40dB measuring amplifier with filters

[blackdog]

Hi

The basis of this measuring amplifier is already I think 10 years old and now posted here with a small upgrade.
And its purpose includes measuring linear power supplies with a scope or mV meter that have an input impedance of around 1Meg.

That input impedance that this amplifier is going to see at its output is important, this if the filters are used that come after the amplifier stage.
Of course you can put an Opamp or another kind of buffer behind the filters, but that makes the whole thing more complex.
My intention was to do it with one good opamp and get as much performance out of the circuit as possible.

And because almost every "normal" scope has an input impedance of 1Meg, and this amplifier is connected via a short piece of RG58 Coax,
then the attenuation by the scope or AC mV meter is then only little.

The explanation of the circuit


R1 of 10 Ohms along with C1 and the diode capacitances of the 1N4007 have two functions.
First the protection of the + input (~+-0.6V) of the LT1037A opamp and the second function is the decoupling for strong HF signals.
C1 can be increased as necessary to e.g. 1nF if very strong HF signals are present.

The amplifier circuit is AC coupled through C2 or C2 and C3.
When S1 is closed, the crossover point at the input is 0.7Hz.
C3 can also be made 10uF, then the crossover point at the input becomes about 1.6Hz.

R2 provides the DC setpoint of the opamp and also determines the crossover point together with C2 and C3.
The gain is about 100x, I measured out the resistors R3 and R4 so that I come out close to the 100x gain in around 1KHz.

R3 and R4 are deliberately chosen low in value to keep their noise contribution low.
That the opamp now sees a low load through R4 is less important with this amplifier, this is because the intention is to measure small signals.
The LT1037a can easily deliver 10mA and therefore this is not a problem.

C6 of 1uF together with the 1Meg input impedance of the scope determines the HP filter at the output and that is about 0.16HZ.
R5 serves to discharge the C6, and the value of R5 is so high, that it has no further influence on operation.
If S2 is in the "Direct" position, then R6 of 100 ohms ensures that the opamp sees no direct capacitive load and remains stable in this position.

No Filters Bandwith


This is a Bode Plot of this measurement amplifier without the use of filters.
At the bottom left of the picture you can see the measured -3dB point which is about 440Khz.
The cursor measurement shows the -1dB point which is about 288KHz.
The green trace is the phase behavior, the small bump at about 15Hz is one measurement error from the scope used.

There is a small rise just before 100KHz (0.15dB), it can be removed, but then the bandwidth is also reduced.
Increasing C4 to 82pF or 100pF removes this small rise in gain from the graph.

Filters

The High Pass filter
At the input via S-1, a 400Hz 6dB/Oct high pass filter can be enabled.
That can sometimes help suppress the hum frequencies a bit, to better look at the noise within the selected bandwidth.

The Low Pass Filters
If one of the filters is selected via S-2, there is a choice between 20KHz and a 100KHz low pass filter.

Some of you will want to note that linear power supplies are always measured at a bandwidth of 20MHz.
I would say, do your best to get that right, more on that in a below. :-)

Building en powering the amplifier.
This amplifier should be built in a metal enclosure and also keep the wiring inside the enclosure short/ twistet input wiring.
This is because most metal enclosures block the magnetic field badly.
This applies not only to this amplifier stage but anywhere you work below say 1mV of signal.

Use BNC connectors to and from this this amplifier.
If you want to measure the interference signal from a linear power supply, use a piece of RG58 coax for that as well and connect it under the clamp of the banana bus.
I will post some pictures later on how I mean this.
In measurement environments with high interference levels, it can also help to place ferite clamps around the RG58 at the input cable and/or the output cable.

Power for this amplifier
My preference is to use two 9V batteries as a power supply.
These are low-noise, but more importantly, there will be no commonmode currents running through the cabling, or at least, much less.

The measurements I usually do with these types of amplifiers go this way:
Connect the power supply to be measured via a isolation transformer.
E.g. this measuring amplifier battery powered and then I also use a battery powered scope to make sure that as little interference as possible is injected into the measurement setup.
The current drawn by the LT1037 is about 3.5 to 4mA from the 9V batteries with e.g. 0.5mA for a power LED that will last the batteries about 100 hours.
Battery power provides many advantages and I do not see the occasional replacement of two batteries as a problem.

Why a 100x preamp and not 20MHz bandwidth.
This measuring amplifier is not designed to measure at the best linear power supplies!
If you want to measure e.g. an LT3041, you will need a completely different amplifier and the gain will have to be +60 or +80dB.

Furthermore, you will then also need a faraday cage!
This is because you will be measuring signals below 5uV at a wide bandwith .
This is all far beyond the capabilities of this amplifier.
And I can also tell from experience, that even if I would design/build such a measuring amplifier, I would almost never be able to fully use it here in my LAB.
The EMC field here is far too large and if you then also want to look at a bandwidth of 20MHz, well I wish you success. 

Most linear power supplies for sale, produce much more interfering signals and noise than the LT3041 and these can be measured well with this amplifier stage.
If you start measuring linear LAB power supplies, you will naturally see that above 10KHz there are few interference signals left.
And if there are plenty of signals, then there are often problems with your measurement setup or you have a modern LAB power supply with an SMPS pre regulator that often still puts a few mV of interfering signal on the output terminals.

OK, I finally tuned with LTspice my basic design below a picture of the amplifier in LTSpice.
The noise plot is related to the output though, so after 100x gain.


