1.8 V reference modelling

Hi Uday,

Regarding generating a 5V reference from a 2.5V reference device:

The LT6658 has two separate buffered outputs. When the reference is configured for 5V as seen in your post, the output resistors actually serve as feedback resistors for the internal amplifier. They are setting the amplifier in a gain configuration of +2; this is how the 5V reference is formed. (Note the additional headroom requirement for VIN1/VIN2).

You can see the internal amplifiers/buffers more clearly in your second image, along with how the resistors can be configured for amplifier feedback.

Regarding the resistor divider with no buffers:

As hinted at above, this is not strictly a voltage divider. These resistors are feedback resistors for the internal amplifier. The internal amplifier is driving the Vout_F pin; current is not being pulled from the feedback resistors and dragging the reference voltage down.



If you're looking for more information and applications of the LT6658, see this article: The Many Uses of a 200 mA Precision Voltage Reference | Analog Devices.

If you provide your use case or key specifications, I better help with any additional concerns regarding LT6654 performance vs LT6658.

Hi Emoon,

Thanks for your reply.

I have already mentioned my use case .

In my current design a stable 1.8V reference is generated by generating 2.048V from LT6654 and then resistor network and then to a buffer. Now I want to replace this configuration with LT6658AIMSE-1.8 and at the same time I do not want to get compromise on performance.

Kindly suggest

It depends on what you mean by performance, which is why I was asking about use case and key specs.

If initial accuracy is critical, I would presume LT6658 would be more accurate than your divided LT664 setup, because of inaccuracies/mismatch in the resistor network.

If noise is critical, then LT6658 would also be a better option, as both parts have similar noise relative to output voltage, but this doesn't account for additional noise from your divider-buffer setup.

But if you're concerned other specs like power consumption, LT6654 is going to be a much lower power solution.

From the information provided, using LT6658-1.8 "out-of-the-box" sounds like a superior solution to LT6654 coupled with a resistor network and buffer. That is, if you just want an output voltage of 1.8V, similar/improved noise performance, and with no buffer or divider, the "Typical Application" configuration on the LT6658 datasheet would be sufficient.

I can help more with specifics if I know what aspects of performance you are looking to not compromise on.

Hi Emoon,

My use case is to have stable reference to high speed ADC IC. By performace I meant noise and precision and you covered both. Is there any other parameter I should consider?

Power consumption is definitely a concern. Can you give an estimate how much it differs?

Thanks
Uday

Hi Uday,

For power consumption, I'd take a look at the supply current datasheet specs for both parts. The LT6658 has an output disable pin, which could be leveraged to reduce power consumption.

But when both references are enabled, LT6654-2.048 has a typical supply current of 350uA, and a max of 600uA. LT6658-1.8 typical is 2.5mA, max 3.6mA.

LT6658 consumes about 6-7x more current when enabled, although this may be negligible compared to your total system power consumption.

Hi emoon, 

Thanks for the details.

LT6658-1.8 consumes 2.5mA at no load under VIN = VIN1 = VIN2 = VOUT1_F + 2.5V = VOUT2_F + 2.5V, right?

by which means if Vout is 1.8V, 2.5mA consumption is at 4.3V. Is this correct?

And how does this consumption varies with load?

WIll it be I_total = I_no_load + I_load ?

And, If I do not need 2nd buffer, can I cut off VIn2 to reduce power consumption?

Thanks,
Uday

Hi Uday,

You're correct, this 2.5mA spec is at 4.3V.

You can expect higher current at higher temperatures, as well as at higher supply voltages, as seen on the plot below.



You're right about I_total, the supply current would be in addition to load current.

We recommend powering all buffers regardless of their use or not, to meet datasheet spec.

Hi emoon,

As I understand, the image is for LT6658-1.2 for conditions  VIN = VIN1 = VIN2 = 4.5V.

Supply Current for LT6658-1.2 is 2mA for TA = 25°C. VIN = VIN1 = VIN2 = VOUT1,2_F + 2.5V. Thus for 1.2V output & 3.7V input

From the graph, for an input of 3.7V input, input current will be 1.7mA or 1.8mA which does not match the above condition.

Can you clarify?

And how to convert a value in ppm to voltage and vice versa? For eg 

1. Output Voltage Temperature Coefficient for LT6658A is Typ 3ppm/C : will it be 3uV for every deg of temp change?

2. Load Regulation for Output 1 Sourcing, ΔILOAD = 0mA to 150mA :is 0.25uV/mA : will it be 0.25ppm/mA

Is there a standard when to use in ppm and when to use in voltage?

Thanks,
Uday

Hi Uday,

Thanks for pushing on this.

Regarding the supply current:

The typical curve section does mention VIN for LT6658-1.2 is 4.5V, but that plot sweeps input voltage, so the 4.5V specification doesn't apply.

I agree with your assessment of the graph for 25C; the input current only appears to be 1.8 or 1.7mA. However, the spec in the electrical characteristics table specifies that the spec applies across the full temperature range (see image below).



This leads me to believe the "typ" spec listed on the table is based on the 125C curve, as it is the "worst typical spec" that was measured. Hopefully that clears things up.

Regarding ppm:

PPM refers to parts per million of the output voltage. So, for a 1.0V reference, 1ppm = 1uV (1.0V * 1e-6). For a 5.0V reference, 1ppm = 5uV (5.0V * 1e-6).

For TC, this would seem to imply for that LT6658-1.2, the output voltage would drift 3.6uV per degree C of change (1.2V * 1e-6 * 3 = 3.6uV).

However, temperature coefficient is generally calculated using the "Box Method" which takes the minimum and maximum Vout values over a specified temperature range, and calculates a linear slope based on (Vmax - Vmin) / (Tmax - Tmin). However, Vout vs temp curves are not linear, so you can expect better or worse temperature performance based on "where you are in the curve". I would recommend checking the Output Voltage vs Temperature curves included in the datasheet.

For more info on TC and the Box Method, see AN82.

There is no true standard for when ppm is used versus voltage. Generally, ppm is used when the specs scale with the output voltage of the reference, and uV or V are used when the spec is generally constant across voltage options.

This is because ppm is a ratiometric spec based on the voltage reference's output voltage. Instead of listing a different spec for each voltage option (i.e. specX = 1uV for LT6658-1.0, specX = 2uV for LT6658-2.0, etc.), a single ppm spec can be used (i.e. specX = 1ppm for all LT6658 options).

Let me know if you have any questions on this, AN82 also includes a unit converter and some descriptions of ppm specs.

HI emoon,

Thanks for the detailed explanation. That was really helpful.

Actually I want to generate 1.6V as well. This can be generated by using LT6658-1.2V option with 1:3 resistor combination in feedback. Right?

My query is whether 1.6 generated using this method will meet all the specifications provided by datasheet just like direct 1.2V?

Kindly clarify 

Thanks ,

Uday