New DIY Project for measuring noise : 10Hz..100kHz LNA // 1MHz RMS-DC converter


As some DIYers have already seen, i love to build measurements tools.
I have build some times ago a very useful tool, based on the AN83 and AN61 from a Linear Technology.

It's a wideband DC..10MHz RMS to DC converter (AN61) and a very low noise, high-gain, 10Hz-100kHz filtered amplifier(AN83).
This two tools together (or not) are intended to be used for electronics noise maesurements.
I had build it to compare noise produced by differents voltage regulators.

For a this use, the two functions are connected one behind the other, and a DC voltmeter is connected at the output of RMS/DC converter to
display directly the noise level measurement at input of the LNA.
This two function are indépendant but located on the same enclosure.

These tools works fine.So the RMS/DC converter described in the AN61 use an obsolete part, the LT1088.
This part is hard to find, and if you find one you risk to get it at very high cost !
Another issue with this part is it's unabality to doing measurements when signal level is too low.
So, it's practical measurement range is limited to about 10 to 100% of full scale.
On the other hand, the major advantage of this circuit is to allowing measurement at frequency up to 50MHz,
with signal with crest factor very high without degrading the precision.

You can see what does look this first version below.It's a self made PCB.

Some weeks ago i had decided to make a new version of this tool.
I want to replace the LT1088 with a modern solution and an easy to find part, to make it more buildable.
Then i think it would be an interesting projects to many DIYers...Isn't it ?

Not many solutions exist to doing RMS-DC conversion, above at reasonable cost.
I eliminate the very known log-antilog converter (like AD736) because they are cosly, doesn't have
wide bandwidth, have big error with crest factor increase.

Nowadays, a good solution is to use a new IC from Linear Tech, the LTC1968 (again they !)
This relatively new IC use an innovative way to do RMS-DC convertion.
It allow very good bandwidth of up to 1MHz, very good precision of less than 1% and low cost.
The measurement precision stay very good with low signal level and medium crest factor.
The LTC1968 is also very easy to do work, only few parts are need to be fully fonctionnal.
The only issue is that it only exist in a very tiny SMD package (MSOP8), to require a good soldering tip.
So, unfortunatly for humans eyes it's quite often the case with modern electronics parts....

The main specifications i would ask for each functions of the new measurement tool are :

Common specifications:

* Powered by batteries (+/-6Vdc) or symetric DC supply +/-6V to +/-12V.
* Housed in low cost aluminium case Hammond 1455L1601.
* Standard Europe 100x160mm 2 layers PCB.
* No exotic parts.

1/ Low Noise Amplifier (LNA):

* High gain of 80dB (x 10 000)
* 10Hz..100kHz build-in -3dB pass-band filter.
* Equivalent input noise of about 500nVrms.
* BNC input and output.
* Overload detection.

1/ RMS to DC converter:

* DC to 1MHz measurement bandwidth
* High input impedance 1MOhms/10pF
* Ooscilloscope probe use for x10/x100 attenuation,
allowing extending measurement range.
* 1v/100mV or 1V/V sensitivity output.
* BNC output and output.
* Overload detection.
* AC or DC coupling input.

You can see below the functionnal synoptic of the new projects ERMSDCV2 with above features :

The LNA doesn't change in the V2, just an overload detection is added, and the possibility to use SMD or DIP parts(dual-pattern).
You can see below the resulting frequency response of the LNA, and it's own output spectrum with shorted input.
(It's not simulation results, but real measurements !)


As you see, the measured output noise level of the LNA is about 5mVrms (-47dBV), so it's about 450nVrms reffered to the input (-127dBV)!
(It's the equivalent noise generated by a 150 Ohms resistor at 300°K !)
This very low noise level in this frequency band allow to do measurements in microvolt range. Like regulators noise, amplifiers noise and more.
To achieve this specs, shielding is very important ! An very good power supply or using batteries is recommanded for measuring this level of noise.
50/60Hz line magnetic field can be an issue, so all transformers must be far to the setup measurement, if possible.

The new schematic of the ERMSDCv2 is done, you can download it HERE (pdf file).
The final design in it's hammong box will look like this ;

And the printed circuit boards will look like this ;

Actually i start to solder parts on the prototype PCB and wait to receive some missing parts.
Of course, i will post results as soon i will try it.
I hope that new projects will interest many electronics fanatic like me. ;-).

As for the "AN67 10kHz low THD oscillator" and "DIY CS5381 ADC" ,
when the project will be ok i will organize a PCB group-buy for this new design
if some DIYers are interested. :)
To follow soon...


I have received this week missing parts and now the protototype PCB has been fully wired.
Below a picture of how it looks.

I've start to do some measurements and seem to work fine.
I will post more precise results in the next week.
Do some DIYers will be interested by a PCB of this design ?
To soon.


This is a very interesting project as usual for you :)
In Linear app note 124 ( Note/an124f.pdf), Jim Williams makes a point that the DC-filtering capacitor on the front end should have very low leakage to avoid adding noise.
What type of capacitor did you choose for this input?
There are some low leakage electrolytic capacitors out there (such as Nichicon KL), but I don't know if their leakage is low enough. Williams used a special tantalum capacitor with is hundreds of dollars!
Hello grenert,

The input capacitor in this design is not so "exeptionnal", it's "only" a Sanyo OSCON type.
The AN124 is a very interesting application note (i've read all, thank you!).
It describe a noise measurement tool with noise floor down to 160nV!
But,it is only in a very narrow frequency band (0.1Hz to 10Hz).
About the input capacitor leakage, you can note that the input resistor is 15 time higher than the ERMSDC design (1.5kR instead of 100R), so the leakage current of the input capacitor is more a concern.
It's clearly not the same tool. :)

ERMSDCV2 : LNA mesurements results.