Bode Plot 20KHz Low Pass filter
Bottom left the -3dB point and bottom right the -1dB point at about 11.2KHz.


Bode Plot 100KHz Low Pass filter
Bottom left the -3dB point and bottom right the -1dB point at about 56.2KHz.


Opamp type
I ended up choosing the LT1037a opamp, this for two reasons, the noise is less and the bias current is much less than the NE5534a that I started testing with.

The DC level on the output is up to about 0.5V with the NE5534a, and that goes off your output range.
Also, the 1/f noise behavior of the LT1037a is a lot better, and I have plenty of them in stock.
And I did not want to include in the circuit an electrolytic capacitor in series with R3.
The LT1037a here has less than 2mV DC on the output, so that doesn't cause any problems with the output.


Some other pictures
Here I am selecting capacitors for the filters, this is because many capacitors in the tray are 5 to 10%.



This is the test PCB I measured on, I will make a final version of that.



Tomorrow I will do some distortion measurements with the QuantAsylum and also try to do some good noise measurements with this at a "clean" EMC site.

SHOOT!
Bram

[blackdog]

Hi,

This is the second part of the explanation of how to build a measurement amplifier for e.g. measuring hum, noise and other signals, which are on the output of a LAB power Supply.
This because e.g. the sensitivity of a scope input is often too low for this.

But of course this measuring amplifier has more applications.
A small summary of the features of the latest version visible below as a schematic.
Gain:                                          +40dB (100x)
Input impedance:                       ~10K
Bandwidth -3dB:                        <1Hz - > 700KHz
Low Noise see OPA1612 data:  <4nV/√Hz depend on frequency
Power source:                           2x 9V Alkaline battery
THD in audio range:                   < 0.002%.

First the latest version of the schematic, not much has changed in it.
The main change is the opamp used, I had made a mistake in the first measurements, I thought I had taken an LT1037a from the bag in my stock, but it turned out to be an OP37, the OP37 is in the same bag....
I switched the OP37 and the LT1037a, and the differences are too small to go back and do the measurements again.
This is also because the two opamps are specified almost the same.

This mistake did prompt me to look further in my stock to see if there are opamp types that are even more useful for this project.

Since I wanted to do THD measurements with the Quant Asylum for nice pictures, I thought about the opamps used in this Quant Asylum ADC/DAC
and I actually had the single version of the OPA1612 in my stock, it is called OPA1611.
The second modification in the schematic is the compensation capacitor C4 in the first schematic, it is no longer needed with the OPA1611.

Latest schematic version


Bode Plot OPA1611
Using the OPA1611 opamp has improved not only the noise but also the bandwidth.
The lower left corner again shows the -3dB frequency, which is over 730KHz.
The -1dB point is indicated in the lower right corner and is also indicated by the right cursor.
The green trace is again the phase behavior of this measuring amplifier.


The pulse behavior
But how good is the pulse behavior now, if an OPA1611 is used.
This is a scope picture of a 100KHz square wave at a large output level of 6Vtt.
This is almost perfect, no abberation is visible around the flanks and I could make the flanks a little better with some compensation, but haven't had time to experiment with that yet.
I was already happy to get more bandwidth with the opamp replacement.


Block Diagram setup
And now some information about the measurement setup, this is because this is very important if you want to start measuring small signals at high bandwith.
Below is a block diagram of what equipment was used to do measurements on this amplifier.
Because a large part of the signals that this amplifier will process at its input are below 1mV,
special measures are needed to minimize the penetration of all kinds of interfering signals into your amplifier and your D.U.T.


There are plenty of function generators here, but they are all connected via the 230V power cables.
This results in all kinds of differential currents through the coaxial cables which, especially for small measurement signals, are regularly stronger than the signal to be examined.

For this type of application anyway, I bought a battery powered generator and it is in the OWON HDS272s.
Furthermore, all connections are made via RG58 coax.
The RG58 test cables I used in the beginning for connecting the in and output of the test PCB are now no longer used, the approximately 2x 10cm cables with their crocodile mouths picked up far too much interference signals.

For these tests, I wanted to have a nice control range for the generator output.
Setting the output level on the OWON HDS272s is not really confortable.
So I grabbed one of my HP attenuators and a extra -20dB BNC in line attenuator.
With this I could adjust the HP attenuator in 10dB steps over the whole control range.
The signal level from the OWON HDS272s is -20dBV then the HP attenuator, then a extra BNC -20dB attenuator and finally a 50Ohm BNC terminator, this because the amplifier has a 10K input impedance.

The Quant Asylum QA403 has an isolated USB connector.
Thus no disturbance is experienced from interference sources in the used computer, top! this is very important if you want to measure smaller signals.

THD measurements
To get an idea of how good the characteristics regarding distortion are, I did a test at 10KHz and an output level of 1V RMS.
The picture shows that both the 2nd and third harmonics are around -100dB.
The signal between marker 0 and marker 1 is an interference signal from the environment.
On the far left the 3rd, 5th 9th harmonics of the Net frequencies are visible.
At a signal level lower than the 1V RMS used here, the distortion drops to almost immeasurable.
The Quant Asylum then distorts -120dB, this I measured with my Audio Precision analyzer which has a generator distortion of almost -140dB.
So I am with these measurements far enough away from the measurement floors of the equipment used.


