EMI-Measurements of a Volt-Nut

[Andreas]

Overview:
=========
In metrology we try to measure precision voltages with stabilities/accuracies in the sub ppm range.
But usually the environment conditions regarding RF-Noise (EMI-Noise) are not ideal.
So readings can differ from location to location depending on RF-sources like:
- switch mode supplies (LED-lamps) (some 10kHz to some 100kHz),
- conducted emissions from USB/network cables, DC-motors, relays (contact bounce "fire" up to 60 MHz)
- and radiated emissions from key-fobs, radio stations, mobile phones (usually above 30 MHz).

https://www.eevblog.com/forum/metrology/metrology-grade-lighting-i-am-in-the-dark/msg2225853/#msg2225853
https://www.eevblog.com/forum/metrology/(ft)-ltz1000a-fairy-tale-or-the-story-of-little-jumper/msg1140683/#msg1140683
https://www.eevblog.com/forum/beginners/oscilloscope-interference/msg2530602/#msg2530602
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg846835/#msg846835
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg369383/#msg369383

Even heavy shielded commercial devices can be affected by EMI and demand for "highly controlled" RF environments:
E.g. Fluke 732b is specced according to the data sheet for 0.18V/m max field strength.
https://xdevs.com/doc/Fluke/732B/FLUKE_732B_734A_INST.pdf
House hold appliances have to withstand > 3V/m and in industrial environment devices have to withstand > 10V/m.
Of course a professional metrology lab will keep all those influences away from sensitive equipment.

But in a hobby lab we usually do not have perfectly shielded rooms and no influence what the neighbour does behind the wall.
So the devices have to be "better" to keep their stability in unknown conditions.
Shielding is one possibility which works good for radiated EMI at higher frequencies (above 30 MHz).
But the power supply/output lines are "antennas" through the housings so they have to be treated with filters.
Filtering possibilities are using capacitors to "short" the RF frequencies and using chokes / ferrites to increase the line impedance to keep the RF out.


cellularmitosis uses EMI-ferrites+capacitors with his "LTZ1000 board (PX-ref v2.4)" to increase immunity against EMI noise comming from supply lines or reference output wiring.

https://www.eevblog.com/forum/metrology/usa-cal-club-round-2/msg1501783/#msg1501783
https://www.eevblog.com/forum/metrology/usa-cal-club-round-2/?action=dlattach;attach=418882

Dr Pyta uses SMD ferrites on his board
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg878662/#msg878662
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg941288/#msg941288

TIN widely uses multi-layer which gives some additional shielding.

To check the effectiveness of filters we need a possibility to test it.
Otherwise we do not really know wether the measures are effective or if they only quiet conscience.

EMC immunity is tested in several ways.
Above 30 MHz up in the GHz range usually a antenna and a capable ~1000W RF-amplifier is used to test against radiated emissions.
In the lower frequency range the test is done either as current injection (BCI Method) or as capacitive injection to test conducted immunity on the power supply and signal lines.

To keep the costs low I decided to use the capacitive injection on signal lines which is usually done from 150kHz up to 80 MHz or even up to 230 MHz in some applications.
So the ideal signal generator has at least up to 20Vpp output voltage and a frequency range from some 10 kHz up to 230/300 MHz, and is able to produce CW + AM modulation (typ. 80% AM @ 1 kHz)
The AM modulation may be demodulated by input protection diodes and give a "offset" on the input signal with the modulation frequency. So AM modulation is usually the test which is harder to survive.

The upper frequency ranges can usually handled by a metal housing and ferrite beads on the signal lines.
So I will try (in the first step) to find improvements for the lower frequency range.
And of course I need to have a possibility to measure the improvements.

Caution: these tests may not be always destruction free!!!

work in progress, will be updated ...

[Andreas]

Setup:
=====

The whole setup is similar to standard IEC 61000-4-6:
As I do only comparative measurements I have done some simplifications.

The table surface is covered with overlapping tinned steel sheets (RS-Components 682-472) to have a common ground.

The coupling decoupling network (CDN) is self built and similar to a CDN-AF2 network according to the standard.
I used NiZn ferrites similar to those from the standard and changed the number of windings to get the necessary inductivity.
According to standard a 6dB attenuator is to be used directly at the RF input of the CDN.
As I feared that the signal level of the signal generator would be too low I decided to include a 3 dB fixed attenuator within the CDN to protect the signal generator somewhat from short cirquits and maintain a higer level.
The CDN has 2 ports:
1 port for the DUT/EUT (equipment under test) where the RF is coupled in.
And another AE port (auxiliary equipment) which is decoupled from the RF by a common mode choke so that the reaction of the EUT can be measured eg. by a DMM.

DUT/EUT and AE have to be placed with 10 cm space on the common ground.
In my setup I use 5 cm card boxes instead to save overall space.
Conection between CDN and EUT is kept short (30 cm).
EUT may be a AD587LW, a LM399 or a LTZ1000(A) reference cirquit.

The AE is either a K2000 or a 34401A which measures the output of the DUT (e.g. voltage reference).
Fortunately I use batteries on my voltage reference devices. Otherwise I would need 2 CDN devices: one for the power supply port and one for the voltage output.

The RF is supplied from a Feeltech FY6800 DDS signal generator.
With sine waveform we can reach up to 60 MHz frequency range.
Typically the sensitivity of analog devices is relatively small band.
E.g. several resonances between (parasitic) capacitors and wiring inductance.

So we need a logarithmic sweep with a small step width of e.g. 1%.
Some devices are more sensitive when a AM modulated signal is used.
So typical testing is done with 1000 Hz 80% AM modulation.
But I can also think that the 1000Hz could be filtered away by the 10 NPLC of the DMM which is used to verfy the voltage reference.
So it might be that a lower frequency like 7 Hz AM modulation harms more in this case.

[Andreas]

FY6800:
======

The Feeltech FY6800 is a entry level DDS signal generator.
For the price we have some limitations.

The frequency range is up to 60 MHz with 1uHz resolution for sine wave generation.
Amplitude is up to 20Vss (+/-10V) without load up to 20 MHz.
Above 20 MHz the amplitude is limited to 5Vss (+/-2.5V).
The generator is controllable by USB (Virtual Com Port) and also has the possibility to modulate with AM.

Unfortunately there are some bugs in AM-Implementation:

https://www.eevblog.com/forum/testgear/fy6800-dds-signal-generator-questions/msg2502990/#msg2502990

- dialing 80% AM results in 66% modulation effectively and also the amplitude is reduced ("down modulation") instead of maintaining the average 100% level
- when changeing the carrier frequency the AM modulation is switched off and has to be reactivated every time the frequency changes.

I decided to ignore the down modulation bug for my measurements (this is also against the standard).

In the mean time there is a successor FY6900 available (with increased +/-12V amplitude up to 20 MHz).

[Andreas]

CDN-AF2:
=======

commercial available CDNs can be got from several EMI sources eg. here:
http://www.schwarzbeck.de/Datenblatt/CDN_AF2.pdf

When looking at the price I decided to build my own CDN-AF2.
According to standard the common mode choke consists of a low frequency part  (>280uH @ 150kHz) and a high frequency part (2-4 toroids).
My local catalog dealer offers following NiZn ferrites which are close:

For the low frequency part I choose a 25 mm core with 13.5 windings (13 outside/14 inside core) which gives ~350uH measured inductance @150kHz.
https://www.reichelt.de/ferrite-core-material-3f3-ferr-tn25-15-10x-p246055.html?

For the high frequency part I choose 4 pcs with 0.5 windings. (only feeding through the core).
https://www.reichelt.de/ferrite-core-material-4a11-ferr-tn13-7-5-5-p246049.html?

Housing is made of the following:
https://www.reichelt.de/metal-shielding-housing-162x68x28-mm-teko-394-p21199.html?

All resistors are 0.6W 1% metal film.
All capacitors are WIMA MKS-2 (polyester foil with 5 mm pin distance)
As already mentioned I include a 3 dB divider into the CDN-AF2.

Attached the "cirquit diagram" including discharging resistors (10 Meg to keep leakage currents low)  for the AE port capacitors.
And some pictures from my build.

16.09.2019 Edit: unfortunately I made a topology fault in the build:

The low frequency common mode choke belongs to the AE side
and the high frequency choke should be on the DUT side of the CDN.
So I built a 2nd device with correct topology. Starting from picture IMG2771w.jpg on.
(Here it is the "right" one which is correct).
So now I can compare if the topology really has an effect (at least up to 60 MHz).

[Andreas]

Comparing some chokes:
================

Real inductors are rarely ideal devices.
The parasitic capacitances lead to resonant frequencies.
Above the resonant frequency the inductor behaves like a capacitor.
So practically the chokes are only capable to filter a relative small frequency band.

Here I measure the inductors with my oscilloscope with integrated frequency generator in a 50 Ohms system.
On the right side the generator output is fed to channel B.
The inductor is connected between channel B and channel A.
Channel A is terminated with a 50 Ohms terminator.

The bode plot/dampening curve is done by a 3rd party software for the scope.
https://bitbucket.org/hexamer/fra4picoscope/wiki/Home
Thanks Aaron.
Since the function generator in the scope is limited to 20 MHz I have to live with that until I write my own program for the FY6800.

First inductor is a Murata BLM31PG601 1206 ferrite bead.
Which is a population option in my AD587LW 10V reference design.
These are specified for 100 MHz.
This explains also that the inductor starts (-3dB) at 850 kHz.
The maximum in the measurement is -17 dB at 20 MHz in a 50 Ohms system.
Usually we have higher impedances up to 377 Ohms (the impedance of "free air") so the real dampening is to be expected even lower at these frequencies. When looking at the result the ferrite beads seem to be better suited for higher frequencies.

In my AD587LW cirquit I filter the GND and 10V output are filtered as a PI-Filter (with capacitors between the lines).
So I also have to regard 2 inductors in parallel. Here the filtering starts at 2 MHz and reaches only -12 dB at 20 MHz.

Single BLM31PG601 measurement:




Dual (parallel) BLM31PG601 measurement:


The 2nd option for population in my AD587LW design is a Würth 51uH common mode choke. (SLM 744242510)
Here we start (-3dB) at 30 kHz and at 20 MHz we have -34 dB dampening. And that nearly independant of the number of signal lines which are filtered (1 or 2).
When looking at the results it would have been better to use both filtering options in series for the AD587LW design.
The Würth common mode choke for the lower frequencies and the BLM for the higher frequencies.

Single line SLM 744242510 measurement.




Dual line (parallel) SLM 744242510 measurement.


Jason uses EMI ferrite cores with 3.5 common mode windings in his LTZ1000.
I tried to get the same cores. But of cause there is no warranty that those are really identical.
Again like on all common mode chokes there is nearly no dependancy on single or double line filtering.
Filtering starts (-3dB) at 50 kHz and reaches -24 dB at 20 MHz in a 50 Ohms system.

Dual line EMI core 3.5 windings.




Single line EMI-core 3.5 windings.


Of course I also measured the NiZn ferrites of the CDN-AF2 device.
Here we need a significant impedance already at 150 kHz. (>= 280 uH).
This is done with the low frequency inductor (13.5 windings in my case).

The inductor starts (-3dB) already at 20 kHz.
At 150 kHz I have calculated 350 uH for the 13.5 windings.
At 2 MHz the maximum dampening of -42 dB is reached (resonant frequency).
At 20 MHz the dampening is reduced already to -28 dB.
This explains why the CDN is built with high and low frequency inductor.



The high frequency inductor of the CDN is built with 4 NiZn ferrites in a row.
(the signal lines are only fed through = 0.5 windings).
Here the dampening starts (-3dB) at 3 MHz and reaches -8 dB at 20 MHz.



And I also tested a 6-hole ferrite bead with 2.5 windings. A Würth 7427503.

https://uk.rs-online.com/web/p/ferrite-beads/2606830/

Filtering starts (-3dB) at 900 kHz and reaches -20.5 dB at 20 MHz.
So just a little bit better than the BLM31 SMD ferrite.



Update: 20.09.2019

In the very beginning when not having the standard available I also thought of using off the shelf single inductors.
But these are relative small band:

Fastron XHBCC 330uH
https://www.reichelt.de/fixed-inductor-axial-xhbcc-ferrite-330-h-l-xhbcc-330-p138551.html?

Starting (-3dB) short above 20 kHz. Peak -65 dB at 2 MHz. And -19 dB at 20 MHz



Fastron HBCC 47uH
https://www.reichelt.de/fixed-inductor-axial-hbcc-ferrite-47-l-hbcc-47-p86464.html?

Starting (-3dB) around 150kHz. Peak -59 dB at 9 MHz. And -30 dB at 20 MHz.



Fastron 09HCP 470uH
https://www.reichelt.de/vertical-inductor-09hcp-ferrite-470-h-l-09hcp-470-p138662.html?

Starting (-3dB) below 20 kHz. Peak -72 dB above 2 MHz. And -22 dB at 20 MHz.



Fastron 07HCP 10uH
https://www.reichelt.de/vertical-inductor-07hcp-ferrite-10-l-07hcp-10-p86398.html?

Starting (-3dB) at 900kHz. Peak -54 dB just below 20 MHz. -47 dB at 20 MHz.



Fastron 11PHC 220uH
https://www.reichelt.de/vertical-inductor-11phc-ferrite-220-h-l-11phc-220-p138677.html?

Starting (-3dB) at 40 kHz. Peak -68 dB just below 3 MHz. -20 dB at 20 MHz.



Fastron 11P 330uH
https://www.reichelt.de/vertical-inductor-11p-ferrite-330-l-11p-330-p72996.html?

Starting (-3dB) at 25 kHz. Peak -70 dB just above 2 MHz. -20 dB at 20 MHz.



Fastron 11P 47mH
https://www.reichelt.de/vertical-inductor-11p-ferrite-47-m-l-11p-47m-p73008.html?r=1

This one has already a high resistance at DC. So it is not really suited for precision measurements.
Starting already at DC. Peak -95 dB at 200 kHz. -24 dB at 20 MHz.



Update 22.09.2019:

Fastron SMCC 47uH
https://www.reichelt.de/choke-coil-fixed-inductor-axial-47-smcc-47-p18207.html?

I use this one in the charger for my LTZ1000 references. Intention was to make the output voltage immune against the switcher noise from the external 24 V DC (swtichmode) adapter. But with low success. The reason is that this choke does not cover the  ~100 kHz range of the usual switchers.

Starting (-3dB) around 200 kHz. Peak -56 dB at 7 MHz. And -23 dB at 20 MHz.



