Problem with MCP1501 (VREF device)

[fc3000]

H all.
I am using a Microchip MCP1501 to create a Vref  0f 4.096V to two ADCs.

My scope shows a 100mV overlay on the output. I cannot figure out what is causing it.

I have read the Datasheet and i am satisfied that i have followed it correctly.
I have also placed a 10uFv cap at the output  without any luck.
I have powered the device two different power sources with no changes.

The datasheet says that the device has an internal chopper amplifier.
If this the source of problem?

I have attached the circuit diagram and the scope output.

Regards.

[PartialDischarge]

Add a resistor in series to the output, before the probe the capacitor or anything

[fc3000]

Add a resistor in series to the output, before the probe the capacitor or anything

Surely that will drop the voltage!

regards

[iMo]

The max output load capacitance is 10nF without a series resistor, the datasheet says (5.1.2)..
Otherwise it will oscillate (as you may see on your oscope)..
Add 10-50ohm, for example, in series with your 10uF. It will not drop your voltage.
Or do use less than 10nF.
Read the DS carefully..

[MasterT]

From my experience, mcp1501 is the worst reference I've ever seen. Oscillation w/o any reason, even 300pF could make it unstable, and in combination with enourmously high noise level  that can't be filtered since IC doesn'y tolerate caps at the output.
Use TL431.

[fc3000]

The max output load capacitance is 10nF without a series resistor, the datasheet says (5.1.2)..
Otherwise it will oscillate (as you may see on your oscope)..
Add 10-50ohm, for example, in series with your 10uF. It will not drop your voltage.
Or do use less than 10nF.
Read the DS carefully..

Thank you for the reply.
I have followed your advice, I have correct q 56R resistor and a 10uF cap in series, between the output and GND.
Nothing has changed. I have also tried with a 1uF cap, same result.

I have also probed at the junction between the cap and he resistor, and i get a much reduced oscillation!
I am confused. I need more advice.

Regards

[eleguy]

I think I had similar problems at past (changed to other VREFs already long ago). But... In my case it did help to have some load resistance and capacitance on out / feedback. Having too much capacitance is a bad idea. If that is the case then buffer or isolate with series resistor. Then to the next question, how is your test setup made? Is stuff on the PCB or breadboard? On the breadboard you have to solder this load (and input) cap next to the pins almost. Do not use through-hole components as "main" input / output cap. That helps a lot. I have seen same behavior with some TI VREFs as well so having a low PDN impedance is a good practice overall.

Hope you get it solved...

[iMo]

Do it step by step - start with none output capacitance, and then increase it until it starts to oscillate.
Try a smaller output cap, like 1nF, you may try to add the series resistor as well to see the change.
Also it depends on the wiring - a star grounding config, a special pcb layout, etc.
Your oscope probe may add to the instability as well. Try to put a, say, 100ohm resistor in series with probe's tip when the oscope probe has been attached (with the shortest grounding leads if possible).

[fc3000]

Hi all. UPDATE.
I have found the cause of the problem.

The MCP1501 device is connected to two DACs and two ADCs.
The connection is via their +VCC to provvide the reference voltage.
Each +VCC has a decoupling capacitor. Overall capacitance is 3uF
I thought the caps are placing a loading too high, and that causes the oscillations.
So I disconnected  those devices VCC from the PCB and the oscillations have disappeared!

Do you believe this is the cause or is this a red herring?

If this is the cause then an OPAMP buffer should do the trick.

What do you think?
Regards.

[magic]

If the output (not input) of MCP1501 was connected with VCC of other chips and each chip had 1μF to ground then yes, you had 3μF on the output of MCP1501, which is more than permitted.

Simple solution: use other reference. Any strong reason to prefer this one?

Weird solution: try to compensate it. Run a direct capacitor from OUT to FEEDBACK and connect both pins to the load (and its capacitors) with resistors. This is utterly undocumented, no warranty and so on, but it ought to work, I think. You will certainly need to see how much current the FEEDBACK pin sinks and size the series resistor accordingly, to avoid excessive error.

