Voltage minor variations. Need to make it stable

I am using 7.4V Li Ion Battery, followed by a 5V (1117) regulator, the output of which is given to 1117 (3.3V)

The input (Li Ion) voltage depending on load is flatuating (Dropping to say 6.8V).

I am connecting a bluetooth module also at 3.3V.

Observations

I closely looked and I saw that 3.3V drops 200-300mV (Approx) for about 7ms and comes back to 3.3V and some spikes (lots of up down lines on CRO) of few mV variations.

This I think is efefcting bluetooth throughput

How can i make sure that the voltage is stable

If I use a saparate supply only for bluetooth with common ground ofcourse, it works fine

I suspect that your 1117 regulators are not genuine American LM117 ICs and instead are poor copies that do not meet the low-dropout spec's. Do you use the input and output capacitors shown on the datasheet?
Or maybe your Lithium battery is poor quality or is too old.

This is called load regulation, which is a function of ESR or Output Impedance, which in the case of linear devices is lowered by the amount of emitter R or RdsOn divided by negative feedback closed loop gain to regulate output.

300mV drop in 3.3 V is equivalent to 9% load regulation, implying your LDO is being used near limit of output power. Chose a better part to handle peak currents.

Example storage capacitor value C and ESR.

If T duration is short enough choose desired drop and estimate n for -nT exponential .e.g. 1T=37%, 5T=10%, 50T?=1%. then choose low ESR <=1/n % of Rload and C from above like 1% C= 50T/R. where T is pulse width of dip, R is effective linear load = 3V/Ip

Note : When Cout has very load ESR where C*ESR<10 us, the feedback stability affected by pole-zero compensation. where datasheet might indicate stable operating point limits for ESR on load cap.

- - - Updated - - -

Since your load is unknown and dynamic, test your LDO for static load regulation, which I see should be 0.5% and dynamic ripple from a step load to represent your actual pulse load , except use a pulse generator to extend the duration of step load to allow regulator to track in 10 to 20 us.

---

After reading LM1117 spec, I suspect your output cap choice or layout is the culprit.

If load is 0.5 A at 3.3V, Rload= 6.6 Ohms. for ripple to be low RC of load needs to >>500us for near ideal ripple rejection or C=100 uF as suggested in datasheet, but since LOW ESR caps are ESR*C=1-10us , more or less, yours may be too high ESR, so a better quality or parallel range of low ESR caps 1uF, 10uF, 100uF to source surge current with low drop

Hi,

This is called load regulation,

Click to expand...

(Maybe off topic here: But that´s exactely why I don´t like to use VCC as VRef of an ADC. VCC is not that precise ... and the ADC readings are not precise, too.)

While it is not nice when VCC bounces some 100mV... on the other hand it should not cause any problems.
But you need to check datasheets of the suppled deices .. the bluetooth module power supply specifications.


*****

desired drop and estimate n for -nT exponential .e.g. 1T=37%, 5T=10%, 50T?=1%.

Click to expand...

It´s not that worse:
1T = 37%
2.3 T = 10%
4.6T = 1%
6.9T = 0.1%
(each 2.3T make the factor of 10)

Klaus

I an using 10uF only. 0805 package ceramic (Output and input caps)

- - - Updated - - -

its a X5R, 10% Tolerance. Part Number: CC0805KKX5R6BB106

poor choice of cap. you need an ultra low ESR cap. X7R's ESR depend on supplier quality, may vary widely with C vs V and ESR.
MEasure pulse width of load ripple please and search for cap with Rated low ESR. Typical e-caps are 100-200 us , low ESR types are typ. 1-10 us for ESR*C product. Regulator bandwidth is limited by internal gain*BW product and may result in step load response times in the xx microsecond range, thus current impulse needs to be sourced by ultra low ESR cap for this duration.

e.g. ESR=1% of load R

It could also be measurement error on your part, with a long inductive ground connection to probe. It should be very short or none using barrel ring to gnd.

Cap p/n provided now looks good for ESR and SRF >1MHz and result should similar to data sheet


This 10uF tant above in datasheet has an apparent ESR of 150mV/500mA or 300 mΩ

while your X5R 805 10uF cap shows an ESR of < 10 mΩ @1.3MHz. thus Zc(f) with load of 3.3V/0.5A =6.6 Ω and a transient response of 6-7 us, would suggest the LDO responds in this time to correct the transient error with an ESR of 1Ω which might correspond to the open loop output driver impedance as I would not expect much gain at 100kHz


So verify your measurement methods and report again with pulse width at 50% or as we say PW50.

If you want less load ripple, then please spec. desired step load ripple, Vp, @ I This defines your supply filter needs.

