The DC Bias on ceramic capacitors varies insidiously even at low voltages and frequencies. Now a days most power supplies are of the switching type running between 50 kHz to 500 kHz (i.e. Buck / step down and Boost / step up). Look at any switching regulator IC data sheet and the example circuit will likely have 2 - 8 ceramic capacitors in parallel at the output for storage and ripple reduction. Typically they are X7R or X5R and often caution against using Z5U or Y5U types. So typically, depending on voltage, the capacitors are 25V or 50V rated units. If the design is anywhere near the rated voltage (i.e. 24V or 48V) the actual value of the capacitance is often < 10% of the rated and marked value of the capacitors. At best the DC output will have a lot of noise, but usually there will be a ton of ripple (even at low current), poor regulation and oscillations. This is "insidious" because it all gets worse at low loading, so bursts of oscillations and noise on the DC output will occur when the load current dips low. Unfortunately most manufacturers of ceramic capacitors do not publish any detailed information about their products. Taiyo Yuden and Murata do have graphs and Murata also has an excellent simulation tool called "Simsurfing" to get exact capacitance value at a given voltage for a specific capacitor. An interesting note about this is that TI makes many ICs for these type of regulators, and often has a PCB for demonstration or evaluation to showcase how they work and help with R&D. The Demonstration PCBs are for internal use to develop the APP Notes, while the Eval boards are available to buy. For one of their ICs they published an APP Note which had a photo of the PCB on the front page showing the parts and output capacitors mounted on a custom designed PCB. The design and calculations called for two, paralleled 4.7 uF MLCC output capacitors to meet the design performance specs. Unfortunately, the measured output was laden with ripple and occasional oscillations, so they kept adding more and more capacitors in parallel to the PCB until it worked as designed and expected. The photo on the cover showed and additional 9 capacitors tacked on to the PCB in an arc to match the existing layout. 😂 After this embarrassing development, TI would always list the exact manufacturers part numbers for the capacitors before publishing the Notes. Summary, when using MLCC X7R capacitors, use the highest voltage parts you have room for and can afford. If this means 100V parts on a 12V DC supply with a greatly varying load, so be it. Hope this helps someone. P.S. all this can be avoided by using COG or, for us old guys, NP0 capacitors, but they are made of unobtanium, cost a fortune and are >10x the size. Go ahead, find some 4.7 or 10 uF MLCC SMT Ceramic COG capacitors and see what you can get 🤣
Pretty insightful, ceramic caps are way more quirky than I was ever let on in university. Wondering if this behavior could be exploited similar to how a varicap diode is used. Might not be the best option or even practical but it's probably plausible.
No. The DC Bias effect of a capacitor is due to the material used for the dielectric, the diode is not. These capacitors also have, in addition to the DC Voltage bias, (compared to other dielectric capacitors) a large variance with temperature and age and the variance is not consistent from part to part and manufacturer to manufacturer.
I'm not sure. If the only goal was to say, design a crystal controlled oscillator that was phase locked to a fixed reference and using this effect to tune it, then is does seem possible. You would need to design the oscillator such that the DC bias is not causing other problems. Certainly doable. If you plan to attempt it, consider posting about it.
These are often treated as a special case in DC circuits referred to as DC transients. The time constant of the circuit will tell you if the transient is relevant or not.
@@joesmith-je3tq, here are my thoughts on it. I'm self-taught, so I'm sometimes wrong about such things, and my apologies if I don't use engineering standard terminologies. This is a bit long-winded, but I don't know a simpler way to describe it. : I think both statements are true. In practice, in the normal application of DC blocking capacitors the currents flowing through them due to noise and slow voltage changes are small enough to have no detrimental effects on the operation of the circuits. But they still exist, because fluctuations in voltage still exist. What matters is the amplitude and rate of those fluctuations from our DC source. We intentionally select blocking capacitors of specific values and types so that they will reduce such voltages and currents to below levels that would cause our circuits to operate incorrectly. But a capacitor is not actually a device that responds to a 'frequency'. That's an incomplete definition. It is a device that responds to a different density of electrons between its 2 plates. A typical applied 'frequency' makes that very obvious since it's a nonstop, typically very fast change in the electron density, so we typically use frequency in our calculations to determine how much current will 'pass' through a given capacitor. But any change in the applied voltage on one plate of a capacitor in a circuit will cause a change on the other plate of the capacitor, a current flow to or from. The amount of current flowing to or from the second plate is just less and less as the rate of voltage change on the applied voltage side is reduced. In most circuits a very low rate of change, such as manually changing the applied voltage from a power supply has no noticeable effect on the output. But in this case where a high-voltage supply is being used, hundreds of volts are changing per second as you crank it between say 1000 volts and 1500 volts for instance, and so a much higher current is flowing to or from the second capacitor plate than one is accustomed to with a low voltage circuit, and it becomes noticeable. As you are adjusting the voltage it is no longer DC, and so the capacitors can no longer fully block it. I'm sure you can measure a voltage across a load while raising and lowering the DC voltage by hand on the HV power-supply even though you have DC blocking capacitors in place. And of course the voltage measured will be substantially higher with a lighter load.
