Jackinnj is plotting the output noise of full and complete voltage regulator circuits, versus frequency. If you are curious about how the BC337 transistor compares, then the thing to do is to build a voltage regulator which uses BC337 transistors in the noise-sensitive positions, and send that full & complete regulator to jackinnj. I am confident he will probably decide to test it and make a plot for you. Naturally it will compete for his rare and precious hobby-hours with other projects, so be patient.
I think Keantoken's cap multiplier is at issue here. Would be happy to put it through the paces if I had one in hand. I haven't any current prejudices, just old ones!
AFAIK, a single BJT has less noise than any of those chips. Theoretically, the total noise of a regulator can be almost as low as the active feedback transistor. So I think it's possible that a simple discrete regulator using a low-noise BJT as the feedback device could still beat any chip in noise. So it isn't about my circuits specifically, although I am thinking about other regulator/PSU circuits I could make.
It would be somewhat pointless to test the noise of the KM except at treble, because it will track the input voltage, so the noise advantage is only at treble and RF frequencies.
It would be interesting to send a KM in for testing, although I haven't committed to that. So many things competing for my attention lately. How important should it be?
It would be somewhat pointless to test the noise of the KM except at treble, because it will track the input voltage, so the noise advantage is only at treble and RF frequencies.
It would be interesting to send a KM in for testing, although I haven't committed to that. So many things competing for my attention lately. How important should it be?
Last edited:
Don't send in any circuits for testing unless and until you think testing them is REALLY important. You don't want to upset Jack, or make him angry with you for wasting his time. By the same token you don't want to waste your own time.
ONLY if and when you think you know how to make a regulator whose noise is significantly below the regulator noise of other candidates, does it make sense to bother with constructing a prototype, and then sub-contracting out the noise measurements. Otherwise, spend your time on other things that intrigue you more. Nobody else is remotely interested in finding out whether regulatorZ replaces regulatorY for the honor of fourth-best noise performance.
ONLY if and when you think you know how to make a regulator whose noise is significantly below the regulator noise of other candidates, does it make sense to bother with constructing a prototype, and then sub-contracting out the noise measurements. Otherwise, spend your time on other things that intrigue you more. Nobody else is remotely interested in finding out whether regulatorZ replaces regulatorY for the honor of fourth-best noise performance.
Kean,
BC337 has a noise voltage of <1nV/rt Hz at a few mA. I'm not sure
a/ What happens at higher currents typical of a reg, I'm presuming it is less and
b/ what the 1/f looks like but I wouldn't be surprised if it is also very low.
Terry
Yeah, it was the 1/f noise that I was wondering about. Some information here:
The transistor doesn't have to carry all the current, it just has to be the feedback transistor and with the right design it will be the main source of noise.
The transistor doesn't have to carry all the current, it just has to be the feedback transistor and with the right design it will be the main source of noise.
Ah - didn't see that, thank you.
If you are after super low 1/f noise then MAT12 / SSM2212 and P ch variants
are best I know. Super low noise.
T
That's a very good point, though these ESR are specified at 100kHz and they should hold up to a few MHz as "capacitors" but I can't find a graph right now. At any rate they should behave far better at a few MHz than a non-SMD non-polymer aluminum low ESR can capacitor but the exact opposite is true in my circuit. That's why I'm asking. The only thing that makes sense is that ESR is too low (the 220uF cap is actually below 9mOhm ESR) but recommended ceramics should also have similarly low ESR as far as I know.
I believe I can hack a board to add a 0805 10uF ceramic in parallel so I'll try that.
My DER EE-500 LCR meter actually can't measure them above 1kHz even though they are specified at 100kHz.
I believe I can hack a board to add a 0805 10uF ceramic in parallel so I'll try that.
My DER EE-500 LCR meter actually can't measure them above 1kHz even though they are specified at 100kHz.
Last edited:
220uf is probably causing the feedback loop to lose phase margin. All of ADIs LDOs have internal ESR compensation and it is possible to shift the ESR to far to the"left" for the loop to properly compensate. This is a case of more capacitance is not necessarily a good thing. Adding a few uF of ceramic caps will keep the loop happy.