Bench setup
Now some pictures how the equipment was here on the bench.
The black box contains the measuring amplifier and everything is connected via coaxial cables.
Also note that the coax cables have ferite clamps around them.
This was necessary for certain measurements to keep the interference signals low enough.
For measuring with the Quant Asylum this was not really necessary, but there will also be some pictures with measurements done with a scope.



Here the additional -20dB attenuator is visible, so -120dB on the HP attenuator eventually becomes -160dB.
This is because the signal from the Owon generator is -20dBV.


Here is the measurement amplifier and the Quant asylum on the floor, this is to get the lowest possible interference level in my measurement room.


Here is an FFT visible with a bandwidth of 25KHz, the signal level at the input of the measuring amplifier here is 1uV!
The noise floor of the Quant Asylum is shifted down due to the +40 dB gain, as this measurement set does not normally have such a low measurement floor.


Just for the fun!
Now how far can I lower the signal so that it is just barely visible in my measurement setup.
And that's at -160dBV which is 10nV!


Measuring this this way and not with a scope, which measurements come later in part three, nicely shows what is possible with actually a souped-up PC sound card.
With no way you can make 10nV visible with a scope without a lot of amplification, bandpass filters, and filtering in the software of the scope itself.

Of course, this Quant Asylum measurement set also has limitations, it is a 192KHz sampling system, so just short of 100KHz measurement bandwidth for some applications.
But this bandwidth is sufficient for many measurements and from 100KHz you can usually apply a "normal" spectrum analyzer.

In part three, I'll show you what you'll run into when you start making broadband measurements at small signals with a scope.
Siglals, Signals, where are my signals!, is see only noise, SMPS, hum etc.? 

I would love to hear your questions and/or comments!

Greetings,
Bram

[blackdog]

Hi,

Now first some pictures about scope noise levels at 1mV input sensitivity and 20MHz bandwidth, the test signal itself is 0.1mV.
These measurements were made to visualize the differences between a number of scope at small input signals.

The scope pictures were made in the following way:
20MHz bandwidth
1mV scoop input sensitivity
Input impedance scoop: 1Meg
Timebase: 1mS/Div
Function generator delivers via an HP 355D attenuator 0.1mV RMS 1KHz Sinus
The scope is terminated at the input by a BNC 50 Ohm terminating resistor
No filtering of the scope software was used with these pictures only the 20MHz setting
Externally, no filtering was used either.
Triggering was performed via the second channel of a Siglent SDG1032x and a isolation transformer was used for the trigger signal to one of the other scope inputs which is not shown in the pictures

1e
This is the display of 100uV RMS Sinus on the Micsig STO1104e battery scope (Sorry, the probe setting is wrong but the picture is OK)


2e
This is the display of 100uV RMS Sinus on the Hameg HMO3004



3e
This is the display of 100uV RMS Sinus on the Siglent SDS2104x HD


4e
This is the display of 100uV RMS Sinus on the Siglent SDS2104x Plus


The two Siglent scope with their 10 and 12Bit resolution are the winners.
But the Hameg and Micsig can make the signal even more visible with some filtering.
But that will come in the next part, my time to post these measurements is up for today.
There is work to be done!

Kind regards,
Bram

[MasterT]

I have a question regarding low freq. end.
Simple math esimation of the impedance 22 uF cap at 10 Hz shows 723.4 Ohm or whatever equivalent for reactance.
Having 3 pA/sqrt(Hz) noise current for OPA translates into ~2nV, and even worse for <10 Hz, where noise current goes up and cap impedance folow same direction basicaly squaring up speed.
 Is there anything wrong with my line of thoughts?

[blackdog]

Hi MasterT, :-)

Thank you for your question!
That was one of the many considerations when developing this amplifier.
On my test pcb is even only 6.8uF, this because of course I had again left too little space on the PCB.... 

There are already two capacitors ready here for the final of the PCB.
See the picture below.
The gray capacitor is a 22uF 63V and the six coupled 10uF are only 50V.


Please note
The maximum voltage of the 22uF capacitor is a concern, though.
Take the value of the capacitor 10, 22, 47, or even 100uF if low 1/f noise is important to you.
Also note the maximum value of the DC voltage of your chosen capacitor, .

Choose the voltage appropriate to the power supplies you will be measuring.
If you need 250V, keep room for large capacitors!
No! do not use ceramic capacitors, they are almost always bad and very voltage sensitive at large capacitance values.

In a follow-up post I wanted to discuss the 22uF capacitor and the 39nF for the 400Hz filter.
But now you can, while thinking about this amplifier I also thought about using a dual JFet with the NE5534a as in an application note from Siliconix.
The nice amplifier with the dual JFet by Jim Williams also came up.

But very quickly you want even more, and more then it is no longer KISS.
What I present here as an amplifier stage some filters around it, couldn't be simpler and even you need the filters if you're going to work on linear power supplies to see their noise and hum behavior.

So the purpose of this amplifier is not to measure the noise at the LTZ1000 from 0.1 to 10Hz.
There are plenty of schematics for that on EEV Blog.

Power Supply Noise Levels
The latest version of "The Art Of Electronics" has some measurements done on Power Supply's regarding their noise (PARD) behavior.
I could not immediately find this this morning, but will post a picture of their measurements posted in this topic.

To get back to your comment, that's right the noise behavior of this amplifier gets worse in the low frequencies due to the bias current because of the XC of the 22uF.
But with most Linear Power, the 1/F noise this adds is completely overshadowed by the 1/f noise from the LAB Power Supply's.