Update 01.02.2020:

Some EPCOS coils which seem to be very promising:
First a common mode signal line choke (CAN choke) with 51uH
https://www.reichelt.de/smd-power-induktivitaet-51-h-epco-b82790-s513-p245682.html?&nbc=1

Starting (-3db) around 35 kHz already going up to -34 db at 20 MHz without resonance.



further some axial coils with self resonance in the 100 MHz range.
first with 40uH
https://www.reichelt.de/inductor-axial-ferrite-40-h-epco-b82111-e-c5-p245652.html?&nbc=1
starting (-3dB) at 200 kHz going up to -40dB @ 20 MHz



2nd  with 22uH
https://www.reichelt.de/inductor-axial-ferrite-22-h-epco-b82111-e-c4-p245651.html?&nbc=1
starting (-3dB) at 400 kHz going up to -34 dB @ 20 MHz

[Andreas]

Automation of test sequence:
=====================

Automation is done by 3 programs:

The first program (FY6800.EXE) controls the FY6800 frequency generator and is responsible
to have synchronized readings between frequency output power and the DMMs.
Frequency sweep is done logarithmic with e.g. 256 frequency steps per decade.
This corresponds to around 1% relative frequency change per step.
After some tests I have decided to start with 10kHz and go to 60 MHz.
Note that the Inductor in the coupling network (CDN) works only above ~100 kHz.
So the 10 kHz signal is dampened significantly. But this way I want to see if there is
some sensitivity to signals below 100 kHz.

After setting the new frequency it needs around 420 ms (CW) - 700 ms (AM)
until the generator output is stable. So I introduced a "ti hold" which is
set typically to 2 seconds (= 3 seconds total cycle time) to allow settling
of the frequency generator and the DMMs.
So each frequency sweep from 10kHz to 60MHz needs ~ 50 minutes.
The latest measured value of the DMMs is recorded together with frequency
and output voltage of the FY6800 into a log-file before the next frequency step is applied.

After each sweep with the same output power/amplitude the next power increment "P Incr" is applied.
I am starting e.g. with 12 dbm. This value is calculated as corresponding power after
the -3dB dampening device which is within my CDN.
Reference level is: 0 dbm = 1 milliwatt into 50 Ohms = 0.2236 V RMS.
So e.g. 15 dbm correspond to 1.26 V RMS after the -3dB dampening device or ~5Vpp as set value
for the FY6800 generator as unloaded (high load impedance) peak to peak output voltage.
The levels might sound somewhat "high" but those are "common mode" coupled on both pins of the DUT.
And the reference or DMM is usually floating against mains earth ground. So the actual influence
on the DUT is rather low as they are usually designed to withstand up to 1000V
common mode voltage at low frequencies.
With 4 power levels the total sweep time is ~ 4 hours.
So that a measurement over night can do 2 total sweeps.

I can also choose AM modulation instead of CW.
I want to use 1000 Hz or 7 Hz with a modulation depth of 80%.
As already mentioned above the output power on AM is too low according to definition
and also the modulation depth is only 66% when setting 80%.

The 2 other programs (K2000.EXE and HP34401A.EXE) can read the results with a 6.5 digit DMM on
the AE port of the CDN from the reference which is connected on the DUT port of the CDN.
To keep the measurement time short but the "noise" of the readings not too high I choose 10 NPLCs.
This leads to 400 ms reading duration (with auto-zero) on the HP34401A and 600 ms on the K2000.
I am using a virtual file to transfer the readings between the programs. So each program
can be used independantly and all results can be logged by the "master" program.

But since even the 30 cm short connections on the CDN can act as antenna the readings
have to be taken with care. You never can shurely know wether the DUT is actually
influenced, or the DMM on the AE port. So the sensitivity of the DMMs has to be tested too.

[Andreas]

K2000 results:
==========

edit 17.10.2019:

The first picture confirms also for the K2000 the "law" that +3dB more EMI-Voltage gives around doubling the drift

3.5Vss: -  7.5 ppm
5.0Vss: - 14.1 ppm
7.1Vss: - 27.8 ppm
10Vss:  - 55.0 ppm

The measurement of 13.10.2019 shows the K2000 on the DUT side and the AD587LW#04 on the AE side both witout any additional EMI-core.

14.10.2019: a additional EMI-core on the AD587LW side shows no improvement -> the K2000 is the device which drifts.

16.10.2019: shifting the EMI-core to the K2000 side gives a little imrovement on the dip.

15.10.2019: again a additional EMI-core on the AD587LW gives no change to the previous measurement.

And obviously the high impedance on the K2000 input is the reason that the improvement of the EMI-core is very moderate on the K2000 side.

the EMI sensitivity of my K2000 is around a factor 3 higher than my HP34401A.
So for future measurements I will use the 34401A.

[Andreas]

HP34401A results:
=============

edit 15.10.2019:

The first picture confirms also for the HP34401A the "law" that +3dB more EMI-Voltage gives around doubling the drift (except for the last step from 7 to 10 V)

3.5Vss: +  3.9 ppm
5.0Vss: +  6.9 ppm
7.1Vss: + 11.7 ppm
10Vss:  + 18.6 ppm

The measurement of 09.10.2019 shows the HP34401A on the DUT side and the AD587LW#04 on the AE side both witout any additional EMI-core.

10.10.2019: a additional EMI-core on the AD587LW side shows no improvement -> the 34401A is the device which drifts.

11.10.2019: shifting the EMI-core to the 34401A side gives a little imrovement on the peak. But the dip gets larger. So obviously there are 2 independant receivers with opposite direction which are dampened differently by the EMI-core.

12.10.2019: again a additional EMI-core on the AD587LW gives no change to the previous measurement.

And obviously the high impedance on the 34401A input is the reason that the improvement of the EMI-core is very moderate on the 34401A side.

[Andreas]

AD587LW results:
=============

Edit: 29.09.2019

I have measured 2 devices of my AD587LW samples:

AD587LW#03 is equipped with Würth SLM 744242510 51uH common mode chokes.
On the measurement of 21.09.2019 we see a dip of -15 ppm @ 5.7 MHz when setting the (unloaded) amplitude of the FY6800 to 10Vss in CW mode (no modulation).
There seems to be a smaller dip at ~2 MHz.

AD587LW#04 is equipped with Murata BLM31PG601 600 Ohms ferrite beads.
The measurement from 26.09.2019 shows -4.7 ppm @ 2.5 MHz at 10Vss.
The surprising result for me is that the ferrite beads deliver a much better overall performance. Even at frequencies (5.8 MHz) where the common mode choke had a much larger dampening than the ferrite beads. Perhaps the self resonant frequency of the 100nF WIMA capacitors which should be somewhere near 5-10 MHz is a possible explanation.

Another result is that doubling the amplitude on the signal generator gives excess change in error voltage.
For the AD587 measurements each factor 1.4 (3dB) signal level changes nearly a factor of 2 in the error.

When looking at the AM modulation (1kHz or 7Hz) then the error voltage is generally lower than with CW mode.
This can be explained by the erroneous implementation of the AM modulation in the FY6800.
The down-modulated signal has less energy than a true AM modulated signal.
The corresponding pictures are marked as AM (for 1 kHz) or AM7 (for 7Hz).
The 7Hz modulated AM modulated files look more "agressive or noisy" as expected (no averaging over integration time).
But the peak error voltage is nearly independant of the modulation frequency.

The most disturbing result was that a measurement nearby (LTZ6 on K2000 with 0.3 m distance between the wiring on the DUT-side and the cable to the K2000) had more error over frequency than the AD587 measured.
So it looks like my LTZ-references are much more sensitive to EMI than my AD587LW references.

But this result also shows: all measurements have to be taken with a grain of salt.
You never know how the coupling of the RF-energy is done: Either over the CDN (where it should) or "over the air" which is not wanted.
Over the time we should get a feeling which instrument/reference is sensitive to which frequencies.
So many of the effects will be explainable if we have enough measurements.

Edit 12.10.2019:
As overview I add the measurements with 10Vpp in ppm for all 4 AD587LW 10 V references.
AD587LW#01 and AD587LW#03 both equipped with Würth SLM 744242510 51uH common mode chokes show dips of -14 and -17 ppm near 6 MHz.

AD587LW#02 and AD587LW#04 equipped with Murata BLM31PG601 600 Ohms ferrite beads have a much larger stray in values of -4.4/+7.6 ppm and -5.8/+1.3 ppm. Note that the peaks above 20 MHz are clipped by reduced output voltage of 5Vss instead of 10Vss.

One result is that the stray from device to device can be rather large even with same population.

[Andreas]

LM399 results:
==========

Hello,
the LM399 results were still missing here the first part of it:

19.04.2020 first measurement of a "naked" LM399 with only a 6K8 resistor to a stabilized (LT1763) 14V supply from 12 NiMH cells.
So there are only bypass capacitors at the voltage regulator with about 40 cm wiring to the LM399.
result with very low level of 0.63 Vss at the FY6800 Generator:
-1150 ppm @ 22.7 MHz and
-554 ppm @ 117 kHz

21.04.2020 same with 2 100nF SMD 0805 bypass capacitors near heater and zener pins:
result with 0.63 Vss level:
-503 ppm @ 52 kHz

so the 22 MHz peak has been filtered by the bypass capacitors but the low frequency peak is still there.
since I feared that the LT1763 voltage regulator also could contribute to the peak I tested with 10 NiMH cells without stabilisation

22.04.2020 same without stabilized supply:
result with 0.63 Vss level:
-548 ppm @ 52 kHz
so no significant change.

Mhm: the optimum capacitor for 52 kHz would be around 400 uF lets see what I have in the drawer ...

Update 28.04.2020
============

I tried two 47uF capacitors one at the heater and one at the zener.

measurement of 24.04.2020 with 0.63Vss showed: nothing!!!!
I first checked if the function generator still works with the oscilloscope: all working fine.

so I increased the level up to 5.02Vss on 25.04.2020
now I see two peaks:
-2.5 ppm @ 77kHz
-2 ppm @ 60 MHz

the 51uH common mode choke (CAN-choke) removes the 60 MHz peak on 26.04.2020
but the -2.5 ppm @ 79 kHz remain
I fear for the low frequency part a buffer OP-Amp like AD4522 may be better to isolate the LM399

on the other side: -2.5 ppm @5Vss is a very good result for the little effort of some bypass capacitors.

Update 16.05.2020
============

Measurement of 01.05.2020 is a repeated measurement of 25.04.2020 with the 47uF on the heater side removed (so only 100nF on heater and 100nF + 47uF on the zener side) this gives slightly worse results on the 60 MHz peak.
So the main effect for EMI hardening is from the 47uF + 100nF on the zener side.

Measurement of 02.05.2020 with a ADA4522 added as output buffer. (47uF still in charge)
Of course I have a 100nF across the output and decoupling against capactive loads with 22R 10nF and 4K7 on the output of the ADA.
The low frequency peak is gone but now the ADA4522 itself is affected by the EMI giving a positive peak at 60 MHz.

Measurement of 04.05.2020 with a additional 100nF directly at the positive input of the ADA4522 buffer.
Practically no change against previous measurement so the ADA EMI is from the already filtered output.

Measurement of 06.05.2020 with 47uF removed from the zener output.
Practically no change against previous measurement so the ADA is isolating the LM399 sufficiently for low frequencies.

Measurement of 07.05.2020 with additional 51uH common mode choke (CAN-choke) on the output.
Practically zero EMI influence.

Conclusion: The LM399 either needs a 47uF capacitor or a ADA4522 buffer on the output. (the 100nF is someting that I will always add for stability). A additional common mode choke (or perhaps some ferrites) improves the behaviour above 20 MHz.

with best regards

Andreas

[Andreas]

LTZ1000 results:
============

Edit: 29.09.2019

First measurement on LTZ#8 (buffered output).

being warned by the "nearby" measurement I thought that I have to reduce the amplitude from 10Vss to 0.89Vss as lowest level.
I wanted to have a maximum ~10 ppm change of output voltage by the RF voltage.
But the first "dry run" showed already a much larger (-42 ppm) change at 60 MHz.
Fortunately the output voltage resumed immediately to the nominal output voltage after changeing to 10 kHz.
(So no permanent damage).

But I had to do further reduction of the amplitude steps starting from 0.32Vss to 0.89Vss since I did not want to kill the buffered output of the LTZ.

See measurement of 28.09.2019:

Again it looks like every 3dB step of the FY6800 output voltage doubles the error voltage.

For 60 MHz the ferrite cores of cellularmitosis or the ferrite beads of Dr Pyta should give a good supression.
It looks like I will have to do a rework for my LTZ1000 PCBs.

By the way: adjusting a voltage reference (JJA) by RF (60 MHz) is now possible for the LTZ1000 too.
(I should make a patent for this). 

Update 02.10.2019:

Time to look how we can get a improvement.
So I tried the core recommended by cellularmitosis with 3.5 windings as adapter outside the LTZ#8.
(see photo img2777w.jpg).

with the same maximum output voltage (0.89V) there was merely a 1 ppm dip visible.
The dip is now near 25 MHz on the measurement of 29.09.2019

Increasing the power carefully to 2.5Vss and finally 5.0Vss on 01.10.2019 now shows the 25 MHz dip at -32 ppm and a -21 ppm value at 60 MHz.
So compared to the measurement without EMI-core (-21 ppm @ 0.63Vss @ 60 MHz) we have a improvement of factor 8 at 60 MHz.

[Andreas]

Reserve 1
=======

[Andreas]

Reserve 2:
========

[Andreas]

Reserve 3:
========

[3roomlab]

for some reference, this interesting 28 page article with noise numbers in E n H field
(some lights can be really noisy!)

https://biblio.ugent.be/publication/8519431/file/8519433.pdf

[SilverSolder]

for some reference, this interesting 28 page article with noise numbers in E n H field
(some lights can be really noisy!)

https://biblio.ugent.be/publication/8519431/file/8519433.pdf

Looks like it is safe to use a Hair Removal Device in the lab? 

[Andreas]

Hello,

don´t know if some one already has recognized this:
the schematics and the build of the CDN differ in one detail.
The schematics is correct but the device is not.

Since I do not know if this has a influence on the measurement results I will have to correct this and repeat the measurements up to now.

with best regards

Andreas

[MiDi]

the schematics and the build of the CDN differ in one detail.
The schematics is correct but the device is not.

You mean that the cores are placed in reverse order?

