LMZ21700's Spec

I will find out more details regarding your request and get back to you soon. -Yang

Hello Sainer,

What an interesting question. The selection of GBW of op amp isn't directly related it turns out, lets try to walk though this.

First, if the switching frequency is exactly 2MHz you need to restrict the bandwidth of the signal reaching the internal PWM from having frequency components above the Nyquist limit of 1MHz, other wise they will alias and fold back. However, the internal IC structure keeps this from ever happening. The feedback input is the inverting input of the internal error amplifier, with the non-inverting input of the error amp being tied to the internal voltage reference. The structure of the error amplifier is often that of a trans-conductance type amplifier. This means that the amplified input voltage difference results in an output current. This output current is applied across the compensation network which is typically a series R-C network to ground. The R-C network averages or filters the amplifier current and this results in the voltage applied to the PWM, because a also connected to the amplifier output and R-C node is the input to the PWM. So the dominant pole determined by the R-C and the error amp characteristics end up limiting the bandwidth reaching  the PWM and this -3dB corner frequency will be in the 1kHz to 4kHz range as a general guess such that unity gain crossover of the loop ends up being around 10kHz. These are all very rough numbers that may not apply to this device but as ballpark numbers are appropriate for the discussion. Assuming you can analyze the loop in a conventional feedback configuration you could employ a Frequency Response Analyzer such as one made by Omicron, Venable, or A/P-Ridley to find a more concrete set of numbers. But this will lead just to an intermediate numeric for the bandwidth of control loop when you try to drive it with the added op amp stage. In other words, how fast can you ultimately command the regulator output voltage using an external mechanism such as you are proposing. This is sort of a reciprocal case of the usual test of the speed of response to a load step transient on the output.

So we have an intermediate frequency numeric for the bandwidth of the actual regulation stage used as an externally controlled power stage instead of the conventional feedback path to create an accurate dc output voltage with fast transient response to maintain accuracy when subjected to fast changes in load current.

This leads to a question about the configuration of the op amp that you are adding. Assuming that this op amp sits insides the complete system loop an has been included as a way to modify what is normally just a resistive divider from Vout, Then the closed loop bandwidth OF THE OP STAGE should be 4 to 10 times greater than the loop bandwidth just determined above. otherwise significant bandwidth limiting in the added stage may drop into the range of significance and cause instability of the whole power loop because of the added circuitry imposing its own phase shifts which will add in directly. Hope that makes sense.

Now to get to your original question of how do you specify the GBW of the selected op amp that is being added? Well that then depends on the topology of the added amplifier stage. Since you haven't included a schematic let me just make something up. Suppose the added amplifier stage has a gain of four and is of the conventional voltage feedback type. Then it needs a GBW that as at least 4 times larger than the bandwidth requirement of the added stage in order to insure loop stability.

Now all this is fairly long discussion and all the numbers are make believe but plausible. The intention is to shed some light on the various interactions of frequency limits as you walk around the loop.

The simple reality, is that for a fairly simple amplifier addition a GBW of 2.2MHz would be a good starting point and above 10MHz might be a waste of money since they often cost more and might have higher quiescent current. But again, just ball park numbers because we don't know your intended schematic.

A few recommendation as well:

1)  From the output of the op amp, to the IC feedback input; place a 1K or 2.2K or so resistor so that if there is ever a turn-on issue or op amp output transient spike that there is a limiting element that limits current flow into some undisclosed section of circuitry on the feedback input internal to the IC.

2) Make sure the sequencing of the op amp power supply is coordinated with application of power to the supply being controlled otherwise there can be startup or shutdown events that can damage the downstream load.

3) There are certain op amps that exhibit phase reversal if the input terminal are driven out of range. This amplifier malady can create a disaster when used inside a control loop because if triggered, the loop may lock up at maximum output voltage with not-so-pretty results so far as the load is concerned.

If you would care to share your intended schematic or system requirements we can make more specific recommendations. From other divisions of TI there are a multitude of op amps of all types. Part number prefixes are commonly TL, OPA(formerly the Burr Brown people down in Tucson), LM, LMV, LMC,(formerly the National Semi people that still sit upstairs here in Santa Clara.) Lots of great op amps to select from and we've be happy to help further. Choosing a TI op amp to go with your TI regulator keeps the circuitry happy as well as the people in purchasing and sales.

