Ripple Current (and fooling around with a HP 204C Sine Oscillator)

[requim]

I'm looking to buy some capacitors for use in a AC to DC power supply I am constructing for my old HP 204C Sine Oscillator.  I'm looking at the various caps that fit the bill (3300uF 35V) but I can't figure out which ones to use.  I've seen some stuff about having to carefully select electrolytic capacitors for power supplies (I'm using LM317 and LM337 chips). 

Do I want caps with high ripple current or low?  Do I want high impedance or low? Do I want high esr or low?  I assume I want low ESR, low impedance, but I can't figure out whether the ripple current rating is what the capacitor can handle or what it outputs.  If it's what it can handle coming in, then I'd think higher is better, but if it's what it outputs, I'd think lower is better.

[IanB]

Ripple current is the AC current flowing through the capacitor. This will cause internal heating and shorten the life of the capacitor. Therefore capacitors that will be subjected to high ripple currents must be rated for this duty.

[alm]

I wouldn't consider a linear power supply a high ripple current application, however, that would be switchers with their high dI/dt. It probably hardly matters in this application, the main purpose of the caps is so that the voltage doesn't get below the dropout voltage of the regulator, not to remove all ripple. You'd need an insane amount of capacitance for that.

[requim]

So I assume an electrolytic rated for 4.22 amps should be sufficient for a +/- 13V DC power supply on 120VAC mains?

[IanB]

You can make some rough estimates from theory.

For instance if you rectify 60 Hz AC you will get a 120 Hz ripple frequency. Your 3300 uF capacitor now has a reactance Z = 1 / (6.3fC):

Z = 1 / (6.3 x 120 x 3300e-6) = 0.4 ohms

That means for each volt of AC ripple voltage across the capacitor the reactive current will be ball park 1/0.4 = 2.5 A.

Now the ESR will reduce this, and the AC ripple is not a pure sine wave, so there are many uncertainties here. But it seems to me that a ripple current rating of 4.2 A does not give a lot of margin.

Maybe an expert can help out with more informed advice here?

[amspire]

It would really help if you connect the circuit up to your power supplies and measure the current. It has to be low, but it would be good to know how low.

It might be as low as 20mA or it could be 100mA.

But at these currents, anything over 1000uF and you don't have to think about it.  you may as well use at least 100uF as the bigger the capacitor, the less ripple the regulators will see, and so the cleaner the regulated output.

In normal circumstances, ripple current from a rectifier mains transformer into an electrolytic is very high, since all of the current has to flow into the capacitor in the two waveform peaks. The diodes are off for the rest of the mains cycle.  So if you wanted 1A output, you could have a peak-to-peak ripple current of 5A to 10A and converted to RMS, it could be 1 to 2A.

But measure the current. Why you need to know is not for choosing capacitors, but you need to work out if you will need to use heatsinks on the regulators.  If the current is 20mA, you shouldn't need a heatsink. If it is 100mA, you probably will need heatsinks.

Richard.

[requim]

@amspire - I haven't ordered the correct transformer yet nor do I have the LM317 or LM337 chips so I can't hook up the circuit yet to make the test.  I was planning on hooking up the oscillator to one of my bench power supplies tonight to verify whether the unit works or not.  I can test for current draw assuming the unit powers up and functions.  Unfortunately for me I don't have the right connectors to hook up the oscillator to the oscilloscope so I may have to try jerry-rigging that until the right connectors arrive.

If I can't do that prior to ordering the parts, then I'd like to be careful since the caps are not inexpensive + shipping + tax. I'd rather not have multiple orders for substitutes.

[amspire]

@amspire - I haven't ordered the correct transformer yet nor do I have the LM317 or LM337 chips so I can't hook up the circuit yet to make the test.  I was planning on hooking up the oscillator to one of my bench power supplies tonight to verify whether the unit works or not.  I can test for current draw assuming the unit powers up and functions.  Unfortunately for me I don't have the right connectors to hook up the oscillator to the oscilloscope so I may have to try jerry-rigging that until the right connectors arrive.

Usually there are components going to ground and the supply rail that you can easily clip leads to.  Otherwise, just solder wires to the PCB for the supply connections.  You don't have to order special connectors just for testing.

