Another small engineering investigation or the cry of Yaroslavna

“I knew it would be bad, but I did not know that so soon” (V. Tsoi)

I don’t know what was the reason for writing this post, probably it was just painful, but two events became the specific motive - one FROM abroad, and the other of domestic origin. It is not known which is worse, but the reader should judge it, so let's get started.

It all started with the fact that one of the products of our company did not turn on at low temperatures. Since the Murata power supply module used in this product was previously used in other products and did not cause any problems, for a long time they pester it on our board. However, after all possible explanations were dropped (as a result of checks and experiments), the last remaining option is correct, no matter how unlikely it seems (Sherlock Holmes method).

That is, it was suggested that the matter is really in the standard module from a well-known company. Indeed, experiments showed that the power module at certain values ​​of the input voltage (located inside the working area) and certain values ​​of the temperature of the module (also within acceptable) cannot turn on when working on a capacitive load that exceeds a certain value (and almost not an order of magnitude reaching the maximum allowable).

It turned out that this is not characteristic of a specific module, but of the entire batch, at least 10 modules that we tested gave the same picture. Moreover, in the process of the above tests, that is, being deep inside the permissible operating conditions, 3 modules are irreparably damaged. Of course, it would be extremely interesting to study modules from old batches in such modes, but by that time they were already unavailable, since they were in long-shipped products, so it was assumed that changes in behavior were associated with any changes in the production of these modules.

Our actions were quite unambiguous - they informed the dealer and waited for the arrival of his representative for conducting joint research. Well, while they were waiting, they decided to try to figure out what was happening themselves. Simplified the scheme to the limit. So, we connect an electrolytic capacitor with a nominal value of 2 * 680 microfarads to the output of the studied power module in parallel with a 56 Ohm resistor and look at the oscilloscope for the voltage shape on it.

We observe the following funny picture (again, some confusion with the flowers, but oh well):

"

So, after power is supplied, the module starts outputting the required voltage. But if he tries to give a voltage jump to the output, then this will require an infinitely large instantaneous current, since we have a capacitor at the output, the voltage at which cannot increase stepwise. In the real world, of course, it will take not infinitely large, but very significant current.

Since the module is designed for a very specific value of the output current, it implements protection against exceeding the output current (usually 110-120% of the maximum), it will work at the moment of issuing this significant current and turn off the source. In order to prevent this from happening, the module is equipped with a soft-start circuit, which at the initial moment limits the value of the output current to a level substantially less than the operating current (and, accordingly, the maximum), and then within a certain time starts to smoothly increase this current, which the graph shows the rising front is clearly accelerating.

When the current reaches the maximum permissible value, the protection circuit will work and the module will turn off. Rise time was 2.5 ms, which is not enough, but acceptable. True, the oscillogram shows that this current was only 5A, which again is not enough for a module with an output voltage of 5V and a power of 50W, but this is the second question. At least with the leading edge of the outfit, everything is clear.

This behavior of the power sources is typical when working at high capacitive load, while we observe characteristic steps - the capacitor is charged to a certain value, the module turned off, paused, tried to turn on again, the voltage across the capacitor did not manage to drop to zero, so the initial current is much less than the first time and the protection is triggered later, while the capacitor is still recharged and in 2-3-4 such cycles the voltage has returned to nominal and then only the leakage current goes through the capacitor, practically no effect Ya to work module. That is how the power modules of a domestic manufacturer, about which later, behaved in a similar scheme, and everything was fine.

But on the oscillogram we see a slightly different picture - the capacitor is charging, the protection is tripping, the module is turned off, but after that the voltage on the capacitor decreases sharply, while the discharge current exceeds the charge current by an order of magnitude - an almost vertical drop. In this case, the voltage decreases to 0, a new charge cycle repeats the first and we never enter the operating mode.

Unfortunately, the company does not provide schemes for its modules, and we are forced to enter into a shaky ground of guesswork and speculation. What could be causing this behavior? The answer that seems logical to me is the following. Currently, it is customary to use transistors (usually field-effect) instead of rectifier diodes in the output stages of transformer power supplies, the so-called synchronous rectifier circuit. This solution can significantly reduce losses on the rectifier element, instead of 1-1.3 volts direct drop on the diode, which at a current of about 10 amps will give 10-13 watts, we have 100-150 mV, which reduces the allocated power by an order of magnitude.