The hardware communicates to the mobile via bluetooth. I noticed that this variation is only for a particular mobile (Intex Make) and not for some others. (I need to check it throughly though to reach this conclusion)

Why would there be this variation only for a perticular mobile. I am checking it as said by sunnyskyguy and will update here.

Meanwhile I am designing a new board (this time its a 4 layer rather than 2 as earlier)

The LDO design is as follows



Part Numbers used:
22uF/10V : CC1206MKX5R6BB226
10uF/10V " CC1206MKX5R6BB106


There are input caps for LDO too. I have a DC to DC that gives 5V, the output of that is connected to LDO




Part Number

47uF/16V : GRM32ER61C476KE15K


1R/1% (0805 package) resistor to maintain ESR. If I remember it right, datasheet says ESR to be between 0.3 to 22R

Am I doing it right??

The load I guess will draw about 500-600mA @ 3.3V (Approximated. I am unable to check on this and hence this approx value)

- - - Updated - - -

I tried having saparate supplies for Bluetooth module and main board. It even works with Intex. This isnt going rt at all...

tiwari.sachin said:

The hardware communicates to the mobile via bluetooth. I noticed that this variation is only for a particular mobile (Intex Make) and not for some others. (I need to check it throughly though to reach this conclusion)

Why would there be this variation only for a perticular mobile. I am checking it as said by sunnyskyguy and will update here.

Meanwhile I am designing a new board (this time its a 4 layer rather than 2 as earlier)

The LDO design is as follows

ΩΩΩ

Part Numbers used:
22uF/10V : CC1206MKX5R6BB226
10uF/10V " CC1206MKX5R6BB106


There are input caps for LDO too. I have a DC to DC that gives 5V, the output of that is connected to LDO




Part Number

47uF/16V : GRM32ER61C476KE15K


1R/1% (0805 package) resistor to maintain ESR. If I remember it right, datasheet says ESR to be between 0.3 to 22R

Am I doing it right??

The load I guess will draw about 500-600mA @ 3.3V (Approximated. I am unable to check on this and hence this approx value)

- - - Updated - - -

I tried having saparate supplies for Bluetooth module and main board. It even works with Intex. This isnt going rt at all...

Click to expand...

This is hard to explain with finger typing and no graphs, but any RC series filter on the output of an LDO is also part of the feedback path and forms a lag/lead filter. This is good for stability to prevent ringing on a step load response, but is poor for peak transient error due to the step. Therefore it is a performance tradeoff. In other words, Too little ESR where the RC break point is higher f than the unity gain bandwidth of the LDO reduces loop phase margin by integrating the current pulses. Too much ESR causes a linear voltage transient dip due to load R/(R+ESR) *100%. Thus assuming a 6Ω load and 1R added ESR you can expect a 14% poor transient drop.

How to optimize both Loop stability and Step Load Response?
This is the crux , assuming no measurement errors in transients probe inductance.
I suggest to also look at load capacitance and determine if there is any series choke and/or low ESR load cap, which changes your model of the step load.

Recommendation,
1) use 0R series R with Cout at 100uF and cap SRF around 1MHz, measure step load response and if resonance is significant.
2) if above response is ringing more than the ESR / LOAD ratio then separate input and output grounds using an additional series ESR R between R27 and 1R on the 100uF cap, see spec figure 12 for details.

3) Or better, add an LC choke outside the feedback loop after the 10uF storage cap to attenuate all transient error pulses using 100uF cap i.e. between 10uF and 100uF but have ADJ feedback from 10uF cap. on Vout.

FIrst define your load sensitivity to transients. Second choose the LC LPF or RC LPF to attenuate the 200-300 mV peak transient drop to TBD level using equivalent fundamental f half cycle and attenuation from filter response for something like > 20 dB and L DCR to cause no more than 10mV drop in 0.5A or 20mΩ 10uH?

Food for thought... You should be able get <10mV transient error.

Let me check that and do the changes. I will update the sch

Meanwhile in old schematic



VDM is battery voltage (7.4V, Li Ion)

I am not using 5V anywhere. Used it just to drop the voltage so that 3V3 LDO works fine.

issues:

3V3 is used for bluetooth MCU etc.

Load is not a constant current consumption device. Its dynamic, variable.

Some boards work and some dont.


I did a few changes

I removed 5V LDO and instead place a 3V3 LDO in place of a 5V and left 3V3 LDO open and so that I could shory PIN 2 and 3 and take 3V3 from there.

Before shorting 2 and 3 of U2, I noticed this

Output of U1 (now a 3V3 LDO), gave 3.3V output

and since PIN 2,4 of U2 are not connected to 3V3, it should show 0V but was showing 0.9V

What might be causing this??