Bigger input capacitor
But there is nothing against putting in a larger value capacitor as far as I am concerned if your application demands it.
The 10 Ohm input resistor together with the two 1N4007 diodes protect the opamp input and ensure that when you connect a power supply the capacitor is for the most part quickly charged.

400Hz filter
When the 400Hz 6dB/Oct filer turns on the bias noise becomes even stronger.
But when do you need this 400Hz filter?
That is when the 1/F noise from the power supply to be measured is so strong that you cannot do any meaningful measurements with a scope, the same goes for mains frequency components.

Scope and/or Soundcard
Measurements with a sound card or with a Quant Asylum will show you the behavior of the power supply.
But I still prefer to do the first measurements with a scope, and then over a wider bandwidth.
And with the latest version I reach more than 700KHz at -3dB with this amplifier.

In many measurements I start, this usually happens at relatively large bandwidth.
Then I get an overview of what all is going on with the D.U.T.
Then I reduce the bandwidth where possible by filters and e.g. averaging in the scope etc, etc.

I hope this explanation clarifies some of the considerations I made.

Sincerely,
Bram

[MasterT]

Right now I'm designing pre-amp for noise measurements in 0.1-10Hz range (possible 0.01-1k), and what I came to conclusion that I can't use super-puper low noise BJT OPA (~1 nV or so) due to the fact that DC blocking cap must have enourmously larger value.
And JFET OPA mostly have not so good input voltage noise <1 Hz. I just have to dig i-net for circuits designed in 70-th (!!!) and combe something usefull with JFET (2sk3557-6).

[Gerhard_dk4xp]

I've measured the noise of some power supplies some time ago:
<      http://www.hoffmann-hochfrequenz.de/downloads/Noise_Measurements_On_Some_Laboratory_Power_Supplies.pdf       >

And yes, the input RC is a problem. It has driven me from 20* ADA-4898 op amps in parallel
to 16 * CPH3910 FETs. The usual FET preamps have a problem of their own because they
present a negative real part of their input impedance in series with a smallish capacitance.
Add 200nH to the input and you have an oscillator.
Even circuits from text books. The cause is usually feedback from the first gain stage via an
op amp to the source of the JFETs. The op amp is too slow. CFB amps are much better here,
but their 1/f noise cannot be masked by the FET stage gain. THS3091 seems to be the best
of the CFBs,
I'm working on a solution with an emitter follower that probably works at the cost of
> 50-60 mA consumption. Current state of the confusion: At least, the sources follow
the gate, so Ciss is harmless and the cascode is an effective Miller killer. Also, the real part
of the input impedance does not fall below + 80 Ohms.

I tried to use the bias servo from the ribbon mike amplifier from AOE ed3, but it could
introduce instabilities at milliHertz frequencies and in actual measurements that takes
forever to track down.

[blackdog]

Hi MasterT, 

Your question is far beyond the design shown by me in this topic.
I know the problem, For measuring voltage references I have an Euler Precision LFLNA-80 available here.

However, this only goes to 0.1Hz and has about 1000uF as input capacitor.
If you want a crossover point of 0.01Hz, be careful for what you wish for... 

How sure are you going to be that what you're measuring is below 0.5Hz and that this is coming from your D.U.T.?
Draft, thermal drift, offset drift of the opamps used, etc, etc.
You will need multiple layers of shielding and then use e.g. aluminum tube so you add some thermal mass as well.
A well regulated oven at say 35 a 38C can also help you, but this will give you a somewhat higher noise number due to the 38C.
An "Oven" cooled with a Peltier to 10C can help you improve the noise number.
There are many possibilities, but all take a lot of time to make and test.

For your impression, my 3458a I bought over from someone who had a stack of them for his work.
He also did maintenance on them, and showed me the drift of several 3458a when paired with a Fluke 732a references.
My 3458A took about 18 hours to get the drift smaller than 0.1PPM from power off.
Most 3458a meters all varied a bit say from 15 to 24 hours and even longer to reach their final stability in a temperature stable environment.

And then when you consider that you want to build an amplifier that has 0.01Hz as a crossover point, so that's going to be very difficult.
Especially since with that you will also have to measure much longer than the 100 Seconds that is there for one measurement.

But I like to see how you are going to want to solve the problem, i like to learn.

Kind regards,
Bram

[MasterT]

Agree, application is different, common part is only low freq. end . Big cap size is not appealing solution since liekage current is proportional to capacity as well.
 I start another thread (not to de-route here) where I will try different approach.
https://www.eevblog.com/forum/projects/measuring-dut-low-frequency-noise-0-01-100-hz-below-noise-floor-of-the-test-equi/new/#new

[Vovk_Z]

I have a question regarding low freq. end.
Simple math esimation of the impedance 22 uF cap at 10 Hz shows 723.4 Ohm or whatever equivalent for reactance.
Having 3 pA/sqrt(Hz) noise current for OPA translates into ~2nV, and even worse for <10 Hz, where noise current goes up and cap impedance folow same direction basicaly squaring up speed.
 Is there anything wrong with my line of thoughts?

I thought a noise makes a problem only with active resistance (noise current has to flow through an active resistance), but not with imagine reactive reactances. I may be wrong. I have a hole in this part of 'noise knowledge'.

