ICE TUBE Q3 TEST: MOSFET or PNP TRANSISTOR Moderators: adafruit_support_bill, adafruit

[Russell 27]

My original post here was to demonstrate several component choices to effectively switch on the tube filament. This gradient issue is a natural effect with a DC source current on the filament. ADAFRUIT has mentioned this. Using a resistor to drop voltage and or current is like robbing Peter to pay Paul. It's a compromise. I'm not criticizing here, it works; to a point. Sometimes better than others. There is really no right way. It's just a matter of consequence using DC. On my own clock I use AC, and the result is really nice. However you personally choose to set up your clock, I highly recommend a little trial if not satisfied.

[jarchie]

There is really no right way.

I agree that there can be multiple solutions to a single problem and that many of those solutions are similarly good.

But I noticed that msenn had "relative noob" in his footer, and I'm sure you remember how easy it is to get frustrated when you're first starting out. Heck, I still get really frustrated sometimes. So in my previous post, I only wanted push msenn in the direction of a proven solution that I was sure would work. I did not mean to suggest that your approach was somehow wrong or inferior, and I apologize if my post implied that. It was not my intent!

Something interesting with Johns firmware on an Adafruit clock (ZVP2110A/11 ohm, & auto dim parts) board, the digit brightness at really low brightness level (clock in dark room) is even.

Last night, I removed the to-spec mod from my Adafruit clock, repaired the cut traces, and installed a ZVP2110A and 50 ohm variable resistor in place of Q3 and R3. The hole in the case was already cut from when I tried this before:



The dim digit problem gets worse at higher brightness, so at the highest brightness setting, I tried increasing the resistance to the largest value while maintaining consistent brightness. At that resistance and at low brightness, a brightness gradient was readily apparent.

My original intention was to test several tubes, but given the significant difference between our observations, I no longer think that would be informative. It seems clear that there's a great deal of variability among tubes when driven out-of-spec in this fashion--i.e., with the filament under direct current at around 3.5 volts with variable anode/grid voltage.

Where on John's Rev D clock (also auto dim parts, tube drive to spec) board the digit brightness level at low brightness (in a dark room) is uneven with the rightmost seconds digit being noticeably dimmer.

Uneven brightness due to manufacturing variation is common at low anode/grid voltages. I'm not certain as to why, but suspect the following: At low anode/grid voltages, the phosphors are not being bombarded to saturation. Minor differences in electron intensity result in differences in brightness. But when the phosphors are being bombarded to saturation, minor differences in electron intensity do not affect brightness.

Regardless, at the lowest brightness levels, the xmas hardware revision does use an anode/grid voltage that is significantly less than what is provided by the boost circuit. To switch high, the MAX6921 has a delay and rise time on the order of 3 us; to go low, the combined time is a little less than 1 us. The PWM signal at the lowest brightness level is only about 1 us in duration, so during each pulse, the MAX6921 anode/grid voltage only rises to a small fraction of that provided by the 50v boost supply. Thus, even in the xmas hardware revision, tube-to-tube variation becomes apparent at low brightness.

I don't know if this explains the dimmer digit on your xmas Rev. D board, but thought it worth mentioning... especially since the brightness was consistent with the pair of Rev. D boards that I built.

[phild13]

I agree that tube construction and variation probably explains it and don't think anything is wrong with the REV D board. I tried different tubes on the clock and there is variation between the 3 tubes but the dim digit look is always basically the same. I am trying remember to get a decent photo of the tube in the dark to post just so everyone knows what it looks like.

[jarchie]

Thanks for looking into this, Phil. I certainly appreciate it and trust your description. So there's no need for a photograph if it's going to be any trouble. It takes me ages to tweak my camera's settings in order to accurately capture the look of these tubes.

[phild13]

Here are a couple photos showing the digits at low and normal brightness. Photos were taken with available light without flash. Some post processing in Photoshop was done. The clock board is John's Rev. D board with auto dim feature installed and running his firmware. The auto dim parameters are -5 minimum brightness and +20 maximum brightness. The drive to spec is turned on in the firmware and the clock is using [edit: FET's] for the AC filament drive.

The parallel red lines in the low brightness photos are the filament which glows when driven to spec. Under normal room brightness the filament glow is generally not visible. The Adafruit clock does not exhibit this as the filament is under driven enough so there is really no filament glow. If you look hard enough in a darkened room at an Adafruit clock you can see the filament glowing slightly.

I did not take any photos, but in an Adafruit clock (with an upgraded fet, 11 ohm resistor and John's firmware) the digits are more even at low brightness levels, but the lowest brightness level is a bit brighter so this would be a factor in evening out the digits. This higher brightness level prevents a direct comparison, but indicates that the tubes I used in testing perform very similar to each other.

I tried 3 different tubes and each exhibited the same characteristics at various brightness levels in both clocks.

There is nothing really wrong in my opinion and the digit unevenness at low brightness. The digits are fully readable at the lowest brightness in a darkened room.

[Russell 27]

Since I took a scolding for asking a question in my own post, I've added voltage drop to the original chart. Something I should have done originally. This should help in understanding the circuit dynamic as a whole.

