POWER DESIGNS PRECISION POWER SUPPLIES

[timb]

Yeah, all the switch resistors in my 2020B are dead on. Even the 5% and 1% tolerance ones are +/- 0.001Ohms. With all the switches in their full positions, I get exactly 10,000.01Ohms on my 6.5 Digit HP'57. So yours seems out for sure.

I might have a replacement here if you want it.


Sent from my Smartphone

Say where are you measuring from/to to get that 10,000.01. I have a 3456A as well, so I can perform the measurement and see what I come out to...

With the supply off, the range set to 0-10V and the vernier fully counterclockwise, look at the board on the back of the switch array and you should see three lugs with a 10K resistor across the top two. Test from the center lug and DC- terminal.

Basically, one end of the vernier is connected to DC- and the other end to the bottom lug, which connects to the switch banks, from there it goes from S5 to S6 then the top lug of the switch board (input of that 10k resistor), the center lug (output of that 10k resistor) connects to RV2 on the back, which straps to RV1 and goes to the main board voltage OpAmp! By the way, that 10k resistor is bypassed when the front switch is in the 10V position.


Sent from my Smartphone

[SharpEars]

Yeah, all the switch resistors in my 2020B are dead on. Even the 5% and 1% tolerance ones are +/- 0.001Ohms. With all the switches in their full positions, I get exactly 10,000.01Ohms on my 6.5 Digit HP'57. So yours seems out for sure.

I might have a replacement here if you want it.


Sent from my Smartphone

Say where are you measuring from/to to get that 10,000.01. I have a 3456A as well, so I can perform the measurement and see what I come out to...

With the supply off, the range set to 0-10V and the vernier fully counterclockwise, look at the board on the back of the switch array and you should see three lugs with a 10K resistor across the top two. Test from the center lug and DC- terminal.

Basically, one end of the vernier is connected to DC- and the other end to the bottom lug, which connects to the switch banks, from there it goes from S5 to S6 then the top lug of the switch board (input of that 10k resistor), the center lug (output of that 10k resistor) connects to RV2 on the back, which straps to RV1 and goes to the main board voltage OpAmp! By the way, that 10k resistor is bypassed when the front switch is in the 10V position.


Sent from my Smartphone

I measure a 5.10539 kOhm 4-wire Kelvin resistance measurement from the center lug to DC- with:

- The power supply turned off
- The voltage range set to 0-10 V
- The vernier turned fully counterclockwise
- All voltage select dials turned to their maximum positions (i.e., )

[timb]

Hmmm, that's weird. You may have to unsolder that white and orange-brownish wire from the bottom and top lugs on the back switch board. That's the way I measured my switches, directly from those two wires after unsoldering them.


Sent from my Smartphone

[SharpEars]

Sadly, I decided to present the 2020B with a 0.5 A load and the voltage dropped significantly. The current limit knob is set to its maximum position (i.e., turned all the way clockwise), so I am not being current limited. Also, the voltage range lamp doesn't flash to indicate that current limiting in effect. It is just misregulating.

At a 5 V, almost perfect setting with no load, I am only getting 4.319 V into the 10 ohm resistive load. I removed the bottom panel and noticed that there are three additional trim pots on the main board. Trim pot R8 controls the maximum current output. I have adjusted it for a 2.1 A output using a setting of 5 V, as instructed in the manual. Trimmer pot R104 is a mystery - I have no idea what it does and its function is not stated on the circuit diagram in the Operating Guide.

At larger voltages, such as 15 V, I am getting a drop of something like a volt and a half into a 1.5 A load (this time), again without any sort of current limiting kicking in.

Also, I noticed that the label above the current knob says, "Pull to set." However, the current limit can be adjusted without pulling the knob (tested with both an ammeter connected directly to the DC-/DC+ terminals and with a load present combined with an ammeter in series). What is even weirder is that if I do pull out the Current adjustment knob all that happens is that the current is limited to half of the available current range (i.e., instead of a 0-2 A adjustment, it becomes approx. a 0-1 A adjustment) further reducing the current output capability of the device while the knob remains pulled out. It doesn't make any sense to me. How the heck is this knob supposed to function when it's pulled out vs. pushed in?

Here is a high quality high resolution picture of the front of the 2020B's main circuit board (To see it in full resolution, click on the image and then click again on the resulting image in your browser to zoom in to 100%):



I would like to get the supply's load voltage to match its no load voltage, so any help would be appreciated. Maybe U1b (current limiting amplifier) is bad, but if that was the case, why is the voltage into no load almost spot on? The shape and color of R38 is suspicious, I will measure it to see if it's good.

