bench testing of incandecent soft start ckt

[NewBie]

As I recall, cells don't sag linearly with load current, so from what I figure, you are what, a little over 2x for the surge current, and probably less than 3x, right?

 

[wquiles]

Not with the WA1319, but with the 3xD M*g bulb I was able to more-or-less measure at least 15Amps peak with no circuit (this with the very low impedance CBP1650 cells). I would expect that with the WA1319 it would be even higher (since the steady state current in the WA1319 is twice as large!).

With the circuit in place, my last measurements (which I was not able to capture properly on the scope ) gave me an indication that with a steady-state current right at 2Amps for the WA1319, the peak pulsed current was about 4-6Amps, very much aligned with your expectations. I really need to buy a few of the TI/Burr-Brown Current Shunt Monitor chips you recommended so that I can make accurate current measurements

EDIT: I just ordered a few of these chips, along with a few precision resistors (0.01 and 0.001) to go with it

Will

 

[NewBie]

Not with the WA1319, but with the 3xD M*g bulb I was able to more-or-less measure at least 15Amps peak with no circuit (this with the very low impedance CBP1650 cells). I would expect that with the WA1319 it would be even higher (since the steady state current in the WA1319 is twice as large!).

With the circuit in place, my last measurements (which I was not able to capture properly on the scope ) gave me an indication that with a steady-state current right at 2Amps for the WA1319, the peak pulsed current was about 4-6Amps, very much aligned with your expectations. I really need to buy a few of the TI/Burr-Brown Current Shunt Monitor chips you recommended so that I can make accurate current measurements

EDIT: I just ordered a few of these chips, along with a few precision resistors (0.01 and 0.001) to go with it

Will

Here is something that will help with that scope you are using, if you haven't done it.

To capture the peak surge, without the soft start, turn on glitch detect, and dial your time base down to 1 nanosecond per division. If you have not done that, you might be surprised.

 

[andrewwynn]

nice (current chips n resistors).. def. a little more resolution than my power supply lead = sense resistor solution (however that is amazingly accurate and simple!)..

I get a 'base' resistance by cranking a known current (3A) through the wire and using my fluke 87III set to 4 1/2 digits to measure the milivolt drop.. I'll get a reading like 120mV.. and simple division will get me 3A/120mV = 40mOhm resistance.

Since the power supply lead is in the equation anyhow, it works very good for me.. and i can measure current spikes otherwise impossible.

I did notice that when i finally got the soft start working on the hotdriver.. that the power supply would no longer click into current limit every time i turned on the light. Even when i could still measure some pretty high startup peaks at 1msec.. there was not enough energy behind that tiny spike.

I'm sure the nsec spikes have to be insane!

-awr

 

[NewBie]

I'm sure the nsec spikes have to be insane!

-awr

You can figure roughly where they get to, if you add the battery internal resistance (and in parallel they don't add, but two cells makes 1/2 the resistance), measuring the resistance of your current path, and the cold resistance of the bulb.

Then just divide the battery voltage by that total. Thats pretty close to where your initial current spike will be.

One of the H4 bulbs I have measures 0.075 ohms cold. The current path in the light has been modified to get it down to 0.050, and the cell resistance for the UPS cell is 0.012 ohms. We have a total of resistance of 0.137 ohms, and the battery voltage is 12V. 12 V / 0.137 Ohms = 87.6 Amps which agrees with a very fast 15 GHz scope, and high bandwidth probe. (It actually measured higher, as the cell was a little over 13V before the switch was closed)

 

[wquiles]

Here is something that will help with that scope you are using, if you haven't done it.

To capture the peak surge, without the soft start, turn on glitch detect, and dial your time base down to 1 nanosecond per division. If you have not done that, you might be surprised.

I have not done that yet. Sure looks like "fun". I will try to sneak to the garage for a little while today if/when I get both of my kiddies to sleep in the afternoon

I did notice that when i finally got the soft start working on the hotdriver.. that the power supply would no longer click into current limit every time i turned on the light. Even when i could still measure some pretty high startup peaks at 1msec.. there was not enough energy behind that tiny spike.

I'm sure the nsec spikes have to be insane!

-awr

Yep, even though not measuring current, just looking at the negative voltage spikes I was able to capture really gives you an indication of how hard the batteries get hit during the initial ON cycle :naughty:



Will

 

[wquiles]

Newbie,
Andrew,

I tried once again today to capture the elusive pulsed current during initial turn ON. I got some graphs, but the values from the scope don't make sense, as I will explain below.

