Simple Voltage Reference

[Sam__]

I'm building a little voltage reference circuit using LTC6655's. I've already build the first board and made some measurements and was wondering if you guys could help me optimise the design.

The circuit design is very simple. Just a voltage regulator section followed by the chips in parallel. Output cap values came from the datasheet.



It's more the PCB design that needs optimising to minimise voltage drops. I think.

The first design was this, I made it yesterday and noticed a voltage drops (~30uV) between the input ground and the ground on the output caps. I would like to get this as low as possible. Along with reducing the drop between the output of the chip and the output header.

I seem to have managed to overwrite the pcb design file so all I have is a picture of it, sorry.



So I've made a few changes and I'm going to see if it's helped at all. I added a copper pour and thickened the output traces.

Here's the second PCB design.



Two questions:

Were those the right kind of changes?

AND

Is there anything else I can do to minimise voltage drops to the outputs and aid the stability of the reference?

If you need any more information then I will try my best to give it.

Thanks in advance.

[Bored@Work]

It looks as if you didn't give any real consideration to the application information section in the data sheet. Linear did put this and other information in the datasheet for a reason, not just for decoration purposes. There is also a reason the IC has a sense pin, again it doesn't happen to be there for decoration purposes.

Read the datasheet. Read it carefully. And then read it a third time. And finally follow the advice in the datasheet. All of it.

[Sam__]

It looks as if you didn't give any real consideration to the application information section in the data sheet. Linear did put this and other information in the datasheet for a reason, not just for decoration purposes. There is also a reason the IC has a sense pin, again it doesn't happen to be there for decoration purposes.

Read the datasheet. Read it carefully. And then read it a third time. And finally follow the advice in the datasheet. All of it.

That's a good shout, I'd forgotten about the sense pin. I'll get on that.

EDIT: There's also good information about the output caps. This is very useful.

[tszaboo]

But why do you need 8 of them?

[Sam__]

The chip comes in seven different values 1.25, 2.048, 2.5, 3, 3.3, 4.048, 5. So I thought it would be useful to have them all on one board so I could easily use any of them.

[Sam__]

After modifying the circuit to include input and output caps, moving the sense wire to the output, and finally adding a ground layer on the back. I made the board. It works! All the outputs are better than the +- 0.025% spec of the chip. Perfect!

Pics coming soon.

[Sam__]

The third revision of the board. Still waiting on two of the references to arrived. Currently on is: 1.25, 20.48, 2.5, 3, 4.096

Front:



Back:




Voltage readings to come.

[Andreas]

Hello,

it would be better to have a separate GND pin at each output voltage.
Routed directly from the GND-Pin of the LTC6655 to the output without any other connection on that side.
Otherwise the GND-Pin will be shifted by several tens of uV against the LTC which is farest away. The LTC6655 consume a relative high current of around 5mA each.

with best regards

Andreas

[Sam__]

Hello,

it would be better to have a separate GND pin at each output voltage.
Routed directly from the GND-Pin of the LTC6655 to the output without any other connection on that side.
Otherwise the GND-Pin will be shifted by several tens of uV against the LTC which is farest away. The LTC6655 consume a relative high current of around 5mA each.

with best regards

Andreas

I put vias to the ground side next to the ground of each chip to try and reduce that problem but I think your idea is a good one. It is something that I did think about, and I think I will do, if I make another revision of the board.


Here are some pictures of the results measured with my two best meters:





Readings:




As can be seen. The 1.25V seems to be the best of the lot. With the 2.5 coming inside the 0.025% spec of the chip. The others are out of spec by a bit as measured by the Datron, but  the Solartron puts them just in spec.

Could this be due to the voltage drop on the ground? Or another reason I may have overlooked?

[Andreas]

Could this be due to the voltage drop on the ground?
Or another reason I may have overlooked?

You can measure the ground voltage drop when putting a tip on / near the GND Pin of the individual reference.

The output voltage + datasheet tolerance is valid for a temperature of 25 deg C only.
Usually this is the chip temperature (measurement is done by a short pulse pulse during production).
The self heating of the LTC6655 is relatively high 5 degrees or more (depending on power supply voltage).

With best regards

Andreas

[Sam__]

Could this be due to the voltage drop on the ground?
Or another reason I may have overlooked?

You can measure the ground voltage drop when putting a tip on / near the GND Pin of the individual reference.

The output voltage + datasheet tolerance is valid for a temperature of 25 deg C only.
Usually this is the chip temperature (measurement is done by a short pulse pulse during production).
The self heating of the LTC6655 is relatively high 5 degrees or more (depending on power supply voltage).

With best regards

Andreas

My room is fairly cold at the moment. ~16C. I might try and warm it up to 20C and retake the readings. See if there's much change. It says the tempco is 2ppm/C.

I will also measure the ground voltage difference between the pins and measured ground. Thank you for your comments.

[mjrandle]

Not sure what voltage is powering the refs, but input voltage on schematic states 15 V.  I would look at minimising heat being put into board and maybe thermally isolate the regulator.


Regards,

Mike

[Sam__]

Not sure what voltage is powering the refs, but input voltage on schematic states 15 V.  I would look at minimising heat being put into board and maybe thermally isolate the regulator.


Regards,

Mike

Sorry I was lazy and didn't enter that value. I'm using a 13.8V linear power supply. Then ~6.2V comes from the regulator.  Good idea about taking the regulator off the board.

[Sam__]

Question: Would filtering the regulator output help at all?

