ESR Meter Adapter Design and Construction

[mariush]

You can always add a dc-dc converter in SEPIC configuration to get whatever voltage you want from 1.2-6v which ranges from one rechargeable aa battery to a crappy phone usb charger.

If needed, there are LDOs which can smooth out the dc output and at those low currents, they only need a few mV above the output voltage to give good output.
Adds $1-2 to the BOM.

A MC34063 in inverter mode to get -2.5v is also just about 20-40 cents added to the BOM, if voltage divider isn't "cool"

[qno]

Nice design.

I always wondered what would happen to the ESR when an Electrolytic Capacitor is biased with some voltage.


The insulation layer between the plates is formed by the voltage between the plates.

[Jay_Diddy_B]

Hi,

Here is a modified concept using some ideas from the forum. It uses a 555 for the oscillator to generate a 50% square wave. The oscillator is also used to generate a negative rail for the op-amp. The positive supply comes directly from the 9V battery. The op-amp is only working at DC in this design. You can use a OP07 with an offset pot to implement the zero adjustment.

In the original design one of the goals was to use single sided construction, so that the PCB can be the front panel.



I have attached the LTspice model

Jay_Diddy_B

[oldway]

Bipolar version of 555 do'nt work very well at 5V. (output not going to rail nor ground voltage)
Output voltage will not be 50%.
Better to use a Cmos version of 555.
For generating the negative voltage, use a pair of bipolar transitors BC847/BC857 driven bij 555 for higher current.
Collector of BC847 can be feed with 9V so it does not overload the 5V voltage regulator.
Bases from the 2 transistors connected together and with pin3 from 555.
Emitters connected together and with C8.
Collector BC857 connected to ground.

[blackdog]

Hi Jay_Diddy_B,

Can you please make some corrections to your schematic.
It is saying 4016 and 4053 as a fet switch...
Can you please put the pin numbers in the schematic for the 4016 or 4053?

Your output from the 555 wil not be RR, the components for the negatif supply rail wil load the 555 to mutch.
Use a transistor/fet buffer.
The 555 oscilator is NOT 50%/50%, it nead some extra components for that, take a look @ the data sheet for that.

Keep up the good work! :-)

Kind regarts,
Blackdog

[oldway]

4066B is far better than 4016B  (much lower ON resistance) but both can't be use here without inverting 555 output.

Output of cmos version of 555 with 5V power supply:
IOH = -1mA     VOH = 4.8V
IOL = 5mA      VOL = 0.13V
Good enough to succeed nearby 50% / 50%
If you need 50% with precision, you have to run the oscilator at 200Khz and use it as clock of a D flip flop. (1/2 4013B)

[Jay_Diddy_B]

Hi Jay_Diddy_B,

Can you please make some corrections to your schematic.
It is saying 4016 and 4053 as a fet switch...
Can you please put the pin numbers in the schematic for the 4016 or 4053?

Your output from the 555 wil not be RR, the components for the negatif supply rail wil load the 555 to mutch.
Use a transistor/fet buffer.
The 555 oscilator is NOT 50%/50%, it nead some extra components for that, take a look @ the data sheet for that.

Keep up the good work! :-)

Kind regarts,
Blackdog

The schematic is an LTspice model of the real schematic so the pin numbers are not shown. There is a real schematic in this thread with the pin numbers.

https://www.eevblog.com/forum/projects/esr-meter-adapter-design-and-construction/msg341177/#msg341177

The switch is the 74HC4053. The switches in the HC4053 are around 140 Ohms. They are in series with a resistor of 470 + 1600 + 200 ohms, the switch only make up 6% of the total resistance.

The advantage of the 74HC4053, over using 4016 or 4066 switches is that I don't need an oscillator with complementary outputs. The inverter is inside the 74HC4053.

The 74HC4053 is considerably faster than the CD4053, but is a lower voltage part.

The 555 is not configured in the normal way. The output is used to drive the RC network. In this configuration to the duty cycle is very close to 50%. It is supposed to be 50% in this configuration.

The 555 output does not need to be rail-to-rail. The analog switches in the source make it rail-to-rail. The CMOS version of the 555 would be better, but there is no LTsice model for the 555C.

