EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: leonababy on April 22, 2022, 03:35:17 am
-
Background
I built a VFD clock with a MM5316 chip back in about 1978. Until recently, I was unaware of Numitron and Panaplex displays. Recently, I have collected several VFD, numitron and Panaplex displays, and a nice set of Nixies. After some research, I decided to not design my own Nixie clock, but rather use Neonixie's chip.
Since the BCD versions of the MM53xx series are hard to find, I am familiar with the MM5316 and it is more readily available, I decided to use it to build clocks with all of the technologies except the Nixie. The 5316 was designed for VFD and LCD displays, and is non-multiplexed. To get it to display seconds, I figured out how to output the needed digits on one half of each 60Hz cycle. A block diagram is attached.
VFD solution
This was straightforward, once the signal to tell the 5316 which digits to output was obtained. The digits are paralleled with each other, and the grids of the appropriate digits are driven on the corresponding half of the cycle. The duty cycle is close to 50%, at 60Hz, resulting in a bright non-flickering display. Back in the 70's I did not know the importance of an AC filament for VFDs, which I have implemented with the LM4871 - it works great.
Numitron solution
There is good data available for numitrons, regarding power requirements, and lifetime as a function of voltage. There is less info on keeping the filaments warm to limit inrush current, but the concept is easy to understand. The hardest part was finding parts in production that could translate from the high voltage low current 5316 outputs to the numitrons, which draw 20-24 mA. Since the 5316 is not driving the displays directly, and the digits are valid for half the cycle, a latch is needed. I provided the ability to dim the display by creating an enable that can have a 0-100% duty cycle.
The last challenge - how to drive a Panaplex display?
The only gas discharge tube I have used in my life is a neon indicator. The "design" consisted of simply adding a resistor in series, and connecting to 117VAC. Many power indicators back in the day were made this way. As a teen in the 70's these were entertaining - a comical thought today.
Constraint - since Panaplex displays are hard to find, I have a mixture of dual displays that have a shared anode, and some that have independent anodes. All have a keep alive cathode. As seen in the drawing and mentioned above, this will
be a non-multiplexed design, so not having separate anodes should not be an issue.
Design parameters - From the datasheets on the parts I have (Sperry/Beckman SP-3xx), the anode supply should be about 180 to 200V, the anode to cathode to voltage drop once fired is 135V, and the cathode current should be 180uA typical. As an application note http://www.bitsavers.org/components/national/_appNotes/AN-0084.pdf (http://www.bitsavers.org/components/national/_appNotes/AN-0084.pdf) states, the driver should have an "on" voltage of 50V, and the driver should provide a constant current - about 200uA. The constant current is important, so that changes in the anode voltage don't result in changes in current, and thus
brightness.
Questions -
1. I have no idea how to create a constant current driver for each segment that would interface from the HCT574 (most straightforward approach considering existing design)or anything else for that matter. And of course I need one for each segment. Any suggestions? I have never used FETs, and was wondering if they would be more appropriate than BJT as a constant current driver. I don't know how critical the current match is, but a vintage NS part for this purpose DS8880 was within 1% for each output of the "programmed" current.
2. Is "interdigit blanking" only an an issue when mutiplexing? If I am turning the cathodes of each pair of digits (hrs, mins, secs) on and off only once each 60HZ cycle or leaving them on constantly, and leaving all the anodes powered always, is this not a factor?
3. Similar to #2, will the Keep Alive Cathodes be needed, since I am not multiplexing? From what I have read it is primarily a concern when the displays are multiplexed.
Thank you for any advice in advance. As you can see, I am pretty much clueless other than realizing that I will have high voltage in the circuit.
Don
-
If you use a regulated HV supply. such as from a boost converter, then the current depends only on the HV and the
anode cathode resistor. Then you don't have to design a constant current circuit.
-
Panaplex display segments must indeed be driven by constant current sinks. The drop across any Anode resistor large enough to be useful will vary the digit brightness depending on the number of segments lit.
