DVR1 Voltage Reference

[branadic]

I wonder if someone here already brought the DVR1 Voltage Reference:

https://www.ab-precision.de/products/electrical-standards/

It is based on a hermetic LT1021 together with an oven. Sure 220.15€ is not cheap and more than the cost for a LTZ1000 circuit, but: "... Every product is factory calibrated, traceable to national standards against a precision voltage standard (732A) with low uncertainty after a minimum of 500 hours of pre-aging for the references and an additional 100 hours for the complete unit...".

-branadic-

[kj7e]

I have one of Doug Malone's LT1021 based 10v references, its good for what it is, but its no LTZ1000 or even LM399.  Even ovenized your still going to be fighting with noise, specs are <1ppm or 10uV for the -10.

http://www.voltagestandard.com/-.html

Out of curiosity I just took some noise measurements. Using 10,000x gain LNA, so 10mV p-p = 1uV.

LT1021-10 - 14 sec 7.36uV p-p;


LT1021-10 - 140 sec 7.28uV p-p;


LTZ1000 - 140 sec (note 5mV/dev) 1.3uV p-p;

[splin]

I wonder if someone here already brought the DVR1 Voltage Reference:

https://www.ab-precision.de/products/electrical-standards/

-branadic-

But oddly there are two versions, both of which are described as non-ROHS and yet:

Please note that non-ROHS prodcuts do not ship into the EU.

Are we to presume that these are built with NOS or used components sourced from Ebay or the like? Nothing wrong with that mind - likely better than new with less drift due to long ageing in storeage.

[branadic]

I have one of Doug Malone's LT1021 based 10v references, its good for what it is, but its no LTZ1000 or even LM399.  Even ovenized your still going to be fighting with noise, specs are <1ppm or 10uV for the -10.

Well, DVR1 uses hermetical version of LT1021.

Are we to presume that these are built with NOS or used components sourced from Ebay or the like? Nothing wrong with that mind - likely better than new with less drift due to long ageing in storeage.

I have no idea...

-branadic-

[David Hess]

Out of curiosity I just took some noise measurements. Using 10,000x gain LAN, so 10mV p-p = 1uV.

Why didn't you have your DSO report RMS noise also?

It would be neat to see the noise versus frequency graph but most DSO FFTs cannot be configured to do it correctly.  It would reveal the effects of flicker noise.

[kj7e]

Out of curiosity I just took some noise measurements. Using 10,000x gain LAN, so 10mV p-p = 1uV.

Why didn't you have your DSO report RMS noise also?

It would be neat to see the noise versus frequency graph but most DSO FFTs cannot be configured to do it correctly.  It would reveal the effects of flicker noise.

Ill play with that some tomorrow.

[quarks]

my very first address to ask anything about this product would be eevblog member ap (the owner of www.ab-precision.de)

[ap]

I have one of Doug Malone's LT1021 based 10v references, its good for what it is, but its no LTZ1000 or even LM399.  Even ovenized your still going to be fighting with noise, specs are <1ppm or 10uV for the -10.

Well, it is a truism that the LT1021 is not like a LTZ1000 in terms of noise and stability. It was never intended as. However the 399, if you look at the datasheet has a factor of 3 worse typical (10uV RMS nominal to 10V versus 3uV) and factor of >10 (!) worse maximum (70iV versus 6uV RMS) noise. The LT1021 is 100% noise tested. Also, there are noise reduction measures within the DVR1 as decribed.
As some have demonstrated here, the LM399 has a relatively bad temperature behaviour. The maximum data sheet spec is 2ppm/K, which is not really low. This is being eliminated in the DRV1 through an internal heater. The use of a thermistor for compensation is avoided, due to the complex adjustment procedure and other technical reasons. Also, the LM399 requires an amplifier for upscale to 10V, adding further temperature/stability drift issues. Exhaustive reading available here on eevblog metrology.

The DVR1 uses a hermetically sealed (actually specific pre-aged version by LT; not an unsoldered/used device of course). Additional pre-aging (on chip level and board level) is applied. The LT1021 is not available in a ROHS-compliant hermetic version, hence the DRV1 does not ship to EU states (except for proven permitted exemptions). Using a non-hermetic version is not an option due to long term drift stability issues as well as short term environmental (humidity) influences. Coating would not solve this problem, as some may think.

