Transistor pairing accuracy

Every time I see a so-called high-current power amplifier with dozens or even twenty pairs of power crystals in parallel, I am secretly surprised: Are all these power crystals paired? Because whether they are paired or not is one thing, and whether they are "accurately" paired is another.

Whether it is American or Japanese crystals, there is an inherent disadvantage: the hFE values ​​of NPN and PNP are very different, especially power crystals. If only one pair of power crystals is used, there is no need to select pairs, because the push-pull action will balance them with each other. But if multiple pairs are used in parallel, then NPN should be paired with NPN first, PNP should be paired with PNP first, and then NPN should be paired with PNP.

Because according to the above pairing, no matter how many pairs you connect in parallel, it looks like there is only one pair. Is this full pairing possible? It is very difficult, very difficult, really difficult. Imagine if ten pairs are connected in parallel, wouldn't the hFE of all 20 crystals be exactly the same? This is not just ordinary difficulty, it is almost impossible.

The hFE value of small signal crystals is much higher than that of power crystals, so the error in the numbers can be allowed to be relatively large. For example, if the hFE of two small crystals is measured to be 180 and 190 respectively, the difference is 10, but the error rate is 5.5%. If the hFE of two power crystals is 50 and 60 respectively, although the difference is also 10 in numbers, the error rate is 20%!

Here are three key points for netizens to refer to: First, it is best to pair NPN and PNP - although it is extremely difficult; second, the pairing of power crystals is more stringent than that of small signal crystals, so as to reduce the error rate and improve accuracy; third, use high hFE power crystals. Power crystals from Motorola in the United States and Sanken in Japan were once the two mainstreams, and other brands could only be selected sporadically, but their hFE was relatively low. Later, Toshiba launched high hFE power crystals, and Motorola in the United States simply purchased them directly from Toshiba; by 2005, Toshiba power crystals had become the mainstream.

How to pair? You can use a curve tracer or an ammeter + voltmeter. The method is not difficult, but you need to have a large amount of them to make accurate pairing. Therefore, imported machines use more than a dozen pairs of parallel connections at once, and the error will never be less than 10%! Because the power crystal pairing error is high, it is easy to be unstable if multiple pairs are connected in parallel, so the emitter resistor value is increased, even to 0.47Ω or more.

In terms of characteristic specifications, there is an obvious difference between power crystals made in the United States and Japan, that is, the leakage current Icbo. American products are much higher than Japanese products. Using a three-meter, some Motorola power crystals have a slight leakage current, especially the 2N3440/5415 pair of small iron shells packaged in TO-5, which is definitely the "leakage king" among Motorola crystals! Don't you believe it? Use the highest resistance level (x10K) of the pointer meter to measure the CE pin to find out.


When choosing the constant current diode CRD and the small signal FET, you will encounter the problem of "one decimal place or two decimal places". If it is one decimal place, such as 4.8mA and 4.9mA, it is enough to prepare two or three hundred pieces. If you need to get two decimal places, such as 4.83mA and 4.84mA, you need to prepare at least thousands of them. FET selection requires not only accurate Idss, but also accurate Vgs. It is guaranteed that you can't pick a few pairs out of a thousand!

At present, the lowest error of electronic components for audio is resistors. The commonly used products are 1% (+/-), and you can also buy grades below 0.5%, so you don't need to spend time to choose. But what about capacitors? Small plastic capacitors have as low as 5%, and large-capacity filter capacitors are not less than 20%. So whether the capacitors should be matched is also a headache. If an electronic crossover system is used, then the RC components used for frequency division, especially capacitors, are best selected!

Vacuum tubes, such as 300B, EL34, and 6550, are sold in pairs, some with two tubes in a pair, and some with four tubes in a set. But is it possible to match dual triode tubes like 6922 and 5751? There is often a 7% error between the two groups inside! There is no way to solve it, and it is also a headache.