IN-8-2 Modular Rev 3 build - feedback and questions

2 years 3 weeks ago - 2 years 2 weeks ago #9466 by Phil_Mod3
Happy New Year everybody!

Thanks to Ian's perfect kit, I successfully finished the Modular Rev 3 kit a few days ago. The tubes are IN-8-2s and I chose the ESP-01 as a time provider. Additionally I've attached an RCWL-0516 for presence sensing. I'd like to provide feedback on a few things and also have some questions for the professionals here:
1. The NeoPixels (NP) have to be mounted on the bottom of the tube holder PCBs (at least for my IN-8-2 holders, see images), which is not really described anywhere. The NP have to be inserted in a cut-out of the PCB. However, the cut-outs were to small and had required a lot of force to insert (at least for my taste), so I chose to use sand-paper and widended the cut-outs a bit. Altogether that worked without problems.

2. I didn't like the lit decimal point of the 10s tube, which indicates display switched on again after motion was detected. I commented out the respective line in the code and updated the microcontroller using an Arduino UNO board. That worked without any problems. Maybe that could become the new standard preset in future code revisions?
3. Using a 9V 1A power supply, the HV calibration did not successfully finish (only counted up and never down). After changing to a 12V 1A supply, HV calibration worked.
4. With the standard 180V target voltage, some IN-8-2 showed flickering (not fast) of the filament areas on top and bottom of the actual number (the part between the disks which separate the numbers). When decreasing the HV, this dissappeard. Currently the clock runs at 165 V. All voltages were confirmed by measuring the HV at the respective test points.
5. When the tubes are lit at the lowest intensity (low/no light on LDR) after being blanked for a longer time, some tubes flicker on a off. The flickering vanishes after a minute run time at lowest intensity or immediately when increasing light level on the LDR. I guess the flickering could be diminuished by changing the low-light PWM settings of the clock? (I think the effect was also there at 180V)
6. I'd like to connect a few things to the 5V Vcc of the clock. Altogether that would result into an additional 500 mA peak power drawn from the respective pins. I understand that this might be to much, although the LM2596T is specified for up to 3A total. In a first test, I could successfully power the additional load without any obvious side effects, of course the LM2596T became hotter than normal, so I attached a heat sink. Any late term side-effects that I should think of? In this context: Is there a drop-in replacement for the LM2596T, which is more effective and thus runs cooler?
7. I generally feel a bit uncomfortable when putting everything in a wood enclosure, since I expect the heat dissipation of all the chips to potentionally become problematic. In order to reduce this, I'd like to invest a few € into an IRF840 replacement with lower Rds(on). For example, I could get an TK31E60W , which has a max. Rds(on) of 88 mOhms. However, I am unsure about secondary parameters (e.g. avalanche enery, input capacitance, switching time), could anybody give me some hints if these things are critical in our case?

OK, that was a long first post. Thank you for reading it and I'd appreciate your feedback/comments.
Last edit: 2 years 2 weeks ago by Ian. Reason: Images in line

Please Log in or Create an account to join the conversation.

2 years 3 weeks ago #9469 by Ty_Eeberfest
Looks good! I'm glad it went together smoothly for you.

I'll comment on a few of your comments and leave the rest for Ian...

2) I don't like the decimals either and modified the code to get rid of them. If it came to a vote I'd side with you on this one: get rid of at least the motion sensor related ones. I also got rid of the AM/PM indicating decimal but I'm not sure that would please most people.

3) Same here. I could never get HV Cal to succeed with 9 volts so I've been running on a 12v supply for about 8 months now with no problems.

4) Every model of tube (and even different date codes within a given model) have slightly different personalities. Tweaking voltage for good brightness without any flash-over like you did is the best solution I know of.

5) This can be fixed by slightly increasing Parameter 27 in the Setting Mode. Check the User Manual for info.

7) I too wanted to see if I could get the HV generator to run cooler. I randomly chose a 2SK2841 because it was the only thing I had on hand with somewhat lower Rds and acceptable voltage & current ratings. I paid no attention to the avalanche, gate capacitance, or any of the other more esoteric specs. It works just fine and does run a bit cooler. The part you mentioned has much lower Rds than my 2SK2841 so I'd imagine it will run much cooler still. I took a quick look at the specs and don't see any huge red flags. If it were me I'd give it a try - about the worst that is going to happen is it just won't workm but I'm pretty confident it will work. Also, here are the specs for the 2SK2481 I used:

Look into it later when the dust is clearing off the crater.
The following user(s) said Thank You: Ian, Phil_Mod3

Please Log in or Create an account to join the conversation.

2 years 2 weeks ago #9472 by Ian
Hi, glad to hear that it is giving you joy... :)

1) The holes are notoriously hard to get right when done as a cut out, and especially with mechanical aspects to them. I have stretched the cut outs to 5.1mm x 5.0mm and that is right if done with a new tool, but it's not always quite right. I small file or a bit of sandpaper is the best answer.

2) Absolutely with you on this. I need to remove that function from the code. It was sort of intended to be an indicator of "I have a motion detector installed" (if you don't have it installed, the normal blanking works as configured, if you do, normal blanking is suppressed). However it just ends up as a "tubes are on" indicator.

3) I have no answer to that. I have never experienced this. I will have a think about it. Running with 12V is perfectly fine because the buck converter is not so prone to heating up.

4) OK, thanks, good to know. Perhaps dropping the standard "factory" voltage to 170V might be an idea?

5) Indeed there is a "minimum dimming" setting made just for this... ;)

6) I can't see anything that speaks against using the regulator as a power source for auxiliary loads.