Update: I thought that R38 (the resistor that looks like a burned marshmallow under the topmost blue trimmer pot) looked a bit weird, but it checks out fine at its rated 0.15 Ohms.

[Vgkid]

That resistor doesn't look good. I would check it.

[timb]

It could be one of the series pass transistors is bad. You can replace any of the TO3 transistors with an easy to obtain 2N3773 and it’ll work fine. Also check Q3, which is responsible for pushing the power transistors.

U1 *could* be bad, but it’s easy to obtain a replacement on DigiKey or Mouser. (You can use the normal plastic DIP package version instead of the $8 metal can package. Though I’ve got a spare metal can version if you want it.)

[timb]

Popped in the new LT1097 and she seems to work fine with one issue: I'm about -26mV at a switch setting of zero, with no more turns on the zero pot left. I guess I could always add some inline resistance to the offset input, right? Or a larger pot.

Either way I'll read up on the LT1097 today and figure out how to modify the 2020B accordingly.

The LT1097 has a much smaller offset null range (+/- 600uV) than the OP-05 (+/- 4mV).  If they were relying on this to correct for other errors which is generally a bad practice, then something else may have to be done.

I took a 100k resistor and hooked it between ground and pin 8 (one of the offset inputs) of the LT1097 and now I’m able to get it zeroed out. Though I’m not sure how this will affect things longterm? Will it hurt the OpAmp? Should I balance things out by sticking the same value resistor on the the other offset input as well?

[David Hess]

Popped in the new LT1097 and she seems to work fine with one issue: I'm about -26mV at a switch setting of zero, with no more turns on the zero pot left. I guess I could always add some inline resistance to the offset input, right? Or a larger pot.

Either way I'll read up on the LT1097 today and figure out how to modify the 2020B accordingly.

The LT1097 has a much smaller offset null range (+/- 600uV) than the OP-05 (+/- 4mV).  If they were relying on this to correct for other errors which is generally a bad practice, then something else may have to be done.

I took a 100k resistor and hooked it between ground and pin 8 (one of the offset inputs) of the LT1097 and now I’m able to get it zeroed out. Though I’m not sure how this will affect things longterm? Will it hurt the OpAmp? Should I balance things out by sticking the same value resistor on the the other offset input as well?

I do not think you will damage it.  You can check the LT1097 schematic which shows how the nulling works.  The change in collector voltage on one side is small.

The reason it is a bad practice is that with these types of operational amplifiers, their input offset voltage drift is minimized when their input offset voltage is minimized so if you null them to a specific offset to correct for other errors, their input offset voltage drift is compromised.  Since Power Designs apparently did exactly this, I assume the increase in drift is not significant.

It still seems a little odd to me so I would try looking for other problems which would cause a large offset but that could be difficult.

[timb]

Okay, this is really fucked up. I accidentally fried the LT1097 (I grounded one of the offset inputs without a resistor attached by dropping a lead; no smoke, just stopped working), so on a whim I stuck the original OP-05 back in and… It’s fucking working. Seriously. Spot on, too.

I’ve honestly got no clue. The only thing that has changed is I don’t have the 10/20V selection switch wired in right now. I wonder if that’s somehow bad? Hmmm.

Or maybe something was messed up with the offset circuit? I un-soldered the adjustment pot, so maybe it was somehow shorting and all this fiddling has fixed it? Without the zero adjust pot even connected, I’m only getting -300uV of offset, which that LT1097 should have easily been able to null out. So, something weird is going on. (Hooking the pot back up with the OP-05 and I’m able get it down to around 5uV of offset.)

This one is truly baffling.

[SharpEars]

Okay, this is really fucked up. I accidentally fried the LT1097 (I grounded one of the offset inputs without a resistor attached by dropping a lead; no smoke, just stopped working), so on a whim I stuck the original OP-05 back in and… It’s fucking working. Seriously. Spot on, too.

I’ve honestly got no clue. The only thing that has changed is I don’t have the 10/20V selection switch wired in right now. I wonder if that’s somehow bad? Hmmm.

Or maybe something was messed up with the offset circuit? I un-soldered the adjustment pot, so maybe it was somehow shorting and all this fiddling has fixed it? Without the zero adjust pot even connected, I’m only getting -300uV of offset, which that LT1097 should have easily been able to null out. So, something weird is going on. (Hooking the pot back up with the OP-05 and I’m able get it down to around 5uV of offset.)