Thanks for this very interesting topic and investigation 

[Andreas]

Hello,

yes you are right the high frequency cores should be on the DUT side,
and the low frequency core at the AE side.
I have updated the previous post with a corrected build:

https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684082/#msg2684082

and also updated the choke measurements with a 6-hole ferrite bead.

https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684085/#msg2684085

with best regards

Andreas

[Andreas]

Hello,

just a note that I have updated further choke  measurements and the description of the test-automation

https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684085/#msg2684085

https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684088/#msg2684088

with best regards

Andreas

[Andreas]

As already mentioned in the AD587LW measurement I got a very disturbing result:
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684097/#msg2684097

A simultaneous measured LTZ6 on K2000 in a 0.3 m distance had much more error voltage than the AD587LW which was tested.
Attached 2 measurements in CW-mode.
You can see that the setup delivers 3 critical (resonance) frequencies: 12.5 MHz, 15 MHz and 43.7 MHz.
An note: above 20 MHz the output voltage of the FY6800 is limited to 5Vss so the 43.7 MHz result would be a factor 4 larger if the amplitude was 10Vss.

Of course the coupling "over the air" is much less reproducable than the coupling via CDN.
But it also shows that we have a very high sensitivity on the LTZ1000.
This is also verified by the first LTZ1000 measurement.

https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684103/#msg2684103

[3roomlab]

the xtal in the K2xxx is 12Mhz
the wire become HF antenna?
dont know the xtal in 34401a ?

do you have a chance to try tiny bead like the picture?

[Andreas]

Hello,

Ferrites are a good idea (especially for higher frequencies) to determine which device is the guilty.

I prefer to attach the EMI-cores outside the device. (so it is reversible).
see updated photo of LTZ1000 here:
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684103/#msg2684103

with best regards

Andreas

[Andreas]

Hello,

this time I tested a 2nd LTZ (LTZ#7) with different population.

LTZ#7 is populated with LT1013  (and 2 additional EMI capacitors) and 7 mm shortened legs of LTZ1000.
LTZ#8 is populated with 2*LTC2057 and long legs of LTZ1000.

LTZ#7 measured factor 1.6 more sensitive to EMI than LTZ#8.

Now the question arises for the reason.
- stray in sensitivity of LTZ1000 with same date code?
- shortened legs more sensitive than long legs?
- LT1013 more sensitive as LTC2057?
- the additional capacitors?

I need more samples.

with best regards

Andreas

Edit:
What also can be seen: the 2 runs of LTZ#7 did not meet the same point at 60 MHz. The temperature changed by about 2 degrees during the measurement. So possibly the EMI behaviour is temperature dependant.

[Kleinstein]

The large EMI effect seems to be really high frequency effect. The early data from the LTZ6 suggest that the that there are resonances and both #7 and #8 seem to have a resonance (just) higher than the frequency range tested. So the difference between the 2 may be just a minimal different resonance frequency.
The interesting part may be a slightly higher frequency (FM radio).

[Andreas]

Hello,

I did some "nearby" measurements in different configurations.

20191006_LTZ06_EMI_CW_nearby30_K2000.PNG shows the configuration without any external filtering.
When looking at the resonance frequencies we have 12 and 15 MHz + ~44 MHz.
The 44 MHz correspond to a wavelength of 6.8m. With a shortening factor of 0.95 and a quarter wavelength antenna we have to look for a 1.6 m long wire.
The connection between LTZ#6 and K2000 consists of 2 parts: one 0.6 m the other 0.7 m so with some wire length within the K2000 and the LTZ#6 we can assume a 1.6 m total resonant length.
The 12 and 15 MHz resonances could result from the 100nF capacitors which are on this line (they have typical self resonant frequencies of 10-20 MHz). One 100nF WIMA MKS02 foil capacitor within the LTZ (across the output) and one 100nF ceramic which is in the middle of the 2part connection.

Measurement 20191004_LTZ06_EMI_CW_nearby30_K2000filt.PNG shows what happens if we use a EMI core with 3.5 windings near the K2000. The 12 and 15 MHz peaks are dampened and the 44 MHz peak is shifted towards 48 MHz. (so the effective antenna length has been shorted somewhat).

20191007_LTZ06filt_EMI_CW_nearby30_K2000.PNG shows the EMI-core on the LTZ#6 side. There is only some temperature drift during measurement visible. (the two 10Vss measurements are 3 hours apart).

20191007_LTZ06filt_EMI_CW_nearby30_K2000filt.PNG is similar with EMI-cores on both sides over night so less temperature drift on the K2000.

The large EMI effect seems to be really high frequency effect. The early data from the LTZ6 suggest that the that there are resonances and both #7 and #8 seem to have a resonance (just) higher than the frequency range tested. So the difference between the 2 may be just a minimal different resonance frequency.
The interesting part may be a slightly higher frequency (FM radio).

I am still hoping that someone with better equipment (frequency generator up to 80 or even 230 MHz) tries to repeat some of my tests at least for the LTZ1000. Interesting would also be a comparison to a LTFLU based device.
Sorry but I for my part do not want to spend several 100$ for a frequency range beyond 60 MHz.
And I am shure that the measures that are done for 60 MHz also do some improvement above 60 MHz.

with best regards

Andreas

[Andreas]

Hello,

I updated the AD587LW measurements with AD587LW#01 and AD587LW#02 result.
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684097/#msg2684097

It is obvious that the stray of the EMI sensitivity can be rather different from device to device even with same population.

with best regards

Andreas

[Andreas]

Hello,

When comparing the EMI sensitivity of my K2000 with my HP34401A then the K2000 shows factor 3 more sensitivity against EMI. -> it is wise to use the HP34401A for further measurements.

https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684091/#msg2684091
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684094/#msg2684094

Interestingly is that both instruments show a increasing sensitivity which is above my 60 MHz measurement limit which can be dampened by additional EMI cores.

with best regards

Andreas

[notfaded1]

This makes me ever more glad I've got a 34401A... 3x less EMI is huge really!  Thanks for the info.  This makes me wonder about other MM now.

Bill

[Andreas]

im curious, are there possible improvements to stacking more filters for the K2000?
sorry wrong question, maybe the right question is, have you tried to stack filters till the K2000 is at the same level as 34401a? and if so what could these filter values be (alot of K2000 owners could be interested  )

no I did not (up to now). When looking on my coil measurements eventually two 47uH chokes could bring a little improvement (7 MHz self resonant frequency filling the gap). But for high ohmic inputs we would need some low ohmic path to the housing to prevent entering the EMI into the device. So I think two Y-capacitors against the housing will give a large improvement. But of course that will reduce the isolation of the floating part of the DMM.

This makes me ever more glad I've got a 34401A... 3x less EMI is huge really!  Thanks for the info.  This makes me wonder about other MM now.

I would not generalize the result. And I would never bet that each device will deliver the same results.
Finally I am only measuring a small window up to 60 MHz. Above 60 MHz I have a blind spot.

On my HP34401A I have added two bypass capacitors on the LM399 reference (to reduce noise).
This could also have some influence on the EMI behaviour.

On the Keithley the soldering side of the PCB is not so easy accessible.
So I did not do this improvement on the K2000.

So for me this only means that I will use the 34401A for trying to improve some other voltage references like AD587, LM399 and LTZ1000.

with best regards

Andreas

[Andreas]

when the 2x 10u are fuzzed, doesnt seem to do much in LTspice but im sure it does EMI things

EMI has much to do with parasitics. (that is shurely one reason why a resistor chain is used).
So how does it look like when you have 2-5 pF input capacitance of the hybrid (for vias, pins,  eventually input protection diodes, etc).

Do your resistor models of the 13 k resistors contain each a ~0.1 pF capacitor across the resistor itself and 0.1pf to some 0.1 pF for the capacitance of the pads on the PCB?

with best regards

Andreas

[Andreas]

but again, on tins picture of 34401a, i could not spot the 6 13k resistors 

According to the Service manual they are 2512 1W types
(CRCW2512 which are specced  pulse proof 1.5W and 500V).
Located left from the front input connector.

I think that your stray capacitances/inductances are a bit on the high side.
Usually I calculate ~1nH / mm so ~5-8nH between each serial resistor.

But of course there is a long ~300 mm input lab chord and the input connector also gives  ~50 mm until the PCB on the 34401A

with best regards

Andreas

[iMo]

I would recommend you to look at the schematics of ie. hamradio transceivers (ICOM, YEASU, KENWOOD) which are easy to find on the web ("service manuals"). You may see there how they do decoupling with 15kHz-60MHz/100W 140-470MHZ/35W inside their boxes.

PS: I can remember an Pi filter ie. Icom used to use at almost any wire inside (except the main signal path and high freq clocks, of course) is 4n7-100uH-4n7.

PPS: there are also other two Pi filter combinations: 47n-100u-47n, 4n7-10u-4n7 (ie in IC-706), most probably adjusted based on required i/o impedance matching..

[iMo]

On my HP34401A I have added two bypass capacitors on the LM399 reference (to reduce noise).
This could also have some influence on the EMI behaviour.

2x100nF ceramics? Did it help somehow?

[Andreas]

Hello,

I cannot tell it for the 34401A since I did no before and after measurements.

But my first references (LM399 based) suffered a lot from EMI by USB-cables nearby.
Here I got a reduction from several uV to below 1 uV change.

with best regards

Andreas

edit:

found a old measurement from 2008:
before it was up to +30/-75uV  (see picture) 
hand sensitivity of difference voltage (169mV) between 2 LM399 references.
(Depending on where you touch the cables or battery packs).

after adding 4.7nF 1206 ceramic capacitors ~ +2/-1uV. (not shure wether it was only noise or real sensitivity).
after changeing to 100nF 1206 Z5U capacitors -> below 1uV (noise limit).

[Andreas]

Hello,

due to the large stray between my samples I decided to build a "EMI-victim" to test what measures give best results against EMI.
So LTZ#9 was born.

In the first step I only populated the components for the "positive reference" according to the LTZ1000 datasheet.
Population is with LTZ1000 (non-A) and therefore the voltage divider is 12K + 1K for the temperature setting as deviation to the data-sheet. R9 (400K) is not populated since I do the T.C. adjustment only if the device survives the "treatment".

There is also no output buffer populated.

with best regards

Andreas

[Andreas]

Hello,

first measurement on LTZ#9 showed a strange behaviour when connecting the CDN.
I first thought there is nearly no capacitive loading because on the DUT side the 10nF capacitors are in series with the 200 Ohms resistors. But I had forgotten the AE side. There we have 2*47nF against RF-Ground so 24nF between the both pins of the unbuffered output. The output voltage reduced from 7177 mV to 7164 mV as soon as the CDN was connected.
See also EMI-Measurement of 17.10.2019.

The scope measurement shows that we have a oscillation with 200 kHz and 100mVpp amplitude.
I never had thought that a slow OP-Amp like the LT1013 can oscillate that fast.
Of course I had to use a floating Oscilloscope for the measurement because the LTZ voltage is centered to the earth ground by the CDN capacitors / resistors. (18.10.2019).

I decided to use my standard buffer cirquit for further EMI measurements with a LTC2057 and capacitive isolating by a 22 Ohms resistor. After removing the power supply from the lap-top connected to the USB scope there was no oscillation visible. (20.10.2019)

The buffered output showed also a very strange effect on the EMI-test. (21.10.2019)
There are very sharp spikes. Even when increasing the number of measurement points from 256 to 2048 per decade (hires measurement of 22.10.2019).

There seems to be some inteference between the chopper and the EMI frequencies.
Fortunately a 100nF capacitor across the output improves the behaviour a lot (25.10.2019).
I tried to reduce the higher frequency part a bit by chosing a smaller capacitor which has a higher self resonant frequency.
100 nf has around 10-20 MHz so 4.7nF should be good for 50-100 MHz. But the result of 27.10.2019 shows only very limited improvement.

So I will have to do further measures to improve the behaviour.

with best regards

Andreas

[Andreas]

Hello,

seeing the "interference" pattern on the response the next logical step is to use a EMI hardened output amplifier like OPA189 or ADA4522-1.
Having the OPA189 in my drawer I tested this one.

First test on 28.10.2019 was with 4.7nF across the output like the previous measurement of the LTC2057.
Output voltage change with same amplitude of 0.45Vss was 5.5 ppm on the LTC2057 and only ~0.7 ppm (with some noise up to 1.2ppm) on the OPA189.

Increasing the amplitude up to 1.78Vss on 29.10.2019 showed a similar frequency behaviour as on the very first unbuffered measurement of 17.10.2019.

Removing the 4.7nF capacitor on 31.10.2019 increases the maximum deviation slightly from 15 ppm to 22 ppm. But also shows some few interference patterns possibly from the chopper frequency.

A 100nF capacitor (01.11.2019) shows similar behaviour (15 ppm) like the 4.7 nF.

So the OPA189 gives a improvement (better isolation of the LTZ1000 against environment) of ~factor 3 to 4 in EMI voltage amplitude than the LTC2057. And is self nearly immune against EMI together with some output capacitor.

I fear I have a new favorite OP-Amp for precision buffers.

with best regards

Andreas

[branadic]

Thanks Andreas, are you planing similar measurements on ADA4522-1 too?

-branadic-

[Andreas]

Hello Branadic,

just digged a bit deeper in my drawer ...
And actually also a few ADA4522-1 showed up.

with best regards

Andreas

[MiDi]

In addition the OPA(2)189 has the nice feature that the inputs do not have back-to-back clamping diodes, which allows to drive the inputs independently between the rails without clamping.

Measurements on ADA4522 appreciated 

[branadic]

Measurements on ADA4522 appreciated   

#Metoo

-branadic-

[Andreas]

Hello,

I also measured a different effect:
During the heater noise measurements (around 4mVpp) I recognized that when someone switched the fluorescent lamp in the bathroom (one floor below my lab) I had some funny spikes on the noise measurement.
And that with having the LTZ#9 within my metal cookies box.

With the fluorescent lamp in my lab I created also several negative jumps up to 1V peak.
Unfortunately these jumps are very different each time. Ranging from "within the heater noise" (below 4uV) then often around 200mVp up to 1Vpeak. So if I want to evaluate this I have to do a "statistical" approach.

I switched 2.5 minutes with a period time of 6 seconds (so 3 seconds between on and off giving 25 switching periods) to catch 20 events above a trigger limit of -20mV.
Average value of the 20 events was -287mV with a standard deviation of 274 mV.

Of course I expect that if the heater power changes the output voltage of the LTZ is also affected.
Especially on the A-type of the reference where the thermal time constant is much lower than on the non-A version.

