# Where did you personally use or see LM317 applied to?

#### q12x

Joined Sep 25, 2015
1,689
I took your advice and I re-make the test but this time at 9V as you suggested.
Here is the data:

I changed the order as it was in the anterior table. Data for 5V is the same as before. I even double check it with this occasion.
What this data means and how I measured:
PSU(V) = I took 2 Voltage measurements, for 5V as before and this new 9V. You can see a long thick line between these 2 Voltages.
ViVo(V)= The voltage across Vreg (see img)
PSU(A) = the amperage I read on my PSU display (while turning POT1 in my cct)
VReg(W) = is the calculated Wattage over the VReg, using ViVo*PSU(A)=VReg(W)
Vo = VReg_Vo to Gnd = for a "constant" 3V
*C= VReg temperature in *C, this time constant reading
Time min = the time I measured temperature in minutes (some values are approximately)
- You were right to insist (but not hard or loud enough) that I should get more close to "operating things within their specifications. LM317 differential voltage operating range is 3-40V; not less than 2". I totally miss it, while concentrated on "bigger" things. Very good observation. It did kicked me in mi balls a bit, but I get up and do it anyway.
- Now, what this new reading is telling me? Well... Line4 is having the same Vreg(W) as Line1. The same goes for Line5 and Line2, the same Vreg(W). That was my intention ! To get the same W, but for these 2 different Voltages(5V and 9V). What this told me? Well, the Vo for Line1 and Line4 is the same. But Vo for Line2 and Line5 is different, in the sense that at 9V I dont get too much derating down as for 5V in Line2. Its still going down at 9V, but less evident as with the 5V.
So this test is telling me you were right......but not completely, because it worked fine in 5V mode as in the 9V. In 5V mode is getting a bit more evident that is going down from the regulation. While in 9V mode is not that evident, but it is still giving clear signs of ramping down, but at a less steep angle than in 5V mode. Interesting how increased Voltage is strenghtening this ramping down.
- One very big issue was again the POT range when using at 9V. It was very dificult to get to those precise PSU(A) values like that 0.10A and 0.18A because it was jumping like crazy. Because I used 1R5W instead of the 5R. At bigger Res, I get a smoother POT travel (but less A range).
- So in conclusion, it is ok at 5V as well, but like you said, getting lower than 3V across VReg, is not correct so I will have to remember this detail for the rest of the experiments over Vregs in general in the future. See? Unwanted lesson ! Hahaha.
- Getting my hands dirty with these experiments, is a very good thing, at least this is the way I learn best ! This is my way, and you can quote me any time, "Q12 said that". Even if I make stupid mistakes, is better to sweat now, than bleed on the battlefield later.
Thanks for the help so far ! Very interesting progress.
---
my setup as it is looking now:

- You can notice the green laundry tongs over the VReg, is there to hold constantly the Temperature probe on the Vreg metal side.

Last edited:

#### dl324

Joined Mar 30, 2015
16,911
PSU(A) = the amperage I read on my PSU display (while turning POT1 in my cct)
VReg(W) = is the calculated Wattage over the VReg, using ViVo*PSU(A)=VReg(W)
I'll keep repeating this until you get it.

You can't depend on the current reading on your power supply. You need to calculate current in the regulator using the voltage drop on the sense resistor.

I've noticed during my tests that the current used by the current limiter is sometimes over 150mA. That doesn't affect my readings because I'm using 2 power supplies. A current of that magnitude would definitely affect your calculations.
One very big issue was again the POT range when using at 9V. It was very dificult to get to those precise PSU(A) values like that 0.10A and 0.18A because it was jumping like crazy. Because I used 1R5W instead of the 5R. At bigger Res, I get a smoother POT travel (but less A range).
Do you still have the 2k resistor I had in series with the 1k pot removed? In addition to limiting the maximum current in the power transistor, it also reduces the voltage adjustment range which addresses the jumpiness.

My original design required a movement of 1/4 of the adjustment range to change current 1A.

The voltage from your power supply varied from 8.92V to 8.85V. That represents a load regulation of 1%. In contrast, the typical load regulation for LM317 is 0.1% (0.5% worst case). Since you're no longer violating the voltage differential spec, the 3.4% you're seeing must be due to the thermal protection circuitry. Without knowing the order of the tests, I can't comment further.

If you repeat your 9V tests using a 3V input to output voltage differential, you'll reduce power dissipation in the voltage regulator and throttling should kick in later.