I have attached the schematic for the Datron 1041 Display board. The main segment drives are off-board (these are DD700 / DS8880) BCD to 7 segment constant current drivers), however you will see in the lower left hand corner, the discrete transistor current sources sinks which drive the +/- and leading 1 segments. These are driven by standard TTL levels. Also on the schematic is the HV inverter, this is pretty primitive and you would need the relevant (small toroidal) transformer - winding details not available.
Questions -
1. I have no idea how to create a constant current driver for each segment that would interface from the HCT574 (most straightforward approach considering existing design)or anything else for that matter. And of course I need one for each segment. Any suggestions? I have never used FETs, and was wondering if they would be more appropriate than BJT as a constant current driver. I don't know how critical the current match is, but a vintage NS part for this purpose DS8880 was within 1% for each output of the "programmed" current.
2. Is "interdigit blanking" only an an issue when mutiplexing? If I am turning the cathodes of each pair of digits (hrs, mins, secs) on and off only once each 60HZ cycle or leaving them on constantly, and leaving all the anodes powered always, is this not a factor?
3. Similar to #2, will the Keep Alive Cathodes be needed, since I am not multiplexing? From what I have read it is primarily a concern when the displays are multiplexed.
1. The attached schematic should sort you out.
2. Yes, interdigit blanking only related to multiplexing, to avoid 'ghosting'.
3. No, the Keep Alive Cathodes are only needed for multiplexed drive, they provide a constant source of ionisation to ensure that the segments light up promptly. The Datron 1041 (non-multiplexed) doesn't use them, the 1051 (multiplexed) does - both use identical Sperry / Beckman panaplex displays.
I hope this helps,
Chris.
P.S. Just a couple of clarifications on the schematic...
- The connection marked '40' goes off to the bases of other transistors on the main board that drive the decimal points (Emitter resistors 27k).
- R12 is there to act as a pullup on /OR as this signal is driven by an open collector gate. The other inputs are driven by an ordinary 7400 NAND.
- The different Emitter resistors are to adjust the cathode currents of annunciators of different surface area to achieve equal brightness.
- The Cathode drive transistors are shown in the BOM as Motorola MPSL01 High voltage NPN. VCEo = 120V, VCBo = 140V, Ic = 150mA, Hfe = 50-300.
- D1, D2 are ordinary 1N1418 diodes.
FURTHER EDIT: I was browsing K04BB today and found the complete 1973 Sperry Panaplex Display catalogue, which includes loads of application data. I've not seen it before, but it is clearly the source of all of the individual display scans on the web! http://www.ko4bb.com/getsimple/index.php?id=manuals&dir=06_Misc_Test_Equipment/Sperry (http://www.ko4bb.com/getsimple/index.php?id=manuals&dir=06_Misc_Test_Equipment/Sperry)
-
I have a bad habit of thinking of things after I post on the internet. I see there are a few replies, which I appreciate and will read later today when I have time, but I did look into current sources, and I thought more about "interdigit blanking" last night.
On interdigit blanking, is this needed any time that a change of the display in a tube display will occur, regardless of if multiplexing is used? The tube needs to be re-fired, to let the ionization "die off" to reduce unwanted effects?
Attached is a basic current source, that has a voltage divider on the input. It was suggested that a zener improves the stability if the voltage supply will vary. So I went down that rabbit hole, but afterwards I wondered if I even need to be concerned since RA will cause the anode voltage to change but not Vref (200V), assuming a regulated supply.
Assuming one of these configurations might work, then I need to turn it on and off. Could one additional transistor, driven by a TTL device, opening/closing the connection to Vref do this?
I appreciate the help already provided, and I will review it today. I just wanted to show I put some more thought into this, that I should have prior to posting - not saying it is correct.
{edit} I discovered Panaplex displays last, after realizing designing a Nixie clock would be a challenge due to the HV drivers and gas discharge challenges. I am observing that I failed to apply the same logic prior to deciding to design a Panaplex clock. I have two Panaplex clocks from the 1970s, and I note they both have HV driver ICs of the day. A Seth Thomas 869, and Heathkit GC-1093, but these are both multiplexed. I'd still like to give it a whack with discrete drivers if I can do it with a few transistors per segment and the existing base design I have.{/edit}
Don
-
I think my reply above will cover your questions.... but you were on the right track with the transistor current sources.