I have a shunt resistor for many months now, but I have not been able to measure accurately its resistance value, supposed to be 0.001 ohms or so, back a million years ago then it was first put into service!:

Along with the scope that I am borrowing from work, I also got a presicion, calibrated bench supply:

I used a wire-wound, 3.9Ohm, 25W resistor (large white block on this picture) along with the bench supply to find out the shunts' value:

Since the bench supply measures the current at any given set voltage, I took some measurements at 1A and at 1.5A, and it seems like the value of the shunt is somewhere around 0.0218 Ohms (averaging the two values recorded). The power resistor is also shown here above my hand-written notes:

I then moved to playing with the shunt resistor in-line with the bulb to see what happened. First, to get a baseline, I first measured the steady-state current without the shunt resistor using the power supply instead of the battery pack since I wanted really steady values (values shown in the power supply's built-in displays):
Vps=5.12V ; Ibulb=1.905A
Vps=5.62V ; Ibulb=2.009A

I then put the shunt resistor in series with the bulb, and I was able to measure the operating steady-state value of current and voltage for the WA1319 bi-pin bulb, again using the power supply until I got the same 1.9 and 2.0 Amps as before (values shown in the power supply's built-in displays):
Vps=5.12V ; Ibulb=1.918A ; Vshunt=52mV (calculated Rshunt=0.0271 Ohms)
Vps=5.52V ; Ibulb=2.000A ; Vshunt=54.2mV (calculated Rshunt=0.0271 Ohms)
These higher values of Rshunt make sense since these were taken on the actual breadboard and the breadboard contains additional small wires to make interconnections, so everything seems good so far .

Now I used the scope to measure the spike of voltage right across this Rshunt (0.0271 Ohms), but this time I used the battery pack instead of the PS, since that is the end goal after all. In all of these screen captures, the top two traces are the instantaneous voltage at each side of the resistor, and the red (lower) trace is the scope's math of substracting these two channels. The top traces are at 1V/div and the red trace is at 100mV/div.

This first one, at 20mS/div, shows the peak value to be 378mV, which would correspond to a peak current of 0.378/0.0271=13.94Amps, which we know is too high since the circuit is operational:

This next one, at 400mS/div, shows the same signals, but also the steady-state voltage at an average 242mV, which would correspond to a current value of .242/0.0271=8.9Amps, which we know its wrong since the steady state current has been measured by many methods/meters, and with a fresh battery pack it is always between 1.9 and 2.0 Amps.

So clearly I am either doing something wrong, or not setting up the scope properly, or the scope is limited in its ability to perform this in-circuit measurement. I am too tired to keep trying for ever, since I want to move on to the simple LDO regulator from AWR, so I will likely leave it as is for now. At least I now have an accurate value for my shunt resistor so that I can use it in the future

Will

 

[NewBie]

I'd try next time, just hooking the scope probe gnd on one side of the resistor, and the other end of the same probe, on the other side of the resistor. This should work alot better than what you were trying to do.

Comments....

Normally you have to use identical probes and do a "calibration", and once done, with no power, you'd have zero volts.

You have about 160mV offset, with no power.

Another thing I see there is alot of noise, so thats going to make measurements tough.

You'd need to subtract the offset from your measurement...
" correspond to a current value of .242/0.0271=8.9Amps,"

0.242 - 160mV offset = 82 mV 0.082/0.0271= 3.03A, which is alot closer, but not exact yet.

But you know the current is 2A, so we'd be able to guestimate a correction factor of 2/3. 2/3 * 3.03 = 2.02A, for the average current.

Looking at the surge, I see 115mV above the offset average.

0.115/0.0271 = 4.24 A * 2/3 (the scaling correction factor) 2.8 A surge.

But the scope is on a very slow 400mS per division, so it is not likely to catch the leading edge of the surge anyhow.


Anyhow, on the lighter side of things, I remembered from the 1970's, a special circuit that was used to make light bulbs last much longer than normal. It was used in areas where there was limited access, accessing them was dangerous, or where you were using a bulb as a flasher, and wanted to get a long life out of it. So I searched high and low, and finally found the circuit. It is another softstart circuit, a little more complicated:

 

[wquiles]

I'd try next time, just hooking the scope probe gnd on one side of the resistor, and the other end of the same probe, on the other side of the resistor. This should work alot better than what you were trying to do.

