+40dB LAB preamp with filtering

[blackdog]

Hi,

Because I have some time left today, I want to show you another project I designed and built to measure the noise / PARD level of power supplies and also voltage references.
But not intended for measuring the best voltage references, this amplifier is not good enough for that.

First take a look at what the amplifier looks like after it's finished.



Functional schematic



And this is the schematic



Architect specifications
Power: 2x9V battery
Input and Output protection +-60V
Compact
RI > 20K
Rout 50 Ohm, unterminated 1M cable
Reasonable immunity to LAB Bench EMC
Switchable High Pass input filter
Switchable 2x Low Pass filter, 10KHz and 100KHz
Perfect square wave reproduction
No special components
Low distortion

As can be seen in the diagram, the amplifier is constructed with two opamps.
After some testing I chose the LM4562 because of the good mix of properties for my application.
TI has stopped producing this IC but still has a lot of work to do here in Europe.
There is also a replacement product available with better specifications in some areas.

Both amplifier steps have an HF trim function for the perfect square wave reproduction I wanted.
I have selected the resistors that set the gain, so that the gain is within a few tenths of a dB.

Where necessary, I have built up the circuit with low capacity for the best frequency response.
The whole circuit is built on the well-known green Chinese circuitsboards and so where it maked sense,
for example at the inverting input of the opamps I have drilled out holes.

There are many opamps usable for this application, but almost all of them have the annoying feature that the pulse response depends on the signal level.
The LM4562 is one of the few exceptions and that is why it is used in this circuit.


Some pictures of building the amplifier



Drilled out holes



Also on this sensitive opamp input...



On top the capacitor bank for the input and at the bottum the capacitors for the output



Some of the first test are done here.



square wave respons of a 10KHz signal.



And now at 100KHz.



1MHz!



A nice fast symetrical puls.



The bottomview



He is finished!



Testsetup outside the LAB for noise measuring



A -40dB reference signal and the noise from the amplifier in the low frequencies.



6Vtt a upgoing signal (10KHz)



a 1.2V signal, this is wat im mean, no abberation



Old Scool frecuency respons... 



400HZ Hig Pass filter.



10KHz and 100KHz filter respons.



Some distortion measurements done on a Audio Precision.
1V RMS Out 1KHz and 22KHZ bandwith.



And now at 4V RMS output level same filter conditions, lower distortion because of the bigger level noise distance.


Same 4V RMS level, but now a 10KHz signal en the filtering is at 80KHz, i do not complaine..



Question, what is the cause of the extra signal on this pulse.



And last picture, the simple low batt indicator that is used.



Shoot at it!

Kind regards,
Bram

[JohnnyMalaria]

That's beautiful craftsmanship 

I just had to check your profile to confirm my hunch about your age. I love that you used a pencil and graph paper. I think it is a skill sadly lacking today.

[Andreas]

Nice project. what housing do you use?

Hmm,

the 20k input impedance is only valid for small input signals.
Above 0.7V the input impedance is much lower.

Usually wideband noise is measured with a high pass of 10 Hz.
So I am wondering why you use 1 Hz and 400 Hz as high pass frequencies.

with best regards

Andreas

[blackdog]

Hi Andreas,

You should know this company 
https://www.reichelt.de/Fischer-Frame-Gehaeuse/2/index.html?ACTION=2&LA=3&GROUPID=7727

I wanted to have a signal conditioning instrument with as large a frequency range as possible, low noise and no square wave abberations.
This amplifier has a -3dB band of around 0.1Hz and 2.7MHz that meets my requirements.

This is only one amplifier in my set of amplifiers, a second one in the same kind of box has +20dB of gain,
some gain set properties and a high quality 50Hz notch filter (> 50dB notch dept)
I also have made some passive low and highpas filters, i can use with these amplifiers.

If you put in 0.7V in the amplifier, it would give 70V on the output 
Be kind to my amplifier please...

I hope that this remarks wil make it more clear.

Kind regards,
Bram

[GEOelectronics]

What you have built looks very nice, thanks for sharing.
I'm trying to specify the noise on this power supply, it is a higher Voltage and operates at 300 kHz:

George Dowell

[blackdog]

Hi GEOelectronics,

This amplifier is not designed for testing of your power supply, but you can use my amplifier set up to design something yourself...
For me, to be shure i dont overlook anything, i would use a amplifier with a bandwith of at least 25MHz if start measuring on switchers,
and even hirger on the modern switchers of 1 to 5 MHZ.
But not to difficult these day with modern opamps.

Kind regards,
Bram

[GEOelectronics]

Thanks Bram, do you have a website?

George Dowell

[blackdog]

Hi George,

Yes and No! 

www.noiseamp.nl

But its empty, i'am buissy on two forums, Circuitsonline in the Netherlands and EEVblog, not mutch time left for a website.
My wife already puls me on my hair out of the LAB.... (and she is very forgiving!)

