That’s what YLP does in the Unicorn 1.0, it lacks precision limited by the ADC’s resolution, and they use an RC filtered PWM signal to drive the mosfet, could be because the MCU doesn’t have a DAC, or because its resolution is not high enough ?
So in the end there is a measurable ripple
It should be across In- and out, quoting an app note I read at some point , with R4 and R5 it forms a compensation network that prevents ringing.
Isn’t not turning the LED blue better? Increased resistance is never ideal but overdriving the LED is rather bad for life expectancy.
Wow! That was fast! I see. O:)
Despite what I said before, the newer version driver will also let a little bit more current if you stack more and more sense resistors, as it reduces the sense voltage burden. The lower modes will grow higher and higher, though, and if you dare to bypass the sense resistor with a blob of solder or by installing copper sheet or wire instead, all modes will top out in output; the lowest one, if set very low by the driver firmware (like 1%), may not (solution: use the most conductive, thickest sense resistor bridge you can afford). In short, you lose modes in this case.
The Convoy “ramping” driver is very solid, really good. Employs a 6788 MOSFET, and just an R005 (5mΩ) sense resistor for up to 8A, so only up to 40mV spent in sense, and the lowest minimum resistance (resistance with the MOSFET “pedal to the metal”, all closed or completely unthrottled) of all linear drivers at Convoy, afaik.
Over time I've learned to appreciate it, despite its weird firmware. It is ideal for gun shooting (hunters), as the mode selection is by series of quick taps: one quick tap, first mode; two quick taps, second mode; three quick taps, third mode. Since the first two modes can be current set with ramping, and the third or special mode can also be set to 100% via the special menu among other options, this means it can be set so that the vibrations of a gun shoot cannot change the output of its flashlight, as all three modes can be easily set equal at 100%.
Here's the sales link of the ∅17/∅22mm ramping driver at Convoy. Oh! :facepalm: I see the chart for the driver mode selection and setting instructions of the driver is missing! Nevermind, the graph is there; it was some sort of connection problem.
Overall I smell a “race for the highest current or output” around here. This is not necessarily a good approach. With the coming of the Osram emitters, with a pretty low Vf and lowest current limits, you need to learn that its drivers must be (or should be) set to limit the maximum current, to avoid any chance of overdriving them. This of course involves looking at the test graphs. However, please bear in mind that not all emitters are equal, and some may have somewhat lower limits. Even if all of them were to stick to no less than the test curves, it is not necessary nor wise to drive them at up to the maximum output current. At these levels, reducing the maximum driving current to let's say ≈80% only renders reliability benefits for the emitter, barely affecting the output. Let's for example see the output graph of a Samsung LH351D 5000K -test- by Texas Ace:
Maximum output current is 9.25A, delivering 2333 lemons.
80% of 9.25 is 7.4, so what do we have at 7.4A? 2220 lemons. This is 2220 / 2333 = 95.15645…% of maximum output at 7.4A, according to test graph information. Do you think this will make a difference to the eye? Go try. x-D
By the way, in the above graph it was easy to see where the maximum output is. Let's take a look at the Reflow conditions tested of Osram KW CSLNM1-KW (White Flat 1mm2) test graph by djozz:
I'll make an small explanation on these duties, just for the sake of it.
So where is the maximum output for these CSLNM1.TG emitters? Is it at 5A? Is it at 6A? Wrong.
djozz only took discrete output measurements each 1A in the latter test (each 0.5A in the former one), and drew straight lines between each discrete value. This is not how the actual output current curve, with continuous measurement or infinite discrete values, looks like.
The actual peak value or maximum in a curve plotted with discrete values and straight lines joining them, is always around the highest discrete value measured for it, and between it and the closest, next highest value. If there are two next highest values, like in the above LH351D curve, then the maximum is spot on the measured point. When there are two maximum discrete values, the peak is between them and higher than both.
In djozz's test the two best emitters peaked a hair above 5.5A, while the lesser performers (excluding the first test's specimen) peaked a little above 5A. In my lightful opinion, I'd probably choose to set my drivers at 4.5A for CSLNM1.TG emitters (the “S” in CSLNM1.TG -1mm²- and CSLPM1.TG -2mm²- means 3030 footprint). There are no tests for the CULNM1.TG or CULPM1.TG ones, so we don't have an idea of the gains for the 4040 footprint. My prediction, though, is that a CULNM1.TG (4040, 1mm²) is barely going to be any better than a CSLNM1.TG (3030, 1mm²). The impact of the footprint may be more noticeable when comparing the CULPM1.TG (4040, 2mm²) to the CSLPM1.TG (3030, 2mm²), but I hardly predict any noticeable gains at stake anyway. Emitter bin is more important in this respect.
Thanks for the technical contributions, thefreeman, it's enlightening. Still, I am @#$% with all of this. I'll see what I do with the sense voltage dividers. I also have a driver from Kaidomain, the P4000, which looked good on paper. But after inspection, I discovered a fairly gross sense resistor stage of an R050 plus 1R0, which is 47.619mΩ, which is efficiency killer for a buck driver driving a 3ish volt load.
P.S.: Now I understand why my latest flashlight build for a brother had such a small brightness difference between the highest mode, and the second highest, as I swapped the sense resistor in its “new downgraded” driver to an R010, like in the older versions of the driver with smaller sense voltage. :facepalm:
I am having a private conversation with Simon right now, this is the last message I just sent him:
I definitely noticed the newer driver with R020 dropping out of regulation quicker compared to an older 17mm 5A driver with R010 and resetting mode group bug in a C8 XP-L HI.
It’s a shame because the 17mm 5A drivers were my go to for modding and Convoy builds, are there other similar alternatives in 17mm besides the LD-A4 and Convoy ramping driver?
Took some time to draw the following picture, peeking at a “new version” (huh?) ∅20mm 12-groups driver, while I was reading the values with my cheap multimeter (maybe gold plated probes can make a difference in this regard) and ensuring all connections were accurate. The SVAKF part is the operational amplifier, LoL. Resistance for R2 would only stick quickly when read with positive probe between it and R3, giving 436.3kΩ. The other way around I get 494.3kΩ after I wait a few seconds for the value to climb up.
Compared to thefreeman's diagram, C2 <=> C3. I also didn't mess with certain stuff (for example didn't make sure if R5 goes to Rsense first, then to ground).
It's enough for now. O:)
Fri, 08/06/2021 - 16:51
Thanks for that, i was wondering about the capacitor values.
Here is a simulation using the 10mR resistor parameters with your capacitor values running 4A in LED. The ripple in the LED current is due to the low frequency of the “PWM” pulse; if the frequency were increased, then that would be filtered out.
Would you be able to check and provide the actual connection of the R4, R5 and C3 on the output of the op amp to the FET—correct any errors in the schema.
The only difference i can see between the 20mR and 10mR version is the R2 value that you measured, and that difference is not really significant. (436 - 494k vs 510k)
Tom E has a pic of another Convoy linear driver with the components desoldered :
So yeah C3 isn’t between the out and in- contrary to my shematic (I just guessed since it was difficult to see) or shematics found in app notes or even in suggested application in op-amp datasheet , for example this one in the OPA333 DS :
Detailed here : http://www.ti.com/lit/ug/slau502/slau502.pdf
First stage is a constant current sink, the Noctigon kr4 driver is exactly like that, with an added resistor between the gate and source of the mosfet.
Note that with Falstad the circuit is stable even without the compensation capacitor, in LTspice there are large oscillations without it.
It could be that it was intended to be connected for compensation as you have shown, but a trace error was made in the layout. As it is now C3 seems to serve no real purpose or provides any filtering.
If i connect C3 for compensation and increase the command level, now i see 6A in the LED and a 60msec delay in reaching the full drive of the FET with a nice ramp up to the desired current. This sort of soft start puts less wear and tear on the parts and is less abusive than a full hard on start; maybe this was intended but the traces weren’t in place.
[edit]
The op amp will drive the output as necessary to cause the feedback at IN- to match the input at IN. The RC combination of R3*C2 causes an exponential ramp up of the input. With C3 wired as shown it is bleeding off charge from the output thru R5 (10k) to cause IN- to match IN. Once the IN+ reaches its final value and is no longer changing, the contribution by C3 gets reduced and the op amp must drive harder to compensate, which raises the output until it reaches the 1.5V gate threshold voltage of the AO6788 SRFET (30V, 80A, 4mR). Now the FET turns on and the feedback is from the Vsense reading. This is the effect of compensation to create the soft start 60msec delay of current in the LED and FET.
Hello again. :-)
Just edited my previous post, I noticed that I am not really sure if the ∅20mm driver I was speaking of in it is actually a new version driver. This in part is because the results I was obtaining with thefreeman's equation didn't make sense, and because the driver was resent to me from Simon maany months ago because of an old firmware faulty unit. I obtained two values for R2, but of course I wasn't measuring R2 with one of its legs unsoldered, something which is a must to do away with any distorting measurement interactions with the remaining parts in the circuit. Because of this, out of the two values I reported for R2 the correct one must be the highest: 494.3kΩ.
Now, here's the figure for R2 out of thefreeman's equation presuming the LDO is a 2.6V one:
R2 = ((Vreg – Vsense) x R3 / Vsense) – R1 = ((2.6V − 0.06V) × 14kΩ / 0.06V) − 100kΩ = 492,666.6̅Ω
This sort of matches my 494.3kΩ measured value for R2.
