|
Arabic
Bengali
Chinese
English
French
German
Hebrew
Hindi
Italian
Japanese
Korean
Malay
Polish
Portuguese
Spanish
Turkish
Ukrainian
Vietnamese|
ENCYCLOPEDIA OF RADIO ELECTRONICS AND ELECTRICAL ENGINEERING Pulse voltage converters. Encyclopedia of radio electronics and electrical engineering
Encyclopedia of radio electronics and electrical engineering / Voltage converters, rectifiers, inverters Myths, fairy tales, legends and toasts about pulse transformers There are many myths around the world about high-frequency power transformers and chokes. Let's try to debunk them. Unfortunately, the least articulate part of textbooks and manuals is associated with magnetic components, complicating generally simple everyday objects and phenomena. Yes, there are many unknown variables, yes, there are many subtleties that you need to know, but the theory is silent about them, and popular literature lies, offering empirical formulas for specific problems as solutions for all occasions. For example. Myth one. The greater the percentage of the core window area filled with copper - ideally 100% - the better. Wrong. In many designs, 100% coverage, compared to say 75% (same number of turns, different wire size) will lead to higher RF losses. You cannot blindly transfer calculation methods from 50 Hz to 500 kHz. Second myth. In an optimal transformer, winding resistance losses and core losses are the same. Wrong. Often one loss figure differs from another by 1-2 orders of magnitude. So what - this is not the main criterion for the designer. This approach is also a legacy of "fifty Hertz" - this is how thermal equilibrium is ensured in massive network transformers. And we have the entire winding - one or two layers, and the heat transfer conditions are much simplified. The third myth. The leakage inductance should be 1% of the magnetizing inductance. Wrong. It should be as low as possible without significantly degrading other important parameters. You can bring it up to 0.1% - great. And sometimes you have to stop at 10%. Fourth myth. The leakage inductance is a function of the permeability of the core. Wrong. The leakage inductance of the winding is practically independent of whether there is a core in the coil or not. More precisely, the whole difference fits into 10% (and this is with mu of several thousand!). You can check. Fifth myth. The optimum current density in the windings is 2A per sq.mm. Or 4A. Or 8A. And the dog is with him. Current density doesn't matter. What matters is the heat dissipation in the wire, and the ability, or inability, of the design as a whole to provide thermal balance at an acceptable temperature. Depending on the efficiency of cooling (from radiation into vacuum to cooling in the boiling phase), the allowable current density changes by two orders of magnitude. Ridley has been building transformers for 20 years, but we never found out the "optimal current density" - only the temperature of the transformer matters to us. Myth Six. In an optimal transformer, the losses in the primary and secondary are equal. Wrong. And if not equal, then what? The main thing is that none of them overheat. Seventh myth. If the wire diameter is less than the skin effect depth, then there are no significant RF losses. A very harmful statement. In multilayer windings, even with a very thin wire, there will be losses. Myth Eight. The resonant frequency of the transformer circuit in the absence of load must significantly exceed the conversion frequency. Wrong. She doesn't matter. In an ideal transformer, the inductance tends to infinity, so the resonant frequency for an open circuit tends to zero ... so what? And the fact that resonance is important is not for an open, but for a short circuit of the secondary circuit. This resonance should be at least two orders of magnitude higher than the carrier frequency. Transformer impedance measurements Device connection option
In this configuration, the analyzer displays the transformer impedance from 10Hz to 15MHz, for load short circuit and load open conditions. For pulse transformers with short windings, it is necessary to provide a short circuit along the shortest path with minimal losses. After all, the closing semi-ring, even with a diameter of several centimeters, already has an inductance comparable to the leakage inductance of the primary. The leakage inductance depends on the frequency! As a ballast Rsense R=0.1..1 Ohm. Measure the ohmic resistance of the windings only with a low-resistance bridge or an ohmmeter with a current generator. Having carried out a cycle of measurements, it is possible to determine: Magnetization inductance - Winding resistance - Leakage inductance - Resonance frequency and quality factor for short circuit and open circuit - Winding capacitance (up to 3 pF per turn).
