NEW PDVS2mini DC Voltage Reference from Ian Johnston - Reviewed.

[TheDefpom]

Hi All,
I thought I would pop this here, I just published a review of a new piece of equipment from Ian Johnston, the PDVS2mini, which is a 0-10V DC voltage reference.

No doubt you will spot things I am doing wrong, I'm not a metrology guy after all, but hopefully the video demonstrates the capabilities of the unit.

[IanJ]

Hi all,

Working on a couple of software bugs at the moment, mostly down to the math.h library which isn't great on these small AVR's (it can't add or subtract properly!). So, having to write (C++) around the issue which is a pain.
Scott's low end non-linearity problem I don't experience on my own PDSV2mini's here.......so some detective work is required.

Prototyping Happy days!

Ian.

[Svgeesus]

So, the Integral Non-Linearity is higher than the PSVS2, but because resolution is also higher it is compensated for in software?

[branadic]

Scott, your meter also has INL which is not zero. You could use your F343 and make differential measurements to get more into that.

-branadic-

[IanJ]

So, the Integral Non-Linearity is higher than the PSVS2, but because resolution is also higher it is compensated for in software?

PDVS2 (18-bit DAC9881) INL = +/-2 LSB typical, +/- 3 LSB max.
PDVS2mini (20-bit MAX5719A) INL = +/- 1 LSB typical, +/- 20 LSB max.

The +/- 20 LSB is scary (when volt-nutting) so there is 11 y=mx+c calib setpoints between 0 and 10Vdc output (1Vdc apart).......to compensate/help minimize any deviation.

An alternative solution would have been to add in an ADC (low INL) and monitor the DAC output and compensate accordingly, but that just pushes the BOM/complexity up..........

Ian.

[MiDi]

Scott's low end non-linearity problem I don't experience on my own PDSV2mini's here.......so some detective work is required.

Assuming the PDSV2mini does no range switch, this simply shows the DNL.
The DS states a DNL of +-1LSB@25°C -> +-10uV@10V FS (This means that this DAC even may not be monotonic).
And it is unlikely that this is only in the low end, the Fluke meter has not enough resolution to show this accurate in 1V/10V range - as Scott already stated.

Ian, are you sure that yours have better DNL?

[IanJ]

I have 2 others sitting here from the same batch and an older one......and on my 3458A no real linearity problems (exception: see below) including at the low end.

However, I had found an issue with the firmware which generated a 1 or 2bit random inaccuracy. It's a well known math.h library problem (it can't subtract/divide properly and shows up as a rounding error!). It's fixed now and makes a nice difference.

This weekend I'll check Scott's video properly and look into the low end issue......but I am beginning to think it's just DNL as we are right down at 1bit step changes......which are magnified in Scott's mV mode on his DMM. My testing has been done on the same 3458A mode across the full range and shows no difference between high or low.

Maybe I will try adding a range selection like I did on the PDVS2, an extremely low RDSon Fet and a low tempco resistor or two. This DAC doesnt have the same Vref range unfortunately, but there are other ways. Hmmmm!

Ian.

[TheDefpom]

Ian did a firmware update to 1.01 which I applied to my prototype unit (Ian corrected a problem the math library, as mentioned above), the unit steps by 10uV without missing a beat now, I barely even noticed the original issue anyway but now it is fixed.

[IanJ]

Ian did a firmware update to 1.01 which I applied to my prototype unit (Ian corrected a problem the math library, as mentioned above), the unit steps by 10uV without missing a beat now, I barely even noticed the original issue anyway but now it is fixed.

Yep, V1.01 is definitely a winner, and subject to some on-going testing being completed I should be going into production in the coming couple weeks.

Maths error:-
On 8-bit Atmel AVR controllers the float type is 32-bits which makes the limit 6 or 7 digits......close to the bone and causes rounding errors when working at 5dp's.
So, rather than rely on the maths library doing any rounding I have taken control of it just before converting and sending to the DAC......thus forcing it the way I want it.

Anyone interested in purchasing please use the NOTIFY ME link and as soon as I have stock you'll get an email.
https://www.ianjohnston.com/index.php/onlineshop/handheld-precision-digital-voltage-source-2-mini-detail

Ian.

