ultra low noise psu from China

Disabled Account
Joined 2002
Yes but haven't you also noticed many uLDO are limited to 5.5V input voltage? These low noise regs are made for cell phones :) 16V is a gift already, 20V is rare and a few can have 36V...

Decentralized design is hot so also no high current types ...From one LM317 powering a complete device to 5 x uLDO regs that each feed a section is progress but harder to DIY.
 
Last edited:
Disabled Account
Joined 2002
Or MIY which will be the future. Modify It Yourself. Stuff is too small and through hole days are long gone. This filters out the ones that see SMD as too difficult and also the ones that have no issue with SMD but do have an issue with the most recent SMD sizes...
 
We are on the same playing field brother :D One of the annoying things in modern electronics is that we are not able to solder the parts by hand which IMO is part of the craftsmanship.

i have just designed a PCB for ADM7150 regulator and already sent it to a PCB factory so it's in making and I have ordered all the parts from Digikey, ...will have it all on Tuesday latest and I will test it and measure it.
i will get back.
 

Attachments

  • Capture.JPG
    Capture.JPG
    76.3 KB · Views: 140
Last edited:
As said, for some reason it sounds a very small tad better. Both my test winners in max. 25VA range are still here and I can not decide which one to keep. I know the LT3045 measures absolute best. As many audio guys only have their ears to judge I had the PSU's tested by them and they also preferred the TPS7A4700 (FWIW, thin ice etc.).

Output impedance of LT3042/LT3045 is not ultra low. TPS7A4700 has lower output impedance (see e.g. here). This may be one reason why TPS7A4700 may actually perform better in practice.
 
Disabled Account
Joined 2002
Good finding! I don't have stuff to measure that precise as an amateur but it may indeed be the explanation. Although.... output/bypass caps in that tests are totally different both in value and in type......Still I would call both regs ultra low output impedance and one of them extreme ultra low output impedance :)

The website is very interesting with many other measurements and projects. Since I was alarmed by the 6 MHz charge pump in LT4320 in this thread I checked all the threads at that website for measurements on LT4320 but can't find any with the 6 MHz horror?! Is it that the 6 MHz is not an issue at all? I know that sometimes you shouldn't look in the kitchen of good restaurants.

* my test winners are a board with 6 x LT3045 in parallel and the other one just 1 x TPS7A4700. I never tried to parallel the latter. Wrong comparison in many aspects, it is just that the load is 0.7A.
 
Last edited:
First, because he gets 0.4 mOhm at 100 kHz, which is 0.6 nH, which is unrealistic.

Second, TPS7A4700 has a "sense" input which is wired to the output connector, as you'd expect. I guess he injects current into the output connector, and TPS7A4700 regulates voltage at this point. However, he measures output voltage at the output of TPS7A4700, not at the output connector. This means the impedance of the PCB trace between the output of TPS7A4700 is not added, but substracted from the measurement.

Example:

Capturex.PNG

R1 is the trace impedance.

Current source I1 is the load, it draws AC current from the output connector, which I labeled "sense". The regulator controls voltage at point "sense": to do so it outputs current that goes through the trace modeled by R1. This creates a voltage across R1.

If I1 draws positive current then voltage at point "sense" goes down. But voltage between regulator output "out" and "sense" goes UP due to the load current flowing through R1. If these compensate almost exactly, you get a false measurement of zero ohm. But if R1 is a bit larger than the LDO's output impedance, you can also get a false measurement of a negative resistance... hence it's important to look at phase too...

This stuff is very sensitive to the setup. And just wiggling the wires will change inductive coupling.

When I tested the LDOs, I made a series of little boards to carry them, with 0.1" headers to the measurement setup. One pin to inject current, one to read the voltage back, both for input and output, and about 20 ground pins, with balanced inputs to read the voltage. With the connector I'd be sure there was no change in inductive coupling between measurements because nothing moved. Then I calibrated it with open-short-load, and the "short" calibration standard was a chunk of copper clad PCB. So at any time I could unplug the LDO and measure the short again to check the measurement floor. I also measured a LDO carrier board with the output caps replaced with wires, etc. It's a bit of work to ensure it's measuring what you think it is...
 
I have zero trust in the 0.1 mOhm output impedance measurement.

First, because he gets 0.4 mOhm at 100 kHz, which is 0.6 nH, which is unrealistic.

I guess you are referring to the sense-line measurements further down the page as in the first diagram the output impedance is 20mOhms@100kHz.

Anyhow the lower output impedance of TPS7A47 vs. LT3045 is quite apparent also in the load transient diagrams in the datasheets.
 
Disabled Account
Joined 2002
That is why I paralleled LT3045. Then the balancing resistors add up :( Ah well, if we are into this small details in already very good regs likely followed in the device they feed by switching regs we are maybe too precise? Maybe I am naive but the PSU they replace has way worse specs so high chance that most linear PSU's that are made to replace the original PSU are better anyway.

Really, even LT1084 based PSU's are already better than the standard cheap SMPS that is delivered with streamers/DACs etc. The loads I use are ugly loads anyway. So it is from ugly PSU feeding an ugly load to a good PSU feeding an ugly load :)

Peufeu, thanks a lot for your verdict on an old PSU design by me. I just found it here and had a smile on my face when I read it was the winner.

https://www.diyaudio.com/forums/power-supplies/371849-power-supply-tests-archives.html#post6642040

Like you already stated, when stuff is good feedback is rare. Contrary to other devices I use this design myself for ethernet switches and routers and it still is competitive with regards to results. Enough egotripping, back to now.
 
Last edited: