Consider a linear regulator having input 50 mVpp ripples at 1 MHz which is coming from DCDC converter. What the linear regulator will do with the ripples. Pass it ? Removes the ripples with output de-caps ?

 

It depends. Show us a schematic.

Your member name implies that you're an Electrical Engineer. Is that fact or aspiration?

 

Consider a linear regulator having input 50 mVpp ripples at 1 MHz which is coming from DCDC converter. What the linear regulator will do with the ripples. Pass it ? Removes the ripples with output de-caps ?

Hi,

It would probably be hard for a regular regulator to pass 1MHz, but this really depends on the voltage regulator IC type.
The input and/or output cap would help with that though. A low ESR input cap could be a good idea, and of course completely tested.

 

Input voltage.
Output voltage.
Output current.
PC board layout (1 MHz is in the middle of the AM broadcast band).
S.C.H.E.M.A.T.I.C.

ak

 

With out knowing what linear regulator ..................
Ripple rejection for LM7805 and 7824. Extend the graph to the right to 1mhz, you can get an idea of rejection. (Cin 0.22uf Cout 0.1uf)

You need to add a CLC filter in the input. The L can be a bead. Choose values that will give good rejection in the 100khz to 10mhz range.

 

 

Hi,

I haven't seen anyone use a capacitance multiplier in years and years. That's since voltage regulator chips became more popular.
A capacitance multiplier is just a voltage regulator/limiter with variable output that adjusts itself automatically over time as the average input voltage changes. When you use a voltage regulator chip you no longer need that kind of fairly old technology because the voltage regulator takes care of that.

 

With out knowing what linear regulator ..................
Ripple rejection for LM7805 and 7824. Extend the graph to the right to 1mhz, you can get an idea of rejection. (Cin 0.22uf Cout 0.1uf)
View attachment 325720
You need to add a CLC filter in the input. The L can be a bead. Choose values that will give good rejection in the 100khz to 10mhz range.

I was also thinking of this graph, power supply rejection ratio vs frequency. This shows that the power supply rejection ratio decreases with increase in frequency.

 

I just have asked a general question. Let's suppose a linear regulator has ripples at the input. The frequency of ripples are 10 kHz and 1 MHz. The power supply rejection ratio of the linear regulator is a function of frequency. Does this means that the linear regulator suppress the ripples more at 10 kHz then ripples at 1 MHz.

 

the video i linked goes over that in depth... including bench tests on a linear regulator using function generator, oscilloscope to show exactly what you are asking.

 

I just have asked a general question. Let's suppose a linear regulator has ripples at the input. The frequency of ripples are 10 kHz and 1 MHz. The power supply rejection ratio of the linear regulator is a function of frequency. Does this means that the linear regulator suppress the ripples more at 10 kHz then ripples at 1 MHz.

Offhand it seems that the 1MHz signals would be hard to pass but it really depends on the regulator.
The error amps in regulators are not going to be able to follow fast changing signals as fast as they might be, and that could allow fast signals to get through. An op amp though would not be able to pass a fast changing signal because of its bandwidth, and a single op amp could be used for a low current voltage regulator.

 

I just have asked a general question. Let's suppose a linear regulator has ripples at the input. The frequency of ripples are 10 kHz and 1 MHz. The power supply rejection ratio of the linear regulator is a function of frequency. Does this means that the linear regulator suppress the ripples more at 10 kHz then ripples at 1 MHz.

As with any other system where the power source was not battery-smooth DC, an inline filter would be the first step. At 1 MHz it would not require much inductance to reduce the ripple. And then shunt capacitors to reduce the remaining high frequency signal. Not a whole lot different from input filtering for electronics in an ICE car today.

 

I was also thinking of this graph, power supply rejection ratio vs frequency. This shows that the power supply rejection ratio decreases with increase in frequency.

That's not categorically true -- nor is it even true for that particular graph. Notice that it improves slightly from 10 Hz to 100 Hz before it peaks and then starts dropping off. Be very careful about extrapolating the behavior of a typical curve for a specific part to draw general conclusions about an entire class of circuits. The manufacture is merely presenting data for THEIR design for THAT part. It may well be that the reason that the ripple rejection peaks midway between 10 Hz and 100 Hz is that the designers of that regulator assumed that the part would frequently be used in circuits powered by 50 Hz or 60 Hz mains sources. Also, the curve is not just for that part, but (probably) for that part combined with specific external components and also under specific operating conditions.

One general statement that is often true of many, if not most, circuits is that many of their performance parameters show a band-like characteristic -- they peak (either at a high or a low) at some frequency and get worse in both directions from there. This is usually because they are designed to operate optimally within some frequency range and performance outside that range is whatever it is, which usually means it gets worse. At higher frequencies, in particular, intentional capacitances get overshadowed by parasitic inductances and vice-versa.