I've viewed it on my (old analogue 20Mhz) scope and can see some 1mV p-p high frequency noise in the high hundreds of KHz to 1MHz region. Would this be effecting the output of the reference much/at all?

[Jay_Diddy_B]

Sam__,

Filtering the supply will not any harm. Do you have any batteries you can use? Either a 9V or 12V battery?

If it doesn't impact the references, it might effect the meters.

Regards,

Jay_Diddy_B

[Andreas]

Would this be effecting the output of the reference much/at all?

Definitely: yes.
If the spikes (e.g. from a switchmode supply) are rectified by the ESD protection diodes of any IC, they can be found as "offset voltage" again in the output measurement.

I would filter the LM317 at input + output with at least 100nF. Better a additional electrolytic capacitor (10uF at the output) and a larger at the input (depending on total current consumption).

Use a linear power supply. (I generally use rechargeable batteries for precision voltage standards).

Additionally filter the LT6655 with 100nF at each input and output (placed directly at the pins).

Without those measures I can measure the influence of the USB lines in my measurement setup with several uV shift if I lay my hand either on the battery pack or any interconnection line.

With best regards

Andreas

[Sam__]

Sam__,

Filtering the supply will not any harm. Do you have any batteries you can use? Either a 9V or 12V battery?

If it doesn't impact the references, it might effect the meters.

Regards,

Jay_Diddy_B

I could cobble together some AAs or 9V batteries but I'd really like to be able to run it from a power supply if I can. As long as it doesn't effect performance too much.

Would this be effecting the output of the reference much/at all?

Definitely: yes.
If the spikes (e.g. from a switchmode supply) are rectified by the ESD protection diodes of any IC, they can be found as "offset voltage" again in the output measurement.

I would filter the LM317 at input + output with at least 100nF. Better a additional electrolytic capacitor (10uF at the output) and a larger at the input (depending on total current consumption).

Use a linear power supply. (I generally use rechargeable batteries for precision voltage standards).

Additionally filter the LT6655 with 100nF at each input and output (placed directly at the pins).

Without those measures I can measure the influence of the USB lines in my measurement setup with several uV shift if I lay my hand either on the battery pack or any interconnection line.

With best regards

Andreas

I should have said before. I already have filter caps (3x 10uF) on the regulator output and 2x 10uF on the input.

I also have put 100nF on the input to each IC, right by the pins. And 10uF on the output, fairly close to the pins.

I was thinking maybe an LC or RC filter?? With a cut off frequency nice and low to get rid of that HF noise.

And to clarify I have a 13.8V Linear power supply powering the board. So there is no switching noise from an SMPS anywhere.

[Jay_Diddy_B]

Sam__

Just to clarify. I was proposing to use the batteries for a test. If the accuracy improves using the batteries, then using a cleaner supply is the way to go.

May be you can find an old style wall transformer, one that actually has a 50 Hz transformer inside as oppose to one that has a switching supply?

Jay_Diddy_B

[Sam__]

Sam__

Just to clarify. I was proposing to use the batteries for a test. If the accuracy improves using the batteries, then using a cleaner supply is the way to go.

May be you can find an old style wall transformer, one that actually has a 50 Hz transformer inside as oppose to one that has a switching supply?

Jay_Diddy_B

Ah yes. I will see if I can find enough batteries to do some tests and see if it improves performance.

The supply I am currently using is transformer based. I do also have a switching one, I may plug it into that just to see how much the noise effects it!

[Andreas]

And to clarify I have a 13.8V Linear power supply powering the board. So there is no switching noise from an SMPS anywhere.

And your computer runs on batteries?

The coupling capacity from primary to secondary side of your linear transformer is not zero.
So every mains line carries (more or less) some switching noise.

Precision instruments use often shielded transformers with a  (capacitive) shield between primary + secondary side.

With best regards

Andreas

[Sam__]

And to clarify I have a 13.8V Linear power supply powering the board. So there is no switching noise from an SMPS anywhere.

And your computer runs on batteries?

The coupling capacity from primary to secondary side of your linear transformer is not zero.
So every mains line carries (more or less) some switching noise.

Precision instruments use often shielded transformers with a  (capacitive) shield between primary + secondary side.

With best regards

Andreas

Oh! I didn't think of that. Good point. Very interesting and good to know.

Would I find a shielded transformer in an old 20Mhz scope by any chance? As I'm currently dismantling one . Are there any obvious visual clues that it's shielded?

[M. András]

very good idea, but please work on the soldering a bit its damn ugly atm. 2 lithium ion cells could power this reference board for months i guess as the power draw is minimal

[Sam__]

very good idea, but please work on the soldering a bit its damn ugly atm. 2 lithium ion cells could power this reference board for months i guess as the power draw is minimal

HAHA thanks. I do need to work on my soldering. Also tinning the ground plane makes it look pretty ugly anyway.

[M. András]

very good idea, but please work on the soldering a bit its damn ugly atm. 2 lithium ion cells could power this reference board for months i guess as the power draw is minimal

HAHA thanks. I do need to work on my soldering. Also tinning the ground plane makes it look pretty ugly anyway.

try for that purpose those chemical tin plating sultions,

[Andreas]

Would I find a shielded transformer in an old 20Mhz scope by any chance? As I'm currently dismantling one . Are there any obvious visual clues that it's shielded?

I do not think that a oscilloscope has a shielded transformer.
On shielded thransformers the shield is connected to one pin of the transformer. Without having connection to any other pin.

But without a proper "guarding and shielding" concept with two (outer + inner) isolated metal housings a shielded transformer will not give significant advantage.

With best regards

Andreas