THS4131: Stability

Hey Stephane, 

Intrinsically, an attenuator is stable if the part is unity gain stable - There are other ways to get into trouble, but that by itself is not one of them. 

Sadly, this question is more common than you would think thanks to an error in the Op Amps for Everyone book - I discussed all that here, 

https://www.planetanalog.com/are-inverting-attenuators-with-voltage-feedback-op-amps-inherently-unstable-and-why-might-you-want-to-apply-them-insight-11/

And, some details on FDA stability sims here - #5 is prelude to this one, 

https://www.planetanalog.com/stability-issues-and-resolutions-for-high-speed-fully-differential-amplifiers-fdas-insight-8/

Hello Stephane,

The answer for this will vary depending on the part, however like Michael mentioned if an amplifier is unity gain stable this shouldn't cause any issues. Note that in the case of an FDA, however, when the gain of the circuit = 1, the noise gain is set to 2 (similar to an inverting amplifier). Because of this, some of FDAs are designed to be stable at gains of 2 and using the parts as attenuators may cause stability issues.

In the case of the THS4131 it seems that an Aol curve isn't provided in the datasheet to confirm the parts phase margin. However, when set to a gain of 1 some peaking can be seen without a capacitive load (figure 4 of the datasheet). Because of this, I believe that using the part as an attenuator will cause further peaking and possibly cause the part to become unstable if no compensation is added into the circuit.

Best,

Hasan Babiker

Morning Stephane, 

Having defined quite a number of these FDA's one of the key designer targets is indeed the target Aol curve intending a butterworth response at some min noise gain. For instance, the THS4541 is aimed at NG=3 for a butterworth, THS4551 at NG =2. 

As a starting point, the TINA models are what you need to work with - those often match the part, but compare to PDS to be sure (those are measured). 

I have test files for most FDA's, here is the THS4131 Aol curve (use the spec load for this). This is a circa 2000 model (the THS4131 was one of the 1st FDA's by the Dallas team, and still one of the best in many ways). I could not find any model updates on this one - but, looking inside, it is a transistor based model with thorough Q parameters - might be a real good one. Macromodels are much harder to get accurate. 

Anyway, GBP is 149Mhz, at Aol = 0dB at 145MHz (NG=1, you can't get that low with just R's) 30deg phase margin. 

and file, 

Aol curve generation for VFA FDA's THS4131 modified Aol and Ro.TSC

Hi Stephane,

can you show a schematic of your planned circuit?

Kai

One more idea 

The original THS4131 device was adopted by the audio group and renamed the OPA1632 with a different datasheet - that group is more proactive in modeling and in fact there is a circa 2009 model for that version (same die). This is  a macromodel by Bill Sands - at that time the modeling guy and later contractor as TI sheds staff regularly. Not sure this would be any better than original transistor level model (those are usually the best) but another option. 

OPA1692 model.TSC

Hi,

Thanks everyone for your contribution.  I'm not an analog guru myself :-)

The circuit is a "floating" current source, but in the microamps or nanoamps range.,,,

It is called "Quad-Feedback Enhanced Howland Current Source with differential amplifiers (QUAD)."

The circuit has been functionnaly simulated.  The input source will be a sin wave of 100Hz to 1.5kHz

I could have made a simple divider before the amplifier but I prefer to use less components if possible...

See attached the circuit...

Interesting circuit, 

What supplies are you using? The main THS4131 advantage is input voltage noise is low - those 10kohm Rf throw that away anyway. Might be a different FDA that is more modern that might be fine 

Incidentally cross coupled feedback gets into another area of phase margin questions.

Where did you find this circuit - there is one like this that is intended to generate an active output impedance

Supply has not been decided yet...

There are a couple of variations of it...

Fig 7 in this article...
https://content.sciendo.com/view/journals/joeb/9/1/article-p59.xml

As it is the "Journal of Electrical Bioimpedance", I assume that currents involved are low, therefore, amplifiers are used at gain of 1 or less.

I did find a resource for this, does not actually get into stability, 

https://content.sciendo.com/view/journals/joeb/9/1/article-p59.xml