PSRR Injector

[Jay_Diddy_B]

Hi,

I would like to share an idea for a PSRR injector. The common solution using only a MOSFET, a resistor and a coupling capacitor, results in an output that decreases with load current. The voltage drop is the gate source voltage of the MOSFET at the operating point.

This can be illustrated with this simple model:



Results



In addition to the nuisance of having to adjust the input voltage as the load current is changed, there can be high power dissipation in the MOSFET

There have been proposals to use the remote sense capability of the power supply to correct for this. However, most common bench supplies have limited ability typically around 1V.

Modification

At the expense of adding a little circuitry and a 9V battery, the average voltage across the MOSFET can be held constant. With the values shown the average voltage across the MOSFET is 1V.



The result is a constant output voltage 1.05V less than the input



The constant offset does have any detrimental impact on the transfer function:



Comments?

Regards,
Jay_Diddy_B

[16bitanalogue]

Can you please attach the LTSpice model?

I am happy with the 100Hz - 1MHz 'flat' performance, but my initial question what is the compliance voltage for the input before current sources start to become crushed?

[Jay_Diddy_B]

Can you please attach the LTSpice model?

I am happy with the 100Hz - 1MHz 'flat' performance, but my initial question what is the compliance voltage for the input before current sources start to become crushed?

See attached …

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi group,

I have continued to work on this idea. I have also being learning KiCAD at the same time.

Here is the hardware schematic:




I have used two kinds of banana jacks. The Keystone 575-4 and the Deltron 571 series. Only one type will be installed.

PCB Design





The board is 100mm x 100mm

Regards,
Jay_Diddy_B

[awallin]

are you trying to do the same things as shown here in fig13?
https://www.analog.com/media/en/technical-documentation/application-notes/an159fa.pdf

for the feedback amp, I guess it's a style choice between two op-amps and a bunch of transistors 

[T3sl4co1l]

I take it, the application is for wideband operation, so that directly coupling it (say, with a low impedance bias tee, possibly using a min-loss attenuator to match impedances) isn't practical?

The circuit also makes a reasonable discrete "variable zener".

Tim

[Jay_Diddy_B]

are you trying to do the same things as shown here in fig13?
https://www.analog.com/media/en/technical-documentation/application-notes/an159fa.pdf

for the feedback amp, I guess it's a style choice between two op-amps and a bunch of transistors 

Hi,

I have built and used the Jim Williams design that was published in the AN159 application note.

Here is a picture of my implementation:





One of the five channels got damaged, so it is running on ' 4 cylinders' instead of five.
I included a modification to add a current sink so that it can sink and source.
The original JW design could only source.

The output voltage of the JW  design is limited to about 12V. The supply rail to the LT1220 op-amp, the VBEs of the output transistors and the ability of the LT1220 to work nears the positive rail.

The JW design uses 2N3866 which are fairly special. I would not trust some of the parts on eBay.

In the JW design the op-amp is in the signal path. It is essential fast LDO with the ability to modulate the output.

The design that I presented here, the op-amp, made out of BJTs, is not in the signal path. The op-amp is used only hold the average voltage across the output device to 1V.

There is a commercial design made Picotest, that is just a MOSFET, a capacitor and resistor.

The design presented here is an improved version of that.

I will compare the two.

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi group,

I made the PCB this morning using my LPKF c60 circuit board plotter.





I will start the assembly …

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi group,

Here are photographs of the PCB assembly:






Initial Tests

The PSRR injector was placed between an HP6554A power supply and 6050A Electronic load.
A DMM was used to monitor the voltage across the injector.
The voltage across the injector was a constant 1V, independent of the load current.



I will do a frequency response measurement later.

Schematic - Rev 2

A few changes were made to the initial schematic:

R1A1 was added this allows the same board to be tested in the commercial configuration.
Decoupling capacitors were added to the input.



Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi,
I an opportunity to do some frequency response tests.

I used a Narda 50 20dB 20W attenuator as a load.

First sweeping with an HP3325B function generator. These are log sweeps.

10Hz to 10kHz Log Sweep



10kHz to 10MHz




These graphs represent 6 decades of frequency.