Richard

[requim]

I was planning on using the clips on the connectors that I got for my power supply.  It's banana on one side and some sort of long skinny shrouded clip on the other.  I figured they should clip right onto the pcb connector without any issues.  As for the oscilloscope I don't have a dual banana plug to bnc connector so I can't get the signal from the oscillator to the oscilloscope w/o some effort.  Needless to say I am ordering the connector tonight along with some alligator clips/leads.

[requim]

I went ahead and connected the wires and powered it up, but nothing happened.  Twice in one day.  First with the Tek 2235 and now with the HP 204c.  I figure I'll test everything on the HP 204C since there aren't that many parts on the board -- starting with the caps first of course and go from there.  Looks like it's desoldering time.  Huzzah!

And yes I probed the unit while it was on and voltages do appear to be flowing through it..  I didn't test for amperage because I assumed it wouldn't tell me much since it's not actually functioning at the moment.

[requim]

I feel like I'm talking to myself..

So I tested the caps on the board (well most of them).  However I don't trust my results.  If someone would be willing to scan through and tell me whether my results are feasible or whether I made some mistake in the process I would appreciate it.  FYI I did desolder one end of each cap I tested.  I used a Fluke 87 V to perform the measurements.  I also double checked some of the really whacky ones.

C1 - fxd cer 0.01 uF +80% -20% 100 vdcw -  11.7uF
C7 - fxd Ta elect 2.2 uF +/- 10% 20 vdcw -  2.16 uF
C9 - fxd TiO2 1 pF +/- 10% 500 vdcw - 4.24 uF
C10 - fxd cer 0.01 uF +80% - 20% 100 vdcw - 11.4 uF
C11 - fxd Al elect 300 uF +100% -10% 10 vdcw - 434 uF
C12 - fxd Ta elect 2.2 uF +/- 10% 20 vdcw - 22.1 uF
C14 - fxd Ta elect 3.4 uF +/- 20% 35 vdcw - 3.39 uF
C16 - fxd Ta elect 0.47 uF +/- 10% 35 vdcw - 0.521 uF
C17 - fxd Ta elect 22 uF +/- 10% 15 vdcw - 20.7 uF
C18 - fxd Ta elect 22 uF +/- 10% 15 vdcw - 20.4 uF
C19 - fxd Ta elect 6.8 uF +/- 10% 35 vdcw - 4.01 uF
C20 - fxd Al elect 100 uF +75% -10% 10 vdcw - 84.4 uF
C21 - fxd Ta elect 2.2 uF +/- 10% 20 vdcw - 2.24 uF

I could not find C13 or C15 on the board.  The schematic showed it but it looked like a diode or resistor was where the schematic indicated it should be.  The others I didn't test because I was going to have to take some stuff apart.  I plan on doing it, but wanted to post my preliminary results.

Also keep in mind this unit was manufactured sometime around 1977 - 1978.

Oh one more question - The jacks to get the waveform from are labeled as 600 ohm.  In order to hook this up to an oscilloscope (Tek 2225) do I need to do anything special?  i put a T on the oscilloscope and terminated using a standard 50 ohm terminator.  Will this work?  And if not, what am I supposed to do?

[amspire]

I went ahead and connected the wires and powered it up, but nothing happened.  Twice in one day.  First with the Tek 2235 and now with the HP 204c.  I figure I'll test everything on the HP 204C since there aren't that many parts on the board -- starting with the caps first of course and go from there.  Looks like it's desoldering time.  Huzzah!

And yes I probed the unit while it was on and voltages do appear to be flowing through it..  I didn't test for amperage because I assumed it wouldn't tell me much since it's not actually functioning at the moment.

The best way to get help on the forum with debugging is take a grab a copy of the schematic, and start writing down voltage that you are measuring in the circuit.  Do most of the measurements with the black multimeter lead on zero volts. If you are measuring between two points, show the points on the you are measuring between.  Good places to measure are the supply rails, and base emitter and collectors of each transistors.

Anywhere else on the circuit as well that you think might be useful.

The problem is very unlikely to be capacitors.

Once we can see the voltages, it is often pretty easy to see a problem.

Richard

[requim]

Can do.  What do I make of the cap readings?

[amspire]

Well C1 and C10 were marked as 0.01uF, but it is possible HP changed them to 10uF capacitors. It would do no harm.

