POWER DESIGNS PRECISION POWER SUPPLIES

[alm]

Yes, tweaking those old differential amplifiers is good fun. There is also the Tek P6046 differential probe that even Jim Williams called 'extraordinarily difficult to trim for CMRR'.

I don't think the 'infinite' input impedance is specified, but I would expect it to be much higher than 100 Mohm, at least in mint condition (dirt on the PCB may significantly reduce it). According to the schematic switch S10 sets the input impedance to 'electrometric', where it is directly driving the gate of a FET. DC impedance should be in the Gohms, AC will of course be much lower.

[muvideo]

I have a 7A13 diff amp to do the 20MHz BW testing. That one was expensive $50.00  it has a interesting infinite impedance mode for 50mV/div and below.

Very nice device the 7A22, I have a pair I bought for around 20€, the same seller wanted 105€
for the 7A13, so I'm still searching for the right one one
The 7A22 can be modded in high impedance removing a pair of jumpers on the board,
it's limited to 10mV to 10uV per divison (since in this range the attenuator is not used)
and the input current in this config is quoted by the manual in the 100pA range, i.e. the
gate current of input fets.

Fabio.

[robrenz]

@ muvideo,

Thanks for the heads up on the high impedance mode. I missed that in scanning the manual.

Do your 7A22's have the awesome machined tuning capacitors like mine in this thread?. The manual I downloaded shows regular tuning caps in the manual.  The jumpers that get unsoldered for the high impedance mode are shown in those pics also.

[muvideo]

@ muvideo,

Thanks for the heads up on the high impedance mode. I missed that in scanning the manual.

Do your 7A22's have the awesome machined tuning capacitors like mine in this thread?. The manual I downloaded shows regular tuning caps in the manual.  The jumpers that get unsoldered for the high impedance mode are shown in those pics also.

It's exactly the same:


By the way one of the mine was broken. One of the problems was
due to a deformed variable capacitor, pushed by the offset pot the
armatures were touching and it was shorted.
The other problem was a resistor cosmetically perfect but electrically
open, before I started checking the resistors I swapped all the
semiconductors and capacitor in that stage.
I was so amused by the fact that it's entirely passive design,
the schematic seem "simple" for the specs, but as everything the
devil is in the details.

In this case the high impedance mod precludes the use of the higher
input voltages, at least trows off the calibrated amplitude since the
attenuator expects to have there the 1Meg input resistor.
One could think to use a pair of switches or jumpers or reed relays
to connect and disconnect easily these jupers, but must be
careful with the eventual introduction of thermocouples with
the contacts.

Fabio.

[robrenz]

That is one kick ass microvolt input scope!  Few can match it today, most top scopes do 500uV/div as minimum, if at all.  Not to mention the dynamic range, 10uV to 10V /div, it took a while to look at that range before I realized it was true!

It does at least 25V/div. in variable.

[Fsck]

So my 2020B has a busted voltage selector switch. (The 0-10v vs. 10-20V & pull to release switch)
Anyone know where I can find a reasonably priced replacement? It's the pull to release switch function that I'm unsure of how to replace. Like a on-off-on dpdt toggle switch?

[robrenz]

AFIK You can use any good quality 2 position (on none on) double pole switch. It needs to be fairly good quality because the controlling resistance goes thru one set of contacts. It does not need to be high power, I think only one milliamp goes thru the controlling resistance and the other contacts just power the neon indicators.

[poodyp]

So I just got a 2020B, I haven't done a full rundown of it yet, but as I was unpacking it I heard something rattling, and found this screw inside:


But I can't seem to find where it's from.  I can't see how it would have got in there, but nothing seems to be missing. Anyone have any idea?

[justanothercanuck]

Ouch!  Good thing you didn't turn it on...  Does it look like it came from the power supply?  Your guess is as good as mine, but maybe when they were assembling them, the lineworker accidentally dropped it in there, couldn't retrieve it and let it run down the line. 

[poodyp]

I figured it out. The rear feet are missing, and this screw matches the screws in one of the front feet. Oddly the other front foot uses Philips screw heads.

I wonder where the screw for the other foot went...

[poodyp]

Interesting, it seems someone replaced the second fuse with a 3/4 amp fast blow fuse. I was  until I realized there was no reason it should be that small, looked in the manual, and yep, should be 3 amp slow blow.

Before it blew I made this graph:


Between 2-10v it's less than 100ppm off. Just for fun I tested with a Fluke 87V. Any point in the graph just means it was one or two counts off. This is all without cleaning any contacts. I need to figure out this deoxit thing and clean them.

[saturation]

That a nice graph.  Chances are if you deoxit the contacts you could extend the 100ppm error to the smaller voltages,  cleansing off oxidation and contact grime play a large role in the reading stability.

[poodyp]

Thanks, I also have a much less impressive one for my EDC 520 and even less impressive for my 2005 . I'm hoping it's just serious corrosion.

