Hi, started again on an old project. I designed the PCB 2 yrs ago.
The circuit - attached.
What values to put for the capacitors HV-C2_ and HV-C3_?
100nf ?
In general it should have been good to parallel HV-C1, and mby the diodes in the bridge, but..
Tried to simulate in PSUD2, but the graph looks worse when I put the small caps.
Also, I must insert them as an RC circuit, with R of 0 ohm.
How to calculate/analyse the value of the caps without PSUD2?
If PSUD2 is to be used, what it could help it the situation?
I entered the transformer data - PSUD calculated impedance of 625 ohms.
I am missing the ESR of capacitors (stays default 2 ohm), and the rectifier leak (default 1 M)
The circuit - attached.
What values to put for the capacitors HV-C2_ and HV-C3_?
100nf ?
In general it should have been good to parallel HV-C1, and mby the diodes in the bridge, but..
Tried to simulate in PSUD2, but the graph looks worse when I put the small caps.
Also, I must insert them as an RC circuit, with R of 0 ohm.
How to calculate/analyse the value of the caps without PSUD2?
If PSUD2 is to be used, what it could help it the situation?
I entered the transformer data - PSUD calculated impedance of 625 ohms.
I am missing the ESR of capacitors (stays default 2 ohm), and the rectifier leak (default 1 M)
> good to parallel HV-C1
Why?
If C1 were "perfect" its impedance would drop with frequency to infinity.
Real caps are not perfect. Older e-Caps had significant resistance and inductance. In some specific cases (mostly radio) it became a custom to bypass them with *small* film or ceramic caps to ensure low impedance in the MHz range.
Modern eCaps are better.
Your tube circuit is high impedance and does not need super low impedance in the power supply.
PSUD is *not* the tool for this. It uses general trends and line frequency harmonics. Like any numeric computation, if you zoom-in small enough you find "random error" due to rounding errors in 32-bit math. This is not real.
Why?
If C1 were "perfect" its impedance would drop with frequency to infinity.
Real caps are not perfect. Older e-Caps had significant resistance and inductance. In some specific cases (mostly radio) it became a custom to bypass them with *small* film or ceramic caps to ensure low impedance in the MHz range.
Modern eCaps are better.
Your tube circuit is high impedance and does not need super low impedance in the power supply.
PSUD is *not* the tool for this. It uses general trends and line frequency harmonics. Like any numeric computation, if you zoom-in small enough you find "random error" due to rounding errors in 32-bit math. This is not real.
Software based simulations depends on models that not necessary are close enough to real parts. It may happen, for example, that smaller caps (100nF) units resonates with the inductance of higher valued e-lytics at some high frequency(es) and causes the circuit to became not properly decoupled. In some instances, a low value resistor in series with the 100nF units make the ciruit be more stable, damping those resonances, and acting as snubbers for the e-lytic's inductance.
If you use a SPICE program like LTspice, you can get fairly accurate SPICE subcircuit models for MLCCs from the manufacturer. Murata (Multilayer Ceramic Capacitors - SPICE Model | Murata Manufacturing Co., Ltd.) and Kemet (http://ksim.kemet.com/Ceramic/CeramicCapSelection.aspx) and probably other vendors provide them on their web site for many of their products, and some of these models are very accurate. They can accurately model their self resonant frequency, and while they do not directly model the effects of DC on the dielectric, you can specify the DC bias and obtain a model that will work well around that bias level.
IMHO, it's worth learning how to use LTspice. It's ideal for this kind of work and is fairly fast if everything is working correctly. Plus, many vendors provide SPICE models for their components, allowing you to use much more accurate components in your simulations. Just guessing a simple ESR is pretty primitive - why not have a cap model that is made from 8-10 components and actually models self resonance and the real ESR value?
IMHO, it's worth learning how to use LTspice. It's ideal for this kind of work and is fairly fast if everything is working correctly. Plus, many vendors provide SPICE models for their components, allowing you to use much more accurate components in your simulations. Just guessing a simple ESR is pretty primitive - why not have a cap model that is made from 8-10 components and actually models self resonance and the real ESR value?
This subject has already been explored exhaustively (and seriously) in a number of threads, this one for example:
paralleling film caps with electrolytic caps
To summarize: paralleling can exceptionally be beneficial, but it is a rare instance.
If the ratio of the paralleled capacitors is >100, the effect is always detrimental (which does not imply that a ratio <100 is always beneficial).
paralleling film caps with electrolytic caps
To summarize: paralleling can exceptionally be beneficial, but it is a rare instance.
If the ratio of the paralleled capacitors is >100, the effect is always detrimental (which does not imply that a ratio <100 is always beneficial).
0nF will be the best value, as Elvee says. Unless, of course, your aim is not best electrical performance but best marketing performance.emosms said:
- Status
- Not open for further replies.