 

That's not categorically true -- nor is it even true for that particular graph. Notice that it improves slightly from 10 Hz to 100 Hz before it peaks and then starts dropping off. Be very careful about extrapolating the behavior of a typical curve for a specific part to draw general conclusions about an entire class of circuits. The manufacture is merely presenting data for THEIR design for THAT part. It may well be that the reason that the ripple rejection peaks midway between 10 Hz and 100 Hz is that the designers of that regulator assumed that the part would frequently be used in circuits powered by 50 Hz or 60 Hz mains sources. Also, the curve is not just for that part, but (probably) for that part combined with specific external components and also under specific operating conditions.

One general statement that is often true of many, if not most, circuits is that many of their performance parameters show a band-like characteristic -- they peak (either at a high or a low) at some frequency and get worse in both directions from there. This is usually because they are designed to operate optimally within some frequency range and performance outside that range is whatever it is, which usually means it gets worse. At higher frequencies, in particular, intentional capacitances get overshadowed by parasitic inductances and vice-versa.

Filtering, like most other parts of designs, is usually a compromise. Much of engineering is the art of compromise, in fact. Perfect designs are often possible but seldom practical. And even less frequently required. A measuring system must provide better than the required certainty and repeatability, but not infinite resolution and exactness. And a power supply, besides being able to supply adequate power and stability, needs only to not add noise to a circuit above it's intrinsic noise level. The sole exception is top-end audiophile equipment, which must impart zero noise or distortion no matter what the cost.

 

My point in #9 was that if the error amplifier in the linear regulator is by default not responding to fast changes present at the input then we can say that higher frequencies in the input will not be pass by the linear regulator ?

If the linear regulator can not do anything with the ripples at 1 MHz (this is the switching frequency of the DC-DC converter whos output is connected to the input of the linear regulator). What is the best way to remove ripples. Adding filter at the input of linear regulator or adding filter at the output of the linear regulator ? What kind of filter ? Common mode chock, EMI filter or just a low pass first order filter ?

I have seen common mode chock used at the input of the DC-DC converter. I am not sure if that also worth for linear regulators.

 

I suggest a series inductive filter in the supply side to the regulator, with an inductance to present an impedance of over 100 ohms at the ripple frequency, but also with a low DC resistance. That should not be too much of a challenge at 1 MHZ. Follow that with a shunt capacitance that will have a low impedance at the same frequency. That all will be in addition to the filtering normally used to block lower frequency noise.

 

You need to add a CLC filter in the input. The L can be a bead. Choose values that will give good rejection in the 100khz to 10mhz range.

C1 can be the output capacitors if the DC-DC converter. I would add a cap right at L.
C2 can be the input capacitors of the linear regulator.

WE-WAFB Sleeve Choke, 7427605 I just pick the first bead with good impedance at 1mhz I could find as an example.
At DC the L atts 0.005 ohms. At 1mhz it adds 10 ohm. The noise is shorted to ground by C1, the wire is opened up t 10 ohms, then C2 shorts out the noise again.
Approximately speaking: At 1mhz L1 and C2 made a voltage divider. If L1=10 and C2=0.1 ohm, then you should reduce the 1mhz noise by 100:1. (approx) Someone will point out the better math, but the simple math can be done in my head.

 

My point in #9 was that if the error amplifier in the linear regulator is by default not responding to fast changes present at the input then we can say that higher frequencies in the input will not be pass by the linear regulator ?

If the linear regulator can not do anything with the ripples at 1 MHz (this is the switching frequency of the DC-DC converter whos output is connected to the input of the linear regulator). What is the best way to remove ripples. Adding filter at the input of linear regulator or adding filter at the output of the linear regulator ? What kind of filter ? Common mode chock, EMI filter or just a low pass first order filter ?

I have seen common mode chock used at the input of the DC-DC converter. I am not sure if that also worth for linear regulators.

Hello again,

Imagine a simple voltage regulator that uses an op amp and series pass bipolar transistor. Now think about how regulation is achieved.

The op amp is using negative feedback so when the OUTPUT changes (note it already changed a little) the op amp detects this and products a drive signal to the transistor to counter that change in the output. The output then changes to a closer value to the reference voltage.
Since there is a delay, that means the output can change before the op amp can detect it. This can happen fast, but the op amp itself has limited bandwidth and slew rate, so it means the op amp will take some time to respond to a step change in input voltage or load current. If the op amp had a bandwidth of 1MHz and the input signal was 2MHz, the output would go up and down 2 million times a second and the op amp would not be able to follow that exactly. That would mean that it would have a hard time developing a new drive signal to counter those faster changes.
If there is any filtering however, that could easily swamp the 2MHz before it even causes the output to change much at all.

It's not too hard to model some of these systems we could look at it in more detail, but it's not that hard to imagine how this could happen.

Now given the op amp alone with no transistor, if the input changes at 2MHz, the op amp just like before has a hard time following that so much less would get to the output of the op amp itself, just like before. It does not have to control anything now though. That would mean the output would be somewhat cleaner. It's not hard to imagine this could be used as a voltage regulator, and in fact it has been used that way in many circuits.

 

Considering that I have had an LM7812 oscillating at several megahertz in the past, it is probably fast enough to cancel out a one megahertz switcher supply ripple. But if the linear regulator circuit uses a very low speed series pass transistor that might not be the case.