[MasterT]

I have a question regarding low freq. end.
Simple math esimation of the impedance 22 uF cap at 10 Hz shows 723.4 Ohm or whatever equivalent for reactance.
Having 3 pA/sqrt(Hz) noise current for OPA translates into ~2nV, and even worse for <10 Hz, where noise current goes up and cap impedance folow same direction basicaly squaring up speed.
 Is there anything wrong with my line of thoughts?

I thought a noise makes a problem only with active resistance (noise current has to flow through an active resistance), but not with imagine reactive reactances. I may be wrong. I have a hole in this part of 'noise knowledge'.

Current is charging a capacitor. If current fluctuates, than voltage on a cap noisy as well.  Higher input bias/ offset current of the op-amp imply higher current noise.
 In general, reactive or not impedance is, it follows the same rules - small capacitors are Noisy by default the same way as big values of resistors.  They are noisy if no current at all and no connection

[blackdog]

Hi,

Thank for al your remarks and Gerhard thanks for the info on your website.

And now back to this topic. :-)
Taking the scope pictures takes a bit more time, so here I show you a add on for a piece of electronics with which you can calibrate the measurement amplifier combined with a scope or sound card.
The diagram below is intended for the measuring amplifier of this topic, this because I assume that the Cal generator will be loaded with 10K, this because the measuring amplifier has this input impedance....


These are the starting points for this calibrator schematic
No really difficult components.
1% resistors
5% capacitors
Low power consumption
Output frequency about 1KHz
Square wave output of 1mV Top Top
Sine wave output of 1mV RMS
For the square wave and sine wave output simply adjust with one trim resistor

It is not necessary in my opinion to make the frequency and output signal of this calibrator better than a few percent.
This is because that is what it is not intended for.
When you are measuring, you sometimes get a bit lost in different signal levels and then it is easy to have a square or sine signal of which you know what the approximate level is.
I got with the trim resistors for the square and sine wave outputs within a few minutes the level within 1%.
For those who like trimpots, do your best to adjust the schematic with that, choose a control range of say +-10%.

Why a square and a sine output?
That's easy if you look at FFTs, a 1mV RMS signal is -60dBV.
And if you look with a scope, a signal with a known Top Top voltage is easier.
1mV Sine RMS is 2.83mv Top Top and if you are busy making measurements, doing extra calculation in your head is not a good plan.

It was relatively easy to go from a 1KHz square 5V TT signal to sine with an approximately 18dB/Oct filter.
This little filter I didn't quite calculate, bit of wet finger work and making sure the output impedance was low enough for a 10K load on the measuring amplifier input.
And the signal also had to be reasonably sinusoidal.

This is what the 1mV and about 1KHz sine wave signal looks like on the FFT of the Quant Asylum.


The fact that there is slightly more than 3% distortion present is irrelevant to me.
It is the ground wave of just over 1KHz that is what it is all about, you know it is within a few tenths of a dB -60dBV, and that is then your reference point for doing your measurements.

This what I'm showing here is not a complete kit, just think of it as a direction on how you can go about making something.
If we take this calibration generator as an example, you can with say two pushbuttons switch the input of the measuring amplifier to the square or sinus output of the calibration schekaling, this to quickly see where your reference is.

But it is entirely up to you whether you like to use this, the diagram is in any case available for the calibration generator.

IC's used
Regarding the ICs used, the TS2950a is a Low Drop TO92 5V stabilizer with a power consumption of less than 0.1mA.
The ones I have here are well within 1% in terms of accuracy.
But be aware that the output signal on the two outputs is directly dependent on the power supply voltage of the LMC555.

Regarding the ICs used, the TS2950a is a Low Drop TO92 5V stabilizer with a power consumption of less than 0.1mA.
The ones I have here are well within 1% in terms of accuracy.
But be aware that the output signal on the two outputs depends directly on the power supply voltage of the LMC555.

It is not possible to use a "normal" NE555 in this schematic, it will draw much more current which is not pleasant for the batteries,
and the output signal is not so nice, causing the voltage dividers on the output of the 555 to be incorrect.

The next post will be some scope photos again or how to best connect test leads to a Power Supply under test that so you get as little measurement error as possible.

Kind regards,
Bram

[blackdog]

Hi,

Today I made and adjusted the Calibrator PCB.
And below a small photo report of the work on this board and some comments.
First I updated the schematic and with a page refresh the changes should be visible.

In principle, this calibrator is meant to be used with the measuring amplifier I show here, but be free to use it for other applications as well, it is your party. 
Sinus output distortion is 3.3% which could be improved a bit, but not necessary.
Why do I bring this to your attention, that is if for another application the THD should be better than say 1%.
In the FFT picture the 2nd harmonic is dominant, if the square wave duty cycle were nicely 50%, then there should be almost no 2nd harmonic in the FFT.

Optimizing the sine output
Also the Sinus filter could be a bit better, but both adjustments again cost me more time and for this project not necessary.
So go wild with Spice, to optimize the specifications if you want.
Make sure that the output impedance of the Sinus filter is not too high for your application, now it is about 150 Ohm.

Below is an Zoomed in FFT picture showing the Sinus output after I tuned the 1KHz frequency it with a small extra capacitor and also brought the output level up nicely with an extra resistor.


