7815 noise and measurement

Noise of the 7815 is supposed to be 60uV or so.

AC CMRR of the 2700 is specified as 70dB.

So to get a 600uV reading on the meter you only need a few volts AC common mode voltage.

--Spehro Pefhany

** You need to explain that one cos it makes no sense.

Why? 100nF is sufficient, same if fully earthed.

** The 100nF cap is for stability, not noise supression.

FSS read the flaming data sheet.

** I take it you isolated the supply safety earth.

** Hi Klaus,

The broad band noise and hum from a 7815 reg is very small, less them 100uV rms.

The only way to measure it with accuracy is to use a pre-amp to boost the level to something a DMM can read.

I use a microphone pre-amp with gain of 60Db ( 1000 times )and a BW of 100kHz to do the job.

Of course you have to AC couple the reg to the pre-amp first.

As with any such testing, you must use a scope to monitor the signal in order to be sure you are measuring what you think you are measuring.

... Phil

I don't quite understand all that. Can you post a schematic?

--

John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

if you want to measure 100uV level signals, you will need a high gain preamp in front of the dvm. Needs to be in a screened box as well. At that sort of level, nose leakage from the dvm internals to the input terminals could have an effect, as will any ground loops, local transmitters etc.

Always used a diecast box for such preampss, with bnc input and output and the internals running from batteries inside the box. Too much noise otherwise...

Chris

I'm not sure about your grounding issues. PhiL H. has a copy of linear V reg. noise (LM317) as a function of load capacitance... Hey I found it here,

George H.

I have measured that 2 years ago. You can assume that the 7815 is voltage-wise 3 times worse than the 7805 because it amplifies the noisy band gap 3 times more.

< > 0 dB is a noise density of 1nV/sqrt(Hz), 20 dB is 10 nV/sqrt(Hz). Most regulators are more or less equally bad. The LT3042 / 3045 / 3094 are in a class of their own.

Load is usually 47 Ohms, circuit like proposed in the data sheet. Moto=motorola, Fair=Fairchild etc.

Pictures left/right are noise of LEDs and Zeners. Everything is measured using the same setup. The preamp voltage noise is 220 pV/rtHz. (20 parallel op amps ADA4898)

regards, Gerhard

On a sunny day (Thu, 4 Jul 2019 14:54:03 -0700 (PDT)) it happened snipped-for-privacy@gmail.com wrote in :

If I understand your circuit right, then you cnange from 15V DC out to about 1.2 V DC out, so about 1/10 the gain and so about 1/10 the noise because normally the 1.2 V reference noise is multiplied 15 / 1.2. ?

** Correct.

** Irrelevant, once the signal has been boosted to the 100mV level.

** Any metal box will do, tin plate or aluminium are fine. ** Nope, my gain box has a balanced XLR input and RCA output with remote PSU.

A simple adaptor converts the input to BNC so I can use a scope probe.

The output of a 3 terminal reg is VERY low impedance so e-field pickup is not an issue.

.... Phil

I like your pic of the SFP guts:

SFPs are amazing. I'm trying to find a non-telecom use for them. Imagine a 1 MHz to 5 GHz bidirectional fiber link for under $20 per end. Pity they are AC coupled, but they have to be for telecom.

--

John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

The problem is not well enough defined to allow us to be helpful. No circuit, no numbers. He may have massive amounts of ambient RF or AC-line common-mode noise. Is the measured noise microvolts or volts?

--

John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Must have got away with that lightly :-). This was in London,

1970's, low noise audio for input and mixing stages. Even the Radford DMS2 boxes used bnc and internal batteries. Must have been a transmitter close by, s rf seemed to get into everything unless well screened...

Chris

Am 05.07.19 um 17:38 schrieb John Larkin:

:-) These are the slightly bigger brothers, XFP. At that time we were dreaming of a selling price of 250$.

It has a DFB laser which draws much more current than a VCXSL. A VCXSL bias works with a resistor instead of choke(s). The flex cable to the laser was a 3D electromagnetics simulation nightmare, but we managed to get pretty optical eyes. The laser is somewhat swimming, slightly moving with the fiber connector. Same for the ROSA.

DC does not work because of 1/f noise. It has a reason that telecom chips spec the noise only above 12 KHz.

cheers, Gerhard

DC is also unsuitable for telecom because the received optical power can very enormously, so the receiver needs AGC. Telecom data is DC balanced, 8B10B or some such, so AC couples just fine.

We tested a bunch of SFPs. They are AC coupled in and out, but there seem to be no standards for the time constants. Some work down to 1 MHz, some down to 10. We didn't investigate receiver AGC time constants; I'm not sure how we could do that.

I could FM or OOK to restore DC coupling, but then I'd get a clock worth of jitter. All that Shannon stuff.

--

John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Scope probe, with the first MHz sent over a subcarrier?

I agree with Gerhard in that you probably wouldn't want to leave DC up to the SFP link even if it was officially supported.

-- john, KE5FX