But the transistor, unlike the diode, must be controlled and there are different methods for this, historically the first of them was direct control. If we have a push-pull circuit, then transistor control from the opposite winding is widely used. This circuit has its advantages (simplicity), but has a greater switching loss than a direct control circuit. Modern microcircuits for building power supplies, as a rule, have additional outputs for implementing this transistor control mode, which is why it has become very popular, we can assume that it is implemented in our case.

But the field effect transistor differs from the diode not only in that the direct voltage drop on it is less, but also in that it is able to conduct current in both directions. This aspect is usually not paid due attention to, but in field effect transistors the drain and source are indistinguishable from the point of view of functioning (a strong statement, but true for a certain class of devices). Therefore, if we turn off the pump transistors, but do not take measures to forcibly turn off the output transistors (like diodes), then the output voltage will short circuit to the ground through these transistors through the drain-source circuit and, accordingly, the output capacitor will be very quickly discharged, which we and observe.

Further steps to verify this hypothesis are obvious: we turn on the diode between the output of the module and the capacitor (again, the field-effect transistor in diode mode, but with direct control) and everything turned out - the voltage for 3 cycles reaches the operating level. Nevertheless, we are not enthusiastic, because we had to modify the scheme, and, as they say, out of the blue.

A dealer representative comes to us, we show him a working diagram, a test bench, oscillograms with and without a diode and offer our version of what is happening. He begins to doubt, proposes to include a resistor of the order of 10 ohms instead of a diode, and this also helps. Well, it still would not help - we reduced the discharge current of the capacitor, as in the case of the diode. Nevertheless, it seems that we did not convince him, but nevertheless we fix the fact that the modules are inoperative in certain modes when applying the scheme recommended by the company. The proposal to put a series resistor and allocate 1W of power on it to compensate for the jambs does not seem successful to me.

Well, now I turn to the promised in the heading crying Yaroslavna. The representative leaves with signed acts, so what? After some time, a message arrives that they are ready to replace us with 3 failed modules at their own expense and EVERYTHING. No changes in the recommended connection scheme, no changes in the circuitry, no newsletter on the manufacturer’s website for other users - nothing. And this is not Pupkin & Co., it is a global manufacturer with a well-known brand, who, it seems, absolutely does not care about the defect found in their standard products (at least in a specific batch of products), which makes it impossible to use it in certain conditions. What is going on with the electronics industry? Did some koker finally win? And on a global scale? Here is such a cry of the soul from a middle-aged developer.

Well, the second part of the story is not so terrible, it just coincided with the first, which is why it got here. After this incident, confidence in the company's products decreased and began to look for options. Since we have already actively used power modules of one domestic company (since I am going to scold her, I will encrypt the name - it produces standard power modules with acceptance 5 and this is not IRBIS - it’s smart enough), and there were no complaints about the quality, we decided to look if they have anything suitable.

The fact is that their series of sources is quite good (and indeed work with this company left a good impression - normal documentation, reasonable technical support, if you know who to call), but these sources have dimensions that are significantly larger, since the specific power is 2- 3 times less. And here on the site a new series of sources is discovered that have a power density even slightly higher than that of MURATA (certainly less than VICOR, but this is a separate song, few compete with them, and they have enough of their cockroaches), has an acceptance 5, and we will carry out import substitution (I don’t know how anyone has it, but we have a sore point).

The layout of the legs, however, is somewhat different, but still standard. I am happy, I inform the management about such nice guys and their products, we rejoice together for a country in which there are still such cool developers, and we are beginning to estimate application options. And then I call this unnamed company in order to purchase samples and drive them in the whole range and it turns out that these products are not yet manufactured, even engineering samples with acceptance of the Civil Code are not available, acceptance 5, of course, will someday later, and orders for they will begin collecting the commercial version EARLY the second half of 2015. So my first question arose - guys, why cheat? After all, whoever needs it, he will know the truth - he will call you. Honestly write that this is a promising development, but your site says that the modules are being released (after a lively discussion with this issue with them, it is now indicated on the site in development). And the second question - is it really necessary to launch power supplies with unspecified parameters for launching even a new series, but still it takes more than six months? If they say to me “not earlier than the second half” of the year, I hear “at the end”, I will be glad to be mistaken. But this is so, a grumbling, but actually the guys are great, and if they withstand the declared parameters, then they will bite off their piece of the market from foreign companies, especially in connection with recent events.

I apologize in advance if I offended anyone, I messed up somehow, but I won’t wash anyway.