You should also be aware of the possibility of the voltage regulator oscillating when protection is activated. I noticed during my testing that the variation in output voltage wasn't monotonic, and that the voltage dropped at some mid-level current before rising back to something closer to what I had set when the current was 2.9A. I'll put a scope on the regulator output the next time I test.

I wasn't concerned about output voltage variations because I expected the protection circuit to be activated at some point. If I let the regulator run at 2.9A long enough, the current starts dropping (below 1A).

Last edited:

#### q12x

Joined Sep 25, 2015
1,689
You need to calculate current in the regulator using the voltage drop on the sense resistor.
you mean over that 1R5W, right?

#### dl324

Joined Mar 30, 2015
16,911
you mean over that 1R5W, right?
Yes.

When you were using a 5V power supply with the regulator output set to about 3V, the voltage drop on the current sense resistor(s) could cause the power transistor to go into saturation. If that happens and you try to increase current further, more current will go through the other transistor to try to establish a higher current in the sense resistor.

If you insist on using such large values for the current sense resistors, you limit how low of a voltage source you can sink current from.

With a 1 ohm sense resistor, you require 1V/amp and you want to prevent the power transistor from saturating (so the voltage on the base of the power transistor needs to be higher than the collector voltage)

If your motivation is to simplify mental arithmetic to calculate current. using 0.2 ohms just means that 0.2V represents 1A. That's what I'm thinking when I monitor the sense resistor voltage ("point 2 volts per amp").

#### q12x

Joined Sep 25, 2015
1,689
Do you still have the 2k resistor I had in series with the 1k pot removed?
I just removed the wire and put in a 2.2K since I have a ton of them scrapped.
----
Ok, i got the voltage across my 1R 5W. Being 1R means the V=A.
New(W) = ViVo(V) * 1o(V=A) = New(W)

What this is telling me: That now, at 1W(1.04) disipation over the VReg, I get 82*C of heat over it ! -Daem.

Last edited:

#### dl324

Joined Mar 30, 2015
16,911
That now, at 1W(1.04) disipation over the VReg, I get 82*C of heat over it
Is your objective to see how much current you can get from the regulator? If so, if you increase the regulator output voltage to 6V, you'll decrease power dissipation in the regulator and will be able to get the same current at a lower case temperature.

The regulator will dissipate less power, but that's good. Any power dissipated by the regulator is wasted. When dissipation in a linear regulator becomes excessive, you use a switching regulator. If switching noise is an issue, you can use a switching regulator as a pre-regulator for a linear regulator.

#### dl324

Joined Mar 30, 2015
16,911
I did some troubleshooting on the voltage dip I was seeing. I thought that it was due to the protection circuitry kicking in and causing oscillations. It turned out that it started happening at less than 100mA, so safe area protection wasn't a factor.

It turned out to be known susceptibility to oscillations. I have ceramic caps on the input and output, but additional capacitance is required in some situations. This is something you should consider in your testing because you have long wires connecting the regulator to the current sink. Wires will add resistance, inductance, and capacitance. A National Semiconductor application note discussed 500-5000pF being bad.

The recommended fix is to use a 10uF tantalum or 25uF electrolytic. I tend to use 470uF because I have hundreds of them. Adding the capacitor on the output fixed the oscillations. The application note mentioned ringing; ringing, singing, and oscillation are all terms for the same thing. The oscillating frequency varied from a few hertz to 6kHz and I could actually hear it when it was in the kHz range.

With the additional output capacitor, the output voltage dropped from 3.01V to 2.95V as the current varied from 10mA to 2.95A. That's about 4X the worst-case spec, but the protection circuitry was activated and it was reducing the output voltage to reduce power dissipation. But, not by much.

Now I need to add a protection diode on my LM317 jigs...

Now I need to take the time to see if the power transistor on the current sink was saturating and preventing me from getting more than 2.95A.

While I was reading AN-178, I saw this diagram of the regulator:

Very similar to the current sink circuit and coincidental that National Semiconductor also used a 0.2 ohm current sense resistor.

#### dl324

Joined Mar 30, 2015
16,911
It is not stopping or blocking all the power through it, like a switch, and I believe it should have do it, as I was imagining/expecting initially. It is still letting power go ! Which sucks.
Do you want a way to limit the maximum current? A current limiter or a fuse?

I've designed a circuit that does both, but I'm not sure I want to give that idea away... I've already posted a bang-bang type current limiter in a thread where a member was trying to build a lab supply. That solution was too simple for him. I liked my solution (it was something I had thought about doing for years) and lost interested in his problem. I think that was around 4 years ago.