The one that makes no sense is C9. It has to be 1pF. It cannot possibly be 4.2uF.

When you were measuring capacitance, did you press the Zero button with the leads open before measuring?

Also you cannot be touching the leads of the 1pF as you are testing - it will be hard enough to take a sensible reading of it anyway.

Just to give you an idea - take two insulated wires and twist them together one turn. That is probably more then 1pF. So just the positions of your multimeter leads can give a big error to the 1pF reading.

I just noticed the schematic already has lots of voltages on it.  That is a big head start. Any voltage that doesn't match is a big "Look Here" sign.

Richard.

[david77]

The measurement for C12 also makes no sense, maybe a typo?
I'd probably replace C19 & C20 they're a bit low.

[requim]

C12 isn't a typo on my part.  I triple checked the value on the DMM.  It may be an issue with the part list / schematic.  C13, C17, and C18 are spec'd as 22.0 uF parts.  Unfortunately I couldn't find C13 on the board and C17 and C18 were both a little low but in spec.

@amspire - No I did not zero out the leads before doing the tests.  Didn't know I needed to do that.  As for my technique I only touching the plastic handles on the probes when the measurements were taken.  I will re-test the caps and compare my new measurements to my old.  I will then proceed to resolder them back into place and do the voltage comparisons.

Thanks for the tips.

[amspire]

Yup. The meter needs to be zero'ed with open leads, so it reads 0 before you start measuring.

Otherwise you are getting a false reading.

[FenderBender]

Higher the ripple current rating the better for this application. It won't heat up and degrade if it has higher RC.

[requim]

So what's a reasonable ripple current?  I saw on 1000uF caps they could go over 10 amps but on 3300uF caps they only went to 4.22 amps or so.  The 10+ amp caps were very expensive (I think somewhere in the neighborhood of > $20 each).  I would have assumed higher capacity caps would handle more ripple current, but from my sample size of two it appears not to be the case.

[requim]

Updated measurements zeroing out the meter.  I've left the original recorded values and added the new ones to the end and put an asterisk next to the ones with significant changes:

C1 - fxd cer 0.01 uF +80% -20% 100 vdcw -  11.7uF - 10.20 nF *
C7 - fxd Ta elect 2.2 uF +/- 10% 20 vdcw -  2.16 uF - 2.16 uF
C9 - fxd TiO2 1 pF +/- 10% 500 vdcw - 4.24 uF - 4.24 uF
C10 - fxd cer 0.01 uF +80% - 20% 100 vdcw - 11.4 uF - 10.04 nF *
C11 - fxd Al elect 300 uF +100% -10% 10 vdcw - 434 uF  - 434 uF
C12 - fxd Ta elect 2.2 uF +/- 10% 20 vdcw - 22.1 uF - 22.3 uF
C14 - fxd Ta elect 3.4 uF +/- 20% 35 vdcw - 3.39 uF - 3.40 uF
C16 - fxd Ta elect 0.47 uF +/- 10% 35 vdcw - 0.521 uF - 0.520 uF
C17 - fxd Ta elect 22 uF +/- 10% 15 vdcw - 20.7 uF - 20.7 uF
C18 - fxd Ta elect 22 uF +/- 10% 15 vdcw - 20.9 uF - 20.8 uF
C19 - fxd Ta elect 6.8 uF +/- 10% 35 vdcw - 4.01 uF - 4.0 uF
C20 - fxd Al elect 100 uF +75% -10% 10 vdcw - 84.4 uF - 84.1 uF
C21 - fxd Ta elect 2.2 uF +/- 10% 20 vdcw - 2.24 uF - 2.24 uF

The only two that are significantly different are the ceramic caps, and I think the two reasons why they changed significantly are that (1) I read/wrote the wrong units down and two, I zero'd out the meter this time which made an obvious difference since they're towards the bottom of the meter's capacitance scale.

So the odd value for C12 is still true.  I just wanted to post the results of the second measurement.  I will now proceed to resolder the caps back in and then probe for voltages.  The journey continues...

[requim]

Before I soldered the caps back on I decided to desolder/resolder Q9 - TSTR: Si PNP 2N3906 - the transistor from my earlier confusion and that had the ugly solder joints.  I tested it according to the following video on Youtube:



While I can't identify which pin is which function (base, emitter, collector) I can say my results didn't match his.  My red lead gave positive results on two different pins while black was on the same pin.  All other combinations came up 0L.  From the video it appears that I should only have received a positive response with red on one position and black on the other two positions.