[niner_007]

Can't wait to get one

You should consider an STM32, not only will it use less power than the AVR, but better software and libraries too

[IanJ]

You should consider an STM32, not only will it use less power than the AVR, but better software and libraries too

Something for the future maybe........but what puts me off slightly is the lack of EEprom, albeit they do have EEprom emulation in flash via paged memory or something like that........well, thats as much as I know about STM32 anyways.

Ian.

[niner_007]

You should consider an STM32, not only will it use less power than the AVR, but better software and libraries too

Something for the future maybe........but what puts me off slightly is the lack of EEprom, albeit they do have EEprom emulation in flash via paged memory or something like that........well, thats as much as I know about STM32 anyways.

Ian.

STM32 have both eeprom (STM32L010F4 for example) and non volatile flash, depending on the model, which will allow you to store calibration constants and so on, but why do you need eeprom specifically? the API for reading and writing to the flash is quite fine, maybe I'm missing something

[TiN]

I don't like STM, their marketing here in TW failed hard when I tried to get some support to consider using STM32 on a project. 

Now let's get back on topic. I got this little cute PDVS2mini in thermal box, running for life first quick and dirty tempco test.



I'm sure somebody will dig a photo with $70K vs $300 DC source .



Design assembly is nice and tidy, with all critical stuff well cared. Graphical LCD is not backlit, just COG panel. It has decent contrast and main output voltage have nice large digits. The menu system is little unusual but not hard to figure out without any RTFM. There is a help string on the bottom strings to show keypad actions.



Access to battery requires removal of 4 screws, and also provide a way to see some internal construction.
There is no separate battery compartment or protection over electronic parts, due to low-cost targeting reason, which is understandable.
PDVS2mini designed to use two 9V batteries in series, with integrated charger. Unit needs to be powered on to allow charging from external DC jack.

I'm working on detailed review/testing, but it will take some time, so here are just some initial dirty results using my secondary equipment (K2002 and Arroyo TECpak controlled box).

PDVS2mini is powered by batteries, placed in DIY chamber. Raspberry Pi controls Keithley 2002 as a digitizer, and TECpak 585 for precision temperature control. 40W noname TEC used as cooler/heater, with water-cooling loop dissipating heat into room elsewhere. The sensor in the box is Honeywell HEL-705-1 100 Ohm platinum RTD. This thermal box setup may not look like much, but its able to maintain any programmed temperature from -5 °C to +65 °C with better than 0.008 °C stability over hours.



Now to the first test results. The first graph is a plot of +10V output from PDVS, with temperature change from +15.00 °C to +55 °C in a slow and controlled manner. There is soaking time at +35°C for 2 hours, and also at peak max temperature, to get an idea of settling and hysteresis behavior. The horizontal axis is time duration, vertical axis is deviation from initial point at +15 °C, which happened to be +9.999994 VDC +/-2ppm.

For illustrative purposes, chart have additional plots of -0.1 ppm/K and -0.2 ppm/K thresholds based on temperature change (using box method) and TEC current readout from Arroyo TECpak 585. Ramp up time for temperature is +0.08333 °C per minute. Measured deviation is -4.0 ppm for +15.0°C to +35°C ramp, and another -5.3 ppm for +35°C to +55°C ramp. Following the TC Box method we can estimate temperature stability at max -0.23 ppm/K which is impressive for $300 USD DAC box!



Now same data, but in voltage/temperature scale representation. Hysteresis on fixed temperature points is around ~2ppm but that will need additional verification with proper setup Unit may need additional time, longer than 30min to get temperature soaked to reach equilibrium. This particular temperature chamber does not use forced airflow, and from datalog looks like it need ~30 minutes to equalize temperature change.



Stay tuned for more

One part I'm not very happy about is banana-only output connectors. Once initial testing and calibration verification test is done, these will be for sure changed to some proper low-thermal EMF hardware  .

If somebody wants particular test setup/conditions - feel free to give feedback.

[TiN]

And battery died after 12.5 hours. 

[chickenHeadKnob]

When you consider that most people would want to keep this device a long time and that the value of the reference to the particular end user increases over time it would be a shame to lose everything to battery leakage. I would extend the battery connector to an external connector as the very first  modification and permanently run from external power.

[2N3055]

When you consider that most people would want to keep this device a long time and that the value of the reference to the particular end user increases over time it would be a shame to lose everything to battery leakage. I would extend the battery connector to an external connector as the very first  modification and permanently run from external power.