Polymer caps don't have much less ESL than a standard cap of the same physical dimensions, but have much lower ESR, which is a problem if you put a ceramic cap in parallel.
For example, the fashionable capacitor of the day could be an oscon like 6SEPC560MW, it has ESR of 6 mOhm and ESL of about 3.5 nH if properly mounted on a 4 layer board with GND as layer 2 and a copper pour as a power plane on layer 1. Of course on a double sided board ESL will be higher, and if traces are used instead of planes, even higher.
The impedance peak that occurs between the this kind of cap and the usual ceramic in parallel can do weird things to the regulator's control loop. It boosts gain while wiggling the phase all over the place. That is a nice way to explore all kinds of instabilities. With luck you'll just get a fugly transient response (and huge noise peak), if you're out of luck it will oscillate...
The use case for these ultra low ESR caps is boards like a PC mobo, where you have hundreds of MLCC in parallel, in this case it works.
FYI, a panasonic FR cap of the same size (6.3 diameter, 2.5 pin spacing) has 4 nH ESL which is not that different, and 80 mOhm ESR (datasheet conservatively says 130), which means you can put it in parallel with any ceramic cap as low as 1 uF and still get perfect damping. And you get a bag of 100 for less than 10€.
For example, the fashionable capacitor of the day could be an oscon like 6SEPC560MW, it has ESR of 6 mOhm and ESL of about 3.5 nH if properly mounted on a 4 layer board with GND as layer 2 and a copper pour as a power plane on layer 1. Of course on a double sided board ESL will be higher, and if traces are used instead of planes, even higher.
The impedance peak that occurs between the this kind of cap and the usual ceramic in parallel can do weird things to the regulator's control loop. It boosts gain while wiggling the phase all over the place. That is a nice way to explore all kinds of instabilities. With luck you'll just get a fugly transient response (and huge noise peak), if you're out of luck it will oscillate...
The use case for these ultra low ESR caps is boards like a PC mobo, where you have hundreds of MLCC in parallel, in this case it works.
FYI, a panasonic FR cap of the same size (6.3 diameter, 2.5 pin spacing) has 4 nH ESL which is not that different, and 80 mOhm ESR (datasheet conservatively says 130), which means you can put it in parallel with any ceramic cap as low as 1 uF and still get perfect damping. And you get a bag of 100 for less than 10€.
Last edited:
That is very useful info. I did not have a ceramic cap in parallel at all though, just a small SMD film cap but a few cm away so I was not counting it. The polymer was not nicely placed though, it is on the opposite side of the board from the ground plane (2 layer board) due to space restrictions. I guess the classic combo of properly placed small smd ceramic and good aluminum cap like those Panasonic HFQ or whatever marketing calls them today (FR? they change every few years) is still the best bet then, 20 years later. I had, however, had great luck with those polymers and used them in many DAC designs without problems. But those were ADP3303 and SOT-23 TPS79133 regulators and the like, not these fancy RF ones.
I am just surprised that any of this would make that much difference at only a few MHz.
I am just surprised that any of this would make that much difference at only a few MHz.
It's the largest cap with the lowest ESR available in that voltage. Going lower only gives you less thickness and costs less, doesn't save any space and normally less ESR is a good thing. Large cap values also help with transients at lower frequencies - load regulation on these isn't perfect after all. So why not take advantage of 21st century technology. That's what I was going for, anyway. Usually when there are ESR restrictions, manufacturer shows a drawing with "not allowed" zones of capacitance vs. ESR so you can't go too low and too high. There's nothing in this datasheet except max 0.2Ohm. There is no oscillation and it looks clean. But connect it to a DAC chip and you get strange stuff. I would have never suspected the cap and regulator until I saw an internet post that blamed low performance on low PSSR and the graphs looked similar to what I was seeing. I spent days trying to debug digital filter and receiver settings only to find the regulator is doing a crap job. Showing up for work but only pretending to work 
.
I am going to place a simple SMD ceramic as you advise and be done with it, though.
.I am going to place a simple SMD ceramic as you advise and be done with it, though.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.