Ground Noise???

- - - Updated - - -

SunnySkyguy said:

2) if above response is ringing more than the ESR / LOAD ratio then separate input and output grounds using an additional series ESR R between R27 and 1R on the 100uF cap, see spec figure 12 for details.

Click to expand...

What are you referring to when you say R27 and 1R. Figure 12???

- - - Updated - - -

Update: In the old circuit that I am explaining. (1117 LDO)

I also was checking the battery voltage (Voltage divider on Battery connected to MCU pins). Removed these resistors and I saw the voltage drop to 0 instead of 0.9 (that I was getting earlier)

I was under the impression that I can connect a voltage diver directly to the MCU analog pin but it seems like it is adding some variations or error if I can call it so

tiwari.sachin said:

Let me check that and do the changes. I will update the sch

Meanwhile in old schematic



VDM is battery voltage (7.4V, Li Ion)

I am not using 5V anywhere. Used it just to drop the voltage so that 3V3 LDO works fine.

issues:

3V3 is used for bluetooth MCU etc.

Load is not a constant current consumption device. Its dynamic, variable.

Some boards work and some dont.


I did a few changes

I removed 5V LDO and instead place a 3V3 LDO in place of a 5V and left 3V3 LDO open and so that I could shory PIN 2 and 3 and take 3V3 from there.

LDO's are aub-optimal solution here for high pulse loads. -Buck would be much more efficient

Before shorting 2 and 3 of U2, I noticed this

Output of U1 (now a 3V3 LDO), gave 3.3V output

and since PIN 2,4 of U2 are not connected to 3V3, it should show 0V but was showing 0.9V

What might be causing this??

Ground Noise???

- - - Updated - - -




What are you referring to when you say R27 and 1R. Figure 12???

- - - Updated - - -

Update: In the old circuit that I am explaining. (1117 LDO)

I also was checking the battery voltage (Voltage divider on Battery connected to MCU pins). Removed these resistors and I saw the voltage drop to 0 instead of 0.9 (that I was getting earlier)

I was under the impression that I can connect a voltage diver directly to the MCU analog pin but it seems like it is adding some variations or error if I can call it so

Click to expand...

critical: determine your battery ESR vs State of Charge SoC, by voltage drop and load R calculations. Maximal power transfer theorem states max power is transferred when load impedance matches source, but this is not max efficiency since internal battery and LDO losses may be greater.

wHen you have a step load such as BT Tx On caps are treated like short circuits with ESR internal so voltage drop from "load regulation error" is due to ESR % ratio with total R load where ESR is degraded by adding R27 and R28. so thats why we dont add it but see if it rings.. If it sags to 0V then your peak load is likely >=1A and LDO's will short out battery

- - - Updated - - -

tiwari.sachin said:

I am using 7.4V Li Ion Battery, followed by a 5V (1117) regulator, the output of which is given to 1117 (3.3V)

The input (Li Ion) voltage depending on load is flatuating (Dropping to say 6.8V).

I am connecting a bluetooth module also at 3.3V.

Observations

I closely looked and I saw that 3.3V drops 200-300mV (Approx) for about 7ms and comes back to 3.3V and some spikes (lots of up down lines on CRO) of few mV variations.

This I think is efefcting bluetooth throughput

How can i make sure that the voltage is stable

If I use a saparate supply only for bluetooth with common ground ofcourse, it works fine

Click to expand...

Using 2 LDO's to drop 7.6V to 3.3V not only wastes 60% of your energy , it needs a heatsink.

Why not scrap the design and use an automated design tool like www.ti.com with free registered email login.
You can dial in any solution for smallest footprint,fewest parts ,,least cost, or highest efficiency (99% vs 94%)

I have taken care of this in the next design but for about 2K boards I have, this really is the issue that I am unable to solve

Hi,

and since PIN 2,4 of U2 are not connected to 3V3, it should show 0V but was showing 0.9V

Click to expand...

Don't care. This comes from the battery voltage divider --> ADC input pin --> internal protection diode -->ADC VCC pin --> (unconnected) VCC supply. The voltage (0.9V in your case) is not predictable. It depends an many items.
As soon as you connect 3V3 to the VCC there is no current flow through the protection diodes. All is normal then.

*****
As SunnySkyguy said...two LDOs in series is no benefit.

Maybe two 3V3 LDOs, both powered from battery, but one supplies the sensitive (ADC) circuit, and the other supplies the bluetooth circuit.

There's no change in efficiency nor generated heat..

Klaus

KlausST... Agree to what you say.

I am designing a new boards and trying to take care of all these issues there.

Thanks everyone for your suggestion