THS4522: Massive thermal drift

Hello Peter,

Perhaps the issue is in regards to the drift in passive devices around the THS4522 causing an imbalance between both sides of the part. What are the temperature coefficient of the devices?  Can you provide a picture of your layout? What is the purpose of adding R137 & R138?

Best,

Hasan Babiker

Hello Hasan,

Passive parts (resistors) are all 25ppm/C.

I did test by touching Opamp package by heat (soldering iron @ 70C) and it drifted right away, while there was no thermal delta on passive parts. So I’m quite sure issue is related to opamp itself.

R137/8 are just to increase crosstalk immunity of opamp output by draining some bias current. Vcm is 1.5V so there is constant 1,5mA. However removal of those resistors was my first test I did, but there was no impact.

I also tried to reduce Rin/Rfb by factor of x20 (to 59R on input and 1k on FB) also no improvement.

thank you,

Peter

Attaching layout:

Nothing obvious here, but you might try dropping in the THS4552 if you are using one of the compatible packages. Intrinsically lower drift. 

Hello Peter,

Do you have an idea of what temperature the device is reaching when in operation? Note that 2uV is just the input offset drift of the device, you will still need to account for gain at the output of the amplifier.To get 1mV of drift at the output at a gain of 22 V/V, you would only need to increase the temperature of the device by about 23 C according to the typical specs. 

Best,

Hasan Babiker

Yea, we spent a lot of time trying to come up the min max drift for the THS4551 and THS4552 - no such effort on the much older THS4522, 

Hi Peter,

never rely on a parameter which is only specified in the datasheet as typical without given a maximum. For instance, if the offset voltage drift is important in your application, then only choose an OPAmp for which the datasheet explicitely specifies the maximum offset voltage drift.

If no maximum offset voltage drift is specified, the worst case can be estimated by dividing the maximum offset voltage over the temperature range of interest by the temperature range itself, which gives 7mV / (40K + 85K) = 56µV / K for the THS4522.

So as you can easily see the THS4552 with its 2µV / K specified maximum input offset voltage drift is an universe better than the THS4522.

Kai

Hello Kai,

Wow, this is really a value. I suspect some devices which I was not able to measure directly on IN+/- to have drift ~10uV/K, now I see it was still relative low.

Issue is that this part was inherited from past design, so we did no invest much in deeper check since in past it was OK, my mistake was that i needed more accuracy, while I have increased gain in parallel, luck I can fix it with simple BOM fix.

thank you for explanation.

Peter

Also Peter,

On centered distributions like offset and offset drift, typ is usually the 1sigma span since actual mean is 0 , from there, most folks use 4 or 5 sigma for worstt case. We wanted to do better on the THS4551, and put the test flows in to get to a more constrained range. There were amazing outliers on Ios drift. 

lucky Peter that we did an upgrade, I am not sure about Kai's max calculation on the THS4522, I would estimate it at +/-8uV using 4 sigma. 

Hello Michael,

yes upgrade was needed. I did test of ~ 12 devices with 2 Dual-Opamps each and my results were often close to those 8uV/C.

For sure typical value is in my case higher. Maximum I've measured directly was ~3.5mV/18C on output ~ 160uV/18C on input, so 8-9uV/C.

In average ~4-5uV/C, therefore I'd like to ask, is there chance to make it worse by suboptimal assembly  (storage, reflow profile, ESD) ?

thanx

Peter

Well Peter, 

It is hard to reconstruct where that +/-2uV/C typ came from on the THS4522. It could just be in error, probably best to just move on to the THS4552 which got a lot more attention on these drift issues (and noise issues)