This one is truly baffling.

Well, it's clear that the laws of physics stop at the door to your house  . Well, except for the ones involved in frying parts that is . It's safe to say we've all been there - on both counts.

On a more serious note, I am beginning to strongly believe in the modified adage: Measure twice, replace once - don't assume anything when repairing electronics... Sometimes, eroded solder joints, bad cable connections, misbehaving pots, switches and op-amps, etc..., can lead to the most bizarre circuit behaviors.

[timb]

I know, right? ROFL

My big problem with part frying is my desk is so cramped and I have no real storage for stuff. That's being rectified this week however as I'll finally have all my stuff unpacked, shelves up and a real lab workspace setup.


Sent from my Smartphone

[SharpEars]

OK, the 2020B not putting out the amps it should was due to a bad measurement cable on my part  . Anyways, with that out of the way I have only one remaining issue after fully calibrating it:

The V/A selector switch makes a poor contact when set to show Amps. I have to wiggle the switch a bit before the current measurement shows on the meter. Is it possible to take this toggle switch apart and clean it or do I need to order a replacement?

[timb]

You can get some DeOXIT inside via the back and shaft. Spray some in and toggle it a bunch.

Otherwise it's just a small DPDT switch (ON-ON not ON-OFF-ON).


Sent from my Smartphone

[motocoder]

Hey timb -

One of my friends who is an engineer that has done some work on HVAC products had a great suggestion for a low-impact way to PWM the heater circuit. The idea is actually a very common approach and I'm sort of embarrassed not to have thought of it myself.

The approach is this: Use the unrectified secondary output (AC) to power the heater. The PWM control is via something that only switches on/off near the zero crossing (ZC). Since the heater is wound in a way to be mostly resistive, this can just be a simple solid-state relay with a ZC feature, or a opto-triac + triac. To control the power accurately, you need a zero crossing detector input to the micro. There are many circuits for this out there, but it's fairly simple to build one with an optocoupler and a few discrete components. Then in the micro, you turn om the SSR, count the on pulses of the ZC detector, turn off the SSR, count the off pulses, etc.

If you decide to go this route, read the SSR data sheet carefully. One of the "gotchas" I found when perusing data sheets is that many of the zero-crossing SSR either don't specify or specify very high zero-cross turn-on voltages. Since most of these things are meant to switch 120 or 240V mains, switching on at 20V isn't a big deal for many purposes. However, I've got a 48V AC signal, so switching on at 20V is a big deal. The units I order have a typical turn-on voltage of 2V, and a max of 5V, so that's at least a little better.

This is the ssr I ordered:

http://www.clare.com/home/pdfs.nsf/0/B76903D991990B0885256A2C006BB7A7/$file/Ps1201.pdf

[timb]

Interesting, I just got in an SSR and was thinking of doing what you said, only directly off the 120VAC mains heater input and not the transformer secondary.

Can you see a downside to this?


Sent from my Smartphone

[motocoder]

Interesting, I just got in an SSR and was thinking of doing what you said, only directly off the 120VAC mains heater input and not the transformer secondary.

Can you see a downside to this?


Sent from my Smartphone

I think this is exactly what the original circuit in my 2005A did. I remember there was a circuit attached to the SCR, and I played around with it in LT SPice a bit, and I believe it's basically a circuit to limit turn-on to near the ZC point (the SCR handles the turn-off at ZC automatically, as that's how they work).

The only downside is you're messing with A/C mains voltage with all the associated safety concerns. Oh, and also you'll obviously need significantly more wire, probably need a resistance around the 1500 ohms of the original circuit.

[robrenz]

This path may lead to destroying the ultra low ripple of these units. All this pwm heater stuff may be reflected in the output

[motocoder]

This path may lead to destroying the ultra low ripple of these units. All this pwm heater stuff may be reflected in the output

go read the other thread. That' what all the discussion is about

[SharpEars]

You can get some DeOXIT inside via the back and shaft. Spray some in and toggle it a bunch.

Otherwise it's just a small DPDT switch (ON-ON not ON-OFF-ON).


Sent from my Smartphone

Anything selling on e-bay that might match?

[timb]

Interesting, I just got in an SSR and was thinking of doing what you said, only directly off the 120VAC mains heater input and not the transformer secondary.

Can you see a downside to this?