If you refer to the Panaplex catalogue I linked, you will see that RA is just there as a protective limiter against transients.
On interdigit blanking, is this needed any time that a change of the display in a tube display will occur, regardless of if multiplexing is used? The tube needs to be re-fired, to let the ionization "die off" to reduce unwanted effects?
No, you're fine.
P.S. The current production ZTX457 is a more than adequate replacement for the obsolete MPSA42 (and MPSL01 used in the Datron).
-
If you use a regulated HV supply. such as from a boost converter, then the current depends only on the HV and the anode cathode resistor. Then you don't have to design a constant current circuit.
It appears from Chris' response that RA is needed, and therefore the HV will be varying.
Thank you for your reply. All input is welcome for me to understand how these things work.
-
You don't have to have an anode resistor if you limit the cathode current.
The Sperry Panaplex notes indicate that the anode resistor is there as protection against rapid HT fluctuations and transients. As the DD700 / DS8880 output transistors and old MPSL01s are only rated at around 110 - 120V, I can see that some level of transient current limiting protection against, maybe capacitive coupling, or the display briefly entering the abnormal glow region (reduced discharge voltage), would be helpful.
With the 300V ratings of the MMBSA42 and ZTX457, they can handle way more than the full HT voltage, so the anode resistor is probably redundant in this case (unless they provide some protection to the dispay itself?). It doesn't hurt to include them anyway.
EDIT: The DS8880 datasheet also makes reference to the anode resistor, as being there to limit the maximum transient segment output currents to 50mA (how you would get to 50mA without some sort of abnormal display discharge, I don't know).
-
P.S. Just a couple of clarifications on the schematic...
- The connection marked '40' goes off to the bases of other transistors on the main board that drive the decimal points (Emitter resistors 27k).
- R12 is there to act as a pullup on /OR as this signal is driven by an open collector gate. The other inputs are driven by an ordinary 7400 NAND.
- The different Emitter resistors are to adjust the cathode currents of annunciators of different surface area to achieve equal brightness.
- The Cathode drive transistors are shown in the BOM as Motorola MPSL01 High voltage NPN. VCEo = 120V, VCBo = 140V, Ic = 150mA, Hfe = 50-300.
- D1, D2 are ordinary 1N1418 diodes.
I can't help but wonder if the solution Chris posted could be expanded to the 30+ segments that will be needed. In other words, on the base side, only two resistors and two diodes will serve all the segments.
In operation, the clock will have roughly from 12 to 32 segments illuminated. I had a guess the base voltage would remain stable, since the current through the bias string is 1.9 mA ( (5-1.2)/2k)with no bases connected, and base current per transistor is only about 2 uA. There is a measured variation of Vb of .03V, only 1%, from 12 to 32 base loads.
Active analog circuit analysis is not my strong suit. It seems like this is suitably similar to the constant current source I posted above, and a similar analysis would work (see diagram). The differences are that the emitter resistors will be at about 0.35v, not ground, and the bias side is shared by all segments. My gut feeling is this should work. Thoughts to the contrary are welcome. If this will work, I am very pleased with the minimal # of components needed to be added.
I see other replies have been posted as I composed this, but I believe they are relative to the anode resistor. I will post this since it stands on its own, I believe. Thank you again for all contributions. I'd rather put the thought in now before connecting 200V to anything, and risking damage to vintage $$$ parts.
Don
-
You don't have to have an anode resistor if you limit the cathode current.
The segment voltage varies from tube to tube and as the tubes age, so a constant current driver maintains more even brightness.
It will also strike faster (not so important for a DC drive clock), and will have more success in striking a marginal segment.
For transistor current sinks, you can use 2 1N4148s or a red LED to establish the base voltage - cheaper and better than a zener. However, if you just use a voltage divider, the VREF will be the (regulated) supply voltage of the 74HCT drivers.