Adjusting the 1mV sine wave output
I did the tuning of the 1mV (-60dBV) sine wave with my Audio Precision measurement set.
Here I selected the Level function and then turned on the "band pass" function "Selective".
This filter has a Q of about 5 and filters out a lot of noise signals so that I can measure a clean 1KHZ signal.
The indicated value in the orange box is much better than necessary, but it was simple to adjust this value to this level with one trim resistor.



Adjusting the 1mV sqare wave output
I adjusted the 1mV TT square wave output with a scope and an external filter(100KHz) and as visible on the scope picture 16x averaging.
This way I can adjust the square signal well enough on two divisions, the scope is here at its maximum sensitivity of 500uV/Div.
And again, it is not necessary to tune it perfectly.
The point is that you have an impression of the signal level when you are setting up the measuring instruments or so if you have already made a number of measurements that you check the level with your Cal. signal just to be sure.

As for the interference signals that you can still see in part in the yellow trace, I'll come back to that when I go over the measurements with the amplifier.
The yellow signal has already gone through a 12dB/Oct LP, this is to ensure that at this high sensitivity setting the ADC in the scope does not clip.
I have a rather high level of interference here on my workbench....


The Calibration PCB.
I made the PCB quite compact, this is on purpose, the wiring is done as short as possible to keep radiation to the other parts of this project as low as possible.

On the top left you can see three capacitors the middle one is C3 in the schematic and the two ceramic capacitors I used to bring the frequency within 1% of 1KHz.
Left below are three 1% resistors visible side by side, the leftmost is R5 of 15K, the middle resistor Is R7 of here 270K and is the trim resistor for the 1mV square output.
Bottom left the white capacitor is C5 0.22uF , the second 0.22uF is C6 and the red capacitor is C7.
The orange capacitor is C4 and I have to adjust the schematic for that because it still says 0.15uF there.

The electrolytic capacitor is C2 at the top right of the IC and the TO92 case is the TS28950a 5V regulator.


On the Dutch forum www.circuitsonline.net, where I can be found a lot, it is a custom to always show the underside of your PCB.

So here is a picture of the underside of the PCB.
This is the unfinished state and this is intentionally left this way for this picture.
The operation is now good and then it is my turn to clean it up.
This includes cleaning with IPA for the resin residue, removing all solder drops, correcting solder that is not nice enough, etc.

By the way, this is a bad piece of PCB from China, those Chinese manufacturers have managed to make such a simple product bad again for their profit.
I have several brands and different sizes of this kind of PCB material, but when you make a solder with this one,
gases come out of the PCM material and it is difficult to make soldering without gas bubbles in the solder, mini brains!


The next post is again about the measuring amplifier itself and how to connect it to e.g. a LAB powersupply.
And, of course, how the amplifier improves the signal, this without adding too much noise.

Comments, I like to hear them!

Greetings,
Bram

PS
I have dyslexia and I use deepl.com as a translation machine.
So my sentences will occasionally or regularly be rather crooked.
Such is life!   

[temperance]

...when you make a solder with this one, gases come out of the PCM material and it is difficult to make soldering without gas bubbles in the solder, mini brains!

The PCB might have absorbed some moisture and the gasses escapes trough voids in the plating. (The amount of voids in the plating can be very high if the holes were drilled with blunt tools.) You can try drying the PCB's. But it takes a while for the moisture to escape because most moisture must escape trough the edges of the PCB.

[shabaz]

I too used OPA1611 for a measurement amplifier attempt.
Not to distract from this great thread (but just in case it gives you any ideas on things to do or definitely not to do! - I'm no analog expert. But perhaps the enclosure stuff or some of the features could be useful add-ons or trigger new ideas in a better design than mine).

Things I would change in my design if I were to do it again:
* Reduce the AC input capacitor to 12uF. Although that would increase the cut-off frequency, I think it would be better than seeing as much 1/f noise.
* Add a 100 kHz roll-off, so that if the extra filter buttons (for 100Hz, 1k, 10k) are not pushed in, then the 100k applies.
* Find a cheaper alu enclosure, the brand boxes are getting too expensive these days
* Move to KiCad (this was one of the last designs I did with EAGLE)

Things that worked out well:
* Surprisingly useful when examining supply rails
* Eliminated almost all wiring (just the battery clips, and a BNC connector to the rear).
* The front panel worked out well, it is a JLCPCB alu PCB - but if I were to do it again, I would use FR4, the difference is so slight.
* Zeroise button comes in handy

[blackdog]

Hi,

Conducted a little experiment today, this because I can't resist. 

I was doing some additional measurements on the test PCB and I had mentioned earlier that it might be possible to get the bandwidth in the higher frequencies a little better.
This despite the OPA1611 there almost no more loop gain above 700KHz.
But you only know when you've tried it and that's what I did today.
One of the advantages of the OPA1611 is, that at higher frequencies it still has a low output impedance,
look at figure 28 in the datasheet and that helps with driving the relatively low value of the feedback resistor R13 of 1K.

The latest version of the schematic.


The Bode Plot with C23
And this is the result, -3dB @ 1.08Mhz, -1dB @ 583KHz.
At the time of making this Bode Plot I had an additional 20dB attenuator turned on, therefore the gain scale is shifted, but that does not otherwise matter to the measurement.


I chose the compensation capacitor so that no abberations occur when using a square wave for large signal and for small signals.
I also tried a higher value for C23, but then I get an increased frequency response around 800KHz, which I tried to compensate with a series resistor, but no succes.
I have a preference for pulse signals that are clean, so no abberations and therefore I found the extra gain with C23 sufficient.