2 extreme samples with measured heater voltage shown in attached screen shots on LTZ#9 (non-A version).

with best regards

Andreas

[Andreas]

Hello,

yes when I extremely zoom in I see some bursts (in kHz distance) with oscillations in the some MHz range (14 MHz in example).
But I have to admit that I use a 20 MHz bandwidth limiter for these measurements. (so cant see GHz).
The norm pulse for those bursts is ~5ns rise time corresponding to 60 MHz with 5 kHz repetition rate.

with best regards

Andreas

[MiDi]

In the LTZ1000 thread I reported some time back about latch up of 7to10V buffer.
In the mean time I was able to knock down the cause after heavy research:
The switching of old fluorescent tube in our loundry room ~10m away from lab.
This happend only when measuring 10V output and with offline linear PSU for Ref.
On buffered 7V or on batteries this never happened even on forcing it with many cyclings of lamp.
Have to do more investigation...

[3roomlab]

here are 2 fascinating papers

according to this paper
https://www.researchgate.net/publication/266201167_Characterization_of_compact_fluorescent_lamp_RF_emissions_in_the_perspective_of_human_exposure

they have some weird range of EM field going up to 380V/m ? maybe this is tube model specific?
they did not say how they got this peak value. sensor touching lamp? or 10cm?
there is also a plot about the noise surge when turning on

and this paper
https://www.researchgate.net/publication/225303019_Study_of_a_fluorescent_tube_as_plasma_antenna
in this paper they use micro tubes for campers, the resonant for that tube is around 5xxMhz
and there is some kind of resonant voltage gain? +16dB?

[Andreas]

Hello,

just to clarify.

In my case I am using a simple old style 36W fluorescent tube (TL T5-Type) with a inductivity (similar to transformer coil) as ballast. Only the starter has been replaced by a modern quick starter type.
So in my case most of the EMI is generated when switching off the inductivity.

with best regards

Andreas

[MiDi]

Same here, 36W T8 with KVG and electronic quick starter.

[Andreas]

Hello,

I know that you are all waiting for the ADA4522-1 measurements.
Especially as also TIN is using those for his FX-reference.
(The question is: intentionally or not?).

But: after starting the measurements and looking after the data-sheet I saw a PCN regarding the metal mask layer of the ADA4522-1. Since this may massively affect EMI-behaviour I fear that I have to repeat the measurements.

https://www.analog.com/media/en/PCN/ADI_PCN_18_0171_Rev_A_Form.pdf

According to the PCN the "new" devices will be from datecode 22-Aug-2019 on.
My parts from the drawer are significantly older (datecode 643 bought in January 2017).

Does anyone know what datecode DigiKey is currently selling for the ADA4522-1 in SO-8 package?

with best regards

Andreas

[Andreas]

Hello,

first results of ADA4522-1 with datecode 643.

The measurement of 06.11.2019 with no capacitor across the output shows in average a significant improvement against OPA189 (31.10.2019).
Around -13 ppm (ADA) against -22ppm (OPA) measured at 1.78Vpp generator setting.

But there are some sharp spikes.
A zoomed measurement (09.11.2019) reveals that the spikes are in multiples of ~4.8 MHz which is one of the internal chopper frequencies of the ADA4522.

Ok on the OPA189 a 4.7 or 100nF capacitor helped to remove those spikes and reduced EMI influence from -22 to -15 ppm.

But WTF! on the ADA4522 the output capacitor increases EMI to
-57 ppm (4.7nF on 11.11.2019) and
-60 ppm (100nF on 12.11.2019)
at 1.78Vpp.

My estimation is: the isolation through the OP-Amp gets completely lost on the ADA when connecting a output capacitor across X4+X5. So the LTZ itself is affected.

The schematics shows the output buffer (R23 is not populated).
I think we would have a better behaviour when we would split R22 (10K) into two resistors and place one before the non-inverting input.

with best regards

Andreas

[MiDi]

My estimation is: the isolation through the OP-Amp gets completely lost on the ADA when connecting a output capacitor across X4+X5. So the LTZ itself is affected.

The schematics shows the output buffer (R23 is not populated).
I think we would have a better behaviour when we would split R22 (10K) into two resistors and place one before the non-inverting input.

Thanks for the results for ADA-4522, very interesting 
Your estimation of loosing isolation would mean the input diodes of ADA start clamping?
Why do you think splitting R22 would improve the results?
If the LTZ1000 is loaded "down" this way, a split resistor would help for LTZ1000, but I do not see how the buffered output could be less affected?

[Andreas]

Hello,

my assumption is/was that the remaining pattern is that what remains from the LTZ.
But in the meantime I have found a Ground connection error in the layout.
The Buffer was added in a revision of the PCB.
So I thought it was a good idea to route the power supply for the buffer directly from the voltage regulator.
(instead first routing to the output buffer and from there routing it to the Star ground from the LTZ1000).

So when I add a capacitor across the outputs (output gnd) it has a very looong way to the ground pin of the output buffer (power gnd).
I fear I have to check what happens when placing the capacitor from the positive output to the GND pin of the output buffer.

The only remaining question is: why is the OPA189 not sensitive to this wiring where all other OP-Amps seem to have some problems.

with best regards

Andreas

[Andreas]

Hello,

first measurement of ADA4522-1 with the 4.7nF capacitor connected to Power Ground (PGnd) instead across the outputs.
-> much better result. -13 ppm @ 1.78Vpp.

with best regards

Andreas

[3roomlab]

i tried a few flavors of models from TDK and bourns
in short, looks like more pi filter stages is better than large C bank, even with small x ~ xx uH L value, large xxx uH values provides even more attenuation
C0G doesnt seem to help with EMI like X7R does, esp if we try to assume -140dB as the point where noise start to interfere with 1ppm of something?
if the simulation scales correctly, then 1 would assume all single C points of attenuations with large enough uF could only affect down to 100ppm to 10ppm scales
so my assumption is ~ very bad noise can be improved to at best 10ppm? by single LC?
then i try some theoreticals, capacitor with micro ohm esr which does not exist, it can attack down -140 to -160. but it cannot do broadband, only peaks

but multi stage LC pi seem to easily overcome -200 -300 db etc, this reminds me of the LISN thread
https://www.eevblog.com/forum/projects/5uh-lisn-for-spectrum-analyzer-emcemi-work/?all
and this
https://www.eevblog.com/forum/rf-microwave/50uh-and-250uh-inductor-design-for-lisn/?all

so in the 2nd pic i ran a rubbish mix of components and its under -160, i guess then 1LC is never enough, maybe 2pi?

[branadic]

Thanks Andreas,

could the ground issue also be a problem for LTC2057?

-branadic-

[Andreas]

but multi stage LC pi seem to easily overcome -200 -300 db etc, this reminds me of the LISN thread

Hello,

I fear only in theory. Practically every piece of a wire is a antenna.
So for higher dampenings than 20-40 dB you will need a shielded housing for every stage of the filter.

Of course you are right a X7R  or a lossy capacitor is more wide band than a high quality capacitor.
And also some ferrites act more as a lossy resistor for RF than as a inductor.

could the ground issue also be a problem for LTC2057?

When I look at the difference before and after it is mostly the 60MHz peak which is gone.
If this is right then the LTC2057 should also profit from correct connection of the output capacitor.

Even the OPA189 should have a little improvement near 60 MHz. But I guess the 20 MHz peak will not change much.

We will see.

with best regards

Andreas

[Andreas]

Hello,

one side effect that I mentioned with my experiments.

The OPA189 always gives a 12-18uV higher output voltage than the other devices. (LTC2057/ADA4522)

This is a bit strange since the offset is in the 3-4uV range typical and also the bias current by the 10K resistor should give a lower offset than 2uV additional.

with best regards

Andreas

[TiN]

Andreas
I have ordered ADA4522 (-2, but PCN affects all flavors anyway) from DigiKey few days ago. Will let you know what arrives once I get package next week.
It was intentional choice why I selected ADA4522 for output stages on my FX and QVR reference designs. 

[Andreas]

Hello,

I also ordered some parts from DigiKey (ADA4522-1ARZ-ND) packaged in tube but they ares still the old version.
(perhaps I should have ordered the cut tape version to get newer parts?)

But I fear the stock of DigiKey has to flat down before they (hopefully) get the new version.

But obviously this will take some time: (CT = cut tape, ST = tube)

Digikey ADA4522-1

13.11.2019
CT: 1395
ST: 449

24.11.2019
CT: 1075
ST: 277

29.11.2019
CT: 1062
ST: 277

with best regards

Andreas

[TiN]

How you decipher AD's/LTC datecodes?

[Andreas]

Hello,

if you look at the picture above
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2780364/#msg2780364

the middle line A#643 looks like a date code = 2016 week 43

The recently bought ADA4522-1 are datecode 751 = 2017 week 51

with best regards

Andreas

[MiDi]

The OPA189 always gives a 12-18uV higher output voltage than the other devices. (LTC2057/ADA4522)

This is a bit strange since the offset is in the 3-4uV range typical and also the bias current by the 10K resistor should give a lower offset than 2uV additional.

A bit strange seems understated.
In such a configuration it is way more than expected maximum for all obvious contributors to offset in sum.

Any idea what could be the cause?
Is the offset independent from value of output cap?

I assume there were 2 different OPA189 in the 2 LTZ?

[rhb]

I quit reading unread posts for a while because EEVblog was eating too much time.  I've only skimmed this thread, so please forgive me if I restate something already said.

I have 2 LED lamps in a 4 ft fluorescent fixture.  This produced a tremendous amount if EMI until I added a large Corcom filter and 1/8th" hardware cloth screen.

I have also put HW cloth over the openings of my Z400 workstations.  But still more to do.

Because of the staggering amount of EMI produced by my Instek MSO2024EA SMPS, I made a custom power cable for all my instruments which is fully shielded with MX and EMT all the way to the back of the instruments.   I posted about the project some time ago with photos.  I also bought, but have not installed a 2.5 KVA isolation transformer.

The Instek GDS-2000E line can be hacked to enable the spectrum analyzer app for the MDO-2000E variant.  This is a very useful tool and far more sensitive than my 8560A.  Above the 200 MHz maximum BW of the GDS-2000E, an SDR is the cheapest solution.  A SDRplay RSP2 will give a lot more dynamic range than and RTL-SDR, but also at 8x the price.

Most of the noise of concern is below 1 MHz.  There is a lot of it,  from a variety of sources.

Even without a good FFT on a scope, a small wire loop from tip to ground on a scope probe is very useful for locating nearby EMI sources.  A proper H field probe of large dimension is needed to locate more distant sources, but easy to make from solid shield RG402.

I went on a rather extreme TEA binge and now have three  5ft stacks of T&M kit on mover's dollies in my dining room for lack of space to set it up.  My current space is part of a 7' x 10' closet.   In order to have the displays  for SA, VNA, scopes, etc at a comfortable level I need 16' of bench, so I am building a 12' x 16' room in my shop which will be lined on all faces with 28-30 gauge galvanized steel sheet soldered at the seams.  I am told by a retired Tempest engineer that should provide about 120 dB of shielding.  The room will have its own HVAC zone on a 3 zone mini-split system.  Counting the cost of the extra HVAC zone, I expect that will run about $3K.  The steel sheet alone is about $800 for a 16' x 12' x 8' room.

A less costly approach and more easily moved solution would be to build desktop size steel box to hold equipment and DUT and extract the data via an optoisolated link.  Galvanized wire screen soldered to the steel sheet will allow airflow while blocking EMI.

To get from the 60 Mhz limit of an F***Tech to 120 MHz try one of these:

https://www.ebay.com/itm/NEW-1-200MHz-radio-frequency-multiplier-module/112082979199

I just saw them a couple of days ago and bought a couple.  Above that an ADF4351 module from ebay will serve to 4.4 GHz with harmonics much higher.

This book:

https://www.amazon.com/Electromagnetic-Compatibility-Engineering-Henry-Ott/dp/0470189304

has a great deal of useful information. however, the author writes poorly in passive voice, so it is difficult reading.

Finally, if you live near a high power broadcast tower, move.  I lived about a mile from a 250 KW FM tower and it drove me nuts.  Even a few inches of wire picked up strong EMI. I built a 120 dB DC audio amplifier.  With a speaker hooked to the output I could listen to the FM station merely by holding my finger 1/2" from the input. Driving a 3" speaker it was loud enough to hear in the next room.

That experience left me with a very strong desire for a quiet room in which to work.

Have Fun!
Reg

[Andreas]

A bit strange seems understated.
In such a configuration it is way more than expected maximum for all obvious contributors to offset in sum.

The values are "typical".

Any idea what could be the cause?
Is the offset independent from value of output cap?

No real idea up to now.
The value seems not to influence the offset. But the capacitor position (OGnd, PGnd) seems to have a influence.
(have to sum up the values).
I have checked for oscillations: but found none on my floating scope.

I assume there were 2 different OPA189 in the 2 LTZ?

yes. 2 different OPA189

The Instek GDS-2000E line can be hacked to enable the spectrum analyzer app for the MDO-2000E variant.  This is a very useful tool and far more sensitive than my 8560A.

Interesting.
But I cannot find any noise specifications for the scope nor any jitter specs.
Or is the 50 ppm/ms the jitter spec = 50 ns which would be rather high for a FFT.

Finally, if you live near a high power broadcast tower, move.
That experience left me with a very strong desire for a quiet room in which to work.

Yes, distance is the easiest way to avoid influence from EMI.
(and at low frequencies like 50 Hz the cheapest option.)

with best regards

Andreas

[Andreas]

Hello,

although I am doing much measurements,  I see not much progress in the "right" direction. (see attached image).

After having reached a good result when connecting a capacitor to PGnd on LTZ9 with the ADA4522 and even 100nF + 100nF (one to Output Gnd = OGnd, and one to Power Gnd = PGnd) did not show much difference.
Column D is a normalized value which makes the drift independant of the actual EMI-Level (with the square root of voltage formula).

My Target would be to get the influence down on a 5Vss level to below 1 ppm.

after reaching  ~100 ppm in the LTZ9 and OPA189 was the "winner" in the OGnd blocking scheme.
I thought that it would be a good idea to exchange the LTC2057 in the LTZ8 with a OPA189 to see what happens when connecting to PGnd.
But obviously in LTZ8 configuration (where I have all EMI capacitors populated so that the Output buffer has a 100nF at the positive input against OGnd) the EMI hardened device does not give any improvement.