Bang-bang is a reference to the fact that current limiting is either on or off and it will toggle on and off very quickly until the overload is removed.

Last edited:

#### dl324

Joined Mar 30, 2015
16,911
Regarding the "high" current I saw the current sink drawing.

My first thought that it was because the power transistor had saturated and additional current I was asking for was coming from the transistor driving the power transistor. It turns out that it's just because the power transistor beta is low.

When I was sinking 2.95A, Q2 (small transistor) was providing 75mA of base current to the power transistor. That works out to a beta of 40 for the power transistor. That's not really that low, but I hadn't internalized that so much current could be coming from Q2.

To make it easier to monitor more than a couple voltages on the current sink when I was concerned about heating from power dissipation, I made a voltmeter board using some inexpensive voltmeters from AliExpress. This lot was a piece of junk because none of the half dozen meters I tested would read 0.00V with the input connected to ground (all read 0.10V, but that was within the seller's specs - I just didn't think any would be that bad). They're only good to within a tenth of a volt for the voltages I was measuring, but I figured that's good enough. For more accurate readings, I followed-up with one of my bench meters (which I found disagreed with each other by ~50mV).

#### q12x

Joined Sep 25, 2015
1,689
I made a voltmeter board using some inexpensive voltmeters from AliExpress.
Check this out:

I made this one LOOOONG time ago, its purpose was to be attached to my testing breadboard and read multiple voltages. Very cool idea I had back then, but these are crap and erroneous voltmeters ! I will explain.
Their problem:
They have a SMD POT on the back of the board. From that POT you set up a limit.
They are measuring ---AROUND--- this pre set limit.
For example, I set mine to 5V limit, so the voltage reading is precise only ---AROUND--- this 5V.
This means that when the voltage is changed, it will read (linearly/progressively) more eroneous toward 0 and also (linearly/progressively) more eroneous towards 30V(maximum in its specs)
The good thing is that they will be spot on, around 5V. So 4.5V to 5.5V range will give you excelent readings.
Again, I set mine to 5V but it can be 9V, 12V and so on. But the behavior is the same.
THeir oficial name is "3Wire Digital Voltmeter"
Here are the specs from the seller page when I bought them:

Hope it helps you !

#### dl324

Joined Mar 30, 2015
16,911
They have a SMD POT on the back of the board. From that POT you set up a limit.
I disassembled one that I shorted and none have any adjust capability. I checked them to around 20V and they were within 0.1V over the entire range.They're good enough for approximate measurements.

#### q12x

Joined Sep 25, 2015
1,689
This movie is more as a summary of all that we discussed here so far.

#### dl324

Joined Mar 30, 2015
16,911
If you still have the 5.6 ohm resistor in the circuit, your current calculations are off by 15%. 1 ohm in parallel with 5.6 is 0.85 ohms.

#### q12x

Joined Sep 25, 2015
1,689
true

#### q12x

Joined Sep 25, 2015
1,689
I was looking over those peltier components you mentioned.
I read the wikipedia page and also looked into all the documentation I could find on the net, from diferent sources.
Im still in doubt how helpful they may be in my applications, also your mentioning they are super power hungry as I read in wiki pages as well. Hmmmm. My PSU can give "so much" - 150W max (30V@5A). But Im thinking on it...
-What name+number you have? I found multiple (Amp) ranges devices on Aliexpress.
Wikipedia page is actually mentioning this exact model name TEC1 and explains its naming and numbers meaning.

From 12715 , 127=internal n-p pairs (not relevant since they are all have the same 127 pairs) and 15 which means 15A
I found at 4, 6, 10, 12,15A .
The bigger in Amperage the better, right? Is my thought.

Last edited:

#### dl324

Joined Mar 30, 2015
16,911
What name+number you have?
The one I have has no labeling. I tested it at 5V and it drew almost 6A. I haven't measured temperatures, but it got uncomfortably cold while I was holding the hot side on a heat sink. Since it draws 6A at 5V, I need to setup a computer power supply before I can do more testing.

I ordered TEC1-12706 from wallmart.com and expect them in a couple days. Then I'll be able to test a 12V/6A device at lower voltages.
Im still in doubt how helpful they may be in my applications
You could cool devices with something that might work better than a slab of metal with no fins. The fins have a significant effect on cooling efficiency.

You don't need any heat sinking for your "naked" tests.

If you get some Peltier junctions, make sure you always use a heat sink or the junctions could be destroyed (quickly).