Square wave responce with C23
And how does the square waveform now look at the enlarged frequency range, judge for yourself.


Cleaned calibration circuit board
Also cleaned the Calibration PCB today, updated the soldering where needed after spending half a day on the heater to get some moisture out of the PCB material.
The PCB I have brushed three times with a hard toothbrush and IPA and still I see a solder ball....
And another point that needs to be touched up....



Maybe tomorrow more info on the progress of this project.
My workflow is almost never linear, so we'll see what I like to post the next few days. 

Kind regards,
Bram

PS,
shabaz and temperance thanks for the info and remarks.

[blackdog]

Hi,

Did someone say more pictures?
Let me see if I can take care of that with some meaningful information in it. 

Small corection schematic calibrator
First, a slight modification of the calibration schematic.
I didn't think it was really neat of me that I hadn't put a little more effort into bringing the duty cycle more to 50% without increasing the number of components.

I used a 555 calcualtor that was available online, to maintain reasonable resistance values and bring the duty clycle as close to 50 percent as possible.
Also, the I wanted to use the component values from the E12 series, so no extraneous values.

I have, using the "on line" calculator
was able to meet the conditions of standard E12 values and finally achieve a duty cycle of 50.25%.

What I pointed out earlier, the 2nd, 4th, 6th, etc harmonics disappear from the FFT at exactly 50% duty cycle.
The reason I finally made it a little better is because I could see on the scope that it was not a nice sine wave.
And now I can no longer see that, the distortion is now just over 2% and was about 3.3%.
But for level reading this adjustment was/is not necessary.
Disadvantages of this modification, the only thing is that due to the lower value R3 is now 1K2 and therefore the current consumption is just over 3mA, for me this is not a problem.
This is because this circuit is only on for a short time anyway.

The modified calibrator schematic for lower THD.


In this FFT image it can be seen that the even harmonics are now much lower in strength this compared to the previous FFT image of the calbrator.


First set of scope pictures when the 100x amplifier is used.
I use the Hameg and one of my Siglent scopes for the following measurements.
The Hameg scope is used to clarify something faster, this is because this scope has e.g. a "Quick View" button and sometimes filtering can be set faster.

These kinds of measurements I show here take a lot of time to make,
if you want to explain something you have to make sure you do it carefully and measurement errors are easily made.
The scoops used are always set to 20MHZ bandwidth.
The input impedance always set to 1Meg, this is because the measuring amplifier is designed for that.
If filtering or averaging has been applied, this is indicated.

This measurement was done with an open input of the 100x amplifier.
No filtering or averaging was applied and the scope was set to "Peak & Hold"



Now the input of the amplifier is terminated with a 50 Ohm termination resistor.
Note the vertical scale, it is now 10uV/Div and in the previous picture the sensitivity was 50uV/Div.


This is a nice check if you have an amplifier that has low noise., input is shorted.
The noise in this measurement is slightly lower than the measurement with 50 Ohm termination resistor.


Scope noisefloor
To show that the noise of the scope is low enough I left everything the same, just removed the 100x amplifier and terminated the scope input with 50 Ohms.
The noise level of the Siglent scoops I also used are lower than the Hameg scope.



Measurement amplifier performance with different signals.
Here it is clearly visible that the measuring amplifier has a large bandwidth, it is not visible here at a 1KHz square wave that the measuring amplifier is between the Generator and the scope.
Only at the rise time is visible that this does not quite match what the Generator provides in terms of rise time.
Also remember that the scope is at 20MHZ bandwidth.

1mV input wide band, so no filters



1mV input 100KHz Low pass
This is with the same input signal only a 100KHz filter at the output.



1mV input 20KHz Low pass
And now with the 20KHz filter on the output of the measuring amplifier.
The signal got a little smaller, which is partly due to the lower bandwidth, but also partly due to the filter used.
The component values of the filter used are slightly different from those shown in the amplifier schematic, but the filter has the same bandwidth.
So only the impedances are a little different, which is otherwise not very important here.



100uV input, No filters
Here one picture at 0.1mV input signal and then without any filters except the scope 20MHz filter.
The broadband noise is now well visible at this low signal level.
But keep in mind, that the measurement amplifier has a -3dB bandwidth of 1MHz.


10uV input, No filters
Now we go one step lower with the input signal, and this picture was taken with 10uV top top and across the full bandwidth of the measuring amplifier.
There is very much noise visible now, and at this bandwidth, at this signal level, little sense will be made of what appears on the scope image.



10uV input, 100KHz Low Pass
The signal can be recognized quite well when the bandwidth is reduced to 100KHz.



10uV input, 100KHz Low Pass + 1024x Avarage
1024 times means is naturally exaggerated and it also takes a lot of time to build this scope picture.
But it shows well what is possible with a modern scope.



10uV input, 20KHz Low Pass
By using the 20KHz filter, it is also quite possible to get a good picture of this small signal.



The first measurements on a LAB power supply!
But doing measurements involves something beforehand, and that is the proper way to connect the measurement cable to the Power Supply.
But first a few pictures of the cable I use for this.
That is a coax cable from Belden RG58 with the type number: 9223



And in this way I prepared the cable for measurements on lab power supplies.