Thank you. I believe I am on a similar page if you look at the diagram in what I posted, which takes Chris' schematic as its inspiration. Can I use the same bias string (2 x R + 2 x D) for all of the segments, and thus save parts?
I see you and Chris are having a discussion about the necessity of the anode R, which appears to be up in the air. I appreciate the discussion.
Don
You don't have to have an anode resistor if you limit the cathode current.
The Sperry Panaplex notes indicate that the anode resistor is there as protection against rapid HT fluctuations and transients. As the DD700 / DS8880 output transistors and old MPSL01s are only rated at around 110 - 120V, I can see that some level of transient current limiting protection against, maybe capacitive coupling, or the display briefly entering the abnormal glow region (reduced discharge voltage), would be helpful.
With the 300V ratings of the MMBSA42 and ZTX457, they can handle way more than the full HT voltage, so the anode resistor is probably redundant in this case (unless they provide some protection to the dispay itself?). It doesn't hurt to include them anyway.
EDIT: The DS8880 datasheet also makes reference to the anode resistor, as being there to limit the maximum transient segment output currents to 50mA (how you would get to 50mA without some sort of abnormal display discharge, I don't know).
Thank you for your added research and comments, Chris. As you say, it is easy to include them just in case. TBD
-
Without going through the maths, I can say that the bias network on the Datron 1045 was supporting 8 transistors, with a maximum of 3 active at any one time.
My gut feeling (and a quick look at your maths) is that one network per (multi digit) display might be a reasonable compromise. Over the decades, these displays have been been manufactured, first by Sperry, then Beckman and then Babcock, it could be that you have a mix of displays and ages. Having one bias network of 4 discretes per display doesn't seem too onerous and might give you some opportunity to balance up display brightnesses if needed. That then leaves you with just a transistor and resistor per segment. It would also help limit any issues to one display.*
* As an aside, I have a Datron 1030 where they fitted the diodes backwards during manufacture! This would have biassed the transistor bases to 5V. Surprisingly, no damage done (in decades!), I wouldn't have found it if I hadn't noticed that the leading '1' and DPs were a bit too bright. The displays operated normally after I corrected the error. It does go to show that the absolute bias isn't too critical though.
P.S. If you do keep the anode resistors, you are only going to be dropping around 2-3V across them, so it seems a small price to pay if there's even a slight possibility of offering some protection to the displays.
-
Wait a second. Don't I need RA in the scenario being considered, to drop about 50-65 volts?
-
At a segment current of 300uA, with the recommended 2k2 RA, it's only a few volts max.
All Neon plasma display (NIXIE or Panaplex) drives worked on the basis that the segments were either off, with the display and transistor passing only leakage current or on, where the display is dropping its glow voltage and the cathode current sources dropping the rest (a significant voltage). They are based on the transistors never seeing anything above their breakdown voltage at enough current to do any harm. It is tight, especially if the HT goes above 200V. More modern 300V transistors take all the excitement out of it.
-
I was out walking my dog in a state park this afternoon, and came up with the realization that RA is not the place to drop the voltage for the segments - primarily because 2 segments vs. 7 segments would make it impossible to drop 50-65 V with one value of RA. Therefore, a separate R for each segment is needed (diagram attached). I came up with 300k or 316k, depending on tolerance. I really do appreciate the assistance of the kind folks here - I feel like I am learning something.
-
Yeah cathode resistors will do it, don't overthink it with CC circuits. I have a clock with a Panaplex with 5 cm tall digits and they use cathode resistors. Your calculated value might be too high. It should be (supply voltage - sustaining voltage) / current.
-
FYI: I was doing a bit of research and came across the following...
1. see attached below for a quick view of different cathode drivers from this Burroughs pdf https://www.shop-tes.com/content/Datasheets/Panaplex_BulletinP101.pdf (https://www.shop-tes.com/content/Datasheets/Panaplex_BulletinP101.pdf)
I'm thinking you can simplify your CC driver by dumping the 3.1V generation and just connecting the base to +5V or driving the base directly from TTL (no base resistor), ie. (d) or (c) in attachment, respectively.