[jarchie]

Here are a couple photos showing the digits at low and normal brightness.

Those pics are gorgeous! Nice job with the camera and touch-ups. Thank you!

I see what you mean about the last digit being not as bright, and you might be correct about Rev D clocks generally having that feature at the lowest brightness. My impression is that I would have ignored that degree of variation, thinking that it was normal tube-to-tube variation resulting from low anode/grid voltage. After seeing the photos, I now suspect that you're noticing something that I simply missed in my testing. But I need to do more thinking and perhaps testing before writing a proper reply.

Unfortunately, I gave away all of my Rev D clocks last Christmas, but I think my 12-year-old neighbor is going to build one soon. Perhaps he will let me play with it to see if I can replicate your findings. :-)

Since I took a scolding for asking a question in my own post,

It really wasn't my intent to scold, and I sincerely apologize if that's the way I came off. I tend to express myself in a very matter-of-fact fashion, and that can easily be misinterpreted as harsh. It's one of my many character flaws, I'm afraid. Next time I write something that seems harsh, please remember that it's probably just my inability to express myself accurately.

I've added voltage drop to the original chart.

Very useful, thanks!

It also shows why using the kit Transistor without a resistor is a very viable option.

From the chart, it looks like you took a current measurement before and after R3, which means that R3 was always in series with the IV-18 filament during the testing. Replacing R3 with a jumper should lower the total resistance, increase the current across the filament, and increase the voltage drop across the transistor. Did you try that? Or maybe I am misunderstanding what you were measuring?

[phild13]

Russell
Hey, how come you have a better Fluke than I do? :D

Thanks for adding the voltage drops to the chart.
It's what is suspected for R3; that it is too big especially when used with the ZVP3306A. Given the data you provided, a really easy no cost way to test or possibly solve for a flaky display would be to jumper or eliminate R3 when using the kit supplied ZVP3306A FET.

A bit interesting is the BS250 (14 ohm) with less voltage drop than a ZVP2110A (8 ohm) as shown in the chart.

John, the second column (Ampere) results are without the resistor in circuit and the third (22 ohm) column is with the 22 ohm resistor in circuit. Edit: At least that is the way I am reading the chart.

Edit:
I should add that the MAX6921 needs a minimum logic level input voltage of 3 volts on VCC (pin 7) which the ZVP3306A can barely meet under the tested conditions. Due to normal manufacturing tolerances of various components used in the clock and the lower average tested voltage (4.7) of the 5 volt rail of the clock, this may make the MAX6921 flaky in its operation causing random display issues with various clocks. The suggested replacements for the ZVP3306A which are the ZVP2110A FET and the 2907 Transistor + 470 ohm resistor will ensure proper VCC levels on the MAX6921 VFD chip.

[jarchie]

John, the second column (Ampere) results are without the resistor in circuit and the third (22 ohm) column is with the 22 ohm resistor in circuit.

Thanks. I'm still confused about the V_t column. Is that the transistor voltage with no resistor?

[jarchie]

I should add that the MAX6921 needs a minimum logic level input voltage of 3 volts on VCC (pin 7) which the ZVP3306A can barely meet under the tested conditions. Due to normal manufacturing tolerances of various components used in the clock and the lower average tested voltage (4.7) of the 5 volt rail of the clock, this may make the MAX6921 flaky in its operation causing random display issues with various clocks. The suggested replacements for the ZVP3306A which are the ZVP2110A FET and the 2907 Transistor + 470 ohm resistor will ensure proper VCC levels on the MAX6921 VFD chip.

Yes, this was my only concern with replacing R3 with a jumper--possible unreliable MAX6921 operation. I'm aware of three or four cases where people have tried to brighten a dim digit by jumpering R3 with the Adafruit-provided ZVP3306A. In every case, jumpering R3 solved the problem. But I also recall when I jumpered R3 on an Adafruit clock with the ZVP3306A; voltage to the MAX6921 was just slightly below three volts. The clock still worked perfectly, however.

[phild13]

The Vt would be the measured voltage drop through the FET or Transistor. This is the column that would indicate the available voltage for the VFD chip and would also indicate the available voltage for the tube filament. The tube was connected. Russell can of course verify if this is correct thinking or not.

Provided the voltage drops measured were with the tube connected, then the VFD chip will see only voltages in the Vt column and not voltages in the Vr column. If a 22 ohm R3 is installed the tube will only see the voltage in the Vr column and not voltages in the Vt column. If a jumper is used in place of R3 then the voltage the tube sees is Vt.

If this is the case and the clock voltage is around 4.7 volts, then the ZVP3306A provides an absolute minimum, or fails to provide enough voltage for guaranteed proper operation of the MAX6921 chip.

The Adafruit clock I have measures 4.6 at the source pin and 3.9 at the drain pin of a ZVP2110A while the clock is in operation. This voltage is within specs for the MAX6921. For the tube the 3.9 volts is further reduced by my using an 11 ohm resistor giving 3.2 volts at the tube.

[jarchie]

The Vt would be the measured voltage drop through the FET or Transistor.