Kind regards,
Bram

[blackdog]

Hi,

Still nobody knows wath te extra signal is on the puls picture?

Give it a shot... 

( I need the cause of this signal, but it is also cleverly annoying in broadband and small signal measurements)

Kind regards,
Bram

[pigrew]

Hi,

Still nobody knows wath te extra signal is on the puls picture?

Give it a shot... 

( I need the cause of this signal, but it is also cleverly annoying in broadband and small signal measurements)

Kind regards,
Bram

It could be oscillations.... not sure. I'd do some AC spice simulations to see if you have a gain peak at around 10 MHz (use your particular op amp's circuit model).

I would also try changing the op amp's decoupling capacitor to be two capacitors, one from each rail to ground.

The magnitude of the ripple is small (about 1 mV)... It could be picked up from anything in the room... you may also want to check if it is there between the first and second stages. Is the noise attenuated when you switch the filter cutoffs?

The enclosure is really pretty. I need to get off this forum and make things like this amp.

[pigrew]

Yeah... the amplifier seems unstable at 17 to 38 MHz if I'm reading the simulation correctly. The reg line is the overall amplifier response, and the green line is the first stage (and filters). Nevermind... I guess it is stable. I forgot I had to split up the amplifier and look at the gain of each stage individually.

I'm attaching a zipped version of the TINA schematic I used for the sim.

(And I forget 99% of the Nyquist plot stuff I learned in school)

[Andreas]

Still nobody knows wath te extra signal is on the puls picture?

Hello,

you make the quizz not easy: perhaps forgotten to close the lid of your cookies box where the amplifier is placed in?
All we can see is a 1us pulse which does not correspond to anything I know.
At least we should know the repetition rate.

with best regards

Andreas

[Gertjan]

Hi Blackdog,

I think your quiz about the ripple is too difficult... It is impossible to guess because nothing is known about the circumstances of your measurement.

Let me give a hint: The timebase of the scope is 1us/div, and I count 10 ripples in one division. So the interfering signal is 10MHz....   

Regards, Gertjan.

[2N3055]

Looks like cca 10 MHz triangle wave superimposed on the signal....

[blackdog]

Hi,

If I had made the question easy, I would have had the whole world in this topic. 

Gertjan, you may not play here 
Why, he is largely the reason of the extra signal on the displayed pulse!

Andreas: this is one of the reasons: perhaps forgotten to close the lid of your cookies box where the amplifier is placed in

pigrew
It is oscilation but not of this amplifier, because it is of perfect stability, look at the squarewave respons, no ringing.
I have a reasonably full LAB with modern measuring equipment, and what is often common, except this guy, yes of course EMC, but this is very stable?


Kind regards,
Bram

[GEOelectronics]

10 MHz? RFI? Stable?
Well after seeing your shop, my suggestion if the signal is stable, start turning off everything with a timebase (some stay running even if device is switched off).

Easier, make a magnetic probe for your scope (or better- Spectrum Analyzer if you have) and "go fishing" around the workbench with everything turned on.

I keep a flashlight on the bench because when tracking down a stray, first thing is turn off all the Fluorescent and LED overhead lighting.

PS your workbench is a lot neater than mine!
George Dowell

[blackdog]

Hi,

As I had already told you is Gertjan is the reason of the 10MHz signal on the pulse 

I Build his 10MHz lab reference and connected many devices to it.
The cables are radiating this 10MHZ signal and if you are measuring wide band amplifiers, it wil always pick this signal up.

And when i switch the 10MHz LAB Reference off, several instruments start complaining...
Not a big problem but just anoying.


A lot of info in the links below, just use google to translate.
Gertjan and i like to talk a lot en also measure a lot, so have fun!

Here a link to Gertjan design of a 10MHz LAB reference:
https://www.circuitsonline.net/forum/view/126987

And this is my 10MHz LAB reference based on Gertjan circuitboard/design.
https://www.circuitsonline.net/forum/view/126898/1#highlight=10mhz

I dit the noise measurements out of the LAB in the middle of another room to lose the most of the EMC problem.

Kind regards,
Bram

[ConKbot]

Time to replace all the RG-58 for the 10 MHz distro with LMR-195UF so the signals stay where they belong.  .

[blackdog]

Hi ConKbot,

I think it's a good tip to look at a coax cable that has a better shielding.

I still want to do some work on the 10MHz LAB reference, because I reversed two outputs, it wasn't the alcohol but I think the tiresiness.
In Gertjan's design I have added an output with a lowpass filter so that all harmonics are well filtered.
And then I can also test whether the floating outputs (symetrical transformers) with a small capacitor to ground will somewhat less radiate.

The 10MHz radiation is annoying when measuring small broadband signals.
Especially if you want to make use of synchronized measuring equipment.
It is on the list to do in the coming summer.
But I don't think it will be the LMR-195U coax, but something that fits well in a BNC connector and is also for sale here in the Netherlands.

But thanks again for the tip 

Kind regards,
Bram