If you keep it normal side up (screen up) it will leak to the bottom of the box, not electronics. Also, it uses rechargeable Li-ion PP3 9V batteries that are not leak prone...

[chickenHeadKnob]

If you keep it normal side up (screen up) it will leak to the bottom of the box, not electronics. Also, it uses rechargeable Li-ion PP3 9V batteries that are not leak prone...

That is good to know, I would still want to keep it powered continuously with a linear supply to age the lm399 for the first three years. There-after I suppose you could switch to intermittent power-ups. 

[TiN]

This device is not designed to be a permanent DC voltage standard, but a portable programmable DC source/DAC. LM399 is not really worthy of proper DC standard anyway due to noise/stability and tempco.
And it can run long time with external charger/PSU connected to the Charge port. PDVS2mini is much cheaper than another 5.5-digit source alternatives, even tho output is limited to only one 0V - 10V range.
Better DAC option are some 30kg+ anchors like Fluke 5440 and alike, for which you need to be in right country/right spot/with cash in hand and will to repair/troubleshoot boards and modules when things go sour.

Handy option for beginner volt-nuts. And if you think ~$300 USD is too much, well, MAXIM's eval board with used DAC, without LCD, case, charger or LM399 already cost $190 . Another popular "volt-nut" DAC with AD5791 with LTZ ref is even more, close to $500 USD.

Don't get me wrong, I'm not implying somebody should calibrate 3458A with PDVS2mini  ,
It's more of educational precision DC source kit, that is fun to have around for some simple tests, without expensive demands for accuracy. And for this role it looks like priced adequate, and 24/7/365 operation is not a requirement.

[TheDefpom]

The unit is pretty damn good for the money, yes it has a dc input jack for charging the batteries, it can run from that in itself, however due to introducing noise via that input, and the extra heat generated during charging Ian recommends that it not be used that way to avoid it affecting the output accuracy.

[IanJ]

Apart from the possible introduction of noise depending on your DC power source......and by example running at 20Vdc at the power jack sees the output vary by about 10uV (1bit !) eventually due to the extra internal heat being generated.
On disconnecting the DC input it takes approx. 30mins for the unit to return to it's normalized output.
I've tried to minimize this heat induced drift by mounting the LM399AH externally, it's under the LCD in it's own little 3D printed conductive PLA thermal cover.

Running off DC permanently then has it's issues per above, but is entitely safe. The MAX1873 battery management IC recognizes an external supply and once the battery is detected as fully charged it will cease charging completely and the DC supply is used to feed the rest of the circuit. The batteries remain in isolation effectively and are not in any way trickle charged.

I haven't had a Lithium-Ion battery leak yet and I have about 40 of them which have been charged, drained and left sitting for months on end in and out of PDVS's.

PS. This unit was never directly designed as a "reference"......it was designed as an adjustable precision voltage source, but with some capabaility of a reference.
Quote from my website "With a user adjustable range from a true 0V to 10Vdc and an accuracy/stability down in the uV’s it has a multitude of uses as a calibrator, reference & precision voltage source."

But TiN is right.........not a great idea to calibrate a 3458A with it........

I'll be in production soon with it.

Ian.

[MegaVolt]

Are you planning to add a remote control?

[IanJ]

Are you planning to add a remote control?

It has a serial header and you can control it remotely from a PC already much the same as the PDVS2 with an appropriate adaptor......however, I will just be supplying the Ascii protocol used and not a Windows app (well, I might supply an EXE unsupported).

Ian.

[MegaVolt]

It has a serial header and you can control it remotely from a PC already much the same as the PDVS2 with an appropriate adaptor......however, I will just be supplying the Ascii protocol used and not a Windows app (well, I might supply an EXE unsupported).

Thank . I did not notice the serial connector.

[IanJ]

Hi all,

I have just made the first batch of units available in my online shop. Here:-

https://www.ianjohnston.com/index.php/onlineshop/handheld-precision-digital-voltage-source-2-mini-detail

Available worldwide with shipping via UPS courier.

Thanks,

Ian.

[oz2cpu]

i am not sorry for kicking this topic back alive :-)
wow amazing product you made there Jan, all looks super professional

I really like to see some closeups of the mecanical implementation of the LM399
you say it is under the display ?
and inside 3D printed isolation case ?
the display looks like it can be taken out, so this detail could be revealed,