7) There's a history to why it's an IRF840, but that reason doesn't exist any more, but the 840 is so robust, I'm loathed to boot it out. I'm starting to move to IRF640 (which has a much lower Rds). The MOSFET doesn't have any "hard" tasks to do, but I want to make sure that I'm not storing up problems for the future by going to the 640.

Anyway, thanks for the post and the feedback!
The following user(s) said Thank You: Phil_Mod3

Please Log in or Create an account to join the conversation.

2 years 2 weeks ago #9477 by Phil_Mod3
Thank you very much Ty and Ian,

1) It was not really a problem, however it left my confused for some minutes. I don't know if it's feasible to update the manual in this regard?!

3) I'd get another 9V power supply anyways and will test with that again. I remember reading about this point here in the forum (was it Ty?!) and luckily had a 12V supply at hand.

4 & 5) I'll most likely change the firmware on my clock, since I am lazy and don't want to adjust this through the menus.

6) OK, thank you, I'll stick with the LM2596 and supply proper cooling.

7) The 2SK2841 was also in my group of potential IRF840 replacements, unfortunately I don't get it from my preferred supplier. Would be intersted to know why the IRF840 was in there originally? ;-) Ian, thank you for the heads up to the IRF640, although reading specs on many MOSFETs I didn't see that one. Too many choices out there. :) Since I could get that from my supplier comparably cheap, I'll give it a try. One question, just to be sure: Is the max. Vds of 200 V of the IRF640 not a bit thight for the HV circuit?

8) I'd like to add some further digital outputs to the clock for switching the loads mentioned under 6) in my original post. From reading the MEGA328 documentation and looking at the clock's schematic, I couldn't find any free IO ports. I guess they are all used already. However, since we have I2C, I thought about adding an MCP23017 IO expander, but with 16 IOs that might me overkill, although the chip is inexpensive. I'd only need one or two additional digital outputs. Has anybody experience with the MCP23017 or an alternative to realize this?

Thank you!

Please Log in or Create an account to join the conversation.

2 years 2 weeks ago - 2 years 2 weeks ago #9481 by Ty_Eeberfest
Just a couple quick comments...

4 & 5) Yep, that's what I do. I've hard coded all my settings and other tweaks that suit my personal preferences. Serves me well, especially since sometimes I like to experiment on my clocks and hate having to restore a bunch of settings. Just be sure to use comments or keep notes in case a new version with some feature you really like comes out and you need to re-do your changes.

6) Just keep in mind that you may need to upsize your power supply if you add much load. The Modular itself uses around 250mA (varies a bit depending on your backlight mode selection) when running normally and a bit more while doing HV calibration after a factory reset or firmware reload.

7) The 640's Vds seems a bit too low to me too but I'd be interested to hear what Ian thinks. As for the 2SK2841 I think Toshiba has stopped making it. I used it because I had one. It's Rds isn't all that great by today's standards anyway, and I mainly mentioned it to make the point that taking some liberties with the non-core specs caused me no trouble. As you say, there are lots of MOSFETs out there that are potential candidates. IMO the "core" specs are Id, Vds, Rds and Vgs(th).

8) Correct - all I/O are used. I have no firsthand experience with I2C I/O expanders. I have read that the MCP23008 (an 8 port expander) is popular with Arduino users. Libraries exist for it and can be downloaded free. If 8 ports is enough I'd cautiously (because I have not tried it myself!) recommend it mainly because there seems to be lots of community support and experience.

EDIT: Looking around some more I see that MCP23017 appears to have as much community support as MCP23008.

Look into it later when the dust is clearing off the crater.
Last edit: 2 years 2 weeks ago by Ty_Eeberfest.

Please Log in or Create an account to join the conversation.

2 years 2 weeks ago - 2 years 2 weeks ago #9482 by Ian
Just the story about the IRF840, to satisfy curiosity:

In the early days, the firmware for the calibration of the HV generator used to approach the target point (180V) from above, meaning that the HV would start way above 180V and come down. This was by setting the Atmega CTC to a minimum count (=highest frequency) and reducing it. This meant that the voltage started usually at about 350V and reduced, one step at a time. Because of the way the frequency generator worked, it took too long to converge from below, but was fine from above (inverse relationship between frequency and timer counts).

I then changed the algorithm to be adaptive: It measures the current voltage and if we are far away from the target from the target, it changes the count by 50, if we are not so far away, it changes the count by 5 and then for the last few volts, it changes by 1. This means we converge more rapidly, and so I can converge from below in the limited time we have during the calibration.
void checkHVVoltage() {
  if (getSmoothedHVSensorReading() > rawHVADCThreshold) {
    int diff = getSmoothedHVSensorReading() - rawHVADCThreshold;
    int inc = 1;
    if (diff > 20) inc = 50;
    else if (diff > 10) inc = 5;
    setPWMTopTime(pwmTop + inc);
  } else {
    int diff = rawHVADCThreshold - getSmoothedHVSensorReading();
    int inc = 1;
    if (diff > 20) inc = 50;
    else if (diff > 10) inc = 5;
    setPWMTopTime(pwmTop - inc);

Because we now never overshoot the voltage, and can't configure more than 200V as the target, the MOSFET and the capacitor don't need to be capable of more than 200V nowadays.
Last edit: 2 years 2 weeks ago by Ian. Reason: middle step count is 5 not 20

Please Log in or Create an account to join the conversation.

Moderators: AccutronTy_EeberfestIan
Time to create page: 0.116 seconds
Go to top
JSN Boot template designed by