THS4275: Temperature-induced Oscillations

Eventually, a schematic will be needed - ideally a TINA file - the folks that put that note into the datasheet are long gone - Kind of unfortunate they proceeded to release a part with this issue.

I went back and read the THS4271 datasheet on this topic, put a lot of effort into describing what was apparently a well understood phenomena, why not fix it - also mentioned the TS4211 as an oscillation free alternate. 

Implies it is a bias setup oscillation, not much you can do externally about that. But at minimum, lets check you decoupling, can you send full schematic with decoupling details. Sometimes if we go to X2Y caps, the supply set up is happier.

Below is a screen-shot of a schematic which includes the decoupling components and a few of the parasitic PCB effects.

 - would have up-loaded the TINA .tsc file, but I can't find a file upload option.

The device mentioned as an alternative has a much worse input voltage noise unfortunately. I've looked at a number

of other parts from TI and others and so far not found one with the same combination of very low input voltage noise

and high speed performance - maybe that's why it is still listed as production status.

It is odd that if they knew all about this problem but didn't just fix it.

As I said at the start, what I really want to know is some indication if this temperature-induced oscillation, should it

occur, is a high frequency effect or a very low frequency - it could be either and the data-sheet gives no clue. If I

knew this I would be able to say whether the high frequency oscillation I can sometimes induce in the re-spun

PCB is due to the chip fault or is being caused by something else which I might be able to fix.

Well Robert, 

1. I assume those 10uH are ferrite beads, if not they should be

2. What are you doing with the disable and the reference pins? 

3. That is a pretty heavy AC load there

4. +/-6.5V is less than absmax 15V (tells me this is a Bicom2 part), but pretty high

5. Yes, 3nV unity gain stable 400MHz true gainbandwidth product is kind of unique. THe OPA820 is in that direction but not there. would need to drop to +/-6V

6. 

Oh I see now your schematic says THS4271, while your lead comment is the THS4275 - so no shutdown and reference pins. 

To attach a file, use the paperclip symbol across the top

The THS4271 model was used as the TI website only offers this one.

The Ref pin on the THS4275 is strapped to analogue ground.

The En pin is actively driven but is connected to a 330pF, other side of which goes to 
analogue ground.

The 10uH is a Coilcraft 1008LS-103-XJL_

Do you know if  the TINA-SPICE model includes anything to simulate the temperature-
 related oscillation mechanism?

In the application, this circuit only has to handle signals in range 100 KHz to 50 MHz but
with a very flat frequency response over that range and very rapid recovery from
transient O.P.overloads.

Thanks for suggestion about OPA820 - I'll revew it, could poss trim down power rails voltages
so long as I can still get enough linear O.P. swing. 

The paperclip is described as 'insert' , below is the result of using it to insert a .tsc file
- if you can recover the .tsc file from here please let me know for future reference.

Regards

Robert Phillips

No robert, you did not get the tina file inserted yet. 

Customer sim models are very simplified from the full cadence design files - Those would probably have shown the marginal phase margin in the supply setup, but the TINA file will most likely not - usually just current sources to set up the amplifier stages. Also, getting a shutdown feature into the TINA model is problematic and often not done. 

Hard overdrive recovery is seldom specified nor modelled. One exception is the OPA688 with output limiting, but its noise is probably too high. 

Oh and just to be clear, the 2002 THS4275 would not be very easy to support now as all those folks are long gone. 

Hi,

Tried upload again using the 'insert' button.

I understand that overdrive recovery is not spec'd for the THS4275, but we have found by trail and error that it
is good enough for our application which is ultrasonic non-destructive testing BTW.

THS4275 Amp1 incl. Decoupling 29_07_20 #1.TSC

Yes, that came into TINA this time, Yea ultrasound NDT is a stress on that first stage, medical does a T/R switch to protect the receiver from the pulse. 

So Robert, 

at the end of the day, we are sort of groping the dark here for answers. Something that goes into and out of oscillation over temperature is by definition low phase margin to begin with. Even if you can convince yourself that your nominal design seems ok, how about that batch of parts shipping 5years from now? 

Anyway, to give you something practical to try, I was working on better decoupling models some years back with this result - 

And what I was looking for here was low resonant peaking in the impedance looking out of the supply pins, this does that nicely, nothing in the amplifier TINA model will show any difference using this, but your board might. I also would suggest a supply to supply x2y cap - long ago, we were testing that and seeing better HD2. 

Here I modified your file to use these supply decoupling models 

THS4275 buffer with better decoupling.TSC