REF70 family voltage reference

[Warhawk]

I just came here to say that TI has pre-released a new voltage reference family REF70.
By no means I am V&N guy but I am wondering how does it stand against the trusted competition?
Noise figures and initial accuracy are quite good. Long term stability is bit worse but not terrible.

https://www.ti.com/product/REF70?jktype=homepageproduct

PS: For those who know me - No, I am not trying to do business here. I am just a curious engineer 

[antintedo]

It's a direct competitor to LTC6655, basically a drop in replacement. In that regard it is very good, they achieved better TCR and don't offer A/B bins like LT. I hope the price will be competitive.

[RandallMcRee]

Analog Devices also seems to be releasing a series of new LCC package Vrefs. They seem to be getting better performance from their existing ADR45x0 series. For example there is to be a ADR4550D at 5V with 0.5ppm max tempco.

https://www.analog.com/en/parametricsearch/11010#/p4997=0.05|3&sort=4997,asc&p4995=5|10&p4007=300n|3u

Following the link just gets you information on the existing 4550B,C variants, however.

Noise is a bit higher than the 070, however. Many would trade a better tempco over a lower noise, methinks.

Edit: ordered four REF07025 units--will update group on noise when I receive them...

[RandallMcRee]

So I did hook up one REF70.

First, there is quite a dependence on the input voltage. Initially, I powered the unit from +15volts which is well within its rating according to datasheet. However, the output voltage was out-of-spec.

So, I lowered the input voltage to 3V and voila, all in-spec.

However, the output has a 10hz oscillation, a nice sine wave, that is immune to either input capacitance or output capacitance and the level is around 3uV. So, superimposed on that nice 0.2ppm noise is this oscillation.

Probably not going to keep characterizing the units given that problem.

[Warhawk]

So I did hook up one REF70.

First, there is quite a dependence on the input voltage. Initially, I powered the unit from +15volts which is well within its rating according to datasheet. However, the output voltage was out-of-spec.

So, I lowered the input voltage to 3V and voila, all in-spec.

However, the output has a 10hz oscillation, a nice sine wave, that is immune to either input capacitance or output capacitance and the level is around 3uV. So, superimposed on that nice 0.2ppm noise is this oscillation.

Probably not going to keep characterizing the units given that problem.

Interesting, thanks for sharing. COuld you eventually share a picture of the setup or a board, final diagram, capacitors used? Maybe our folks should be aware of the problem. Which package did you use? Any flux residues on the board?

[Kleinstein]

A 10 Hz oscillation is pretty odd - very much at the low end.  The only process I can imagine to be that slow would be some thermal coupling. If this is the case changing the supply would likely change things - however more like getting worse with a higher supply and 3 V is already at the low end.

[TimFox]

A strange suggestion for the cause of your 10 Hz problem:
Many years ago, I built a vacuum-tube theremin that had a conventional regulated power supply (pass triode, pentode error amplifier, and VR tube in cathode of the pentode).  After curing a 70 MHz parasitic oscillation with the usual series resistor at the triode grid, I became paranoid about oscillation and carefully searched the output.  I found a 10 Hz waveform of maybe 100 microvolts pk-pk that was line-synchronous.  It turned out to be due to variations in the AC mains caused by a soldering station plugged into the same outlet strip.  Apparently, the soldering station servo used zero-crossing switching and was in a limit cycle while idling in the holder.  Turning the station off cured the problem.
Does your 10 Hz oscillation synchronize with the mains?

[RandallMcRee]

Thanks for the suggestions.

If I can muster the energy I will try to quantify this better. I realized after posting that osc is over 10Hz and less than 60Hz, the oscilloscope (rigol 1054) was probably giving me the wrong number--it tends to do that on waves that are not rock solid. I did notice, in support of your suggestions that the oscillation was larger in magnitude with high Vin. Like I said, I put the maximum datasheet Cin, Cout on it and it still persisted.

As to the line causing these issues that is quite possible but I believe this would, in any case, be an issue only for the REF70. I have numerous other Vrefs that work fine--Ltc6655, max6126, max6350, ltc6657, ltz1000, lm399, adr4525. Also measured noise on 731B plugged in to same line. None of these show a problem (tested on the same day were only the ltc6657, lm399, ltz1000).

[uski]

If I can muster the energy I will try to quantify this better. I realized after posting that osc is over 10Hz and less than 60Hz, the oscilloscope (rigol 1054) was probably giving me the wrong number--it tends to do that on waves that are not rock solid. I did notice, in support of your suggestions that the oscillation was larger in magnitude with high Vin. Like I said, I put the maximum datasheet Cin, Cout on it and it still persisted.

Did you try looking at the signal at different timebases and/or sampling rates ?
I think of a possible aliasing issue here. You're probably aware of this - but just checking.