So my next planned steps are to add the EMI capacitors on LTZ9 step by step
(moving away from datasheet cirquit) to see where something goes wrong.
I fear I still have a long journey until the final optimized cirquit can be drawn.

with best regards

Andreas

[Andreas]

Hello,

Is the offset independent from value of output cap?

The value seems not to influence the offset. But the capacitor position (OGnd, PGnd) seems to have a influence.
(have to sum up the values).

After reviewing my measurements the statement for the OPA189 here is:
the value of the capacitor (4.7 nF or 100nF) makes no change. (at least in Output Ground position).

For the position of the cap.
Output Ground gives the lowest Offset drift.
Followed by capacitors in both position.
Highest drift is performed with capacitor from output to power ground.

with best regards

Andreas

[TiN]

I've got ADA4522-2 today from DK.
Attached photo.

[Andreas]

Hmm,

840 -> week 40 in 2018 from the date code.

so still the old version before the PCN

with best regards

Andreas

[notfaded1]

What do you think the change might be with the product change?  Has no one gotten a new one yet?  You are talking about the die revision at end of 2018 correct?

Bill

[Andreas]

https://www.analog.com/media/en/PCN/ADI_PCN_18_0171_Rev_A_Form.pdf

According to the PCN the "new" devices will be from datecode 22-Aug-2019 on.

The change is described within the PCN.
two metal layers of the design are changed.
the new die will not be delivered before 22-Aug this year.

with best regards

Andreas
 

[3roomlab]

My Target would be to get the influence down on a 5Vss level to below 1 ppm.

http://www.giangrandi.ch/electronics/anttool/tx-field.shtml
on this website, it has a conversion which spits out V/m field strength

this is another site with V/m numbers
https://www.emf-portal.org/en/emf-source/110
somewhere in the middle is a line stating some measurement of 241V/m when 10cm from a tube

using the giangrandi "tool"
a TX of 1W @ 0.0001km (10cm) 0dBi is approx 54V/m
from the 732A pdf, it states it can go out of spec @ more than 0.18V/m, which means any WIFI 25mW within 10m
0.18V/m according to the giangrandi "tool" is 10uW from 10cm.

edit : im curious as to how much power is injected by the 5V, since the voltage is fixed, the way to know how much EMC enters is by current? do EMC testing also measure current?
by treating the LTZ circuit like an antenna, the amount of EMC (RF power) it could impede going in is purely by the right RF chokes? by having caps it allows shorts, but i think in RF where a cap "shorts", it could be creating a resonance. this scenario could be similar to the input front end of the 34401a vs K2000. i think the crucial part are the L, a number of chokes which makes nodes high in impedance to RF like the string of chokes in LISN
maybe some RF nut can shed some more light? as im not very RF inclined. i assumed many many many things

[Andreas]

The OPA189 always gives a 12-18uV higher output voltage than the other devices. (LTC2057/ADA4522)

Hello,

today I reduced the 10K feedback resistor (R22) by paralleling a 1K resistor. (giving 909 Ohms total feedback).
Cirquit see here:
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2784834/#msg2784834

Result of the 9.1K resistor change was a 21.1 uV reduction of the offset voltage.

So the OPA189 seems to have a effective input bias current of 2.3 nA
which is much higher than the datasheet value (with a weird measurement condition of 500pF and 100k).
For me the OPA189 seems to be only suitable for source impedances below 1K at the inputs.

with best regards

Andreas

[Andreas]

edit : im curious as to how much power is injected by the 5V, since the voltage is fixed, the way to know how much EMC enters is by current? do EMC testing also measure current?

The amount of current which enters is dependant on "parasitic capacities".
So it also depends on the distance to the next metal plane.
That makes these measurements a bit difficult for the repeatability.

The day before yesterday I forgot to remove the charger (switchmode supply) before starting the measurement.
The result was much worse than only with the battery powered voltage reference.
(more parasitic EMI current entering through the device).

with best regards

Andreas

[Andreas]

Just a little summary what I did in the mean time.

As preparation for further hardening measures I added C8 + R19 from my cirquit diagram to the data sheet population of LTZ9.
Without these components I cannot add C9 to the cirquit (as this would lead to oscillations).

The comparison of the measurement from 17.10.2019 and 02.12.2019 shows that adding C8+R19 has no visible impact on EMI behaviour at the unbuffered output.

So I increased power level and added C9. (03.12.2019 + 04.12.2019)

Interestingly the deviation at 60 MHz remained at -88 ppm when adding C9. Whereas the 20 MHz deviation is now suppressed.
So I assume that the 20MHz deviation belongs to the LTZ part of the cirquit and the 60 MHz deviation to the output buffer.

But a comparison before and after adding C9 (01.12.2019 + 06.12.2019) on the buffered output shows that for the output buffer C9 increases the EMI sensitivity.
The sensitivity is even worse (and more wideband) when the switchmode charger is attached to the LTZ9. (05.12.2019)

with best regards

Andreas

[MiDi]

The OPA189 always gives a 12-18uV higher output voltage than the other devices. (LTC2057/ADA4522)

today I reduced the 10K feedback resistor (R22) by paralleling a 1K resistor. (giving 909 Ohms total feedback).
Cirquit see here:
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2784834/#msg2784834

Result of the 9.1K resistor change was a 21.1 uV reduction of the offset voltage.

So the OPA189 seems to have a effective input bias current of 2.3 nA
which is much higher than the datasheet value (with a weird measurement condition of 500pF and 100k).
For me the OPA189 seems to be only suitable for source impedances below 1K at the inputs.

Would not have expected that the offset voltage of a modern device suffers significant from this typical input impedance 
Wondering if an input capacitance at the inverting input to gnd would improve this - I know this is going a bit offtopic...
Andreas, you could populate R23 with ~100pF-1nF (leakage <100pA@10V) with R22=10k, C22=10nF to verify.

These measurements (incl. charge injection with circuit by TI) are on my TODO-list, but for now it had low priority as LTC2057 and ADA4522 are used for this kind of application and no problems documented so far.
Odd that these effects are not even mentioned in datasheets and there are only few AN/documents mentioning it, afaik none does describe or gives details on it.

Edit:
There is another report of this effect in the TI Forum

[Andreas]

Would not have expected that the offset voltage of a modern device suffers significant from this typical input impedance 
Wondering if an input capacitance at the inverting input to gnd would improve this - I know this is going a bit offtopic...
Andreas, you could populate R23 with ~100pF-1nF (leakage <100pA@10V) with R22=10k, C22=10nF to verify.

Hello,

I did a measurement on LTZ#8 with OPA189 from the worst case scenario: (28.11.2019 with 100nF to PGnd R22=10k, C22=10nF)
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2810142/#msg2810142

And added a 470pF (500pF datasheet recommendation) 1000V FKC3 capacitor (to PGnd = directly to the OPA189 Gnd instead of the long way)
Result: another 5-5.5 uV increase of the "offset" when the 470pF capacitor is populated.
I had expected "no change" since C22 already gives a path to the output of the Op-Amp which is low impedant driven.

with best regards

Andreas

[Andreas]

Another test with LTZ#8:

starting from setup of 28.11.2019 with 100nF to PGnd R22=10k, C22=10nF
I replaced the OPA189 with the EMI-hardened low power version OPA187.
Of cause the 1/f (0.1-10Hz) output noise increases somewhat (I measured ~0.1uVpp).

The good news:
- The output voltage between unbuffered and buffered output measured within 1uV. (so no abnormal offset due to bias current).

The bad news:
- against the measurement of 28.11.2019 with OPA189 (-48 ppm @ 1.78Vpp @ 60 MHz)
  there is a large increase of EMI-sensitivity for the OPA187 (-1392 ppm @ 1.78Vpp @ 60 MHz)
- And additional some distortions on the OPA187 below 100kHz (obviously interference with the internal chopper frequency).

I will have to look if a additional capacitor across output ground gives some improvement.

with best regards

Andreas

[rhb]

You might want to make an RF sweep with a spectrum analyzer.  I strongly suspect that the radio and television broadcast bands are the biggest source of EMI.  That's certainly true for me living in the woods.  Otherwise, most of the noise will be below 100 kHz from nearby sources which can be tracked down and squelched.

Recent Chinese T&M kit is especially prone to producing a lot of EMI from the SMPS.  I shielded my entire mains feed to the instruments because of my concern about radiated EMI being picked up by adjacent instruments.

As I need to set up a new bench for the consequences of a severe TEA binge I am going to build a  12' x 16' x 8' room in my shop lined on all faces with galvanized steel sheet with soldered seams.  That will get covered with plaster board.  I plan to experiment with spraying wall board mud mixed with coal or charcoal  and iron or steel filings as an RF absorber if I can find a cheap source of iron or steel filings.

If you don't have the luxury of doing that a largish box would do for holding an instrument and some references.  I have to build a room to keep out wood dust, so lining it with steel sheet and RF absorber is a small frill. 

I have a 3457A in transit and am looking for a 44492A 10 input two wire relay board for it.  Once I have that all in hand I'll put the 3457A/44492A in a non-rectangular metal box lined with RF absorber and its own Peltier based temperature control system.  The lab area will have its own HVAC zone, but tight temperature tolerances work best with small spaces and masses.

For someone living in an apartment, a completely closed metal rack with a dedicated temperature control system seems to me the only way to avoid EMI.  Put a shield between each section of the rack, use proper feed thrus and EMI will not be an issue.  For hobby metrology I had assumed this was the norm, more for temperature control than EMI.  It's the cheapest solution to tight temperature control.

For cooling a rack, strip the mechanicals from a small refrigerator to build a small cooling system for the box with a holding plate with phase change media to minimize compressor cycling.  For a single instrument, use a Peltier travel cooler.

If you build an enclosure, make all the surfaces non-orthogonal.

Designing for high CMR is vitally important in precise work.  But avoiding high common mode fields is more important.

Have Fun!
Reg

[Andreas]

Hello,

after adding a capacitor from output to output Ground I get a dramatically reduction in EMI-Sensitivity.
(so I actually fear that the Capacitor to power ground is broken somehow?)

So If I am able to handle the 80 kHz interference that will be the best result up to now.

with best regards

Andreas

[Andreas]

@rhb:

my intention is not to avoid EMI but to harden my devices so that I get the same results independant of environment.

Here it is not easy to "live in the woods". The population is simply too dense.
And I have no intention to dig myself into a abandoned salt dome on battery supply which would be the only possibility to avoid EMI here.

with best regards

Andreas

[Andreas]

Hello,

I also did some tests on LTZ#9 with a different approach at the frequency of maximum sensitivity:

I fed in 60MHz 3.56Vss (unloaded generator output) signal into the buffered output of the reference.
Then I touched different parts like battery the LTZ itself and the plastic OP-Amps with a finger.
Of course the output voltage changed a lot by changeing the parasitic capacitances.
The "deltas" are recorded to find a position of maximum sensitivity.

Further I used a "probe" to touch the pins directly with a "defined" parasitic capacitance:
https://www.reichelt.de/miniatur-klemmpruefspitze-0-75-mm-schwarz-pruef-mps-2-sw-p106196.html?

The "deltas" between untouched and touched pins are also recorded together with the "absolute" value (without EMI signal).
It soon got clear that even the housing of the LTZ has some EMI relevance.
Since the housing carries around 0.5V against the ground pin it cannot be connected directly to ground.
I used a capacitor to do a RF short to ground. Which influences somewhat the other sensitivities (especially pin 3).
The soldering caused a permanent shift of the output voltage by -4 .. -5 ppm.
So I will not use this measure (see picture) for my calibrated LTZ´s.

Further from the measurements it gets clear that pin 6 is the most sensitive pin to EMI.
(the base of the temperature sensing transistor).
So by adding C11 between base and emitter of the sensing transistor the sensitivity can be largely removed.

with best regards

Andreas

[Andreas]

Hello,

during my weekly ageing measurements for the LTZ references I mentioned that LTZ#8 (with OPA187 buffer) had a large noise together with my ADCs (ADC13, ADC15, ADC16 and ADC17). This happened only on the buffered output of the LTZ#8 reference. The unbuffered output was unaffected.
Further examination showed up to 150 uVpp on the 2:1 divided output of the LTC1043 dividers. (so 300uVpp on the 10V input side).
Normally the LTC2400 ADCs show only around 10uVpp on the 5V input or 20uVpp on the 10V input.

So obviously the OPA187 has interference with the switched load of the LTC1043 dividers.

I made a measurement to show the difference of the buffered and unbuffered output of LTZ#8.
After exchange to an ADA4522 the noise was in the normal 10uVpp range
(except for some time immediately after connecting to the unbuffered output).

so for me another reason not to use the OPA187 as output buffer.
On my K2000 (10 NPLC) and 34401A (100NPLC) there is no interference visible.

with best regards

Andreas

[Andreas]

Hello,

I updated some EPCOS coil measurements on the coil comparison post:
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684085/#msg2684085

with best regards

Andreas

[Andreas]

Hello,

I repeated the heater noise + Lamp switching experiment on LTZ#9 see:
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2776212/#msg2776212

This time with C11 populated between base + emitter of the heater sensing transistor.
Under same conditions I measured  much lower excursions of the heater voltage.
Instead of up to 1Vp I measured maximum 0.2Vp excursion.
If I look only on the low frequency part (e.g. with a 1 MHz filter) the maximum is  around 80mVp
So C11 gives a factor 5-12 improvement.

From statistics side
20 measurements in average with 108mVpp (against 287 mVp peak only in formerly measurement)
Standard deviation 44 mVpp (against 274 mVp before)

when filtering with 1 MHz we get
20.5 mVp average
and 21.7 mV standard deviation.

with best regards

Andreas

[SilverSolder]

[...] design from a EMI proof power supply first [...]

One or more large 12V lead-acid battery/ies?

[Echo88]

Regarding suitable power supply here are some suggestions:
LT1533-based DCDC: https://www.analog.com/media/en/technical-documentation/application-notes/an70.pdf
Sine-Wave-Driven-DCDC: https://www.analog.com/media/en/technical-documentation/application-notes/an29f.pdf Page 4
https://x.artofelectronics.net/wp-content/uploads/2019/11/9xp14_low-noise_isolated-pwr.pdf

[iMo]

FYI - a bit harder EMI measurement, hopefully not applicable to your labs

[Andreas]

One or more large 12V lead-acid battery/ies?