Per-tact current limitation, correctly implemented, allows you to create an unkillable PN. To do this, the current sensor must be fast (latency of a few nanoseconds), and be loaded directly on the control input of the controller IC. False tripping of the protection against parasitic bursts is suppressed by the RC low-pass filter. Here it is necessary to decide on a compromise between speed and noise immunity, so that excessive filtering does not miss the real excess current. Controllers with protection off at the leading edge of the pulse are also not a panacea. Those 100ns delay (or so) during which the defense is blind can also kill the PN. Therefore, it is advisable to forcibly limit the switching speed of the transistor (which also reduces the level of interference and radiation both into the current sensor and into space). How to test current protection? Short-circuit the PN output - after the rectifier and output filter. Unfortunately, with a short circuit in the rectifier itself, no current protection will help your transistors. Connect the probe to the current sensor. Gradually increase the supply voltage until the controller starts to generate a carrier. On the oscilloscope, you should observe narrow peaks - the protection circuit should quickly turn off open transistors. The amplitude of the pulses must correspond to the protection threshold. Raise the supply voltage to the maximum. The duration of the pulses should narrow. The amplitude can grow (due to delays in the propagation of the current feedback), but not significantly. And if it grows in proportion to the input voltage - stop, your OS is too slow. Then - this is important - the measurement cycle should be repeated at minimum and maximum air temperatures This is important: the parameters of the ferrite, on which the current transformer is wound, can float away with temperature so much that it will not seem small. Snubber (snubber - damper) - RC circuit parallel to the winding - for shunting high-frequency ringing. The ringing must be suppressed, otherwise failures, excessive pickups and instability of the converter are possible. Typically, an RC shunt is sufficient to quiet unruly windings if the ringing frequency exceeds the carrier by about two orders of magnitude or more. And if not, then you need to look for workarounds, because then a significant proportion of the carrier and its nearest harmonics will fall into the shunt bandwidth. First. Determine the frequency of parasitic oscillations. To begin with, run the circuit at a low load current. The oscilloscope probe - in order not to make changes to the circuit - must have a minimum self-capacitance. If not, try bringing the probe close to the ringing circuit with no electrical contact. Please note that the ringing frequency floats with the primary circuit voltage. Second. Calculate the equivalent RLC circuit for frequency and Q factor. From the primary side, the leakage inductance is known (should be known!) On the secondary side, the capacitances of the diodes are known. Characteristic impedance Z = 2 * Pi * f * L (for known L), Z = 1 / (2 * Pi * f * C) for known C Third. For starters, let's try just the R-shunt, R=Z. Let's calculate the heat losses on the shunt. If they are indecently high, we supplement the link with the capacity C=1 / (Pi * f * R). Increasing the capacitance is useless - losses increase, ringing suppression does not improve (capacitance at RF is completely conductive). Fourth. Let's recalculate the loss power by R : P = 2* C * V * Fcarrier - this is the loss of only the carrier without heat generation on the ringing. We check in a real circuit. The first approximation - as a rule - is immediately suitable for most cases. Component placement and routing near the IC is critical! This is repeated in each datasheet, but it does not interfere with repeating again. First of all - the frequency-setting capacity of the generator. Place it at the very foot of the IP. Not five millimeters, but the closer, the better. Otherwise, inexplicable phenomena are possible - for example, a circuit designed for 100 kHz will generate at megahertz, a mermaid will come out of the Yauza, etc. Moreover, it may not emerge on the prototype, but in the serial board it will appear in all its glory. Secondly, the capacitances in the power circuits should also be soldered as close as possible to the legs of the IC. The output of a generator saw (where it is accessible from the outside) does not like to be loaded (as do I). Therefore, when selecting a signal from this output, be careful - even a 100 kΩ load can change the shape of the saw. It is most correct to generate the saw in parallel, without connecting to the primary circuit of the generator. IS 3842, 3843 allow you to set a pause between pulses from 5% to 30% of the period. 