Is it safe for me to assume the transistor needs to be replaced or should I continue to resolder everything and gather voltages?  Or should I desolder two pins from all the transistors and test them in the same manner?

[david77]

I'd say the transistor is ok.

The 2N3906 is a PNP transistor. If you lay it on the table, flat side on top and the wires facing you the left pin is
Emitter, middle is Base and right is Collector.

If you now attach the negative (black) lead to the middle (B) and touch the left (E) or the right (C) pins your meter
in diode test mode should show you the Uf of the diode you're measuring (roughly 500...800 on the display).
If you reverse this setup the meter should always show overflow.
When you connect the probes to the left (E) or right (C) pin your meter should show overflow, same when you switch
the probes around.

You have to imagine a bipolar transistor like two diodes in one case. The base is the point where both diodes are
connected. In a NPN Transistor B is the anode side of both diodes and E and C are the cathode side of each diode.
In a PNP transistor this is the other way round. B is cathode and C and E are the anode sides of the diodes.

[amspire]

Is it safe for me to assume the transistor needs to be replaced or should I continue to resolder everything and gather voltages?  Or should I desolder two pins from all the transistors and test them in the same manner?

Put it back and start testing the voltages. Nothing disastrous will happen even if the transistor is faulty, and the voltages will tell the story.

The very first voltages you want to test are the supply voltages on the PCB, to make sure that the +13/-13V you think you have applied has actually been applied.

If you cannot find components, like some of those capacitors, you can work out where they are by looking at the circuit. If they connect to a switch and are not on the PCB overlay, they will either be mounted on the switch, or they will connect somehow to the wire that comes off that contact on the switch. Just follow the wires and connected PCB traces and you will find it.

As I said before, the schematic tells you many of the voltages that should be there.  If you are not sure about which lead is which on a transistor, look at which other components a lead connects to on the PCB, and then compare it with the schematic. So of a transistor lead connects to R123, and R123 connects to the base of the transistor in the schematic, then you have found the transistor base lead.

I think that 2N3902 was in the Automatic Gain Control (AGC) circuit. The job of that circuit is to adjust the gain in the oscillator circuit to exactly 1.000 . If it is exactly 1, you have a beautiful stable low distortion sinewave.  If it is more then 1 , you have a badly distorted sinewave. If it is less then one, you have no oscillation. The best way to see what is happening in the AGC is to measure voltages.

Richard

[requim]

So as it turns out the reason why the Sine Oscillator seemed to be broken was because when I was outputting the signal to the oscilloscope I had the channel mode set to Ground.  Oops.. Still learning how to use the basic equipment.  The unit appears to draw 410 mA, which is just slightly above the spec'd amps in the manual (400 mA). 

The quality of the sine wave generated is poor.  If you would like I can take a video of it and post it to youtube or something.  I get what I would assume are transient spikes in the output and the sine wave is anything but smooth.  I probably need to go over a tutorial again on using an oscilloscope since I can't seem to get clean output.  The best I can do is get the wave to move backwards in a relatively smooth manner.  Any other combination of settings I choose seems to make it go way to fast and blurry.

My primary question is this -- Are the transient spikes fixable?  And can I get a smoother sine wave out of this, or is it due to the design?  I have yet to record all the voltages and match them up against the schematic.  I can do that if it will help.  At this point I did what I wanted to do which was to figure out whether the unit worked or not, and it does appear to work though it obviously has an issue or two.

I have yet to test the second 204C that I received so I guess now I'll pull it out and try to turn it on and see what happens.

Another thing which I am confused by -- On my two HP e3610a's that I used to test this, I set the voltage to 13 V for each unit, which it appears to be doing fine, however the amperage that shows on the display of each unit only shows 0.01 amps, however when I tested it using the Fluke 87 I got .410 Amps.  I thought the amperage flowing through the power supply would be show up on the power supply display.

[amspire]

400mA is way too high. This its an oscillator that can run for 300 hours from an internal battery, so I would guess that less than 50mA is necessary.

If you compare the voltages to the schematic, you will find something wrong.

Richard