I guess I'm looking for confirmation that Vt was measured with a jumper in place of R3, because the voltage drop across a FET changes with current. For example, I just disassembled my clock with a ZVP2110A for Q3 and 50 ohm variable resistor for R3 and took some quick measurements. At the minimum of 0.27 ohms, the voltage out of Q3 is 3.78 volts; at 10.72 ohms, 3.90 volts; and at 22.50 ohms, 3.99 volts.

[phild13]

Well since I was not there I can only state what was posted. From the first post, the tube was measured out of circuit with 5.1 volts applied to the filament. Tests made with and without the R3 value.

I'm not good at this but if you look at the spec sheets Vgs graph for the ZVP2110A I think it shows that the FET is operating in what I think is called linear mode (not fully on) not switch mode at 4.7 volts and so voltage drop would change with current. Total available current on the drain is also limited in linear mode though it is plenty for the clock. I think if you were to drive the gate with 10 volts then you would not see any or much difference in the voltage measurement on the drain. The fet was tested at 10 volts so I think that is the level that it will be fully on which puts it out of the linear mode and into the switch mode.

[jarchie]

Well since I was not there I can only state what was posted. From the first post, the tube was measured out of circuit with 5.1 volts applied to the filament. Tests made with and without the R3 value.

I think I understand that much. My initial question was directed at Russell, although I sincerely appreciate your efforts to answer it. The crux of my question is really this: Was Vt measured with a 22 ohm resistor in series with the filament?

I'm not good at this but if you look at the spec sheets Vgs graph for the ZVP2110A I think it shows that the FET is operating in what I think is called linear mode (not fully on) not switch mode at 4.7 volts and so voltage drop would change with current. Total available current on the drain is also limited in linear mode though it is plenty for the clock. I think if you were to drive the gate with 10 volts then you would not see any or much difference in the voltage measurement on the drain. The fet was tested at 10 volts so I think that is the level that it will be fully on which puts it out of the linear mode and into the switch mode.

I'm not good at this either. My point was only that FET behavior can change with current--at least under some circumstances--so in Russell's chart, Vt might differ depending on whether-or-not a 22 ohm resistor is used in series with the VFD filament. Without knowing how the Vt measurement was taken, I don't know how to interpret it. That's what's confusing me, but I suspect only Russell can set me straight--unless I'm missing something obvious.

[jarchie]

I was able to play with an unmodified Adafruit Ice Tube Clock this evening. I think ESD damage to Q3 is unlikely, as it was assembled in my garage. We haven't been using the heat, and the cracks in the concrete floor are visibly wet from moisture seeping through, so humidity must have been quite high. Normal "hobbyist level" ESD precautions were taken (e.g. placing parts on the ESD bag, touching the bag before the parts, holding the case instead of the leads, etc.).

The crux of my question is really this: Was Vt measured with a 22 ohm resistor in series with the filament?

With the 22 ohm resistor installed in R3, the voltage out of Q3 (the ZVP3306A) was 3.32v; out of R3, 2.21v. When I jumper both sides of R3 with IC clips, the voltage out of Q3 falls to 2.94. So in Russell's chart, Vt must have been measured with the 22 ohm resistor in series with the filament.

Oddly enough, jumpering R3 also resulted in the single-flashing-segment variant of the flaky segment problem (examined below)!

...the flaky segment problem--unlike the dim digit problem--is not necessarily due to voltage or current. It seems likely that the flaky segment problem is due to Q3-related power ripple. There's a more in depth discussion later in Frank's thread.

Without the jumper across R3, the display worked perfectly, but the output of Q3 showed significant noise spikes at irregular intervals:



And those individual spikes look like this:



Jumpering R3, which resulted in the single-flashing-segment variant of the flaky segment problem, looked like this:



By reducing the resistance of R3 to about 4 ohms, I was also able to replicate the classic flaky segment problem. I tried to examine the output of Q3, but simply attaching my oscilloscope was sufficient to fix the display!

Replacing Q3 with a ZVP2110A eliminated the noise:



After installing the ZVP2110A, the clock worked perfectly. But, I've played with a couple dozen tubes over the years and put one aside because it was especially prone to the dim digit problem. With the ZVP2110A installed as Q3 and the Adafruit-provided 22 ohm resistor installed as R3, the dim digit problem was noticeable with my dim digit prone tube--but not too bad. The several other tubes I tried worked flawlessly. (I don't have pictures, because I'm not as skilled with my camera as Phil is. It's really hard to accurately photograph these tubes!)

After swapping R3 for an 11 ohm resistor, I saw no evidence of the dim digit problem, even with my dim digit prone tube. With the 11 ohm resistor in place of R3, the output of Q3 (the ZVP2110A) remained nearly unchanged:



So my conclusion is that the flaky segment problem really is due to noise. Furthermore, the flaky segment problem is probably caused by marginal behavior of the the Adafruit-provided ZVP3306A--and not ESD damage. I hope Adafruit reviews their choice for Q3 and revises the kit accordingly.

For now, I can recommend a ZVP2110A for Q3 and an 11 ohm resistor for R3. That combination seems to work well for the vast majority of tubes, although some have a slight brightness gradient at lower brightness levels.