I added a little solder to both ends of the cable.
And I pinched the shield a little flat with the pliers visible in the foreground, this to secure the cable under the terminal blocks/connectors (if any)
It is of course also possible to use small connectors that you can clip under the banana connectors.
Keep in mind to keep the measuring cable a little longer, but not too long, because of EMC signals.
It is best to take the length into account because this cable is not very sturdy for the way I connect it.



This is the right way with these banana connectors that most power supplies are still equipped with.
This picture shows what is almost the best way to connect the measurement cable.
At a 90 degree angle with the twisted banana wires running to the load.
And so connect the measurement cable in the way shown in the photo!



I have seen many videos on "junk tube" of people testing power supplies and dynamic loads not knowing how to do it properly.
So a few more pictures on how NOT to connect the test lead. :-)
Seems convenient, but it is not the right way, think about connector resistance and inductance of the connections, especially if you're going to do dynamic testing.



Haha, You are a comedian by profession? 



And now some real measurements on an HP 6237A power supply.
And first, the +output which can go from 0 to +20V and can be loaded with a maximum of 500mA.
The output is loaded with about 300mA in this measurement.
I show measurements without filtering, at 100KHz bandwidth and 20KHz bandwidth.



This is a measurement with 100KHz bandwidth.



This is a measurement with 20KHz bandwidth.



I draw the conclusion that this channel is nice and clean.
Only a very small portion of the noise at higher bandwidth comes from the power supply itself.
I have done tests with and without load on the power supply and the differences in noise are very low, otherwise not worth pictures.
But not everything is so nice about this power supply, I now let you see a measurement of the 18V channel.

18V @ 1-Ampere 100KHz bandwidth.



At the 18V output, the mains frequency is doninand.
But with 0.8mV top top, this power supply still meets modern requirements.
The other measured channel with 100mV top top is really very good.


More More More
By the way, there are many more measurements to do and a few I did, but what you see here today in this topic, is 8 hours of work. :-)

I want to tell more about wiring, so how to connect things and what to look out for when doing P.A.R.D measurements.
Just connect a second measurement cable, a coaxial cable to trigger something and you suddenly find yourself with a completely different signal on your scope, why?

You think you are wise and you connect the ground connection of your power supply to ground of your measurement system and you have  a horror signal on the scoop, why?
cell phones of your own and/or colleagues, transmitters in the neighborhood, etc, etc.
So many things to tell, to make as few measurement errors as possible.

And of course, what is of importance of the PARD signal you see on the scope.
The noise up to 100KHz is probably local filtered by decoupling in the D.U. T. you are testing.
Connecting cables that are too long and not twisted are wonderful antennas for both your power supply, but also for your D.U.T.

And I now only showd measurements of linear power supplies.
The fun starts when you start measuring on anything related to SMPS power supplies.
I'll will also show some picture's of an SMPS power supply in a future post and I'll call this power supply an EMC Lighthouse. 

Time to make dinner for my Girlfriend! 

Shoot!
Bram

[blackdog]

Hi,

Today a small post about an SMPS power supply that I have described as EMC Lighthouse.

I needed a power supply that could deliver a little more current and I thought I'd buy a modern SMPS power supply for the few times a year I need it (I should know better, actually).
I ended up with the reasonably priced OWON SPE3102, 30V at up to 10-Ampere.
See the picture below


A few things are certainly very nice about this power supply, such as the display and a 5V USB output and some more nice menu options.
The first thing you notice when you turn it on is the good looking color display and then comes the... they built in an F16 jet engine that runs at full power!
How is it possible that some things pass quality control at OWON....

But then the horror begins....
Connect this power supply to your D.U.T. and everything instantly finds itself in an EMC fog.
I hadn't mentioned that the power supply was already turned on with the On/Off button on the front panel.
It is still off then, let's see what is on the output terminals of the power supply that is on but on the front the standby button is still off.

On standby
Look at the scale 2Vpp! you gotta be kidding me.


Power On 12V 750mA
Now the power supply is on, set at 12V and 750mA, hello World can you here me!
Yes indeed probably over 100 Meters away with some longer cables to the D.U.T.


This result was a bit too much for me, I spent 15 minutes crying in a corner of my LAB. 
Now what could I do about this, let's build a commonmode filter in a steel case with an extra capacitor with a low ESR.

External Common Mode Filter
A toroid that came from a UPS was used for this filter, as well as a Rubicon low ESR capacitor and finally a ferrite clamp.


Stand By
And now some pictures of the result when this filter is used.
This is when the power supply is on stand by with the filter, look at the scale.



Power On 12V, 750mA
And now some pictures of the result when this filter is used.
This is when the power supply is On with the filter, look at the scale.


Conclusion OWON SPE3102
When using an additional external filter, this power supply is usable without injecting too much interference into your D.U.T. or other equipment.

That leaves us with the F16 Jet Engine, ooo sorry the fan....
This is an SMPS basically a low loss circuit, so why put in a fan running at full power?
What were they smoking when developing this product?

If you look inside the box at the PCB you will see that it is built modern with good products as far as I can tell,
there were some strange decisions made when developing this product, probably the marketing department and bean couters had too much influence.

Of course I am aware that 95€ is not a lot of money, but if I can buy a very nice PC power supply for the same money that provides more power,
and a much lower interference level, something is not right.
So be mindful when you buy an SMPS power supply, what does the interference level do to the rest of your environment.
Do you charge batteries, control relays and/or motors without sensitive electronics, then I don't see any obstacles, except to protect your ears. 

Kind regards,
Bram