I hope to convince you with this to try the simple solution instead...


edit
I actually have this chip, but in 3.3V. I played with it, it's just a reference plus an opamp, connected together in the most obvious way. There is a bizarrely incomprehensible "block diagram" in the datasheet, it makes no sense to me, I think it can be ignored. FEEDBACK drives an internal feedback divider of the opamp and sinks not entirely negligible current. I also tried putting various things in the feedback loop of the opamp (NPN emitter follower, even LM317 IIRC) and it seemed to work as expected, although I'm not 100% sure if I checked for oscillation now that I think about it.

I wanted to use this as a single chip (+ external transistor) constant current regulator, but the project never progressed past a crude breadboard proof of concept.

[SiliconWizard]

Yes. But as an additional thought, you may want to use the voltage reference strictly to drive the Vref input of your DACs/ADCs rather than power these entirely (which I assume with your connection to "Vcc").
That'll allow you to separate both, will be cleaner and will allow to use much smaller bypass caps at the Vref inputs, while removing bypass caps at the Vcc pins is not a good solution IMHO.

Unless maybe your ADCs and DACs don't have a separate Vref input, in which case, either change them or change the reference voltage chip.

[fc3000]

Yes. But as an additional thought, you may want to use the voltage reference strictly to drive the Vref input of your DACs/ADCs rather than power these entirely (which I assume with your connection to "Vcc").
That'll allow you to separate both, will be cleaner and will allow to use much smaller bypass caps at the Vref inputs, while removing bypass caps at the Vcc pins is not a good solution IMHO.

Unless maybe your ADCs and DACs don't have a separate Vref input, in which case, either change them or change the reference voltage chip.

Yes the DACs use the VCC as voltage reference.
Unless i redesign  the PCB i am stuck with this choice.
This is what happens when you rush and do not read the datasheet properly. 

I think i will use a tiny PCB with an OPAMP and then add it to the circuit. It will be  a cut-and-paste job.
If this does not work i will have to redesign the PCB.

Regards
 

[PartialDischarge]

I suspected you had something else at the output that's why I told you "Add a resistor in series to the output, before the probe the capacitor or anything"
You should have used an ADR360, they are made just for that, to power ADCs.

I think i will use a tiny PCB with an OPAMP and then add it to the circuit. It will be  a cut-and-paste job.

Don't add more mistakes to the board, get rid of the MCP and bug solder a AD364 in place.

[eleguy]

The MCP1501 device is connected to two DACs and two ADCs.

Which parts exactly you are using as a DAC and ADC?

[fc3000]

The MCP1501 device is connected to two DACs and two ADCs.

Which parts exactly you are using as a DAC and ADC?

Hi.
I am using MCP4725 and MCP3201,

Regards

[PartialDischarge]

I am using MCP4725 and MCP3201,

Regards

Those are two completely different parts, the DAC is not a precision part and that's why it doesn't have a separate Vref input. It should be used in closed-loop calibration systems, or open loop calibration with some user feedback, so that its offset and non-linearities don't really matter.
The ADC does have it because its specs are better.

[iMo]

Powering ADC/DAC chips off a voltage reference could be tricky. The power consumption of the entire chip may vary significantly during its operation and that may overload the load regulation capability of the voltage reference.
Even when you power the Vreference input of an ADC/DAC only (if such input is available) the current pulled off the reference is changing based on the conversion process.

[fc3000]

Powering ADC/DAC chips off a voltage reference could be tricky. The power consumption of the entire chip may vary significantly during its operation and that may overload the load regulation capability of the voltage reference.
Even when you power the Vreference input of an ADC/DAC only (if such input is available) the current pulled off the reference is changing based on the conversion process.

Hi.
On that I have to disagree a bit.
The power consumption of the devices is low.
The problem is that the devices require large caps thus heavily loading the MCP1501.  And that was my mistake, not taking that into consideration.
Regards

[magic]

One more thing you could try is putting some inductor (with not too high DC resistance to avoid voltage drop) between the reference and the chips with their capacitors.