I generally use NiMH-cells for my references.
Mainly to avoid one stage (inner housing) of guarding and shielding.

with best regards

Andreas

[iMo]

Below what I see when I switch my wifi router on/off (confirmed, I can send morse keying the router on/off ).
The router is ~4m off the plastic box with the 10V Vref and 34401A.
Notebook 1m off the box, wifi on all the time.

The measurement taken on a standard LM399AH wiring (dead bug 78L15+OP07+LT5400), the heater blocked 4u7 tantalum, the zener with 100nF ceramic close to its 4 pin socket plus 4k7/3u3_foil low pass towards opamp. 78L15 blocked and output of the OP07 blocked 25ohm/100nF_foil.
34401A with 1m long twisted cable rolled to a small 15cm dia bundle.
An stm32 MCU on the RS232 with the HC-05 sending data to my notebook.

The router on/off difference is always aprox -1.5ppm as depicted.
Interestingly - when the wifi router is ON the Vref output is "more" stable and quiet.

PS: y axis is ppm (filtered), blue is the temperature inside the Vref box (not related to the issue).

PPS: a clip-on ferrite bead put on:

a. RS232 cable - no change
b. 34401A close to its inputs - no change
c. close to the Vref output - the on/off ppm difference halves (-0.7ppm)

[dietert1]

Another nice example that a precision reference isn't "ready" if it needs to be treated like a raw egg.
Maybe your Wifi isn't off completely when you think it should be off. Nowadays those things have their own life.
Can you do something similar with a cellphone?

Regards, Dieter

[iMo]

The wifi router is switched off by pulling its power cable off - thus no power off the antennas. I have closed my phone in a metal box and put in a different room during the experiment (actually this is NOT an EMI experiment, but my measurements with 399s I run - I had been thinking it is a popcorn or a defunct 399 or something like that, until I discovered the relationship between the time I switched off the router and the jumps in my data).
The HF energy enters the Vref output via cables, the 100nF foil does not work at those frequencies and it messes up with the OP07 opapm. I will add ceramics and common mode choke at the Vref output tomorrow..

[MegaVolt]

I had been thinking it is a popcorn or a defunct 399 or something like that, until I discovered the relationship between the time I switched off the router and the jumps in my data).

Refrigerator?

[Andreas]

@IMO
thanks for the experiment (I have also ordered some LM399 for that purpose but you were faster).

the 78L15 is blocked with 100nF both at input and output?
(I have seen a 3.3V regulator with about 400mV (>10%) off when using electrolytics according to the data sheet on EMI-Tests (30MHz-1Ghz 3V/m). A additional 100nF keramics at input and output helped).
The LM399 is also sensitive to heater supply changes.
And I would also spend a 100nF in parallel to the heater pins. (additionally to the 4.7uF).

Where does the power supply come from (battery / wall wart)?
Eventually you will also need a common mode choke here.

Does exchangeing the op07 against a ADA4522 give a improvement?

with best regards

Andreas

[iMo]

@Andreas: my power source is a "low noise" 723 well blocked with ceramics. Then follows 78L15 (as a feasibility experiment only) with 220u input and EDIT:100nF ceramic output. I will add 100nF ceramics to the heater. I do not plan chopper there, OP07 perhaps an OPA277 is enough. Chokes and UHF ceramics at input/output are a must, it seems.

[Andreas]

there are also some EMI-hardened other precision OP-Amps (non chopper).

with best regards

Andreas

[iMo]

I had been thinking it is a popcorn or a defunct 399 or something like that, until I discovered the relationship between the time I switched off the router and the jumps in my data).

Refrigerator?

It is the wifi router, powering it off/on changes my Vref output by 1.5ppm, I repeated that many times after the discovery and it works nice

[iMo]

there are also some EMI-hardened other precision OP-Amps (non chopper).

OMG, an LM399 reference does not need this kind of gold plating, imho - it requires a well made decoupling, that is all.. 

[Andreas]

Hmm,

it is not only the reference which gets affected by the EMI.
Also the OP-Amp is affected.
So I think a ADA4177 is shurely not too much.

with best regards

Andreas

[iMo]

I meant 399 reference as the "board" .
I would opt for an opamp with low offset drift, something like <0.5uV/K.
OPA277 looks best with +/-0.1uV/K.

[iMo]

Added in/out filters (2T on a ferrite bead plus 4n7 ceramics) and clip-on ferrites on the RS232 and Temperature cables.
Still 0.4ppm difference between router on and off.

[Andreas]

Hello,

I fear this will end in a metal housing with feed through capacitors for all in/outputs.

with 4.7nF: what did you do so that the Op-Amp does not oscillate with capacitive loading?

with best regards

Andreas

[iMo]

with 4.7nF: what did you do so that the Op-Amp does not oscillate with capacitive loading?

There is the 25ohm resistor at the opamp's output (then there is 10nF foil and 4n7 ceramics).

[dietert1]

You could try resistors on the supply side, too. Sometimes resistors make better filters than inductors. There should be enough headroom to loose some 0.5 V DC on the resistors.

Regards, Dieter

[iMo]

Is it only me affected by the wifi virus ?? Could it be people do not see it because a) they do not switch their routers on/off during such a measurements, b) their measurement resolution is not enough to see the jump?

[MegaVolt]

Is it only me affected by the wifi virus ?? Could it be people do not see it because a) they do not switch their routers on/off during such a measurements, b) their measurement resolution is not enough to see the jump?

I checked the Fluke 732a along with the Keithley 7510. Turning the Wi-Fi or computer on or off did not give noticeable steps. But the inclusion of an ancient refrigerator, I see But the connection was made quickly and the wires were not shielded.

[SilverSolder]

Is it only me affected by the wifi virus ?? Could it be people do not see it because a) they do not switch their routers on/off during such a measurements, b) their measurement resolution is not enough to see the jump?

Router 2 floors away from the lab?

[Andreas]

Hello

some further measurements:

LTZ#9 populated with ADA4522 as output buffer.
Capacitor 100nF across the output.
And a 100nF between Output Ground and ADA4522 (power) ground which fixes the layout error.
On the input of the ADA4522 there is a 4K7+100nF low pass filter.
And C9+C11 of my LTZ1047B schematic are also populated.

Measurement of 04.04.2020 shows -240 ppm deviation @ 60 MHz with 5Vpp Output level on the FY6800.

Measurement of 03.04.2020 shows improvement by the 3.5 windings on a EMI core suggested by Jason.
so -7.4 ppm @ 24 MHz as max deviation with 5Vpp output level.
At 20 MHz the ouput amplifier is switched so that the level is slightly different after switching.

And finally Measurement of 02.04.2020 shows best result with the Epcos 51 uH common mode choke for signal lines (CAN bus choke)
This choke already looked very promising when being measured here: (3rd last section)
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684085/#msg2684085

Result: -1.4 ppm @ 60 MHz with 5Vpp: so my personal target for EMI reduction is reached here.
Since the choke is now outside the LTZ#9 housing I will have a slightly worse result with the choke inside the housing due to parasitic capacities increasing.

With best regards

Andreas

Edit: picture of external CAN coil attached.
picture of output buffer ADA4522 with capacitors attached

By the way PCN 18_0171 has been updated on the Analog Web site: the new parts with the mask change will be available earliest mid of this year.

https://media.digikey.com/pdf/PCNs/Analog%20Devices/PCN_18_0171_RevB.pdf

[Andreas]

Hello,

I have updated first results of LM399 here:

https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684100/#msg2684100

with best regards

Andreas

[notfaded1]

Is it only me affected by the wifi virus ?? Could it be people do not see it because a) they do not switch their routers on/off during such a measurements, b) their measurement resolution is not enough to see the jump?

I should really check this... I have a wifi router on my bench.  I do see some jumps I had though might have been from AC mains but with lot of filtering still seem to be there.

[Andreas]

Is it only me affected by the wifi virus ?? Could it be people do not see it because a) they do not switch their routers on/off during such a measurements, b) their measurement resolution is not enough to see the jump?

I should really check this... I have a wifi router on my bench.  I do see some jumps I had though might have been from AC mains but with lot of filtering still seem to be there.

Hello,

You might have more EMI sources than that what you think:

It took a while until I noticed that my network printer connected by ethernet also has a WIFI and additionally a NFC source.

with best regards

Andreas

[Andreas]

Hello,

a further update on the LM399 here:
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684100/#msg2684100

with best regards

Andreas

[dietert1]

Thanks Andreas, since i am trying to "tune" our HP 3456As.

I was wondering though, how much might be the effect of the bypass capacitor leakage current. For example if leakage gets increased somewhat due to heating to let's say 1 uA, times zener differential resistance would give us maybe 10 uV, so that would be 1.5 ppm - as an upper limit of an unstable contribution. Leakage of MLCC caps seems to be of similar magnitude, so it gets a bit nasty, if one needs to use foil caps.

Also i remember a claim (MisterDiodes?) that a direct bypass would accelerate aging due to increased noise current. So the bypass could include a series resistor, for example 5 or 10 Ohms, with the output taken from the capacitor. I remember from my LTFLU experiments that a direct 100 uF bypass on the 10 V output caused a lot of noise current inside the 7 -> 10 V loop. Improved with 100uF + 3 Ohms, but i don't have numbers for that.

Regards, Dieter

[Kleinstein]

For a voltage regulator of any kind a large low ESR capacitance is difficult. This is because for principle reasons the output impedance will be inductive over a large frequency range and often very close to a low loss inductor. In this case a low ESR capacitor promotes ringing and only helps for frequencies well above the resonance. A capacitor with some extra resistance can help.

For just RF interference one should not a really large cap, at higher frequency the LM399 is no longer so low in impedance. So chances are one could get away with some 100 nF or 10 nF. 10 nF are even available as C0G with really low leakage.

[Andreas]

For example if leakage gets increased somewhat due to heating to let's say 1 uA, times zener differential resistance would give us maybe 10 uV, so that would be 1.5 ppm - as an upper limit of an unstable contribution.

So the bypass could include a series resistor, for example 5 or 10 Ohms, with the output taken from the capacitor.

Zener differential resistance of a LM399 is below 1 ohm -> 1 uA will give a 1uV change.
But from my LNA I know that if you constantly keep the capacitor charged at room temperature the leakage will be much lower than 1uA.

With a series resistor you will increase the differential resistance so leakage has a increased effect.
And remember: most of the current on a LM399 Zener is shunted with a parallel regulator.

And what is still open is what happens when you use a buffer amplifier. It might keep the low frequency peak away (if the EMI does not travel by the ground pin). So a 100nF would do the job for the higer frequencies.

with best regards

Andreas

[dietert1]

Thanks, so the LM399 includes a buffer loop and i would try to avoid a large capacitor over the output, but put a small resistor in between, maybe with a 1 uF PP capacitor to Gnd. I also have 10 uF MKT that would fit. For the heater in the HP3456A reference it's probably two capacitors, since the heater gets fed from +/- 15 V. And maybe again two small resistors in series with the supply.

Regards, Dieter

[Kleinstein]

The capacitor directly at the LM399 reference section would be more about keeping RF out, so no need for really large capacitance. Some 100 nF should be large enough. Because of the internal control loop some series resistance is probably a good idea.  Due to the low output resistance even 10 µF would not give much filtering action. In the 3456 filtering would be more effective at the other side of R502. One may just see the effect of reduced noise in the 5-30 kHz range. The next frequency range of interest would than be at some 25 Hz (1 PLC mode) or 2.5 Hz (10 PLC mode).

[Andreas]

Thanks, so the LM399 includes a buffer loop

For the heater in the HP3456A reference it's probably two capacitors, since the heater gets fed from +/- 15 V. And maybe again two small resistors in series with the supply.

Hello,

not yet: it is still on my todo list wether a buffer can avoid a large capacitor directly on the zener output.

On the heater side: I am not shure wether the 47uF capacitors are really needed there since I usually put those anyway at the heater. In any case I would not put series resistors on the heater side unless I really need to limit inrush current. The datasheet already mentions stability issues with series resistors. (a > 2uF Tantalum is needed).
And yes: if you want to use larger capacitors with a split supply you might need 2 capacitors depending on the voltage regulators. (some regulators might not start up without protection diodes across the output when there is a larger capacitor across +/-15 V).

The capacitor directly at the LM399 reference section would be more about keeping RF out,

unfortunately the LM399 is very sensitive for lower frequencies too. So you need some measure which also works at 50-100kHz where many switchmode supplies are working.

with best regards

Andreas

[dietert1]

As mentioned before, a good measure at lower frequencies may be an external buffer. The HP 3456A has on its reference module the 7 V to 12 V amplifier, so the LM399 output signal stays inside that module. Maybe a metallic shield around the whole reference module can still improve isolation from external fields.

Regards, Dieter

[Andreas]

Hello,

I have updated the LM399 measurements with a ADA4522 (EMI hardened) buffer
So this one actually filters the low frequencies:

https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684100/#msg2684100

with best regards

Andreas

[MegaVolt]

Can add a little more schemes? A verbal description leaves room for misunderstandings.

[Andreas]

Hello,

got some ADA4522-1 from Mouser with Datecode #A946
so shurely after PCN change.

Lets look wether the EMI behaviour has changed or not.

with best regards

Andreas

[Andreas]

Hello,

Measurement of 12.04.2020 shows LTZ#9 with already a "newer" datecode DC751 against the measurement on 04.04.2020
(additionally the 100nF on +input of the OP was removed).
Result: -219 ppm @ 60 MHz and 5Vpp

On 25.08.2020 I did a repeated measurement to see the "stray" of the measurements.
Result: -259 ppm @ 60 MHz and 5Vpp

So as expected there is some stray between the measurements (Temperature, slightly different parasitics due to slight position changes).

Measurement of 26.08.2020 shows LTZ#9 with the Datecode DC946 (after PCN)
Result: -238 ppm @ 60 MHz and 5Vpp

So the EMI behaviour of ADA4522-1 obviously has not changed by the mask change.

with best regards

Andreas

[dietert1]

This may be a strange proposal, anyway: Do you have a spectrum analyzer? Did you ever check your references for RF emission? A LTZ1000 probably has enough gain inside to be a RF oscillator. Frequency would depend on length of signal tracks, probably in the 100 MHz - 500 MHz range. I am asking this because i read about the high capacitance of zeners and how similar their construction is to varactors.

Another question i was thinking about: You show how external RF can shift the reference voltage. Did you ever have a look at how it affects noise/stability? Maybe we should add some RF to make the zener "glow better".

Regards, Dieter

[Andreas]

This may be a strange proposal, anyway: Do you have a spectrum analyzer? Did you ever check your references for RF emission? A LTZ1000 probably has enough gain inside to be a RF oscillator. Frequency would depend on length of signal tracks, probably in the 100 MHz - 500 MHz range. I am asking this because i read about the high capacitance of zeners and how similar their construction is to varactors.

