5 0.48 2.48
    1.208 3.78 3.14 1.88 0.64 2.20
    1.199 4.11 3.28 2.22 0.83 1.86
    1.189 4.63 3.65 2.69 0.98 0.86
    1.176 4.94 3.82 3.02 1.12 0.03

    The worst case is when OUT2 = 1.225V.  The transistor dissipates 3.14W.

    • We would like to know on how the ITEC was computed. Is it by VTEC divided by the TEC resistance?
    • Or is it by the measured voltage on the ITEC pin? Please help advise.
    • Same thing with transistor power dissipation computation and how it was computed.

    We have here sample computation that based on your values.

     But in here, we got ITEC by VTEC divided by the TEC resistance (assumed to be 0.25ohms).

    Based on the given circuit VDS of the PMOSFET can be determined.

    By KVL:

    VDD – VDS – ITEC*Rsense – VHIGH = 0

    Therefore,

    Consider the given values

    VDD = 5V ; ITEC = 1.5A ; Rsense = 0.02 ; VHIGH = 2.86V

    VDS = VDD – (ITEC*Rsense + VHIGH)

    VDS = 5V – (1.5*0.02 + 2.86)

    VDS = 2.11V

                  Power Dissipation of transistor on this condition will be:

                  Pdiss = VDS*ITEC

                  Pdiss = 2.111*1.44

                  Pdiss = 3.04W

                   Having these values, we can now predict the Temp Rise by multiplying it to thermal resistance.

                  TFET = 3.04*52 = ~150degC

    Thank you and best regards!

  • Please know that this curve is valid only during open-loop setting, without load connected between LFB and SFB.

    That explains why the voltages are not what I expected.

    What is the current limit setting of your circuit?

    My I-Sense is 5 mohms.  

    V[ILIMH] = 1.24V --> 0.10A Heating

    V[ILMIC] = 2.08V --> 6.63A Cooling

    I'll send the schematic next.

    Power Dissipation of transistor on this condition will be:

                  Pdiss = VDS*ITEC

                  Pdiss = 2.111*1.44

                  Pdiss = 3.04W

    That's about what I'm getting.  I wish it were a little less.  I have an SOIC-8 N/P transistor pair (SI4501) with a steady-state thermal resistance of 110C/W.  

    BTW, I would be interested in a transistor optimized for lower thermal resistance rather than R[DS,on].  I think, though, much of the J-A thermal resistance is between the package and ambient, and therefore outside of the IC developer's control.  I may have to use a bigger transistor.

    Do you know any tricks I can do with VLIM and ILIM to get the part to saturate the linear transistor at lower currents?  

  • If it is ok with you, can you share to us your schematic so we can help you?

    Click on image or follow this link for enlarged image.