I'm looking for a low noise LDO linear regulator for an RF project, I've seen a few mentoned and one is mentioned as being especially suitable for my application, the TPS793xx series but...
The TI TPS793xx datasheet claims output noise as low as 32 microVolt with suitable output capacitor.
Comparing this to the ones in my parts bin, LT1085 and LM2940, I can see how that's low noise as they specify 150microVolt (LT1085 is 0.003% of output, 150 microVolt and so is the LM2940 at 5V)
But, and this is what's confusing me, the TA78L05 datasheet claims only 40 microVolt.
So, am I wasting my time with the 'low noise' regulators as my bubblegum TA78L05 is near as damnit the same spec?
Am I missing some nasty that's lurking in the datasheet, the noise figures quoted for instance are specified under 100KHz?
(edited because I thought I was being smart inserting the micro symbol but the forum SW doesn't show it)
Over what bandwidth?
Noise density is what matters. More to the point, how much does it actually affect your RF circuit? If it's a proper RF circuit, noise outside of the passband is no noise at all!
And how accurate does the voltage actually have to be? You're probably better off with a capacitor multiplier (i.e., RC filter + emitter follower) than an LDO.
Tim
In The Art of Electronics (Table 9.3) they have a big comparative chart of LDO regulators, including noise 10Hz-100KHz. It seems that the champion is the
MAX8510 with typical 10uV RMS, at a reasonable 0.58$ price. But it only gives 120mA current. In the table, the TPS793xx is quoted at 16uV RMS with NR cap. I have always been impressed by the claimed 7uV RMS of the venerable 723 (10Hz - 10KHz) with filtering cap, but that's not LDO at all. Anyway, nothing in that table comes even close to the magnificent 0.8uV rms of the LT3042 linked by EmmanuelFaure.
I'm not sure that low frequency noise is not important for RF circuits; can't that noise get mixed into the passband, for example in an oscillator, in the from of phase noise?
Your TL78L05 may have roughly the same noise spec as the TPS793 family, but the rest of the specs will tell the real story. Take a look at the PSRR, dropout voltage, and line and load regulation. The TPS793 beats it all hands down.
Now, if you have a pretty clean supply coming in (don't need high PSRR, or good line reg), don't care too much about power dissipation (high dropout is ok), and your load is fairly constant at low frequencies (only used for biasing RF components, no current transients, etc), then you don't really need the TPS793xx and can try out the TL78L05 regulator.
Over what bandwidth?
Noise density is what matters. More to the point, how much does it actually affect your RF circuit? If it's a proper RF circuit, noise outside of the passband is no noise at all!
And how accurate does the voltage actually have to be? You're probably better off with a capacitor multiplier (i.e., RC filter + emitter follower) than an LDO.
Tim
I'm experimenting with AD8307 chips so it's DC to >500MHz but there will eventually be a fairly tight filter to bring that down to a narrow bandwidth.
No matter what detector I eventually choose, at the moment I'm trying to make sure I've got the supply as 'clean' as I can reasonably make it so I'm not introducing measurement errors by adding avoidable noise.
Voltage accuracy, as long as the supply can source ~10mA, doesn't seem to be that important in this particular application.
In The Art of Electronics (Table 9.3) they have a big comparative chart of LDO regulators, including noise 10Hz-100KHz. It seems that the champion is the MAX8510 with typical 10uV RMS, at a reasonable 0.58$ price. But it only gives 120mA current. In the table, the TPS793xx is quoted at 16uV RMS with NR cap. I have always been impressed by the claimed 7uV RMS of the venerable 723 (10Hz - 10KHz) with filtering cap, but that's not LDO at all. Anyway, nothing in that table comes even close to the magnificent 0.8uV rms of the LT3042 linked by EmmanuelFaure.
I'm not sure that low frequency noise is not important for RF circuits; can't that noise get mixed into the passband, for example in an oscillator, in the from of phase noise?
Good grief, I'd forgotten about the '723, I#ve used so many of those in the past.
LDO isn't really necessary, the reason for quoting the TPS793xx device is because I've seen it specified as 'better' than the 78lXX devices for use with the chip I'm using.
Seems not everything you read on the 'net is accurate, who knew...
Over what bandwidth?
Noise density is what matters. More to the point, how much does it actually affect your RF circuit? If it's a proper RF circuit, noise outside of the passband is no noise at all!