Hello,

I fear that is not easy. The buried zener is no "power" device so the chip area is relatively small giving only some pF capacitance below the breakdown voltage.

We are operating the zener above the breakdown voltage so the capacitance is near zero.

In the standard cirquit there is a 22nF in parallel to the output transistor which dampens any RF.
Additionally I have on my devices 100nF across the zener voltage output and 100nF across the BE junction of the transistor, giving effectively 50nF across the zener. So how should the some pF in parallel at a very constant voltage give a RF signal in the MHz range?

Besides this: I usually measure 1/f noise 0.1 .. 10 Hz and also wideband noise 10Hz .. 100kHz which are according to the data sheet.

Sorry no spectrum analyser available. I would also never put a 50 Ohms input in parallel to a unbuffered LTZ1000 (it ages very fast under this condition).
When using the Oscilloscope with FFT I see more or less the FM radio stations nearby. So without a very good shielded room you will measure only garbage.

Another question i was thinking about: You show how external RF can shift the reference voltage. Did you ever have a look at how it affects noise/stability? Maybe we should add some RF to make the zener "glow better".

Stability will be worse since you never get the same RF energy into the zener depending on where the capacitance of your body is located versus the reference. So this is no option for me.

with best regards

Andreas

[Andreas]

Hello,

Strange: today I got a updated PCN revision C notice from Mouser for the ADA4522-1.

https://www.analog.com/media/en/PCN/ADI_PCN_18_0171_Rev_C_Form.pdf

Now I am somewhat confused.

Did I do my verification of EMI behaviour too early? (and need a DC later than Aug 20th this year?)

Whats going on there?

with best regards

Andreas

[MiDi]

Andreas, I do not think it is relevant, the reports are updated.

Material will be separated by date code for conversion to the new die. ADA4522-2 cutover date was 1918.

[Andreas]

Hello,

I did some measurements with the ferrite that Frank used in his latest LTZ6+LTZ7 references.
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg3263778/#msg3263778

The advantage of such a ring ferrite is that one could use a force sense terminal by using 4 wires on the common mode choke. Against a ready made CAN choke with only 2 wires like used here:
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg3002670/#msg3002670

Since I feared that one winding and ~2 uH would not be enough for a useful filtering (the CAN common mode choke has 51 uH) I first tried the maximum possible with the 0.5 mm wire that I had on hand giving 20.5 windings (20 outside 21 inside the core).

The diagram 20201023_EPCOS_B64290L38X830_20_5_wdg.PNG shows a good filtering of -46 dB but a rather low resonant frequency of 5.5 MHz which might give insufficient filtering at higher frequencies. (And of course I would have to use a thinner wire to get a common mode choke). At 1 MHz the measured inductivity gives ~560 uH.

So I tried 2x10.5 windings as common mode choke hoping that the resonant frequency would double.
20201024_EPCOS_B64290L38X830_10_5_wdg.PNG shows the result. Resonant frequency shifted to 7 MHz and of course a reduced maximum filtering (-34 dB).  Measured inductivity @1 MHz ~ 140 uH.

Next try with 2x5.5 windings so having ~50uH measured inductance like the CAN choke.
Again 7 MHz self resonant frequency with maximum filtering -24 dB.
The CAN choke had -34 dB @ 20 MHz and showed no resonance up to 20 MHz.

Finally 2*1.5 windings with maximum filtering of -8.25 dB at 6.5 MHz self resonant frequency.

So whats wrong with this ferrite?
When looking up the Material data sheet of N30 which is used by this type the description is: Wideband pulse transformers up to 400 kHz. That may be a explanation why the dampening decreases above 7 MHz.

Ok the more interesting thing like the measurement in a 50 Ohms system will be how the inductors behave together with the LTZs compared to the CAN common mode choke.

with best regards

Andreas

[kleiner Rainer]

hello all,

contrary to popular belief, EMC work is NOT voodoo or black magic, but the correct application of well known engineering principles. I have been doing this for the last quarter of a century, successfully bringing many millions of € worth of devices through CE and FCC testing and certification.

First principle: know your enemy. In this case: which disturbance frequency? Many of the older equipment we collect and use was designed in a time, when AM and FM radio and CB radio was the main threat. Often it was simply assumed, that precision lab equipment would only operate in temperature and humidity controlled, well shielded rooms. Using a calibrator of 70s or 80s vintage in a home environment exposes it to strong fields in the 900, 1800, 2400MHZ and 5150 to 5700MHz range. Was it designed to cope with that? Maybe it has filtering against AM and FM radio, but not much more - mobile phones and WLAN were not in every household back then.

So what can we do? Throwing any ferrite we can find in our junk box at the problem does not help, as Andreas found out. Using the wrong ferrite can make it even worse, no joke! It must be the RIGHT ferrite material at the RIGHT place. How do we find out which one?

For a first look at the problem, it is useful to look what radio amateurs do in such a situation. I found a useful introduction to RFI problems and solutions:

http://audiosystemsgroup.com/PAARA-RFI-2011.pdf

The guy who wrote it, Jim Brown K9YC, was the EMC expert of the AES and has done extensive tests to assess the susceptibility of audio equipment to RFI.

Another great resource is the Book "Electromagnetic Compatibility Engineering" by Henry W. Ott. I have it on my lab bench during EMC work. Saved my ass on some occasions. It should be called "The Art Of EMC", really...

And now some hints: if you want to solve EMC problems, you should be able to measure/see whats going on - preferably up to 1GHz. I do quick checks with a handheld transceiver (Yaesu FT1XD) in scan mode. It also helps if you can listen to the disturbance - is it white noise, some kind of buzz or does it sound like radio or a data modulation (you all remember the sound of a fax or modem, do you?). Another low-cost option is the new TinySA. If it works as well as my NanoVNA, is a must for hobbyist EMC work.

For a broadband suppression choke, there is still the good old six-hole ferrite bead:

https://www.we-online.de/katalog/datasheet/7427503.pdf

As you can see, useful up to at least 1GHz. I keep a bag of them handy in the lab. Due to their construction, they do not show the parallel resonance of the classical choke. They are also very useful in filtering power supplies and brushed motors.

I would also recommend to have a look at the application notes on the Würth homepage:

https://www.we-online.de/katalog/de/pbs/emc_components  (switch to "english" left of the shopping cart symbol).

Hope that helps.

Greetings,

Rainer

[Andreas]

For a broadband suppression choke, there is still the good old six-hole ferrite bead:

https://www.we-online.de/katalog/datasheet/7427503.pdf

Hello,

good idea since I have them already here.
And they measured somewhat better than the Murata BLM31 which behaved better than the "small" CAN chokes which I used in my AD587LWdesign. See Measurement here:
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg2684085/#msg2684085
(unfortunately the forum software has reordered all pictures to the end of the post).

And finally I have the NiZn Ferrites from 4A11 material used as high frequency cores for the AF2 network (Ferroxcube TN13/7.5/5-4A11).
They look very promising with 15.5 windings and are only a little larger than the cores from Frank.
I measured -40 dB @ 20 MHz and ~112uH @ 1 MHz. Starting already (-3dB) at 50 kHz.

with best regards

Andreas

[Andreas]

Some measurements on LTZ#9 with the EPCOS/TDK B64290L38X830 ferrites with different number of windings.

LTZ#9 has ADA4522 buffer with latest date code so around -238 ppm @ 60 MHz and 5Vpp without additional external filtering/coil.

a) with 2 x 10.5 windings
20201025_LTZ9_Buf_ADA4522_DC946_100nF_PGND_GNDs_C9_C11_TDK_10w5.PNG

around -12 ppm deviation @ 60 MHz so a improvement of factor 20 against the buffered reference alone.
So in practical application the ferrite behaves better than expected from the 50 Ohms measurement.

The EPCOS 51uH CAN choke had -3 ppm deviation in worst case under this conditions so only a factor 4 improvement.

b) with 2 x 5.5 windings
20201026_LTZ9_Buf_ADA4522_DC946_100nF_PGND_GNDs_C9_C11_TDK_5w5.PNG

around -38 ppm deviation @ 60 MHz so a factor 3 worse than above.


c) with 2 x 1.5 windings
20201026_LTZ9_Buf_ADA4522_DC946_100nF_PGND_GNDs_C9_C11_TDK_1w5.PNG

around -85 ppm deviation @ 52 MHz which seems to be a resonance.
Giving only a improvement of factor 3 against the naked buffer.

with best regards

Andreas

[Andreas]

And some further measurements with ferrites.

d) the Ferroxcube TN13/7.5/5-4A11 with 2x15.5 windings together with LTZ#9
nearly as cheap as the TDK ferrite core but a little bit larger allowing 15.5 windings on each side.
https://www.reichelt.de/ferritkern-material-4a11-ferr-tn13-7-5-5-p246049.html?&nbc=1

Diagram 20201028_LTZ9_Buf_ADA4522_DC946_100nF_PGND_GNDs_C9_C11_4A11_15w5.PNG shows around the same performance as the 51uH CAN choke. with -4 ppm @ 60 MHz and 5Vss.

e) the Würth 6 hole ferrites 7427503 (2 needed for the LTZ#9)

Diagram 20201029_LTZ9_Buf_ADA4522_DC946_100nF_PGND_GNDs_C9_C11_2x6hole.PNG shows -43 ppm @ 60MHz and 5Vss.
So for me it looks as if the 6 hole ferrites start working at around 20 MHz and may be a additional option wide above 60 MHz where I have my "blind spot" at the moment.

And now some hints: if you want to solve EMC problems, you should be able to measure/see whats going on - preferably up to 1GHz.

Agreed but what do you recommend for generating sine waves with at least +15 dBm up to 1 GHz (or at least 230 MHz) with computer interface and for a price which is adequate for a "secondary hobby" (< 200 EUR).

With best regards

Andreas

[dietert1]

Your TN13/7.5/5-4A11 common mode choke has about 80 uH (AL = 0,358 uH, *15*15). I measured the WE bead choke to be about 15 uH. So when you use two in parallel, you have 7.5 uH. This gives a factor ten and roughly explains the ratio of ten you find in your measurements. Your measurements are in a frequency region where this simple math still applies. It does not really tell you what happens with WLAN and the like.

We have a 7 KW photovoltaic generator on our roof with two switched mode power converters and i think i can see on my reference logs when the sun was shining. The effect is about 0.2 ppm on the daily averages. The problem is i can't turn it on or off, so it's a bit of try and error.

Regards, Dieter

[kleiner Rainer]

...

And now some hints: if you want to solve EMC problems, you should be able to measure/see whats going on - preferably up to 1GHz.

Agreed but what do you recommend for generating sine waves with at least +15 dBm up to 1 GHz (or at least 230 MHz) with computer interface and for a price which is adequate for a "secondary hobby" (< 200 EUR).

With best regards

Andreas

NanoVNA together with NanoVNA-Saver as PC freeware. The firmware also supports using it as a signal generator. You should still have enough cash left over for adapters, a set of good quality SMA cables and a Tiny SA. The output power of -18dBm is low enough for unlicensed use (+15dBm is QRP transmitter power!), but high enough for EMC work IMHO. Remember: Disturbance levels are commonly specified in dBuV... Optionally a power amplifier (MMIC-based) could be added. But as already mentioned - keep the licensing requirements in mind!

In an emergency, the second (receiver) port can be used as a crude SA - I demonstrated this at a club meeting with a rubber antenna connected to the input and showing all the GSM activity around 900MHz in the room. This use would also benefit from an LNA.

Introduction:

https://www.pvrc.org/Powerpoint/NanoVNA%20IntroductionCombined-2.pdf

https://www.bwcelectronics.com/articles/NanoVNA%20User%20Guide.pdf

Wahlweise in Deutsch:

http://www.gunthard-kraus.de/inhalt_de.htm

But be warned: VNAs and spectrum analyzers are another rabbit hole...

Greetings,

Rainer

[Andreas]

but high enough for EMC work IMHO. Remember: Disturbance levels are commonly specified in dBuV...

The standards to this theme tell something different.

If you have a look at DIN IEC 61000-6-1 together with the referenced standard IEC 61000-4-6 the level for household devices is 3V RMS on signal and power supply lines. (129.5 dBuV). Industry levels according to DIN IEC 61000-6-2 are even higher with 10V RMS and 140 dBuV.
Measured at the corresponding input with CW by replacing the device by a 150 Ohms load.
Additional you should have a 80% 1kHz AM modulation.

Most of the generator power is lost in the adaption network (CDN) to generate 150 Ohms common mode.

So it is no wonder why commercial test equipment for that test uses > 7W (38.5 dBm) linear amplifiers.
https://www.teseq.com/products/NSG-4070.php
https://www.theemcshop.com/16-emc-cdn-s-coupling-decoupling-networks

So +15 dBm generator power are in comparison at the low edge to keep the costs low.

with best regards

Andreas

[kleiner Rainer]

Andreas,

from your measurements and questions I assume you are a hobbyist with limited budget. In EMC work, 200€ is nothing - or the price for a single IEC standard. Luckily, the Indian government put their relevant standards on their server for everybody to use:

https://law.resource.org/pub/in/bis/S04/is.10052.1.1999.pdf
https://law.resource.org/pub/in/bis/S04/is.10052.2.1999.pdf

This is as close to the basic EMC standards CISPR 16.1 and 2 as you can get without paying.

IS10052.1 describes measuring apparatus and Annex N describes the coupling units for current injection. This part should be of interest for you.
IS10052.2 describes methods of measurement of disturbances and immunity. Again, Section 3, part 3.2 should be of interest to you.

With the help of these two standards you should be able to determine what to measure and with which equipment. DIY is possible, but then the question of calibration arises, as usual. Connections to EMC people or a university lab could help. The Nano VNA I recommended would at least help to measure the insertion and return loss of coupling networks.

Greetings,

Rainer

PS: more Indian Standards:  https://law.resource.org/pub/in/bis/manifest.litd.9.html

[Andreas]

Hello,

I have checked another 4A11 ferrite core.
TN 10/6/4 which is nearly as small as the TDK core.

With 0.5 mm wire I can get only 2*11.5 windings.
So with ~the same AL-value than than the TN14/7.5/5 core somewhat less inductivity.
Starting (-3db) near 200 kHz it reaches -32 dB at 20 MHz.