@oldway

The 2F divided by 2 with a F/F will give an exact 50% duty-cycle.  I would probably use that technique if was building a bench meter. But I was building a small adpter.

The LTC6990 timer IC can also be used to provide an exact 50% duty cycle. There is a Frequency divider inside this chip. It has a single output, so an inverter is required.

Total power consumption of the circuit is less than 10mA, so there is no need to added any transistor buffers.

I did post the LTspice model if you want to play with this design.

There are many ways to do this, some is personal choice, some decisions are driven by the parts that you have. The fun is giving building your circuits and testing them.




Regards,

Jay_Diddy_B

[oldway]

Total power consumption of the circuit is less than 10mA, so there is no need to added any transistor buffers.

Output current of the Cmos version of 555 is very low (only 10mA source, 100 mA sink), this is the reason why i recomand to add a transistors buffer.
Bipolar 555 can source and sink 200 mA and buffer is not necessary.

[Jay_Diddy_B]

Hi group,
Here are some accuracy measurements made from the prototype.

The ESR Adapter and a HP 3457A in the 4-wire mode were used to measure a General Radio 1433-W decade resistor. This decade resistor can be set in 0.01 Ohm increments.

The graphs show the HP3457A on the X axis and the ESR adapter on the Y axis.





The equation for the graph is displayed in the y=mx + C

No attempt was made to adjust the ESR Adapter. You can see that the ESR Adapter is reading 4% high with 0.1 Ohm offset. I am not sure where the 0.1 Ohm offset comes from. This was after shorting the leads and using the REL feature on my Fluke 189.  Later I will measure some SMD resistors.


Jay_Diddy_B

[c4757p]

The 555 is not configured in the normal way. The output is used to drive the RC network. In this configuration to the duty cycle is very close to 50%. It is supposed to be 50% in this configuration.

I came to this conclusion about that configuration as well, but I tried it a few times and have found that to be very far from the case. The last time I used that method I had a duty cycle around 40%! Worked fine in SPICE though. Are you sure it's working properly?

[Jay_Diddy_B]

I came to this conclusion about that configuration as well, but I tried it a few times and have found that to be very far from the case. The last time I used that method I had a duty cycle around 40%! Worked fine in SPICE though. Are you sure it's working properly?

@c4757p

I have not built the 555 version of the ESR adapter. I was simulating different configurations of the circuit proposed in the forum.

Using the LTC6990 will produce a very accurate 50% 100kHz oscillator.

What I realised afterwards is that the symmetry in this configuration is a function of the loading on the output pin. If the output pin is unable to swing to the rail or ground, this will increase the time spent in the that state.
The timing capacitor is being charged and discharged between 1/3 and 2/3rds of the supply voltage.

I am very pleased with the version of the circuit using the LTC6241 dual op-amp.

The accuracy measurements speak for themselves. A trimming resistor in series with the 1600 Ohm resistor in the source can be used to adjust the gain. In my unit the gain needs to be lowered by 4%.

I think the 0.1 Offset that I measured with the General Radio Decade resistor, is probably going to be something subtle like skin effect, the different between the DC resistance (HP3457A) and the AC resistance measured with the ESR Adapter.

Later I will be exploring the source of the offset.

But since the main purpose of the ESR adapter is to find dried out caps, the performance already exceeds the goals.

Thank you for your interest.

Jay_Diddy_B

[oldway]

The 555 is not configured in the normal way. The output is used to drive the RC network. In this configuration to the duty cycle is very close to 50%. It is supposed to be 50% in this configuration.

I came to this conclusion about that configuration as well, but I tried it a few times and have found that to be very far from the case. The last time I used that method I had a duty cycle around 40%! Worked fine in SPICE though. Are you sure it's working properly?

This happen because output voltage (high) of a bipolar 555 is only 3.3 V with 5V power supply.
With Cmos version, output voltage is 4.8 V and duty cycle is then very close to 50%.

[kripton2035]

No attempt was made to adjust the ESR Adapter. You can see that the ESR Adapter is reading 4% high with 0.1 Ohm offset. I am not sure where the 0.1 Ohm offset comes from. This was after shorting the leads and using the REL feature on my Fluke 189.  Later I will measure some SMD resistors.
Jay_Diddy_B

can you do the same measurements at 10KHz ?

