Some time ago, I used a pair of 7808 and 7908 to generate symmetrical reference voltage in my circuit. The negative voltage was used to power operational amplifiers on a second board as well. The whole thing worked about a year or so. Then, during a small modification which had to be measured with the second PCB disconnected, a video A/D converter suddenly died.
After a closer examination, I have found -14V at its reference input, instead of expected -8V. Well, I bought a new L7908 by ST. This time I have checked the voltage before placing in a new A/D converter. Again, there was -14V.
The rest of this article analyses the fact, that:
The 79xx voltage regulators series made by ST have to be connected to a load consuming at least 5 mA. Otherwise, their input voltage (decreased by about 8V) appears right at the output. So, according to our measuremets, you can get up to 20V at the output of a L7905 without a load.
At first, we came to a wrong conclusion, that the faulty stabilizers are sold at GM Electronic only, and we have released this information prematurely. Again, we do apologize to GM Electronic for releasing this information. The fault is definitely caused by ST, the manufacturer.
Then, we have studied carefully the datasheets. There, they do clearly state that the output voltage is guaranteed in a certain load current range; however, there is NO warning, that a voltage four times higher than the nominal output can appear at the output pin.
Specifications of the L7900 serises voltage regulators by SGS-THOMSON, according to an April 1997 datasheet:
We have contacted local subsidiary of the Motorola company since they develop and manufacture similar series of IC's here in the Czech Republic. After their advice we have tested whether this problem can be caused by oscillating of the stabilizer.
Recommended operating circuit is shown. We have measured using digital multimeters Metex M-3850 (input voltage) and Digitek DT930F (output voltage). A resistor TR505 1k/A (measured resistance 1.0142 kOhm) was used as a load. As a supply, we have used a BS 525 (DC power supply) and a transformer with one-way rectifier.
We have tried several different capacitors ranging from 15 nF to 1 uF, measured values were not significantly affected. Values listed below were measured using a 150nF capacitor.
Conclusion is, that oscillations are not the cause of our overvoltage problem.
The comparison in this article is not totally objective. In most cases, only a single sample of a given type by a given manufacturer was available; exceptions are MC7908 (3 samples measured) and, of course, stabilizers by ST (5 samples of each type measured; samples purchased at different times in different stores). The comparison is only an example what a developer can buy in an electronics components store.
Two stabilizers were used for comparison. The first one is L7908CV W990A9731 manufactured by ST Microelectronics Group or SGS-Thomson (abbreviation 'ST' is used in this article), and a MC7908CT QWG524 made by Motorola. Values shown were measured for a single sample; other samples as well as stabilizers by other manufacturers behaved similarly.
|Figure shows dependence of the output voltage on the input voltage for a L7908 and a MC7908 with no load, and with approx. 1k load. The ST stabilizer is clearly distinct.|
|Output VA characteristics of a L7908. Input voltage 26V DC. |
The marginal current 4mA corresponds to a 2k load.
It was not necessary to graph such a characteristics for a MC7908 since the output voltage remained almost constant.
|Current consumptions of both stabilizers, with and without a load. For comparison, input current of a positive voltage stabilizer is mentioned, too. Uin=20V, Rload=1k.|
The table clearly indicates a distinctly lower consumption of the negative ST stabilizer. This is probably the cause of the problem.
Since the 7905 stabilizers are used more often, several other parts are shown in the graph as well. The TA79005S is interesting too - it has the same fault. However, the fault does not occur for commonly used input voltages.
|Figure shows the dependencies of the output voltage on the input voltage for a L7905, TA79005S, KA7905; both with and without a 1k load. Again, the ST (and the TA79005S to a certain degree) behave differently.|
|Current consumption of individual stabilizers with/without a load.|
The situation is similar with other stabilizers. Stabilizers by ST differ, beginning from a certain voltage. So, we examined the datasheets more closely. The internal schematics of 79xx and 78xx are very different (you can see both internal schematics in datasheets, which are included at the end of this article). If ST really does use such an internal structure, the problem is most likely caused by unsuitable values of resistors around the 6.2V Zener diode at the output.
DOWNLOAD .PDF datasheet of a 79xx by National Semiconductors. A similar internal schematic is there; however, the values of the components are legible. Unfortunately, we could not get a sample to measure.
Since the internal structure is different, such a fault does not appear.
Possible cause of the problem:
The ST company can argue that this is possible; however, that the voltage from -4.5V to -5.5V is defined for load currents from 5 mA to 1A. This is true, BUT:
- The company says, that they make compatible parts. However, components by other manufacturers behave differently.
- In the LM317 datasheet by the same company, minimal load current is specified (see picture below). There is no such information in L79xx datasheets (see picture way above).
- Of course, the output voltage may fluctuate; however, it can't be four times the nominal output voltage.
- If this is an intention to reduce internal consumption of the circuit to a minimum, at the expense of problems with output voltage with no load, it should have been mentioned in the datasheet. Nothing like that is there.
LM317 stabilizer specifications by SGS-THOMSON. Here, minimum load current is specified:
Download & Links:
Please note the differences in internal sctructure. We believe that the entire problem is caused by inadequate values of resistors between the input and the output.
As I have said at the beginning of this article, I believe this fault is significant enough that we should inform about it.
The fact that we found this fault in ST parts only does not signify that other manufacturers' ICs are free from it. So, keep this possibility in mind when designing your own applications.
We will be glad to publish a response from the manufacturer, as well as readers' comments to this issue.
At the very end of 1998, the Czech subsidiary of ST has informed us that information about this behavior will be included in a future update of the respective datasheets.
Translated by: Joe Hlavac