It gets better with thinner (0.28 mm) wire and 2*24.5 windings.
Starting (-3dB) near 20 kHz it reaches -45 dB at 20 MHz.

The picture shows the TDK/EPCOS core in comparison with the TN10 cores.

https://www.reichelt.de/ferritkern-170-nh-material-4a11-ferr-tn10-6-4-4a-p246048.html?&nbc=1

with best regards

Andreas

[Andreas]

Hello,

in the mean time I measured the smaller TN10 with 4A11 material cores on the now temperature compensated LTZ#9

with 0.5 mm dia wire I get 2*11.5 windings on the core.
with the smaller 0.28 mm dia wire there are 2*24.5 windings.

The 11.5 windings give somewhat weaker result near 20 MHz.
On 60 MHz the results are very similar.
So I believe that here the parasitic capacitance between input and output of the common mode choke plays more a role than the actual inductivity.

with best regards

Andreas

[Andreas]

Hello,

to put some light behind my "blind spot" I got one of those cheap NanoVNAs.
At least the frequency response of inductors above 20 MHz (the limit of the frequency generator of my scope) can be shown.

For each component I made 2 measurements to increase frequency resolution at least at the lower frequencies:
100kHz - 250 MHz and 100kHz- 3000 MHz.

the candidates are some of the previously known common mode chokes:
a) Ferroxcube TN10/6/4 4A11 core with 2*11.5 (0.5 mm dia) windings.
b) Ferroxcube TN10/6/4 4A11 core with 2*24.5 (0.28 mm dia) windings.
c) Ferroxcube TN13/7.5/5 4A11 core with 2*15.5 (0.5 mm dia) windings.
d) Epcos/TDK B64290L38X830 core with 2*10.5 (0.5 mm dia) windings.
e) the Würth 6 hole ferrite (2.5 windings)

with best regards

Andreas

[dietert1]

Today i found this (german) video on EMI measures with power supplies.


And  how these elements look like in the 1 to 2 GHz region:


Regards, Dieter

[Andreas]

NanoVNA together with NanoVNA-Saver as PC freeware. The firmware also supports using it as a signal generator.

Hello,

some questions to this?
How are the commands to get a sine wave out of the NanoVNA2?
At least there is no option on NanoVna-Saver to switch from rectangle to sine.

with the supplied PC-SW I always get only pulsed outputs and never a CW signal.
Is this possible with some commands?

Can I also do 1 kHz 80% AM modulation with the NanoVNA2?

with best regards

Andreas

[dietert1]

Can confirm this. Our nanoVNA V2.2 at 100 MHz to 100 MHz produces the expected peak at 100 MHz and about -10 dBm minus cable loss and a second peak at 300 MHz about 10 dB below, so it's probably a rectangular output signal. Don't know whether it's a problem though. As far as i understand the two receiver channels in a VNA are selective and even phase sensitive.

Regards, Dieter

[Andreas]

Don't know whether it's a problem though.

Hello,

I think its no problem for the usage as VNA. But if I want to use the signal generator for EMI measurements I would like to have a sine wave since the "receiver" in this case is not frequency selective.

with best regards

Andreas

[Andreas]

Hello,

after having added the 100nF capacitor between Output Ground (connector) and Power Ground (Buffer OP-Amp) I have no longer a significant difference between the ADA4522-1 and LTC2057 as buffer. (LTZ#3 and LTZ#7 shown as example).
This capacitor gives a improvement of about factor 6 (ADA4522) - 10 (LTC2057) against only having the output capacitor.

So the assumption that the EMI hardened ADA4522-1 is much better than the LTC2057 was only due to a "wiring fault" in my 2 layer layout with a long route from output connector to the buffer ground.  (so in this case it is really "better").

Attached a table with measurement results of all my buffered LTZs after adding the "ground shorting" capacitor.
So the "typical" result is a deviation of 60-90 ppm when 60 MHz and 5Vss at the unloaded function generator output is applied to the output of the LTZ via coupling network.

On LTZ#9 where the deviation is 213 ppm I still do not have all possible EMI capacitors populated on the PCB.
I will have to examine which of the missing capacitors contributes to the missing factor 2-3.

with best regards

Andreas

[kleiner Rainer]

Andreas,

the output cap soldered to the banana jacks looks like a class 2 MLCC in disguise to me - 100nF is in that range of size. Be careful - most class 2 ceramics are piezoelectric: if you tap them, they give you a voltage impulse, nothing you want in a precision reference. Better use a class 1 ceramic like C0G/NP0 or a film cap. They are recommended for that application. Bob Pease wrote about this problem in "Troubleshooting Analog Circuits", pages 45 and 111.

In my opinion, the board could use more ground pour around the opamp. And dont get me started about ESD...

Greetings,

Rainer

[kleiner Rainer]

Andreas,

sorry, I overlooked your question regarding the nano VNA. Here my answers:

- if you need a CWsignal, open the menu on the VNA, go to "Stimulus", then "CW freq", then enter the frequency you want. This I meant with "firmware". I use an older version of nanoVNA Saver, that cannot do it - will have to check the newer versions.
- yes, it is a square wave generator. The receiver part is tuned and only looks at the frequency it wants to measure in amplitude and phase, it ignores the harmonics. Think SDR.
- no, you cannot amplitude modulate it - at least not without external help (PIN diode modulator?).

Greetings,

Rainer

[Andreas]

the output cap soldered to the banana jacks looks like a class 2 MLCC in disguise to me - 100nF is in that range of size.

Hello,

of course I know that. On the first 2 samples I just added a capacitor for testing. (and "forgot" to replace it).
On the other side: the output is rather low impedant so the effect of a parallel capacitor should be low.
I think I should make some tapping tests before I replace them on LTZ#3 and LTZ#5.

with best regards

Andreas

[Andreas]

On LTZ#9 where the deviation is 213 ppm I still do not have all possible EMI capacitors populated on the PCB.
I will have to examine which of the missing capacitors contributes to the missing factor 2-3.

Hello,

attached the current schematic of LTZ#9 with the 213 ppm EMI influence.

The usually populated but missing in LTZ#9 components are circled in the diagram.
The dashed components are optional (not populated in my LTZs)
Additionally the 100nF output capacitor (GNDO) is added.
And the 100nF capacitor between the grounds (GNDs) which are connected to each other via R10 star ground at LTZ.

So the non populated parts are:

C12 (Zener base emitter)
C13 (Heater AMP feedback)
C14 (Heater AMP inputs)
C15 (Current AMP inputs)
R13 (startup-resistor)
R17 (Op-Amp stability with FET)
T2  (FET for low voltage operation)

so the next test will be with C12 added which is shurely the next sensitive point at the LTZ1000 after C11 which is already populated.

with best regards

Andreas

Edit: attached pictures of current population.
what is missing in the schematic is the capacitor from LTZ9 housing (internal substrate) to ground.

[Andreas]

Hello,

C12 100nF 0603 added (directly at the LTZ-pins 4+7 so in the center of the LTZ pins).

Result is reduction from 213 ppm deviation to 95 ppm @ 5Vss and 60 MHz.

I think there will be not much improvement by adding the further capacitors (but who knows).

with best regards

Andreas

[Andreas]

Hello,

next test with C13 + C14 populated.
As expected there is not much improvement from 95 ppm to 87 ppm.
So this could als be due to measurement variations and is no clear sign for improvement.

with best regards

Andreas

[Andreas]

and finally C15 ...

resulting in only 34 ppm deviation at 60 MHz and 5Vss.

This is much better than all my previous LTZs where the FET T2 is populated. (around 80 ppm)
So the question is: is there a sensitive point at T2?

Further tests will be necessary ...

with best regards

Andreas

[Andreas]

Hello,

further tests.
Added T2 in 2 steps. First R13 + R17 with no change as expected.
Then T2 the FET which results in lower loading of the LT1013 and giving more headroom for low battery voltage.

Result: increased deviation from 34ppm to 101ppm (see 02.01.2021)

Made several changes like Gate to Source, Gate to Drain and Drain Source capacitors in variations which gave a slight change but no real improvement. So obviously T2 is not the reason for the EMI-sensitivity but the changed loop gain of the current amplifier makes the EMI visible which is rectified at the base emitter junction of the LTZ.

Then I got the idea to add the "feed through" capacitors C16-C19 which connect the output signals and the power supply from the LTZ cirquit to the inner metal housing. Since in my drawer I was missing enough 4.7nF 0805 capacitors I used 10nF instead.
This lead to a dramatically improvement of EMI behaviour below 5 ppm change @ 60 MHz 5Vss.
The downside is that I saw sporadically oscillations ~34kHz on the buffered output immediately after switching on of the LTZ.
And a additional resonant frequency with 4 ppm at ~8MHz.
The sporadically oscillations where no longer seen after removing the Gate Source capacitor (10nF) from T2. But who knows ..

So I will do further checks (perhaps smaller "feed through" capacitors).

with best regards

Andreas

[Kleinstein]

The BF245C JFET is a rather small one and it can add quite some extra effective resistance and this reduces the loop gain for the LT1013. It may be better to use a larger fet, more like J112/PN4392  (the J111 / 2N4391 may have too high the treshold and would have to be checkes before) or similar.
Added capacitors at the FET aremainly loading the LT1013 and this usually reduces phase margin.

Edit: As shown in the plan C17 would also load the amplifer and can thus cause trouble. If filtering is needed, there should be some series resistance (e.g. > 50 ohms) beteeen the reference and C17.

[Andreas]

Hello,

Added capacitors at the FET aremainly loading the LT1013 and this usually reduces phase margin.

That is what I feared and thus removed the FET capacitor(s) first.

Edit: As shown in the plan C17 would also load the amplifer and can thus cause trouble. If filtering is needed, there should be some series resistance (e.g. > 50 ohms) beteeen the reference and C17.

I think that C8+R19 should do its job in this case.
C17 (in series to C16) is switched in parallel to C9 which is much larger.
But obviously on power up C18 and C19 are moving the shield to a part of the supply voltage.

So in the first test I removed C17 to fix eventually instabilities.
Hoping that C9 will have enough filtering for the 7V output.

The result is to 13 ppm increased sensitivity. This time at 9.7 MHz.
So removing the filter from the sensitive end (output) of the shielded area is obviously not a good idea.

with best regards

Andreas

[Andreas]

Hello,

Just an update over lengthy measurements:

You remember that after adding T2 (the BF245C FET) the sensitivity of LTZ#9 increased from   ~33 ppm to ~100 ppm @ 5Vss and 60 MHz generator output setting.

First results with 10nF to the inner shield where promising (~5 ppm) but showed a secondary resonance at lower frequencies.
I tried several combinations of C16-C19 including direct connection (0 R) of the shield to input or output ground or leaving the more critical capacitors unpopulated.
Some of them (the better ones) are listed in the attached table.
The best behaviour is obviously with all capacitors the same value and somewhere between 10nF and 100nF.

Now 2 questions arise for me:

- will the output voltage have increased (broadband) noise from the low noise voltage regulator (LTC1763) if the shield is not grounded and there is a capacitive path from the 14V supply to the 7V zener output? So I will have to do noise measurements to answer this question.

- up to now I had the mindset that the FET increases the loop stability since the overall amplification of the LT1013 is increased when the load current is reduced and the input to output voltage difference is decreased. Besides this the LT1013 is kept cooler by a external power stage. But do I have really improved stability of the output voltage by the FET or is it better to remove it (which would also decrease EMI sensitivity)? But I fear that my 6.5 digit measurement instruments are not stable enough to answer this question.

with best regards

Andreas

[Kleinstein]

The JFET as a source follower only adds current gain, but not voltage gain. So the loop stability in the 100 kHz range for the OP gets more like a little more difficult.
The loop for the OP is anyway quite complicated already with the original LTZ circuit and the added caps add to this.

For the more low frequency part (e.g. < 100 kHz) the parasitic capacitive coupling to the ouput should not be so bad, as the impedance is still relatively low. AFAIK the LT1013 should have like an emitter follower in the output stage, so more like a moderately low ouput impedance also at 100 kHz or open loop. The output from the BF245C should be also not so extemely low and essentially resistive (which is good as it can't resonate). So I would not expect much noise at the output from not having a shield. I would still expect that a connected shield should help with RF interference. It should not make much difference at 100 kHz, but maybe at 10 MHz or 1 GHz.

The point I am missing is dissipative elements, so to absorb Rf energy that is coupled in from the outside. A capacitor alone does not make a good filter. A capacitor and long trace as inductor makes a a low pass filtering, but will show a resonance at some frequency. Without some absorbers I see no way to avoid resonances - one can shift them around with extra caps, but they will allways the there at some point or the other.  A low ESR cap is nice to provide power to a circuit, but it is not a good way to stop RF signals, if there are high Q inductors (copper lines) at both ends.

[Andreas]

The JFET as a source follower only adds current gain, but not voltage gain.

Hello,

see not only the single tree but the whole wood:
every OP-Amp suffers from voltage gain when he is loaded with current.
So why should a FET not help to support the poor LT1013?

regards

Andreas

[Andreas]

Hello,

I did another test with a common mode choke.
This time a Wuerth SL2 744227 with 51uH.
This one is better when we regard only the 20 MHz point.

And as comparison the already known EPCOS B82790-S513 with 51uH.
This time with additionally VNA measurements above 20 MHz.
This is showing clearly that the EPCOS is better over a wider frequency range well above 250 MHz.

with best regards

Andreas

[Andreas]

Hello,

first EMI measurement on a ADR1000A. (#02)
The result was rather low -13.6ppm@60MHz with the open loop voltage set to 5Vpp on the function generator.
So I feared a fault in the setup and made a repeated measurement on LTZ#7

The conditions are comparable to LTZ#7 (non-A version) with ADA4522 output buffer and same EMI capacitors populated (C14+C15 missing) on LTZ1047B PCB in aluminium housing.
The repeated measurement on LTZ#7 showed -116 ppm @ 60MHz so slightly higher than at the last measurement (-92.3 ppm).
So the setup is principally in order. (and within usual EMI tolerances).
Cirquit diagram of the PCB see here:
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg3395070/#msg3395070

So it seems actually that the ADR1000A is less sensitive to EMI than the LTZ1000.

with best regards

Andreas

[Andreas]

Hello,

further comparison to ADR1000A#1.
I have added a buffer to the zener output (the 10V output already was buffered) with a ADA4522.
And added some EMI coils + capacitors to the output. All within a aluminium housing.

As result:
no visible EMI influence on 6V6 and 10V output at 5Vss setpoint of the frequency generator.

With best regards

Andreas