2. other Panaplex projects are using the classic multiplexed nixie design; MPSA42 cathode driver with single cathode resistor, and MPSA42+92 anode driver.
Project 1: http://tubetime.us/index.php/2008/10/03/panaplex-wall-clock/ (http://tubetime.us/index.php/2008/10/03/panaplex-wall-clock/), schematic here: http://tubetime.us/wp-content/uploads/2012/04/PFClock.pdf (http://tubetime.us/wp-content/uploads/2012/04/PFClock.pdf). Note the D1 zener + D2,3,4,5 diode "pre-bias".
Project 2: https://github.com/74hc595/Panaplex-Clock (https://github.com/74hc595/Panaplex-Clock) using https://www.ti.com/lit/ds/symlink/tpic6b595.pdf (https://www.ti.com/lit/ds/symlink/tpic6b595.pdf) driver (outputs clamped with built-in 50V zener) and 22K cathode resistor. ie. functionaly the same way Dave's video #950 explains how to use a ULN2003 with 48V zener on the common pin to drive a nixie.
-
I was out walking my dog in a state park this afternoon, and came up with the realization that RA is not the place to drop the voltage for the segments - primarily because 2 segments vs. 7 segments would make it impossible to drop 50-65 V with one value of RA. Therefore, a separate R for each segment is needed (diagram attached). I came up with 300k or 316k, depending on tolerance. I really do appreciate the assistance of the kind folks here - I feel like I am learning something.
You're over-thinking it (and possibly not reading replies) compared to your previous calculations - probably best not to take the dog for any more walks without taking a schematic and calculator.
- RA was never dropping the voltage for the segments. It is purely protective. At a typical value of 2k2, it drops hardly anything, as previously explained.
- You don't need the the additional cathode resistors RC1-7 that you've now added. The 50-65V is dropped across the cathode transistor (It's a current source - it has compliance).
- At 300uA and 65V, the transistor is dissipating <20mW.
PS. You still haven't said whether you intend to use through hole or smd components, as edavid asked (although he seems to have deleted his replies).
-
Without going through the maths, I can say that the bias network on the Datron 1045 was supporting 8 transistors, with a maximum of 3 active at any one time.
My gut feeling (and a quick look at your maths) is that one network per (multi digit) display might be a reasonable compromise. Over the decades, these displays have been been manufactured, first by Sperry, then Beckman and then Babcock, it could be that you have a mix of displays and ages. Having one bias network of 4 discretes per display doesn't seem too onerous and might give you some opportunity to balance up display brightnesses if needed. That then leaves you with just a transistor and resistor per segment. It would also help limit any issues to one display.*
{interesting snip}
P.S. If you do keep the anode resistors, you are only going to be dropping around 2-3V across them, so it seems a small price to pay if there's even a slight possibility of offering some protection to the displays.
Yes, I calculated 0.8 to 3V or so if the 2.2k resistor is included. I am wondering since the 2.2k resistor appears to be a need created by the concern for the DS8880 incurring a 50mA transient, if it would not apply to my circuit with discrete parts.
I also considered that using a biasing network for each display might be the way to go. Knowing that it was used for 8 segments in the Daytron is helpful. It will be interesting to see how relative brightness shakes out - in one datasheet posted in Dieter's website, it mentions segment current relative to the b segment (eg 1.25 times, 1.5 times, etc for various segments), and the DS8880 datasheet makes reference to this designed in as well. I also believe I need dropping R's for each segment when biased at 3.1V as I mention in another post. Thank you again, Chris.
-
Yeah cathode resistors will do it, don't overthink it with CC circuits. I have a clock with a Panaplex with 5 cm tall digits and they use cathode resistors. Your calculated value might be too high. It should be (supply voltage - sustaining voltage) / current.