@@johnwest7993 I have an old battery charger for example that is nothing more than a transformer with various taps, a meter, shunt and rectifier. By your thinking, my batteries are charged with AC, which again is not true. Here's a Wiki article. I did not take the time to read it but it should help you. en.wikipedia.org/wiki/Direct_current
![NoCap - Yacht Party [Official Music Video]](http://i.ytimg.com/vi/rm_lvDln8kE/mqdefault.jpg)
The DC Bias on ceramic capacitors varies insidiously even at low voltages and frequencies. Now a days most power supplies are of the switching type running between 50 kHz to 500 kHz (i.e. Buck / step down and Boost / step up). Look at any switching regulator IC data sheet and the example circuit will likely have 2 - 8 ceramic capacitors in parallel at the output for storage and ripple reduction. Typically they are X7R or X5R and often caution against using Z5U or Y5U types. So typically, depending on voltage, the capacitors are 25V or 50V rated units. If the design is anywhere near the rated voltage (i.e. 24V or 48V) the actual value of the capacitance is often < 10% of the rated and marked value of the capacitors. At best the DC output will have a lot of noise, but usually there will be a ton of ripple (even at low current), poor regulation and oscillations. This is "insidious" because it all gets worse at low loading, so bursts of oscillations and noise on the DC output will occur when the load current dips low. Unfortunately most manufacturers of ceramic capacitors do not publish any detailed information about their products. Taiyo Yuden and Murata do have graphs and Murata also has an excellent simulation tool called "Simsurfing" to get exact capacitance value at a given voltage for a specific capacitor.
An interesting note about this is that TI makes many ICs for these type of regulators, and often has a PCB for demonstration or evaluation to showcase how they work and help with R&D. The Demonstration PCBs are for internal use to develop the APP Notes, while the Eval boards are available to buy. For one of their ICs they published an APP Note which had a photo of the PCB on the front page showing the parts and output capacitors mounted on a custom designed PCB. The design and calculations called for two, paralleled 4.7 uF MLCC output capacitors to meet the design performance specs. Unfortunately, the measured output was laden with ripple and occasional oscillations, so they kept adding more and more capacitors in parallel to the PCB until it worked as designed and expected. The photo on the cover showed and additional 9 capacitors tacked on to the PCB in an arc to match the existing layout. 😂 After this embarrassing development, TI would always list the exact manufacturers part numbers for the capacitors before publishing the Notes.
Summary, when using MLCC X7R capacitors, use the highest voltage parts you have room for and can afford. If this means 100V parts on a 12V DC supply with a greatly varying load, so be it. Hope this helps someone.
P.S. all this can be avoided by using COG or, for us old guys, NP0 capacitors, but they are made of unobtanium, cost a fortune and are >10x the size. Go ahead, find some 4.7 or 10 uF MLCC SMT Ceramic COG capacitors and see what you can get 🤣
Pretty insightful, ceramic caps are way more quirky than I was ever let on in university.
Wondering if this behavior could be exploited similar to how a varicap diode is used. Might not be the best option or even practical but it's probably plausible.
No. The DC Bias effect of a capacitor is due to the material used for the dielectric, the diode is not. These capacitors also have, in addition to the DC Voltage bias, (compared to other dielectric capacitors) a large variance with temperature and age and the variance is not consistent from part to part and manufacturer to manufacturer.
I'm not sure. If the only goal was to say, design a crystal controlled oscillator that was phase locked to a fixed reference and using this effect to tune it, then is does seem possible. You would need to design the oscillator such that the DC bias is not causing other problems. Certainly doable. If you plan to attempt it, consider posting about it.
Just a thought…. How about using a signal generator and set it to the resonant frequency of the crystal.
That is basically what I have been showing. Read up on what a VNA is and how it works.
An interesting thought: When you are raising or lowering a DC voltage you no longer have a true DC voltage.
We can exp[and on your logic to suggest that because all things have noise, there is no such thing as DC. Of course, both statements are false.
These are often treated as a special case in DC circuits referred to as DC transients. The time constant of the circuit will tell you if the transient is relevant or not.
@@joesmith-je3tq, here are my thoughts on it. I'm self-taught, so I'm sometimes wrong about such things, and my apologies if I don't use engineering standard terminologies. This is a bit long-winded, but I don't know a simpler way to describe it. :
I think both statements are true. In practice, in the normal application of DC blocking capacitors the currents flowing through them due to noise and slow voltage changes are small enough to have no detrimental effects on the operation of the circuits. But they still exist, because fluctuations in voltage still exist. What matters is the amplitude and rate of those fluctuations from our DC source. We intentionally select blocking capacitors of specific values and types so that they will reduce such voltages and currents to below levels that would cause our circuits to operate incorrectly. But a capacitor is not actually a device that responds to a 'frequency'. That's an incomplete definition. It is a device that responds to a different density of electrons between its 2 plates. A typical applied 'frequency' makes that very obvious since it's a nonstop, typically very fast change in the electron density, so we typically use frequency in our calculations to determine how much current will 'pass' through a given capacitor. But any change in the applied voltage on one plate of a capacitor in a circuit will cause a change on the other plate of the capacitor, a current flow to or from. The amount of current flowing to or from the second plate is just less and less as the rate of voltage change on the applied voltage side is reduced. In most circuits a very low rate of change, such as manually changing the applied voltage from a power supply has no noticeable effect on the output. But in this case where a high-voltage supply is being used, hundreds of volts are changing per second as you crank it between say 1000 volts and 1500 volts for instance, and so a much higher current is flowing to or from the second capacitor plate than one is accustomed to with a low voltage circuit, and it becomes noticeable. As you are adjusting the voltage it is no longer DC, and so the capacitors can no longer fully block it. I'm sure you can measure a voltage across a load while raising and lowering the DC voltage by hand on the HV power-supply even though you have DC blocking capacitors in place. And of course the voltage measured will be substantially higher with a lighter load.
@@johnwest7993 I have an old battery charger for example that is nothing more than a transformer with various taps, a meter, shunt and rectifier. By your thinking, my batteries are charged with AC, which again is not true. Here's a Wiki article. I did not take the time to read it but it should help you. en.wikipedia.org/wiki/Direct_current