About the deoxit, I'm probably going to get the d100l 2ml tube and a lot of wooden cotton swabs and go to town. I wish there was a less complicated way to take the front panel apart, it's going to be a pain to clean.

[robrenz]

IMO don't take it apart. Get the white handle Microbrush to apply the Deoxit. Cotton swabs will waste a huge amount of the very expensive stuff. Just wet each each contact with D100L and then cycle the switch a lot. Wash off with your favorite electronic cleaner and blow out. Then lightly lube each contact with Deoxit G100L with another white microbrush. I competely agree with using the liquid (not spray) versions of deoxit.  HERE is my setup.

[poodyp]

IMO don't take it apart. Get the white handle Microbrush to apply the Deoxit. Cotton swabs will waste a huge amount of the very expensive stuff. Just wet each each contact with D100L and then cycle the switch a lot. Wash off with your favorite electronic cleaner and blow out. Then lightly lube each contact with Deoxit G100L with another white microbrush. I competely agree with using the liquid (not spray) versions of deoxit.  HERE is my setup.

Yeah, I was lamenting because I tried and realized it'd be a nightmare with a lot of blood, tears, and probably destroying something. Makes me wonder how they put it together.

Thanks so much for the pointers. I now have some brushes and deoxit d5 and gold on the way.

Just for fun, here's a look at how bad my 2005 is:

I didn't bother going beyond 100mV because it was worse than 100%. All the lamps are dead too. And it killed my dog. And gave me indigestion. And set my car on fire.

[saturation]

Its worth the labor, PSU with these types of stability specs are  sold today as 'source units' by Agilent, and the starting price is > 10x more than you'd pay for even the most expensive PD supply, which unfortunately, as I still track may units on eBay, the base price is now 2x what I paid for it.  So maybe eBay folks read eevblog.  Enjoy.

[poodyp]

So here's my progress:


So I'm kind of confused. As you can see, after I cleaned it all the peaks and valleys are gone, which I thought made sense, but then when I calibrated it, they came back, and it's nowhere near as accurate as I thought it would be. Zero didn't really need to be adjusted all that much, it was at about -0.8mV, but I got it to within 0.5uV, then at 20V I got it to 20.005V.

[robrenz]

May I suggest first doing a repeatability test of each decade independently. Maybe 5 to 10 data points thru each position and measure the min-max and SD of each of the 10 positions.  Do this separately for each decade, don't move the other decades while doing the test and cycle 0 thru 9 and back to 0 this may show difference in value for approach direction. If you have another very stable supply you could use it to offset the higher voltages to your meter for increased resolution. In this test we are only looking for repeatability  not absolute accuracy.  These are 1 ohm per mVolt units so that is a mOhm per microvolt so the switches need to be really well cleaned and lubed. You also may want to check the phasing of each decade with the detent to make sure the switch contacts are well centered when in the detent position. The detent system itself needs to be in good shape, some of mine had some very hard caked grease that was particularly hard to remove.

[saturation]

The green line is expected, the blue line is erratic and suggestive of dirt; the yellow line now appears similar.

I suspect the calibration pot has oxidized grit; once manipulated the dirt is catching on the pot wiper arms.  Try turning the adjust multiple times back and forth to clear out the grit and see what happens.   I can't recall if the cal pot is hermetically sealed or not, and amenable to dexoit cleaning.

So here's my progress:


So I'm kind of confused. As you can see, after I cleaned it all the peaks and valleys are gone, which I thought made sense, but then when I calibrated it, they came back, and it's nowhere near as accurate as I thought it would be. Zero didn't really need to be adjusted all that much, it was at about -0.8mV, but I got it to within 0.5uV, then at 20V I got it to 20.005V.

[poodyp]

So here're the results of a quick test Rob suggested:

Avg: 5.002337
Min: 5.00231
Max: 5.00235
SDev: 0.0000125166555704632

Avg: 0.5005721
Min: 0.500546
Max: 0.500592
SDev: 0.0000135273385737515

Avg: 0.04995493
Min: 0.0499536
Max: 0.0499577
SDev: 0.00000138246960987263

Avg: 0.004990644
Min: 0.00498626
Max: 0.00499526
SDev:0.00000330608260970913

It's interesting that the 10 and 1 mV decades are all below where they should be, while the 1V and 100mV decades are all above. Something I forgot to mention: the contact for the second set of resistors on all decades only appears to make contact between each setting. Is that normal? The entire rod is indexed, so I don't see how it could be off that much. Otherwise the contacts appear to line up right.

[robrenz]

Lets put this in perspective.  The spec for the 2020 is 0.1% of setting plus 500µV. 
Lets not even consider the additional 0.001% regulation, 8 hour stability of .001% + 100µV, and TC of 0.001% or 50µV whichever is greater/ deg C.

Min: 5.00231   that is 40µV variation (.36% of a .011V error band)     Max error is .00235V spec allows .0055V  less than 42% of spec.
Max: 5.00235      +.047% of set value error.

Min: 0.500546   that is 46µV variation (2.3% of a .002V error band)    Max error is .000592V spec allows .001V  less than 59% of spec.
Max: 0.500592    +.118% of set value error.