[grenert]

Could part of the 4% high reading be related to the fact you are feeding a square wave into the cap, rather than a sine wave?  You have high frequency harmonics accompanying your fundamental.  So when you think you're measuring ESR at one frequency, the effective measurement is actually at something higher?

Obviously, 4% is already very accurate!  This is more of a curiosity question. 
Thanks for sharing the design with us!   

[PA4TIM]

When I designed my ESR meter the biggest problem was getting a 50% dutycycle. I solved that by making an in frequency adjustable comparator oscillator (20kHz to 200kHz) followed by a 4013 devide by 2 to control the action of the 4066 ports.

I used a very easy way to zero the reading. The minus probe of the multimeter (or panel meter) is taken from a Voltage divider between 5V and ground. You measure the offset and then calculate the fixed resistor and potentiometer so that the wiper sees the offset voltage as it is set in mid position.

What is the lowest capacitance you still can measure ESR ?
 
There are a few reasons I made it adjustable.
-the first is because datasheets give ESR for frequencies between 100Hz and 10 kHz as a function of D (and D at 1 kHz gives after calculation,  ESR at 1 kHz and that is not the same at 100 kHz)
-At 100 kHz they give Impedance.
-Big caps in classic lineair supplies can score high on ESR because at 100 kHz skineffect and ESL polute the measurement and because they work around 50-100 Hz it is useless to measure them at 100 kHz
- and because almost nobody does it. It is just adding a potentiometer so it is not much work. (and it is a good test. If ESR goes sky high at 10 kHz the meter is measuring impedance instead of ESR)

[dannyf]

Later I will be exploring the source of the offset.

Depending on your set-up and capacitors used, there could be many potential reasons.

The esr meter is no different than the ones proposed here earlier: it measures the ac impedance on the dut. The difference is that a synchronous detector was used so you could measure low (ac) impedance - thus low ESR. The diode detectors used in other esr meters would not be able to do that due to the Vfwd threshold.

Another approach would be to use a precision rectifier but that requires considerably higher slew rate.

A synchronous detector in my view is a nice solution for its low/zero threshold and linearity.

[Jay_Diddy_B]

Could part of the 4% high reading be related to the fact you are feeding a square wave into the cap, rather than a sine wave?  You have high frequency harmonics accompanying your fundamental.  So when you think you're measuring ESR at one frequency, the effective measurement is actually at something higher?

Obviously, 4% is already very accurate!  This is more of a curiosity question. 
Thanks for sharing the design with us!   

In the source the following resistors add up:

470 ohm
140 ohm 74HC4053 switch
1600 ohm resistor
200 ohm resistor
=============
2410 ohms (nominal)

With +/- 2.5V supply this should put 207.4 uA pk-pk through the device under test.

with a 10 ohm resistor this becomes 2.074mV

This is multiplied by the gain of the amplifier stage:

The amplifier has a nominal gain of 1.8 MEG / 33.2K = 54

This gets us to 2.074 x 54 = 112 mV

So the design should read high.

5% resistors were used in parts of the prototype, the 5V regulator tolerance also contributes.

So some pretty simple things can explain the 4% error. Some form of adjustment is needed.

Jay_Diddy_B

[Jay_Diddy_B]

Hi,
I have done a little investigation.

The General Radio 1433-W decade box is a calibration laboratory grade decade resistance box. It is made with non-inductive resistors. I measured the resistance box on my HP 4274A LCR meter. I found that decade box a residual with the knobs set to zero is around 30 mOhms. I also discovered that the decade resistor had an inductance of 700nH.

700nH has an impedance of 0.45 ohms at 100 kHz.

So the ESR adapter has dealt with some of the reactive component.

The challenge here is that the ESR adapter uses square waves for the source. The effect of the inductance is not straight forward.

Resistance Standard







I decided to build a resistance standard using 1% surface mount resistors. I have two terminals at each end of the resistors to allow 4-wire measurements.

The measured inductance of these resistors is between 7-20nH


With this resistor I got good measurements.

The synchronous detector is helping with the accuracy, but for it to work properly it needs a sine wave source.