Good idea - easy to try indeed. Can I also move the shunt to the "ground" side of the lamp's circuit path, from the source of Q1 to ground, or would (as I think more about it) affect the bias of Q1 enough to affect the circuit's behavior?

Comments....

Normally you have to use identical probes and do a "calibration", and once done, with no power, you'd have zero volts.

You have about 160mV offset, with no power.

Another thing I see there is alot of noise, so thats going to make measurements tough.

You'd need to subtract the offset from your measurement...
" correspond to a current value of .242/0.0271=8.9Amps,"

0.242 - 160mV offset = 82 mV 0.082/0.0271= 3.03A, which is alot closer, but not exact yet.

But you know the current is 2A, so we'd be able to guestimate a correction factor of 2/3. 2/3 * 3.03 = 2.02A, for the average current.

Looking at the surge, I see 115mV above the offset average.

0.115/0.0271 = 4.24 A * 2/3 (the scaling correction factor) 2.8 A surge.

But the scope is on a very slow 400mS per division, so it is not likely to catch the leading edge of the surge anyhow.

Totally awesome - you are d'man :bow:. Now things start making sense. I will try to look into this aspect for sure since it does affect "all" measurements on the scope a little :thumbsup:

Anyhow, on the lighter side of things, I remembered from the 1970's, a special circuit that was used to make light bulbs last much longer than normal. It was used in areas where there was limited access, accessing them was dangerous, or where you were using a bulb as a flasher, and wanted to get a long life out of it. So I searched high and low, and finally found the circuit. It is another softstart circuit, a little more complicated:

Interesting circuit - not all that complicated. However, not as good as using a true MOSFET, right? Maybe the transistor(s) were heat-sinked also?

Will

 

[NewBie]

Since your light is battery powered, wherever you connect the scope ground becomes ground, if you don't have the other probes connected... Since you just need a current measurement, you could hook the ground to the sense in the high side, and the tip on the other side of the resistor.

You don't get all the other traces this way, but what you are really after is current anyhow.

The No.44 bulb draws 0.25A and the No. 47 draws 0.15A.

With a little thought, you could modify for a 2N3055, or more thought would get you to a MOSFET for the drop transistor.

 

[andrewwynn]

I'm not sure you measured your shunt resistance correctly.. if you are calculating based on the 3.9ohm power resistor (i.e. subracting voltage).. not the best way.. just measure with 4 1/2 digits and a multimeter the voltage drop on the Rsense.. hooking up separate leads right on the shunt terminals for the meter.. it looks like you might have the leads down to the breadboard in the equation which will throw off your measurements dramtically.

My bench supply will output a preset current regardless of voltage.. i actually set the current by shorting the leads and dialing the current.. If you don't know if yours can handle that it's best to stick the 3.9ohm in the ckt as well.. but just put your two resistors in series and with a known measurement of current (your bench supply displays the current).. now measure the mV drop on just the shunt resistor.. I'm pretty confident you'll have a much lower measurement of resistance.

Even with the fancy driver in the hotdriver and using slow start.. there is an initial spike of current that can be double-digits. it has no energy in it just voltage akin to a static spark can be 15,000 V but won't hurt you.

I think you need to re-calibrate your shunt.. it should be like a 100 or 200A shunt based on the apparent physical size, but the value you are saying calculates to more like a 2A shunt.. (should have 35mV at full scale).. 17.5mohm = 2A shunt.

You will be quite amazed to see the difference when you measure the shunt directly w/o wires.. you can't have any add'l wires in the measurement.. you probably have at least 3-4mohm of wires.. and each contact on a breadboard can be at least 10-20 more mohm.. a shunt that big typically has a resistance of 3-4mohm.. a 100A shunt is only 35/100ths of a mohm! consider that an entire inch of 18ga wire is merely 1/2 a mohm of resistance.. and now look at that big block of metal..

If i had to guess you have a 50 or 100A possibly 200A shunt there.. it should be on the order of 1/3 mohm to 3.5mohm of resistance.

put your 3.9ohm and the shunt in series with the power supply set to a precise current.. 2A or 3A is good.. hook up some leads separate to the shunt terminals and measure the voltage... you will find you introduced a LOT of resistance you didn't realize.. looks like you have a switch or two possibly in the loop.

