THS3061 ringing in audio (hopefully includes schematic)

Hi, Mark,

Let me have one of my colleagues answer your question.

Due to our internal org structure, this part is in a different group, so I am going to move your post to their e2e.

-d2

Hello Mark and Don,

Sounds to me like you are dealing with excessive phase because of the presence of U502 and U504 within the loop to form a composite amplifier.

May I propose you review the section on Composite Amplifiers and Figure 8 here. You may have to devise some compensation to make sure you have adequate phase margin:

Feedback Plots Define Op Amp AC Performance Analog & Mixed-Signal SBOA015 - TI.com

Regards,

Hooman

Thanks for the reply Hooman.  I think that the solution may be a simple one, so I pose this question;

How critical are the 0.1uF decoupling caps to the stability of the circuit?  I thought that they were mostly for high frequency stabilization ( MHz range) and would not be too critical for our use in the low kHz range.

I ask this because I had a working, stable version of the single op amp design (AD8610, THS3061) using Wima 6.8uF MKS-2 caps paralled with Wima 0.1uF PET SMD's.

I just learned that the company who builds these for me ran out of Wima caps and substituted Cornell Dubilier 0.1uF SMD's.  Unfortunately I too am out of those caps and before I purchase another 250, I would like be have some reasonable assurance that the caps may indeed be the culprit.

Since the single curcuit was only mildly affected by the new caps, but the dual circuit (AD8620, THS3062) was totally unstable, my other question is:

Are the dual package op amps more finicky than the single package?  FYI: both are SOIC.  I don't think that the PC board design is a factor.  If anything, its more compact than the single package design as all in's and out's are on the same side and require less trace for feedback paths etc.  In fact, the dual circuit takes the same footprint as the single circuit.

Hi Mark,

I have mostly experienced oscillations whenever there is an amplifier (THS3061) within the loop of another amplifier (AD8610 or AD8620- does not matter). This is a classic instability problem as the phase shift of the two active devices (and any additional phase shift due to parasitic capacitance, etc.) erodes your phase margin. That's why I pointed you to the Application note below which discusses some remedies:

Feedback Plots Define Op Amp AC Performance Analog & Mixed-Signal SBOA015 - TI.com

Most likely your issue is not the decoupling cap (0.1uF) manufacturer. Although the capacitor placement (as close to the supply pins as possible), return path (ground current should not pass close to the inputs), and type (resonance frequency as the upper limit of usability) do matter greatly.

I recommend you review the Application Note above and see if you can quell the instability by applying compensation. You can also review the simulation technique I have shown here using TINA-TI, and let me know if you need more help:

Using Pspice to analyze amplifier loop stability (Part 1 of 2) 

Using Pspice to analyze amplifier loop stability (Part 2 of 2)

Regards,

Hooman

Hello Hooman:

Thanks for the additional info on Pspice sims.  My original Tina-TI sim worked fine, but I guess that these sims are not able to include PCB design.  Can they differentiate between a ceramic and mica cap of the same value?  I was never certain of this.  It would seem important that they can.

While waiting for a reply, I read  SLOA069 – September 2001-"How not to decouple.." This article has an emphaiss on knowing your capacitor.  

The fact remains that I have 10 working, stable circuits on PCB's using single op amps, and the only variable is the 0.1uF cap.  The dual op amps(a new design)  are not stable.  Perhaps there is some stray capacitance on the PCB, but I doubt it.  We take great care in PCB design. We have selected the feedback resistors for maximum stability, while sacrificing bandwidth as we only operate in the 5Hz to 100kHz range.

I have ordered the Wima 0.1uF SMD caps and will let you know if this helps.  Regarding decoupling, I am unsure as to how much is necessary.  Is it only for the bandwidth in use or is it for the entire bandwidth of the op amp?

Thanks

Mark

Hello Hooman:

Thanks for the info and your hard work. FYI, we use rail supply of +/- 12V.  About a year ago I imported the 8610 spice model and ran simulations to 100kHz,  It looked perfect.  Never thought of  looking at 55MHz.  as we're using it with input signals to 20kHz.

Like you, I too noticed that decreasing the overall local feedback of the THS3062 made for more stability, but liked the fact that with a gain of 4V/V the burden was on the heftier 3061 and not the AD 8610

Do you think that some slight ringing at 55MHz will adversely affect the output of the 3061 at <=20kHz?

If so, perhaps as is shown below, I could sacrifice some bandwidth for stability by increasing Rf to 750 ohms, and Rg to 250 ohms.

From pg.13 THS 3061 White Paper SLOS394A – JULY 2002 – OCTOBER 2002

For most applications, a feedback resistor value of 750 Ω is
recommendeda good compromise between bandwidth and phase margin that yields a very stable amplifier.
Table 1. Recommended Resistor Values for Optimum Frequency Response
GAIN RF for VCC = ±15 V RF for VCC = ±5 V
1 750 Ω 750 Ω
2, –1 560 Ω 560 Ω
5 357 Ω 383 Ω
10 200 Ω 200 Ω
As shown in Table 1, to maintain the highest bandwidth with an increasing gain, the feedback resistor is reduced.

And while we're on SPICE.  I've often wondered how the simulation looks at or rather defines the rail supplies with respect to regulation and supply decoupling.  Does it define it as a perfect world?  Are the op amps considered to be precisely decoupled?  Is the supply perfectly filtered and regulated? Does TINA assume a perfectly laid out PCB, with minimum trace lengths and optimum widths? Can I input my own supply design and run my sims using my supply design?  Can I use my decoupling scheme of 6.8uF // 0.1uF?  As you know, not all caps are created equal, so are there spice models for each manufacturers passive components, or does TINA see all passives as equal?

Thanks for the help

Mark

Hi Mark,

Some responses to your questions:

1. Change to RF= 750, RG= 250: Simulations results point to much more instability! So, I am not sure you are going in the right direction. Remember that you have a Composite Amplifier and while the statements about raising RF for a Current Feedback Amplifier like THS3061 are valid, in this case simulation is pointing to the composite amp phase margin as an issue.

2. Simulation Limitations: TINA-TI models are macromodels (not device models). They have certain characteristics built-in, and only those are simulated. Power Supply decoupling, grounding issues, capacitor characteristics are usually not modeled.

In your particular case, as I recommended yesterday, I  suggest you try reducing the loop gain by raising RG on the bench (using a series pot for easy implementation) and let's see what you get. Another factor with high speed amplifiers is the resistance on the inputs. I'd also suggest adding a series resistance (say 20ohm) with the non-inverting input of the THS3061 (or THS3062 - no difference).

Regards,

Hooman