Jay_Diddy_B

[dannyf]

the ESR adapter uses square waves for the source.

You can run it through a filter to produce a fair good sine wave. R/C filters would be sufficient since you use opamps to amplify the signal.

[PA4TIM]

How you are gonna measure ESR with a sinewave ?

[oldway]

@Jay_Diddy_B: I think you are going in the wrong way.
This is a simple, cheap and clever project: it's not intented to compete with 5000 US$ instruments.
Your ESR meter will be very useful for repair, not for quality control and we don't need high precision.
I have made some modifications for my own use :
I will not use the four terminals option, i don't need this.
But i provided a switch for 10Khz/100Khz and also a trimpot for calibration so it is not necessary to use 1% precision resistors.
No worry about offset, i will use the relative function of my Fluke 87 IV.
Nb: schematic only for explaining my ideas, it is not been tested.

[Jay_Diddy_B]

Hi,
Here is the concept for measuring the ESR with a sine wave. I have attached the LTspice model for people who want to try the model.



The square wave source has been replaced by a sine wave generator. In addition to the sine wave you also need a square wave that is in phase with the sine wave to operate the analog switches in the detector. The AD9838 would be handy, because it has a sign output, which indicates if the output is positive or negative.

The AD9839 would require the addition of a micro to set the frequency etc..

I am going to try and do an analog design.

This modification improves the performance for small value capacitors and greatly improves the performance if there is inductance in series with the capacitor.



The design retains the protection feature develop earlier.

Again, these improvements are largely academic. The original design will sort the good caps from the bad ones.

Jay_Diddy_B

[Jay_Diddy_B]

@Jay_Diddy_B: I think you are going in the wrong way.
This is a simple, cheap and clever project: it's not intented to compete with 5000 US$ instruments.
Your ESR meter will be very useful for repair, not for quality control and we don't need high precision.
I have made some modifications for my own use :
I will not use the four terminals option, i don't need this.
But i provided a switch for 10Khz/100Khz and also a trimpot for calibration so it is not necessary to use 1% precision resistors.
No worry about offset, i will use the relative function of my Fluke 87 IV.
Nb: schematic only for explaining my ideas, it is not been tested.

I had a look at your schematic. The only concern that I have is the Vos of the op-amp in the U1-A position. The Vos is multiply the gain.

The LM324 is 2mV typical, 7mV max. Multiplied by the gain this becomes 110mV typical  385mV max.

The LM324A is better at 3mV typical, 3mV max.

Jay_Diddy_B

[Jay_Diddy_B]

You can spend countless hours and dollars trying to build a quality LCR meter. Thats why they cost several thousand dollars. Given the avalibility of low cost DAC,DDS chips and fast 32b micro's to process the signals (phase /amplitude) it would be intresting to do if you had the time.

I know the original design is good for the intended purpose of finding bad capacitors.

Is there any interest in improving the performance so the design will work with very low ESR parts?

Is 20 Ohms full scale the right value or would 5 or 10 Ohms be better?

How important is it that the board is 5x10 or 10x10 cm so that it can be made at itead?

Thank you for your interest.

Jay_Diddy_B

[oldway]

I had a look at your schematic. The only concern that I have is the Vos of the op-amp in the U1-A position. The Vos is multiply the gain.

The LM324 is 2mV typical, 7mV max. Multiplied by the gain this becomes 110mV typical  385mV max.

The LM324A is better at 3mV typical, 3mV max.

I have allready answered to this:

No worry about offset, i will use the relative function of my Fluke 87 IV.

This is not a mass product project, you can also select the best LM324 (lowest offset) if you want.

Is there any interest in improving the performance so the design will work with very low ESR parts?

I don't have interest.
20 Ohms full scale seems to be a good option.
I will do my own board.

NB: good quality 9V battery are expensive, for this reason, it's very important is to reduce power consumption.
I took great care of this:
- TLC555 (Cmos version of 555) has a very low power consumption of 1mW at 5V. C8 must be choosen of low value to minimize power dissipation in R7 and R21.
- 4013B has a low consumption and no output load.
- TL431 is working with min. current for regulation. (1mA)
- power supply is low loss
- Led is in serie without loss.
- And, of course, LM324 is a low power quad OP with very low supply current drain.