I noticed when i tried to take voltage measurements in my hotdriver on the breadboard.. that i couldn't get accurate voltage measurements on the FET until i SOLDERED the test leads.. so when i need accuracy that's exactly what i do.. tack on some 30ga wire-wrap wire to my terminals.. the only way to be sure.

-awr

 

[wquiles]

I'm not sure you measured your shunt resistance correctly.. if you are calculating based on the 3.9ohm power resistor (i.e. subracting voltage).. not the best way.. just measure with 4 1/2 digits and a multimeter the voltage drop on the Rsense.. hooking up separate leads right on the shunt terminals for the meter.. it looks like you might have the leads down to the breadboard in the equation which will throw off your measurements dramtically.

My bench supply will output a preset current regardless of voltage.. i actually set the current by shorting the leads and dialing the current.. If you don't know if yours can handle that it's best to stick the 3.9ohm in the ckt as well.. but just put your two resistors in series and with a known measurement of current (your bench supply displays the current).. now measure the mV drop on just the shunt resistor.. I'm pretty confident you'll have a much lower measurement of resistance.

For the actual measurement I setup the simple circuit as in my hand-written notes, outside of the soft-start circuit, so I have minimal stuff in the way. I actually setup the PS (which measures the output current) to output exactly 1Amp and then 1.5Amp and used my Fluke 87 in high resolution mode to read the mV drop at the shunt. I am fairly confident that the average value of 0.0218 Ohms is pretty close, although of course it "does" include the two short wires coming out of it.

I think you need to re-calibrate your shunt.. it should be like a 100 or 200A shunt based on the apparent physical size, but the value you are saying calculates to more like a 2A shunt.. (should have 35mV at full scale).. 17.5mohm = 2A shunt.

At least on the little sticker (barely visible) it says that the shunt (when new?) was supposed to be 0.00065 Ohms, and it is rated at 5Amps max.

You will be quite amazed to see the difference when you measure the shunt directly w/o wires.. you can't have any add'l wires in the measurement.. you probably have at least 3-4mohm of wires.. and each contact on a breadboard can be at least 10-20 more mohm.. a shunt that big typically has a resistance of 3-4mohm.. a 100A shunt is only 35/100ths of a mohm! consider that an entire inch of 18ga wire is merely 1/2 a mohm of resistance.. and now look at that big block of metal..

Maybe I am missing the point, but a shunt with no wires attached to it means it is a nice piece of metal in my bench. The only way I can insert a shunt this big into a circuit for measurements is to actually have some wires attached to it, just like I show in my pictures. That is why when making my measurements for the shunt I included the value of those small wires into my measurement. True, 0.0218 Ohms is not the "real" value of the shunt by itself, but I honestly don't care what is the shunt's value by itself. Am I trully missing the point you are trying to make here? Please help me understand what you meant.

I noticed when i tried to take voltage measurements in my hotdriver on the breadboard.. that i couldn't get accurate voltage measurements on the FET until i SOLDERED the test leads.. so when i need accuracy that's exactly what i do.. tack on some 30ga wire-wrap wire to my terminals.. the only way to be sure.
-awr

That is a great tip indeed. I must do this next time - the little interconnect wires in the breadboard can't be condusive to accurate values, right?

Will

 

[wquiles]

Since your light is battery powered, wherever you connect the scope ground becomes ground, if you don't have the other probes connected... Since you just need a current measurement, you could hook the ground to the sense in the high side, and the tip on the other side of the resistor.

You don't get all the other traces this way, but what you are really after is current anyhow.

Thanks. I seem to vagely remember from my EE days (10-15 years ago!) that I had to be careful with conecting the ground of the scope right across a resistor when the actual circuit ground was somewhere else. However, you are right - since the circuit is battery powered, as long as I don't have any other instruments connected while making this ONE measurement, the whole circuit should really be "floating" regarding the "real" ground of the scope

The No.44 bulb draws 0.25A and the No. 47 draws 0.15A.

With a little thought, you could modify for a 2N3055, or more thought would get you to a MOSFET for the drop transistor.