Sent from my Smartphone

I think this is exactly what the original circuit in my 2005A did. I remember there was a circuit attached to the SCR, and I played around with it in LT SPice a bit, and I believe it's basically a circuit to limit turn-on to near the ZC point (the SCR handles the turn-off at ZC automatically, as that's how they work).

The only downside is you're messing with A/C mains voltage with all the associated safety concerns. Oh, and also you'll obviously need significantly more wire, probably need a resistance around the 1500 ohms of the original circuit.

That’s interesting. I was wondering about the SCR.

Here’s my results from iCircuit:









iCircuit: 2005_Thermostat.cir

I thought they were just using the SCR so the thermostat didn’t have to handle any current.

Anyway, I’m comfortable with 120VAC and don’t mind if I’m using an SSR anyway.

On my 2005, I’m rebuilding the heater coil and need a new thermostat, so my plan was to simply use one of those thermostat ICs and an SSR. I’ve found one that can do 1c of hysteresis which is a lot better than the 10c of the original. I was also going to add one of those small pill shaped bi-metal thermostats set at, say, 85c as an emergency shutoff. I might just wire it inline or have it trip a resettable fuse so the user is aware. (Most likely the latter.)

This unit is for resale, so I’m just looking for a simple, reliable option that will last.

Now, on the 2005A, doing PWM with an SSR is looking like a very viable option.

By the way, I know this is a PD thread, but I thought I’d share some pictures of a heated zener reference from a card in my Fluke 382A Calibrator:







The two power resistors on top heat the metal casing with a thermostatic cutoff in-between them, there’s also a diode buried inside the metal that acts as a temperature sensor I think (black wires going INTO the metal frame in picture two). Interesting setup.

[timb]

You can get some DeOXIT inside via the back and shaft. Spray some in and toggle it a bunch.

Otherwise it's just a small DPDT switch (ON-ON not ON-OFF-ON).


Sent from my Smartphone

Anything selling on e-bay that might match?

I’ll find a suitable match for you, give me a few.

[MarkPalmer]

WoW, that Fluke reference is an exotic looking thing!  I worked on cleaning up the 5020 over the weekend, the top facing side of the supply with the trace side of the board and switches was pretty dirty and grime covered with everything able to flow right down in there.  They cleaned up good.  I cleaned up the component side of the board also, all the IC's look to be dated around mid-1984.  C1, 3, 5 and 34 are out for replacement.  This thing looks to be in pretty good overall shape. 





-Mark-

[David Hess]

I think this is exactly what the original circuit in my 2005A did. I remember there was a circuit attached to the SCR, and I played around with it in LT SPice a bit, and I believe it's basically a circuit to limit turn-on to near the ZC point (the SCR handles the turn-off at ZC automatically, as that's how they work).

I took a detailed look at the circuit and agree.  I wonder why they only used half-wave power.  I assume the heater is shielded well enough from the reference not to couple low frequency noise.

[motocoder]

timb -

I went over this before, but maybe you missed it. I already made a thermostat that you can set the hysteresis to whatever value you want (currently set to around 0.6C). However, that won't get you temperature regulation to within 1C. The problem is, when the thermostat shuts off, a big load of heat has already been transferred from the heater coil to the metal of the tube. This heat continues to soak in, and the heat within the tube continues to rise. I don't recall the exact numbers, but it was very significant - way more than the 1C.

There are at least two ways to fix this: 1)lower the thermal "mass". 2) Introduce a control system that takes this into account and which manages the heat on/off cycles to reach some sort of steady state. That's the whole point of the PID controller, but there are other circuits that will work.

Robrenz - Regarding the PWM, the current plan is for the "PWM" to be very long cycles, and for the on/off transitions to be at or near zero crossings on the AC waveform. I am 100% positive if I ran that off the AC mains, they would be indistinguishable from the original circuit in terms of the disruption they add to the output voltage. And I think even running them off the secondary voltage, the same one powering the output regulator, it's not going to cause issues. I say this because I already have my replacement heater and thermostat running that way, and although I do see fluctuation in the output (but still < 100uV), it is correlated with the overall heat cycle, not the point where the heater turns on/off.

Anyway, that's my theory, and we'll see how it works out if I can ever get my oscilloscope back from Rigol and that nichrome wire that Tim submitted to the black hole that is the US Postal Service.

[David Hess]

Why not just use a linear control loop?  It is not like you can get better efficiency with pulse width modulation.