Now I just have to rig up the driver to some led load, and raise its voltage to see if it goes beyond ≈4.2A (it has an R020 in parallel with an R050 as Rsense). It should not…
Sun, 08/08/2021 - 09:56
What about the ∅17/22mm ramping driver? Did someone around here bought some lately? Does it come home as seen in the pictures?
It's been been a while since the last time I bought one of these, hope all is good. The ∅22mm one is easily grinded down to ∅20mm, btw. ;-)
Got mine for less than 2 months ago, 5 of them (17mm) pretty much the same like the older ones…
Finally spent some time assembling the head of my custom red L21A… for @#$%. Tested it on my power supply, and it went well above 4.2A, which should be the limit with an R020 and an R050 as sense stack.
:facepalm:
After all what has been discussed here, do you need such a thing? I have links, but if your drivers are new their sense voltage is a tad above twice the old one (more voltage drop loss), and causing the absolute current limit to also be more than doubled with an older value resistor… this is why I am dumping mine. Sheesh!
Had to test this again for some :GLASSES: reason…
By the way, the holes in the retaining rings are too tiny. Even my finest tips round nose pliers are too thick for them. After some struggle in part due to me having tightened the ring, I decided to make use of my drill plus a 1.5mm bit to enlarge the holes. Problem solved, but again I have to say this sort of stuff should not be needed.
So I dismantled the driver from the pill, took out my heat gun and proceeded to desolder the smaller resistance sense resistor (R020), leaving just an R050 for sense. Moments afterward I assembled it together with the pill again. I wanted to measure the current limit the driver allowed, and since somehow I was expecting it to be a new version unit, I removed the R020 sense resistor to make sure my SST20 emitter could deal with the current.
I am using a Wanptek 30V 10A precision PSU, I first set a current limit (just in case), raise the output voltage enough for the resistance of the output wires and alligator clips not to be a limit (≈6V), and then proceed to stick the alligator clips in the spring and over a pill border by hand, using this latter connection to switch modes. Here are the tared to zero numbers of the PSU's digital amperimeter for mode group 1 (0.1%, 1%, 10%, 35%, 100%, strobe, biking, battery check), with strobe, biking and battery check ignored; all values in mA:
So I obtained two matching series of measurements, and according to the last 100% figures of ≈1300mA, sense voltage must be 65mV.
Thank God I tested the driver again instead of dumping it, this means I need to solder the sense resistors more carefully.
With this data, thefreeman's equation gives out a more or less matching R2 value with my measured values if presuming the use of a 2.8V LDO. For other drivers we need component measurements.
Oh now i see why—that is a very specialized op amp with proprietary internal chopper circuits to reset the offsets which limits the Unity Gain BW to 350kHz, and with ultra low input offset voltages. So it would be prone go unstable and oscillate at the slightest sneeze, similar to some of the older op amps that had no internal compensation.
The OPA2333 basically is 2 amps in 1—it has another op amp built inside it in addition to the one that is available to the user…too damn smart for me.
I’ve just received some Convoy 6A drivers (Biscotti-clone firmware) that have R020 sense resistors. The image on AliExpress shows a R010 sense resistor.
With my bench power supply at 4.2V the driver only pulls 3.5A, but at 5.2V or higher the current tops out at 6.5A.
Based on my very limited understanding of this thread it seems like on the driver rather than doubling the sense voltage the output has been (approximately) halved.
Does this make sense? It seems surprising that Simon would be selling a 3.5A driver in pace of a 6A driver.
Can you confirm that once it’s installed into a flashlight?
When you set your power supply at 4.2V, there are 4.2V between the power supply output connectors. In fact, and to be precise, these 4.2V are “where the measurement is taken”, this is somewhere inside the power supply close to the output connectors.
Once the circuit is closed and current starts to flow every resistance in the circuit causes a voltage drop. This voltage drop is V = I × R.
Physical connections themselves have some small resistance, this is why battery cells are soldered or spot welded in battery assemblies, and not set inside battery holders. You could for example assemble a big battery out of battery cells using battery holders, but each resistance in the contacts between each battery holder and its own cell would severely limit the effective current output of the cell. Resistance in the contacts versus cells' own internal resistance is significative.
Then, PSU output probes have resistance; the alligator clips in them have resistance too. And the contacts you make between alligator clips and the driver… have resistance.
Once all of the above is considered, you understand that in a closed circuit the actual input voltage at the driver when fed from the power supply is lower than what the PSU's voltmeter is saying. For you to know what is the actual input voltage at the driver when this is done, you could hire some assistant for him/her to peek right at the driver with the probes of a multimeter.