3844, 3845 - up to 70%. If you need to lengthen the pause, you can work around these limitations by changing the timing of R, C. Then adding another resistor from the RTCT pin to the power plus will speed up the charge and slow down the discharge, lengthening the available pause time. IC UC3825 - the minimum pause time (absolute, in milliseconds) is rigidly set by the capacitance Сt, see the documentation. But it is possible to do as described above - by connecting a resistor to Ct. That's just the time it will float all the time with the supply voltage. IC output drivers do not like inductive loads - such as isolation transformers - which cause gate bounce. Moreover, if it does not manifest itself in the laboratory, then in real life it will definitely come up at the most inopportune moment. After all, the parameters of the transformer float ... Therefore, it is recommended to protect the gate with diodes, and in parallel with the primary of the transformer - with a resistor. First-generation controllers, especially older ones, are extremely unstable both in terms of reference voltages (you can live with this) and in terms of timing, up to an incorrect sequence of triggers and excessive drift of the carrier frequency (depends on the stability of the reference levels). If you want - use IS either a recent year of manufacture, or with suffixes indicating "improved" options. Those. TL594 not TL494 etc. For example, an undocumented feature of the Bryansk ICs KR1156EU2 (analogue 3825) - with 12V power supply, correct wiring, with an inhibiting level at the ILIM input, output 14 is at a low level (normal) and short, approximately 11 ns peaks creep through output 100 - "undercut" fronts of the carrier amplitude up to 9V. Somewhere the trigger is not working properly. But these cuts are enough to open the shutter and (and suddenly) kill the circuit. About the OC loop cutoff frequency About measuring the gain of a closed-loop FN - it is best to measure it as described in the next section, using a spectrum analyzer (an oscillator is not enough). For forward and reverse PV with voltage control, the cutoff frequency should be no more than a quarter of the zero frequency of the transfer function on the right half of the complex plane. If the fulfillment of this condition does not allow to reliably stabilize the output, then it is necessary to redo the output filter. For all PNs, the cutoff frequency should not exceed 1/8 of the carrier frequency. The increase in the cutoff frequency is limited by the inevitable noise, ringing and other parasitic phenomena in the PN to a level of about 15 kHz. If for any reason you need to understand it, the inevitable complication of the circuit is the introduction of an external, high-speed error amplifier in the OS loop. Most importantly, the cutoff frequency of the OS is not an end in itself. Important output impedance in the frequency range required by the load, suppression of input voltage instability, and suppression of input noise. Be sure to measure the behavior of the feedback loop before putting the instrument into service. The device, which is discussed below, introduces a voltage source (sweep generator) into the OS circuit break (points 1-2). Then the signal spectra are recorded at any two points of the circuit and the frequency response of the ratio of these spectra is displayed. The ratio of the output spectrum to the input spectrum is the transfer characteristic (in amplitude). You can repeat the device qualitatively using a generator with a transformer output and voltage stabilization on the secondary winding, and an oscilloscope. Measurement of loop parameters by spectrum analyzer АР102В - PN with optocoupler decoupling
The connection points of the probes of channels A and B allow you to measure various transfer functions
Measurement of loop parameters - PN without galvanic isolation
A-1 B-2 : loop gain A-3 B-2: amplification of the power node and modulator A-1 B-3 : boost (cut) of equalization circuit Always ground the circuit being measured. If its primary circuit is galvanically connected to the network, connect the measuring instruments to the network through an isolating 1:1 transformer (not LATR). If grounding is not possible, isolate the analyzer inputs. Better not just with a capacitance (it can fly out), but through a special decoupling amplifier. At lower frequencies, use the maximum output of the oscillator, and when passing through the feedback cutoff frequency, it is worth reducing it, while making sure that the circuit does not go into overexcitation. Above 30 kHz, measurements are not very reliable due to grounding and interference problems. In any case, the oscillator signal must be injected into that part of the circuit in which there are few variable components from both the carrier frequency of the PN and the mains frequency. Device frequency response example
Switching power supply failures Very unpleasant events. Many components of a pulsed voltage regulator operate at the limit of the safe operation area, and when one element flies, others die behind it, destroying the very reason for the failure. And looking for her in the dark is not fun. Here is a short list of the main reasons known to professionals (which, however, are silent ...). A. Current overload of the key - either the transistor crystal dies, or the wire between the crystal and the leg burns out. Therefore, operational current protection is necessary, regardless of power. The lack of current protection often shortens the life of the device. Knowing the construction of the PN of car amplifiers, which, as a rule, do not have per-cycle current protection (IS TL494), the reader has the right to be indignant! The dog, it seems to me, is where he rummaged. On the one hand, PN with current protection imposes higher requirements on the accuracy and coordination of all components of the path, and performing them in the automotive temperature range will lead to an increase in the cost of the amplifier. On the other hand, at 12V primary power supply and a real (short-term) MIS current limit of about 50 ... 