A couple of weeks ago, I stumbled across an eBay listing for an EDC 521 Programmable DC Voltage/Current Source for $150. After a little research, including reading through a recent
EEVBlog forum thread on the similar EDC 522, I bought it.
For those who aren't familiar, this device was made in Boston by Electronic Data Corporation, which is now a division of Krohn-Hite. It's like the
EDC MV106J that Dave uses, a precision, low-drift, high-accuracy DC source for use in calibration of meters and sensors (and anything else that requires a precise, stable reference voltage). Like the MV106, it uses an aged, selected, 1N829 temperature compensated reference Zener to achieve its basic stability and accuracy.
It has some key differences from the MV106 too. The EDC 521 adds a 100v DC range to the 10v and 100mv ranges on the MV106, and then goes a step further by offering 100mA and 10mA constant current ranges. It also supports remote control over GPIB. The digital control requirement is probably partially responsible for the use of a different method of adjusting the output voltage compared to the decade divider used in the MV106. The EDC 521 uses a fixed resistor divider network to produce 10 output voltages. These voltages are switched using a CD4052 digitally controlled analog MUX CD4052 ICs to obtain a value for each decade. The value for each decade is buffered, and the values for all the decades are summed using 6 more precision resistors before being fed into an output amp. The MV106 uses somewhat more precision resistors + six low-resistance, manually-controlled mechanical switches to make a kelvin divider for setting the output voltage. It is also physically larger, with a depth of 20".
When the
EDC 521 arrived, the unit was a little worse for wear than I expected. The bottom panel had big dent and the red filter over the LED display was partially torn. Inside was a little better, but still needed attention. the plastic supports for the back-side of the PCB, where the electrolytic filter-caps are placed, had slipped free of the case due to deteriorated adhesive.
I banged out the dents, cleaned up the board and re-affixed the supports. Along the way, I noticed that the board had some revisions in place. After cleaning it up and inspecting it, I powered it up to do an initial check. After leaving it running for a few hours a spot check showed the output readings seemed well within the specified tolerances of both the source and the meter. I powered it down overnight.
The next day, I hooked it up to one of my Keithley 2700 DMM via the front-panel binding posts, set up a script to log readings at 5s intervals, and powered the 521 back up.
The above graph shows measured voltage for a nominal 10.00000v setting from the time the output switched-on through the two-hour specified warm-up period, and another two hours past that. At the end of the specified two hour warm-up period, it was 9.99986v, well within the specified tolerance of 0.000258v (0.002% of setting + 0.0005% of range + 3uV), as well as the DMMs uncertainty of 0.000350v 30ppm of reading + 5ppm of range.
Over the next 3 hours, the output continued to rise. The graph above overlaps with the previous one, and shows the measured output over 5 hours from the completion of the 2 hour warmup. It ended up leveling off at 9.99989v.
Over 4 days following warm-up, the voltage varied over about 80uV, or 8ppm of 10v, including the steep rise that continued after the warm-up period. If we just consider just the period starting at ~4.5 hours after power-on the measured voltage varies by just 40uV, or 4ppm. This is better than both the specified individual (10ppm, 14ppm) and combined (~17ppm) 24-hour stability specs of both the source and the DMM.
Looking at the min/max std and avg for a 10 minute sliding window over a ~4 hour period shows that the short-term difference in minimum and maximum readings is about ~6uv, or less than 1ppm. This is below the nominal precision of my DMM, in the noise of the added 7th decade available when collecting readings by GPIB.
Based on this evidence and analysis, I conclude I scored a deal on this.
I didn't have a temperature log to accompany the initial data I collected, but for the last few days, I've had a temperature logger going along side of the voltage measurements, so I can look at the temperature coefficient of the test system. Based on spot checks, it seems to be about 1-2ppm/C°, not bad considering that is about what the temp coefficient for the DMM. I'll post an update once I've charted and analyzed that data.
After that, I'm going to make a script to cycle the 521 through all or most of its full range and record voltage readings.
For a bit more detail, I've got two blog posts up:
Let me know if you have a critique of my interpretation of the data, or there are any details you'd like me to explore.
Home
Help
Search
Login
Register