I post some measurements done this week with the new ERMSDCv2.

First, the output noise of the 80dB LNA.
This measurement was done with 50 Ohms termination at input.
You can see that the output noise is about 5mVrms, corresponding to
500nV at the input.


Now, it is the transient response of the LNA.
The input signal come from a square wave generator, with a 1Mohms/1ohms resistors divider to get the right level (the full scale of the LNA is below mV !).


You can see below the frequency response of the LNA (normalized: 0dB=+80dB)). Amplitude and phase are plotted.
These measurements were done with a Gain-Phase Solartron analyzer.


Same as above with zoomed view.

I need to make some changes on the RMS/DC section, i'll post results on that soon.


Mother of Pearl, after some reflexion i think too that adding a logarythmic output can be interesting.
I've made some tests with a AD8307 and the results are very good.
I don't know yet how i will add it in the original design...
Maybe a switch in front panel for choosing with LOG or LIN output, or an other separate output...

I continue to do other tests about that and i work too to rebuild the front-end of RMS-DC converter.
I think add log converter before RMS-DC converter, in AC coupling mode (10Hz-1MHz).
DC measurement in LOG mode is problematic with the AD8307 and also not very useful in practical use.
I'll post soon a new synoptic.

This Google pulls many hits:

linkwitz shaped tone burst testing - Google Search

and this page shows exactly what I'm on about:


It's funny to see that someone else has done exactly the work I did on this back in the mid-80s.
Now, if someone would come up with a "normalizing" time base for a 'scope's external input that would keep one, or five, or fifteen cycles the same width on the screen while frequency is varied that would be just ducky. Note that it's not that hard to do so as the "burst generator" Linkwitz & Co. designed so long ago starts with a continuous tone, and with a little glue logic, a zero crossing detector and a 5-steps-up/5-steps-down attenuator generates a very useful signal that has some interesting higher, odd order spectral components that are supremely useful for measuring what I call "resolving power."
News from the projects.


These days i've working on the ERMSDCv2 design.
The final schematic is very close, and all my tests results are concluant.

I've finally add, as requested by "Mother of Pearl" an logarythmic output.
This one used an AD8307 logarythmic amplifier from Analog-Devices.
The log output allow to measure extremly low signal level, down to 50uV(-85dBV).

To know more, you can see the new synoptic of the design below ;


You can download it as pdf HERE.

The x1/x10 swtich has dissapear on this new version(2.1).
I prefer keeping low signal level to ensure good signal integrity in all measurement bandwith (0-1MHz) and staying away from supply rails.

The input sensivity level of the RMS converter is 350mVrms or 3.5Vrms with standard 10X scope probe.
For the log converter, the max input level is 1Vrms (0dBV) with direct connexion to input or 10Vrms ( 20dBV) with 10X scope probe.
Because input impedance is well know, 1MOhms 10pF, using X100 scope probe is also possible to extend input voltage range.

The input include a compensated divider network wich can be also use (or not) depending if you prefer an highest sensitivity level.
If you prefer use external scope probe, the network is shorted with a jumper.

The full schematics of this new version can be downloaded HERE. (pdf file 250ko). IT IS NOT THE FINAL VERSION !

I've doing many measurements on the RMS converter section and on the LOG section of the design.
You can see all results below;

1/ LTC1968 RMS voltmeter error vs input level (signal sine wave at 10kHz).


2/ LTC1968 RMS voltmeter error vs input frequency (signal sine wave at 250mV).


3/ AD8307 LOG converter passband, 10Hz-10Mz view (signal sine wave at 250mV).


4/ AD8307 LOG converter passband, zoomed view 100Hz-1MHz (signal sine wave at 250mV).


5/ AD8307 LOG converter linearity vs input level (signal sine wave at 10kHz).


5/ AD8307 LOG converter dB error vs input level (signal sine wave at 10kHz).


Note: All measurements where done with input network disable (x1).

As you can see, results aren't ridiculous.
When seeing RMS error graph, I think it could be better tuned.

Now schematics is near to be fully validate for me, and i think to report all of these modifications in the final PCB work.

Don't hesitate if you have any question about the design.

Mother of PEARL, i think it will be more interesting for you to add the external circuitry you need for your own need, adding all of that in the ERMSDC design does not seem to me very relevant.

Some news.


I still work on the ERMSDCV2 design and improvements.
I'm very happy with last results.

Because LOG output has been added, i modify the input sensitivity of the RMS/LOG input.
For the log output, input can be up to about 1.5Vrms ( 3.5dBV) and can measure level as low as about 25µV(-92dBV).
The dynamic range of the LOG converter is very impressive, and it's because i've
decided to add it on the final design.
You must note that the minimal sensitivity of the LOG converter are directly related on the noise of the analog input stage.
To obtain the noise floor og -94dBV, i use an input stage with OPA627 and AD817.
First OPAMP (OPA627) is FET input low noise.It allow high input impedance with very low noise contribution.
The 2nd OPAMP AD817 is more noisy but have greater bandwidth for doing ~1.8MHz 2nd order low pass filtering.

I've also modified the RMS/DC design.
Now, the max input level (before input OPAMP saturate) is about 2.5Vrms.
The input/output gain is now 1(instead of 2).
So, without input probe it can handle 2.5Vrms (7.1Vpkpk) and 25Vrms(71Vpkpk) with a 10X scope probe.
The input stage has been improved using FETs instead of 1N4148 diodes for over-voltage protection.

The overall precision of the RMS converter is very good in all bandwidth and with any waveform type (sine,square,triangular).

The new PCB is now ready, i'm checking all gerber files.
I'll post final schematics and more documents in my next post.
I open also a new wiki, for interested DIYers who would want build one.
To soon.