If you care

the baThe voltageThe output of the error amp (is usually load to ground.

Hello Alan,

 Thanks for your explanations.

Here I reattached my schematic from my past post 

As shown, I am designing a DAC controlled current source right now, but has been still in grief selecting the Op Amps. We tested some of Op Amps in the past but most of them ended up with instability caused by the Op Amps. Following is events what I was going through.

1. We have tried to use Chopper-stabillized Op Amp by other IC maker (AD8629ARMZ) in purpose to get the zero drift characteristics, but for some makers it seems like it needs compensation capacitor on its feedback to remove the noise  made by "the chopping" on a certain frequency. Yet, when we put the capacitor on the loop it became unstable and ringing. We found that there was a trade-off between the RF network inside the DC/DC converter and the compensation capacitor of the Op Amp.(We used different DC/DC Converter at that time)

2. Recently I tested the LMP2012 (by reason of it is an Auto-zero Op Amps, not a Chopper-stabillized).    To select the Op Amps, I am using two simulation tools right now, one is TINA simulator (to test the Op Amps only),    and the other one is PSpice simulator (to test the transient level of the loop).    I test the Op Amp by cutting off the differential Amp circuit only from the loop,  and in case of LMP2012, I got that it was ringing when the load Voltage went around 3.4~3.6V. I benchmark it, and confirmed that it was ringing when the voltage went around 4V.    I don't know what the cause but anyway, I think it didn't fit to my system.

3. We also test OPA2330 in the past, but found that there is a switching/changeover of PMOS/NMOS of the Op Amps on a certain temperature.

4. Up to this point, I ran all the TI's Op Amp with GBW below 1 MHz to the simulators, and got that some some was ringing on TINA simulator, and some went divergent on PSpice Simulator. This morning I found that OPA2344  went well on TINA, and PSpice simulator (after running it for about a day; exceptionally slow!!). Do you think it fits to this system?. I had still not benchmark it, but it doesn't have a PMOS/NMOS range changeover characteristic as OPA2330 I described on No.3 above, right?.

My system requirement is following.

1. Load resistance range is 30~40 Ohm with constant Power range(240 mW~ 340 mW),    so Load voltage range is 2.68~3.68V. But considering margin, if possible we want Vout can go till 4.75V.

2. Our operation temperature is around 0~85 degreeC, and we want current change due to temperature on the Load below 5 mA, which means we need a low drift Op Amps.

3. Because of limited spaces, the Op Amp size should be 8VSSOP:3x3 mm:15 mm^2 or smaller.

Thats all for now, and thanks again for your attentions.

 

Sainer

Hello again. It's 8PM here and I've worked in the lab all 3 days of the holiday weekend. I'll dig through your description and we'll see what we can suggest. I'll reply in the morning.

Alan

This just in... CSP/BGA op amp part numbers and basic specs.

LMV881… 23MHz…1.5x1mm uQFN…Shutdown…

LMV712TL – 5MHz, R-R I/O, High Output Drive, 2x1.5mm DSBGA

OPA2376 – 5MHz, R-R I/O, 1.3 x 2.3mm DSBGA

OPA1612 – 40MHz, Low Noise, 3x3mm SON

OPA2316 – 10MHz, 3x3 WSON

OPA2322 – 20MHz, low noise, 3x3 WSON

OPA727 – 20MHz, low Vos, 3x3 VSON

OPA835 – 56MHz


Thanks Alan

Alan,

Thanks for your thorough explanations and calculations.
Unfortunately, we can't change our current "Current driver circuit" architecture to the new one right now.

This architecure actually proposed by TI in the past with different DC/DC converter and Op Amps combination.
But the proposed DC/DC converter needed external components (especially inductors), while on the other hand
we have only limited spaces, so we started considering other DC/DC Converters and ended up to LMZ21700.

Anyway,I also got a suggestion from Op Amp teams as following.

e2e.ti.com/.../423119

Still digging the Op Amps.
That's all for now.


Sainer