Ahhhh - that explains the smallish transistors and not using a MOSFET like I do. In the circuit I am using my currents at at least 1Amp, 2Amps with the WA1319, and possibly up to 3-4Amps steady-state with other WA bulbs, so the MOSFET was more of a "requirement"

Will

 

[photo2000a]

this was agreat post, i noticed the 3 transistors in the pix and said himm sure enough modern electronics those were great magazine i had a ton of fun building those projects even build this soft start way back then i was only a kid guess i just loved lights

 

[wquiles]

photo2000a - Glad this post was able to revive some interesting memories


Andrew,
You were right about Rsense being hard to use due to contact resistance. I had to re-arange the place where I was inserting Rsense since I was not getting consistent reading due to variable contact surface on the breadboard!!!. Once I was able to test with the battery and with the bench PS (at the same steady-state current), I was able to measure again and came up with Rsense = 0.0265 Ohms, which of course includes the short wires from the shunt to the breadboard.

Newbie,
Your idea worked great. I was able to measure the Voltage drop across Rsense with the scope (with nothing else attached!) and got the following shots with less noise and now at only 50mV/div (and 100mS/div). The first one shows the measurement cursor on the peak value (82mV) and the second shows the average value (54mV). Based on the accurate measurements above, these are about 10% higher than they should be, so once I account for this, the average value was about 1.851Amps (the battery has discharged some, so I am not at the 2.0Amp rate) and the peak would be about 2.9Amps or so. So even with a full pack, a max. of about 3 to 3.5Amps (with the soft-start circuit in place, of course) beats the 15Amps that I measured earlier with no circuit in place. Of course, at some point I will have to re-do the direct drive peak pulsed current now that I have a working Rsense :

Will

 

[NewBie]

That looks much better:

When you do the deal without the soft start in there, get some shots at 1nS per division, and 1uS, with glitch detect on...

Glad to be of help.

I think what Andrew is trying to say, in engineering terms, is called a Kelvin connection (for the sense resistor). That should ring some bells and shake out the dust upstairs...or what do they call it up north...oh, permafrost on the brain.

 

[andrewwynn]

that little hump is exactly what i'd expect from the ckt.. absolutely wonderful.

obviously to include the Rsense in the ckt you need to use the short wires to the BB.. if they are 24ga wire, that would be about 1/2 the resistance of your whole Rsense.. it would be better to just use a loop of wire... 4.67 inches of 24ga wire is 10 mohm.. very easy to calculate from that.. whatever the mV measurement is.. divide by 10 and get A.. the key is you have to measure right on the wire, not through the contact.. if you are taking the measurements from another lead that is across a breadboard contact you can estimate 5 to 10mohm PER CONTACT.

you should get a measurement right to the posts on the shunt.. it's ok that it uses the wires but you can't get an accurate measurement if using them.. though you basically made your own resistor including the wires etc.. if you want accurate you have to measure right no the posts of the shunt.. just add two more pigtails going nowhere that you can use with your meter leads.

-awr

 

[wquiles]

that little hump is exactly what i'd expect from the ckt.. absolutely wonderful.

Thank you . I could have not gotten here without your help :bow:

obviously to include the Rsense in the ckt you need to use the short wires to the BB.. if they are 24ga wire, that would be about 1/2 the resistance of your whole Rsense.. it would be better to just use a loop of wire... 4.67 inches of 24ga wire is 10 mohm.. very easy to calculate from that.. whatever the mV measurement is.. divide by 10 and get A.. the key is you have to measure right on the wire, not through the contact.. if you are taking the measurements from another lead that is across a breadboard contact you can estimate 5 to 10mohm PER CONTACT.

you should get a measurement right to the posts on the shunt.. it's ok that it uses the wires but you can't get an accurate measurement if using them.. though you basically made your own resistor including the wires etc.. if you want accurate you have to measure right no the posts of the shunt.. just add two more pigtails going nowhere that you can use with your meter leads.

OK, now I get what you were trying to tell me . I will add the extra pigtails and will measure the voltage drop right on the shunt this weekend. Thanks

Will

 

[wquiles]

That looks much better

When you do the deal without the soft start in there, get some shots at 1nS per division, and 1uS, with glitch detect on...

Glad to be of help.

I think what Andrew is trying to say, in engineering terms, is called a Kelvin connection (for the sense resistor). That should ring some bells and shake out the dust upstairs...or what do they call it up north...oh, permafrost on the brain.

Thanks very much Newbie :bow:. I have been very frustrated until now trying to get this important measurement, but you "really" helped me understand how get a good current measurement with the scope

Yes, when I run the Direct Drive, I will use the glitch function of the scope (which I finally found, by the way ).

Will

 

[NewBie]

Tektronixs does a great job of hiding their features under menus, don't they?