Last edited:

#### dl324

Joined Mar 30, 2015
16,911
Using 100 ohms for R10 was limiting the maximum current I could sink to less than the 4A that the circuit was designed for.

High junction temperature increases the power transistor beta until around 1A. After that, it drops off faster than it does at lower junction temperatures. At a current of 3A, beta for 2N3442 is around 35. At 4A, it's under 20.

To get 4A, base current needs to be 200mA. I changed R10 to 14 ohms by putting 2 33 ohm 2W resistors in parallel with the original 100 ohm resistor and was able to get over 4A. I didn't operate at that current level for long because I need to do dissipation calculations for Q2.

I'm glad I used BC337 instead of substituting 2N3904 for the 2N2222A. I never internalized the fact that base current for the power transistor could be in the 500mA range if I decided to go for currents higher than 4A.

#### q12x

Joined Sep 25, 2015
1,689
If you get some Peltier junctions, make sure you always use a heat sink or the junctions could be destroyed (quickly).
ooooooooooh, I didnt know this detail ! Very good that you tell me ! Excellent.
It occurred to me when I wake up, that I can actually slap the transistor --directly-- onto the Peltier device !!! On its Cold side. And the Hot side, like you said, to a heatsink. But there is a bit of detailing I was imagining further. Since the peltier module is 40x40mm, meaning 4cmx4cm, its pretty big even for a TO-3PF case (15x24mm) measured. Not mentioning TO-220 or TO-126. So to distribute the Tr heat more efficiently over the Cold side, and the cold side to actually influence at maximum the high temp from tr, I was thinking of a 'slab' of Al heatsink between the tr and the Cold side of the Peltier. Like a sandwich.
So it will look like this:

Cool, yes?
But after my lecture yesterday on wikipedia and your very good points, I imagine only the peltier alone will need it's own dedicated power transformer to about 100W most probably. Hmmm. Hahaha.
- The positive effect that I personally see from all of this, is considerable smaller size (heatsink space) of operating the tr at quite high wattage !
- The negative effect is very high power consumption.

#### q12x

Joined Sep 25, 2015
1,689
Using 100 ohms for R10 was limiting the maximum current I could sink to less than the 4A that the circuit was designed for.

High junction temperature increases the power transistor beta until around 1A. After that, it drops off faster than it does at lower junction temperatures. At a current of 3A, beta for 2N3442 is around 35. At 4A, it's under 20.

To get 4A, base current needs to be 200mA. I changed R10 to 14 ohms by putting 2 33 ohm 2W resistors in parallel with the original 100 ohm resistor and was able to get over 4A. I didn't operate at that current level for long because I need to do dissipation calculations for Q2.

I'm glad I used BC337 instead of substituting 2N3904 for the 2N2222A. I never internalized the fact that base current for the power transistor could be in the 500mA range if I decided to go for currents higher than 4A.
To be honest, I dont understand everything you explained. I get you are talking about your project and that you -probably- update it from how it was originally? You should start making some movies, 1or 2, or how many necesary, like I am, showing the tests you do, how you probe things, and all the good stuff. You can keep the movie for 24 hours online and then delete them, if you feel your personal space is too sensible to public exposition. Or leave them for posterity, like I am. Im full of good ideas, or what? haha

#### dl324

Joined Mar 30, 2015
16,911
So to distribute the Tr heat more efficiently over the Cold side, and the cold side to actually influence at maximum the high temp from tr, I was thinking of a 'slab' of Al heatsink between the tr and the Cold side of the
You've got the right idea. The ceramic on the cold side of the Peltier junction won't conduct heat very efficiently, so using a slab of metal will use the full junction area.
The positive effect that I personally see from all of this, is considerable smaller size (heatsink space) of operating the tr at quite high wattage !
- The negative effect is very high power consumption.
It's not efficient energy-wise, but it will let you test transistors at their maximum junction, current, power dissipation levels.
You should start making some movies, 1or 2, or how many necesary
I believe well drawn schematics are the clearest and most concise way to communicate circuit information. With that schematic and a few sentences or paragraphs, I can convey where to take measurements more concisely than a multi-minute video.

While I'm not fond of how my voice sounds, I'm a much better public speaker than a significant number of people who post videos on YouTube. What's with all of these people saying so many 'ands', 'ums', and 'buts'. For the cases where those words don't make sense, silence would be much more preferable.

There's a guy who posts some informative videos on YouTube, but his speaking skills are poor and I can only stay engaged for a few minutes before all I hear are the 'ands' and 'ums'.