250A per arm (1 ... 4 good transistors), the current - taking into account all the resistances of the circuit - is simply not able to reach destructive values long-term work on a short circuit, which will lead to fatal overheating). Compare this with a network power supply, where the primary is 300V, and the current limit (at the same power to the load) is 5 ... 25A. B. Gate-drain overvoltage. MOS transistors from good houses - IR, Motorola (let's add SGS-Thomson and Infineon to the list) are not so easy to kill. They hold current and drain-source voltage overloads, but gate overloads will kill them too. The gate driver must be guaranteed to keep the voltage in a safe zone, if necessary, install zener diodes. We do not recommend the use of integrated high-side drivers in high-voltage circuits. Better - transformers, they are more resistant to interference. B. Most often, the circuit dies when turned on. After all, when you turn on the output capacitance is discharged - the circuit "sees" a short circuit. Your current protection should work quickly enough even with extremely high input voltage. "Soft start" controller does not save from this scourge! D. The built-in "anti-parallel" diode of the MIS key is a source of problems. He is slow. Let this diode conduct current, this is not fatal, but during the conduction of the diode, a rapid change in voltage to the opposite is unacceptable, if at the moment of change the gate voltage is not applied. A similar failure often occurs in a full-bridge circuit. Upon completion of the conducting state, the leakage inductance generates a bounce, and at its first peak, the source voltage may exceed the supply voltage - the diode will turn on. Well, okay, now these transistors open anyway. But if at the second - negative - peak of the bounce - and on the opposite shoulder, the diodes also open, breakdown cannot be avoided. Solution - put snubbers. E. Check if the controller's power-on undervoltage protection works properly. In controller ICs, it is quite reliable. And in other components (comarators, drivers, etc.) - it is not known. The requirement is simple - when the power is turned on, the controller as a whole must be set to the standby state, on the gates of all power switches - a strictly locking level. E. Failures of high-voltage capacitors at high temperatures. G. Failure of Schottky diodes due to excessive reverse voltage (assuming sufficient heat dissipation). A reduction factor of 80% for voltage is a useful safety net. I explain. A feature of the LH is the exponential growth of the reverse current with temperature. In many applications, reverse current dissipation is comparable to forward current losses (up to 20%)! Next comes the chain heating and the diode dies. Therefore, power LHs are more critical to heat dissipation than conventional diodes. 10. Use the right tool. You need a high-speed storage oscilloscope that captures single pulses. After all, the MIS key can collapse in XNUMX nanoseconds, and one must be able to see this. It is important to properly connect the oscilloscope ground as well. If there are a couple of transistors, a trans and a rectifier in the circuit, why not take it and model it head-on? It's no more difficult than modeling an LSI for a million transistors. That's a good question, but that's all - there is simply no suitable software, and the data for calculating transformer models will still have to be taken manually. Of those known to science and practice, the best for our purposes is an analog computer that you will have to build yourself - a Breadboard. And nothing compares to him. Firstly, no simulation will take into account many parameters that are critical for PN, especially those that go beyond the boundaries of real wires and components (heat exchange processes, EM radiation). After all, many of these factors are determined by the location of components and traces on the board - they cannot be taken into account without building it. The same resistance and inductance of the wire from the key to the winding is a critical component of any PSU. And, secondly, the models inside the traditional CAD system are not designed for the correct processing of large-amplitude impulses, and often they simply do not converge to a solution. The role of modeling in the design cycle. Is it worth getting involved with modeling then? It's worth it, but you always need to remember (and know, of course) the limitations of CAD models. Here's how to use them >Publication: klausmobile.narod.ru
Machine for thinning flowers in gardens
02.05.2024 Advanced Infrared Microscope
02.05.2024 Air trap for insects
01.05.2024
▪ Helium discovered outside the solar system ▪ The fastest jaws in the world ▪ Creating the Perfect Chocolate Texture
▪ section of the site Data transfer. Article selection ▪ article Vyazemsky Petr Andreevich. Famous aphorisms ▪ article Who wanted to create a helicopter powered by gunpowder? Detailed answer ▪ Article Eucalyptus blue. Legends, cultivation, methods of application ▪ article Range 160 meters in Radio 76. Encyclopedia of radio electronics and electrical engineering
Comments on the article: Michael Great ![[!]](https://diagram.com.ua/comments/smilies/excl.gif){kind=small}
Home page | Library | Articles | Website map | Site Reviews
www.diagram.com.ua |
See other articles Section 
Latest news of science and technology, new electronics:
All languages of this page