Min: 0.0499536    that is 41µV variation (3.7% of a .0011V error band)    Max error is .0000464V spec allows .00055V  less than 9% of spec.
Max: 0.0499577   -.093% of set value error.

Min: 0.00498626   that is 9µV variation (0.89% of a .00101V error band)    Max error is .00001374V spec allows .000505V  less than 3% of spec.
Max: 0.00499526     -.275% of set value error.

This is all well within the original specifications.    Those microvolt variations are milli ohm variations in contact resistance.  It is fun to try to improve it further. 

[poodyp]

Lets put this in perspective.  The spec for the 2020 is 0.1% of setting plus 500µV.

Actually I forgot to point out this is my 2005, my 2020B is already pretty good. I'm hoping it can get better after being cleaned.

Lets not even consider the additional 0.001% regulation, 8 hour stability of .001% + 100µV, and TC of 0.001% or 50µV whichever is greater/ deg C.

Min: 5.00231   that is 40µV variation (.36% of a .011V error band)     Max error is .00235V spec allows .0055V  less than 42% of spec.
Max: 5.00235      +.047% of set value error.

Min: 0.500546   that is 46µV variation (2.3% of a .002V error band)    Max error is .000592V spec allows .001V  less than 59% of spec.
Max: 0.500592    +.118% of set value error.

Min: 0.0499536    that is 41µV variation (3.7% of a .0011V error band)    Max error is .0000464V spec allows .00055V  less than 9% of spec.
Max: 0.0499577   -.093% of set value error.

Min: 0.00498626   that is 9µV variation (0.89% of a .00101V error band)    Max error is .00001374V spec allows .000505V  less than 3% of spec.
Max: 0.00499526     -.275% of set value error.

This is all well within the original specifications.    Those microvolt variations are milli ohm variations in contact resistance.  It is fun to try to improve it further. 

So, don't worry about the accuracy, just be happy the repeatability and drift are good (I measured it at 6uV at 12 hours so far)?

The green line is expected, the blue line is erratic and suggestive of dirt; the yellow line now appears similar.

I suspect the calibration pot has oxidized grit; once manipulated the dirt is catching on the pot wiper arms.  Try turning the adjust multiple times back and forth to clear out the grit and see what happens.   I can't recall if the cal pot is hermetically sealed or not, and amenable to dexoit cleaning.

The pots aren't sealed as far as I can tell, but I'm not sure how I'd open them without destroying them. I tried wiping them back and forth several times, but it doesn't seem to have made any difference.

[robrenz]

Lets put this in perspective.  The spec for the 2020 is 0.1% of setting plus 500µV.

Actually I forgot to point out this is my 2005, my 2020B is already pretty good. I'm hoping it can get better after being cleaned.

That makes it even better! The spec for the 2005 is 0.1% of setting plus 1mV.

So, don't worry about the accuracy, just be happy the repeatability and drift are good (I measured it at 6uV at 12 hours so far)?

Not at all,  %-B just realize it is way better than its spec and improving it is going to take some very precise tongue angle

I suspect the calibration pot has oxidized grit; once manipulated the dirt is catching on the pot wiper arms.  Try turning the adjust multiple times back and forth to clear out the grit and see what happens.   I can't recall if the cal pot is hermetically sealed or not, and amenable to dexoit cleaning.

The pots aren't sealed as far as I can tell, but I'm not sure how I'd open them without destroying them. I tried wiping them back and forth several times, but it doesn't seem to have made any difference.

One way you could check them is to disconnect one end of the pot and connect your ohm meter between there and the wiper and see if there are jumps in the readings as you travel the zone you have been using. If the ohm range crosses two of your meter ranges make sure you lock your meter into the higher range so you dont think the jump when auto ranging is a bad pot.  I did that earlier in this thread    You can also do the standard contact resistance variation test shown in the bourns potentiometer handbook

[robrenz]

If the pots do seem bad and you want to try cleaning them before you replace them you can use the method described on page 2 of this thread on posts #18 thru #22.  look at the typical construction of this type of pot on page 8 of this download of the  Bourns trimmer primer.  Guesstimate the location of a small hole at each end that will allow you to blast stuff through the pot with a syringe. Either manual or air powered (solder paste dispenser) I suggest a sequence of lots of IPA folowed by just air then a small drop of deoxit D100L. Cycle the pot multiple times to work the  deoxit.  Then wash again  with lots of IPA and air dry. Then a small drop of Deoxit fader lube to lube the pot and cycle again to spread the fader lube. Deoxit G100L will work if you don't have the fader lube.

You can do all this without removing the pots from the base board since they are laying on thier sides.

[terabyte]

Im pretty happy.  After over a year of wanting another PD PSU I finally got a 2020B;
http://www.ebay.com/itm/Power-Designs-2020B-Variable-Power-Supply-Precision-DC-Calibrator-2AMP-20V-/221195966773?ssPageName=ADME:L:OC:US:3160