And how accurate does the voltage actually have to be? You're probably better off with a capacitor multiplier (i.e., RC filter + emitter follower) than an LDO.
Tim
I'm experimenting with AD8307 chips so it's DC to >500MHz but there will eventually be a fairly tight filter to bring that down to a narrow bandwidth.
No matter what detector I eventually choose, at the moment I'm trying to make sure I've got the supply as 'clean' as I can reasonably make it so I'm not introducing measurement errors by adding avoidable noise.
Voltage accuracy, as long as the supply can source ~10mA, doesn't seem to be that important in this particular application.
Hi
The rated noise floor of the AD8307 is only 1.5 nV/Hz. They only rate that as "typical" without a frequency limit. One would *guess* that it is a HF measurement rather than a LF number. You do not need to go to crazy lengths for the power supply on the part.
Bob
Hi
The rated noise floor of the AD8307 is only 1.5 nV/Hz. They only rate that as "typical" without a frequency limit. One would *guess* that it is a HF measurement rather than a LF number. You do not need to go to crazy lengths for the power supply on the part.
Bob
Thanks Bob, In this instance it's only curiosity to see how much difference, if any, the regulators make to the measurement output but that information could well become useful in the future when I start to play (I hesitate to say work, it's a hobby project for fun) with different RF building blocks and other log amps/detectors.
Plus, I don't think it'd hurt me to 'do it right' even if 'right' is overkill for this particular part if it only costs a few pence/cents extra.
Out of interest, could you recommend an alternative Ic or other system for RF level measurement purposes? I'm looking for small signal measurements rather than large ?
Hi
The rated noise floor of the AD8307 is only 1.5 nV/Hz. They only rate that as "typical" without a frequency limit. One would *guess* that it is a HF measurement rather than a LF number. You do not need to go to crazy lengths for the power supply on the part.
Bob
Out of interest, could you recommend an alternative Ic or other system for RF level measurement purposes? I'm looking for small signal measurements rather than large ?
Hi
A lot depends on what kind of signals you are after. The most common problem is measuring / observing small signals in the presence of a lot of noise. Having a 1 nV / Hz, 10 Hz to 1 GHz gizmo sounds cool until you notice it overloads on everything from power lines to power supplies to cell phones. For most things that get done, some amount of selectivity is needed. Again, it depends on what you are after.
Bob
Hi
A lot depends on what kind of signals you are after. The most common problem is measuring / observing small signals in the presence of a lot of noise. Having a 1 nV / Hz, 10 Hz to 1 GHz gizmo sounds cool until you notice it overloads on everything from power lines to power supplies to cell phones. For most things that get done, some amount of selectivity is needed. Again, it depends on what you are after.
Bob
The building blocks I'm experimenting with are aiming towards a homebrew swept SA, I've been reading and sketching ideas around the various homebrew designs on the web.
I'm not expecting performance to equal that of a 'Proper' SA but it's an interesting learning experience, seems the AD8306 and '8307 are regarded as fairly good in a homebrew design, I just wondered if there was a better solution.
Hi
A lot depends on what kind of signals you are after. The most common problem is measuring / observing small signals in the presence of a lot of noise. Having a 1 nV / Hz, 10 Hz to 1 GHz gizmo sounds cool until you notice it overloads on everything from power lines to power supplies to cell phones. For most things that get done, some amount of selectivity is needed. Again, it depends on what you are after.
Bob
The building blocks I'm experimenting with are aiming towards a homebrew swept SA, I've been reading and sketching ideas around the various homebrew designs on the web.
I'm not expecting performance to equal that of a 'Proper' SA but it's an interesting learning experience, seems the AD8306 and '8307 are regarded as fairly good in a homebrew design, I just wondered if there was a better solution.
Hi
For a "old school" spectrum analyzer, you are doing log detection at the output of an IF strip. That is very much a narrow band (as opposed to a wide band) situation. There is no need to go crazy about 500 MHz bandwidth detectors. Your (likely) widest bandwidth on the analyzer will be < 1 MHz. Exactly what the conversion scheme of your SA turns out to be depends a lot on what sort of microwave "stuff' you have lying around. A simple 1.5 GHz design will still be using an oscillator well above 1.5 GHz and a first IF somewhere in the microwave range. The second IF (if you have one) will depend on the filtering you rig up for your first IF ...
Lots of variables.
Bob