In post #2 the 64dB ripple rejection is from a TI datasheet.
64dB typical at 120 Hz Vo=10V.
I did not use the graph dB versus frequency.
Where they bugged, the curves for Cadj=0 and Cadj=10uF are in the wrong order.
Furthermore, I guess the raw figure, 64dB is for a 120 Hz saw tooth shape ripple, while the graph is for a sine AC ripple at Vin.
64dB typical at 120 Hz Vo=10V.
I did not use the graph dB versus frequency.
Where they bugged, the curves for Cadj=0 and Cadj=10uF are in the wrong order.
Furthermore, I guess the raw figure, 64dB is for a 120 Hz saw tooth shape ripple, while the graph is for a sine AC ripple at Vin.
The 'ripple reduction' graph isn't plotted for a real ripple waveform (which contains plenty of harmonics) rather for a simple 120Hz sinewave. So its optimistic in practice because the rejection falls above 100Hz. Meaning the actual output of the reg sees a higher proportion of higher harmonics than the input.
Sorry - the query was directed at the OP. Not everyone has accurate uV measurement capability, whether a meter or cro, although a soundcard spectrum analyser would likely be easier to calibrate at a higher level, and then assume accurate scaling.
Another interesting thing on the LM317 datasheet is that, with 500 mA and 15 volts of Vin to Vout, that the 120 Hz ripple rejection goes all to hell when the output voltage ventures beyond 25 volts---getting 24db worse by the time the output approaches 30 volts!! This may, indeed, call for a MUCH higher value cap on the output than otherwise expected. I wonder if this also applies when the Voltage I/O is a more reasonable 5 to 7 volts.
This sounds awfull, may be, there is a mistake in the datasheet.
I've used a LM317 at Vo = 48v ...........!!!!!!! Microphone +48 phantom power.
I do not understand their Vin - Vout = 15V.
Usually we don't want such a high voltage drop.
At 500mA this asks for a heatsink to evacuate 7.5 W
It regulates with a voltage drop as low as 3V; So, even with safety margins this 15V doesn't make sense to me. Am I missing something
I've used a LM317 at Vo = 48v ...........!!!!!!! Microphone +48 phantom power.
I do not understand their Vin - Vout = 15V.
Usually we don't want such a high voltage drop.
At 500mA this asks for a heatsink to evacuate 7.5 W
It regulates with a voltage drop as low as 3V; So, even with safety margins this 15V doesn't make sense to me. Am I missing something
Last edited:
Indeed, 15V doesn't make sense.
That is a typo. It is clear from other graphics nearby.
Read Vin - Vout = 5V.
5V is a usual operating choice.
The OnSemi datasheet gives a much better ripple rejection 80dB typical, instead of 64dB typical on the TI datasheet.
Beware of the noise figure. Both datasheets give 0.003% Vo noise rms.
So a 15V power supply is expected to be 0.45 mV noisy. So the 0.1 mV ripple I calculated do not make much sense. From this I think 10 000uF is overkill, 3000uF would't make any difference.
Don't assume that the current TI part is entirely similar to the original NS part, changes in process and design improvements often occur quietly over time and the improvements are reflected in the newer data sheet. While I can't vouch for the accuracy of this particular data sheet we do use a lot of TI parts in our products at work and they generally meet or exceed the published specifications if used with care.
Ripple rejection and output noise are not synonymous, the noise is measured over a specified bandwidth, the ripple is at a specified frequency. 100uV of ripple at 100Hz or 120Hz will not be lost in the total 450uV of noise over say a 20kHz bandwidth.
Nope, they're not referred to the same measurement bandwidths, one is simply uncorrelated broadband noise, and the other is a combination of highly annoying spectrally discrete harmonics of the AC line frequency and in addition the ripple is much more annoying than broadband noise over a 20kHz bandwidth. Take a look at the noise spectrum using an FFT and you should see the discrete tones of the ripple and its harmonics are dominant. Neither tbh are particularly great performance levels, but better the older fixed voltage regulators they were intended to replace.
I thought you might find this interesting if you are looking at ways of reducing noise on the output of inexpensive monolithic voltage regulators. (It's pretty neat stuff)
Finesse Voltage Regulator Noise! |
Finesse Voltage Regulator Noise! |
Hmmm....maybe. But the datasheet under "Electrical Characteristics" shows the maximum current at Vi-Vo ≤ 15 volts to be 1.5 A min, 2.2A typical. That would be Pd = 22.5 watts min, 33 watts typical. So if it IS an error, it's done in two places. Your application of Phantom Power draws VERY little current (5 mA usually), so it may not be relevant. Another poster correctly noted another error on the datasheet---the reversal of the Cadj =10uF and Cadj = 0 uF on the "Ripple Rejection vs. Frequency" graph; so perhaps the proofreading here is not so great.
The On Semiconductor datasheet seems to indicate better performance than the T.I. one. The ripple rejection is ≥80 db w/ 10uF adjustment cap; and that stays constant up to 35 volts output. This would give me about 13.2 uVolts ripple from a OnSemi LM317 with 10,000uf input cap, at 1 ampere of current. Far below what I can hear.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.