The purpose of these regulated variable load drivers is to take the input voltage and regulate or limit it in a way that the current for the different modes matches what is claimed, this is done by “sensing” the current flow as the current goes through the driver's sense resistor, something which causes a subtle voltage drop at the sense resistor which, after magnification (amplification), is seen by the driver microcontroller, and the driver microcontroller uses this information to “regulate the gas valve” at the MOSFET (it adjusts the MOSFET's VGS), causing the output voltage to match what is needed for the exact mode current to flow. With the recent changes which resulted in a more than doubling of the sense voltage, the voltage drop in the sense resistor stage is higher. The problem with this is that the sense voltage is an “inevitable loss” which mathematically speaking is V = I × R. This loss is more noticeable in the higher modes, and despite not being very noticeable, is comparable to replacing a high current battery with a not so high current battery.
The actual problem (for sense resistor modders) is that we made our math presuming the older sense resistor and voltage value, which is very important to precisely limit the current for smaller emitters like SST-20's and Osram white flat 1mm² units.
With all of the above which has been said, I hope for you to understand that the sense voltage has been actually more than doubled in fact, and that the actual maximum output (when input voltage is enough) is now higher than what is claimed. The wrong part in all of this is that maximum output current is a tad higher than what is claimed (this matters a lot for those of us who do sense resistor modifications), but also that to attain this current at the emitter the driver needs more voltage than the older versions of the driver with lower sense resistors (they cheaped out on the driver).
The reasons why you could only measure 3.5A with the power supply are explained above. The maximum driver output has not been halved, it has been raised a tiny bit instead, but at the price of a slightly higher internal resistance now versus the old version (higher sense resistance), resistance whose effective recovery is unavoidable without laborious modifications. The sense resistance is important because the current regulation of the different modes is proportional to the inverse of its value (in other words, current regulation of the different modes is proportional to the sense resistor's conductance), and thus if you bridge the sense resistor you lose regulation, this means the driver will let all the voltage and current to go through because it will be unable or barely able to sense the sense voltage, i.e. all modes will be maxed out without driver limit except maybe the lowests (0.1% - 1%) modes which will just look very bright or nearly as bright.
So, are you measuring correctly if you set the power supply output at 4.2V and stick the wires at the driver? No. In practice the PSU output wires and contacts have a lot more resistance than the flashlight internals (springs, switch, and wires), this is the reason you must raise the output voltage of the power supply until it hits a current wall; this current wall is the point at which the driver regulates or limits the current. See what you said (5.2V), and also what I said (6V). Since each time you move the output wires and connectors you also vary their resistance, this is the reason I used a high figure: to ensure that these movements would have no chance of reducing the input voltage at the driver too much while I was switching modes and etc.
This explanation also reveals why people who measure current by dismantling a flashlight's tailcap and sticking multimeter probes attached to a multimeter between the cell cathode and the flashlight body, do it very wrong. Among contacts, probes and the multimeter itself have a lot of resistance, way more than that of the flashlight's tailcap by itself. No wonder why they measure low values.
To measure the current in a flashlight you must use a clamp meter around the flashlight body, this works because the flashlight body is used as a current path. This measures the current which goes from the battery to the driver, and this current value is also equal to the current which hits the emitter in flashlights without switching drivers (boost, boost-buck, or buck drivers).
I actually took my measurements with the driver installed in the head of a Convoy L2 with a white flat. I don’t have a clamp meter so I connected the power supply to the to the driver spring and flashlight head using alligator clips.
As Barkuti pointed out this is not an optimal setup as there is the potential for a lot of voltage drop due to the test leads and poor contact. However, I have checked the voltage between the ground ring of the driver and the use of the spring and it’s the same as the power supply voltage so I don’t think that’s the issue. I have also tested a different driver (led4power A4) using the same method and it’s pulling the expected current.
I’m going to set up another one of the Convoy drivers outside a flashlight for another test.
Edit: After writing this post I realised that I wan’t measuring the voltage drop properly. This could have been avoided if I had used a clamp meter, or perhaps just thought about what I was doing a bit more carefully. Thanks Fantastic and Barkuti for helping me understand.
Thanks Barkuti!
I was in the middle of typing a reply that said your explanation made sense but it didn’t match my measurements when I realised that I hadn’t been measuring properly.
I was checking the voltage at the driver when I first turned the power supply on and the driver was in the lowest mode. When the driver was in higher modes the voltage drop due to the test leads was higher, but I wasn’t re-testing the voltage at the driver (not enough hands). :person_facepalming:
So why was the led4power driver pulling the expected current? I expect it’s the lower internal resistance in the driver, but I’m also no longer sure what current I had it configured to - I’ll have to look at the manual and check. 
So now my issue is that the high mode on this driver is actually >6A which isn’t great for a white flat, but I can focus on understanding sense resistors so I can adjust it. 