[splin]

I wonder if someone here already brought the DVR1 Voltage Reference:

https://www.ab-precision.de/products/electrical-standards/

-branadic-

But oddly there are two versions, both of which are described as non-ROHS and yet:

Please note that non-ROHS prodcuts do not ship into the EU.

Are we to presume that these are built with NOS or used components sourced from Ebay or the like? Nothing wrong with that mind - likely better than new with less drift due to long ageing in storeage.

Oops - just looked at the datasheet - a much more likely explanation is that the TO5 (hermetic) versions of the LT1021 are not ROHS  compliant which is a bit unfortunate. There really should be exceptions in the ROHS rules for specialist equipment like this which is never going to be sold in large volumes.

[kj7e]

Out of curiosity I just took some noise measurements. Using 10,000x gain LAN, so 10mV p-p = 1uV.

Why didn't you have your DSO report RMS noise also?

It would be neat to see the noise versus frequency graph but most DSO FFTs cannot be configured to do it correctly.  It would reveal the effects of flicker noise.

Just for you, though not sure its setup the way you would prefer.

LT1021-10,


LTZ1000A;

[David Hess]

Out of curiosity I just took some noise measurements. Using 10,000x gain LAN, so 10mV p-p = 1uV.

Why didn't you have your DSO report RMS noise also?

It would be neat to see the noise versus frequency graph but most DSO FFTs cannot be configured to do it correctly.  It would reveal the effects of flicker noise.

Just for you, though not sure its setup the way you would prefer.

Thanks.

The RMS results look reasonable but it is difficult to tell without knowing the bandwidth and shape factor.  Some DSOs cannot make measurements of RMS noise which is odd because it is so easy to do.

The FFT result also look reasonable but of course are not calibrated for resolution bandwidth and do not extend to a high enough frequency to see the 1/f corner.  Besides resolution bandwidth calibration, the trick is to average multiple FFTs but *not* average the acquired waveform which will just screw everything up.

I have been looking for inexpensive DSOs which can do both of these things.

[kj7e]

Part of the problem is the LNA I'm using has a 10Hz bandwidth and falls off quickly beyond that, so I set the upper limit to 20Hz.

[fcb]

I have one of Doug Malone's LT1021 based 10v references, its good for what it is, but its no LTZ1000 or even LM399.  Even ovenized your still going to be fighting with noise, specs are <1ppm or 10uV for the -10.

Well, it is a truism that the LT1021 is not like a LTZ1000 in terms of noise and stability. It was never intended as. However the 399, if you look at the datasheet has a factor of 3 worse typical (10uV RMS nominal to 10V versus 3uV) and factor of >10 (!) worse maximum (70iV versus 6uV RMS) noise. The LT1021 is 100% noise tested. Also, there are noise reduction measures within the DVR1 as decribed.
As some have demonstrated here, the LM399 has a relatively bad temperature behaviour. The maximum data sheet spec is 2ppm/K, which is not really low. This is being eliminated in the DRV1 through an internal heater. The use of a thermistor for compensation is avoided, due to the complex adjustment procedure and other technical reasons. Also, the LM399 requires an amplifier for upscale to 10V, adding further temperature/stability drift issues. Exhaustive reading available here on eevblog metrology.

The DVR1 uses a hermetically sealed (actually specific pre-aged version by LT) . Additional pre-aging (on chip and board level) is applied. The LT1021 is not available in a ROHS-compliant hermetic version, hence the DRV1 does not ship to EU states. Using a non-hermetic version is not an option due to long term drift stability issues as well as short term environmental (humidity) influences. Coating would not solve this problem, as some may think.

Nice unit.

I agree that the noise is better than the LM399 - although I'm probably more in the 'compensate' than 'ovenise' references at the level of LT1021.  The hermetic packaged LT1021 was no-go for us as it wasn't RoHS and it cost more. We also found heavy coating with silicone reduced humidity related effects.

[ap]

Well, both products really target different applications, which makes a lot of sense. Not everybody has the same needs. The DVR1 was rather designed with limited consideration of costs (based on the LT1021, intentionally not the LTZ1000) to minimize aging and temperature drift by electrical design, extended pre-aging, related parts selection and calibration based on a traceable Fluke 732 / Datron 4910 voltage standard bank using the 3458A as a transfer meter only to exploit the sub-ppm-level accuracy of the voltage standards in use. Also, the DVR1 is available in two versions, one with 3 buffered 10V outputs, and one with 10V, 1V and 0.1V outputs.