Measurement with the HP3577A VNA



The frequency response is nice and flat and just like the LTspice model.

Regards,

Jay_Diddy_B

[Jay_Diddy_B]

Hi,

I had an opportunity to test the PSRR injector today.

I used a power that I found on the website for the test.

https://electronicprojectsforfun.wordpress.com/injection-transformers/

Model

First a model to get an idea of what the PSRR should look like:



and the results:



The answer is about 58dB of attenuation at 10kHz.


Hardware Preparation

The power supply was fitted SMA connectors on the input and output. There are some very small signals in this test.



Here is the complete test setup with the PSRR injector:



Measured Result




The measured result is 55dB of attenuation at 10kHz.

The measured results agree with the LTspice model.

Regards,
Jay_Diddy_B

[16bitanalogue]

are you trying to do the same things as shown here in fig13?
https://www.analog.com/media/en/technical-documentation/application-notes/an159fa.pdf

for the feedback amp, I guess it's a style choice between two op-amps and a bunch of transistors 

Hi,

I have built and used the Jim Williams design that was published in the AN159 application note.

Here is a picture of my implementation:



(Attachment Link)

One of the five channels got damaged, so it is running on ' 4 cylinders' instead of five.
I included a modification to add a current sink so that it can sink and source.
The original JW design could only source.

The output voltage of the JW  design is limited to about 12V. The supply rail to the LT1220 op-amp, the VBEs of the output transistors and the ability of the LT1220 to work nears the positive rail.

The JW design uses 2N3866 which are fairly special. I would not trust some of the parts on eBay.

In the JW design the op-amp is in the signal path. It is essential fast LDO with the ability to modulate the output.

The design that I presented here, the op-amp, made out of BJTs, is not in the signal path. The op-amp is used only hold the average voltage across the output device to 1V.

There is a commercial design made Picotest, that is just a MOSFET, a capacitor and resistor.

The design presented here is an improved version of that.

I will compare the two.

Regards,
Jay_Diddy_B

Had to check the box for notify.

Out of curiosity, the voltage across the FET remains 1V across load. Do we still not have an issue of voltage drop from the PSRR injector OUTPUT to the DUT input?

If you are connecting the source of the nMOS to the DUT through banana cables then you will still see a I*R drop. Maybe not big deal for low currents and particular measurements, but let's say for efficiency, line reg, and load reg. A 100mV drop across those cables can lead to a couple of % reduction in efficiency. Line and Load regulation will also be dinged.

[Jay_Diddy_B]

Had to check the box for notify.

Out of curiosity, the voltage across the FET remains 1V across load. Do we still not have an issue of voltage drop from the PSRR injector OUTPUT to the DUT input?

If you are connecting the source of the nMOS to the DUT through banana cables then you will still see a I*R drop. Maybe not big deal for low currents and particular measurements, but let's say for efficiency, line reg, and load reg. A 100mV drop across those cables can lead to a couple of % reduction in efficiency. Line and Load regulation will also be dinged.

16Bitanalogue,

Yes there will be a small IR drop between the PSRR injector and the power supply being tested.

The PSRR injector is only used when doing PSRR measurements. It is removed when measuring line regulation, load regulation, efficiency etc.

It is possible to measure PSRR versus load regulation or PSRR versus the input-output differential voltage.

You should have very short leads between the PSRR injector and the power supply being tested.

The bench power supply that I use most of the time is an HP6554A 60V 9A linear supply. I use its remote sense leads to compensate for voltage drops. I would connect the sense leads to the input of the PSRR injector.

Regards,
Jay_Diddy_B

[16bitanalogue]

The PSRR injector is only used when doing PSRR measurements. It is removed when measuring line regulation, load regulation, efficiency etc.

You are correct, I woke up in the middle of the night thinking the same thing. What does that say about me? 

A really good connection, short as possible, one could design the injector and DUT board with terminal connections and connect them right on top of one another. Keystone 7006/7 or the 575-4 for a less expensive option from the other thread.

[Jay_Diddy_B]

Hi,

I have allowed for two kinds of 4mm banana jacks.

I have used the Keystone 575-4 many times.

I have ordered some of these from eBay:



I haven't used these before.

Regards,
Jay_Diddy_B