It sounds like I am headed in a somewhat reasonable direction with the help of you guys. The circuit Chris came up with, whereby the switch is also the current limiter was key to keeping the parts count low. I assumed that I would need two transistors per segment, and additional resistors might be needed. I understand the values will need to be refined once I start testing - which looks like I may be ready to do. I have a medical research HV power supply here for bench testing.
-
I think your confusion might be because you're conflating the two approaches. Either you use a cathode resistor or a constant current circuit, you don't need both. I prefer the cathode resistor because if the HV is well-regulated then the current will also be controlled. It's up to you.
-
I was out walking my dog in a state park this afternoon, and came up with the realization that RA is not the place to drop the voltage for the segments - primarily because 2 segments vs. 7 segments would make it impossible to drop 50-65 V with one value of RA. Therefore, a separate R for each segment is needed (diagram attached). I came up with 300k or 316k, depending on tolerance. I really do appreciate the assistance of the kind folks here - I feel like I am learning something.
You're over-thinking it (and possibly not reading replies) compared to your previous calculations - probably best not to take the dog for any more walks without taking a schematic and calculator.
- RA was never dropping the voltage for the segments. It is purely protective. At a typical value of 2k2, it drops hardly anything, as previously explained.
- You don't need the the additional cathode resistors RC1-7 that you've now added. The 50-65V is dropped across the cathode transistor (It's a current source - it has compliance).
- At 300uA and 65V, the transistor is dissipating <20mW.
PS. You still haven't said whether you intend to use through hole or smd components, as edavid asked (although he seems to have deleted his replies).
1st I apologize for not replying to the question of through hole vs. SMD. I can use either. Also, I did not know the MPSA42 is obsolete. I bought some 2 years ago from Mouser or Digikey and did not see a warning this is the case.
2nd, I made an assumption that was obviously incorrect - that the transistors would be saturated. While you say "don't take the dog for any more walks," it is true that on the walk I identified an issue based on my assumption - even though the assumption may be incorrect. It sounds like this is not true. By "compliant" I assume you mean it is in the active region. Is it the case that once it is saturated, it no longer acts as a constant current element?
I am a bit lost now - I admit it. Are you saying that the Rc resistors can be eliminated?
Don
-
I think your confusion might be because you're conflating the two approaches. Either you use a cathode resistor or a constant current circuit, you don't need both. I prefer the cathode resistor because if the HV is well-regulated then the current will also be controlled. It's up to you.
Yes, I am definitely confused. Thinking I understand something, but in fact am clueless on a key point. A bit embarrassing but that is OK. Reading Chris' post and yours, it is clear I did not understand the difference, and the root cause is I assumed that the transistor in the current sink would be saturated. Perhaps if it were, they would not act as a constant current regulator.
Looking at the recent post by pqass with citations from the Burroughs App note 101, that appears to suggest a one transistor approach with a cathode dropping resistor (thought the R values seem low). I believe you are suggesting I am looking at an approach like that to drive from TTL with a cathode resistor, OR a constant current sink that does not need one. And the current sink would not need a voltage dropping resistor.
I need to regroup. I said up front that I don't understand analog well. That is coming back to the fore.
Don
-
1st I apologize for not replying to the question of through hole vs. SMD. I can use either. Also, I did not know the MPSA42 is obsolete. I bought some 2 years ago from Mouser or Digikey and did not see a warning this is the case.
2nd, I made an assumption that was obviously incorrect - that the transistors would be saturated. While you say "don't take the dog for any more walks," it is true that on the walk I identified an issue based on my assumption - even though the assumption may be incorrect. It sounds like this is not true. By "compliant" I assume you mean it is in the active region. Is it the case that once it is saturated, it no longer acts as a constant current element?
I am a bit lost now - I admit it. Are you saying that the Rc resistors can be eliminated?
Don
Yes, that's right, the transistor is operating in linear mode, dropping whatever voltage is necessary to maintain a constant current, so yes the Rc resistors are redundant. It is only if you use saturated switches that you need the Rc resistors.
Yes, MPSA42s are no longer manufactured, although as edavid previously pointed out, they are probably still easy to find. If you do want to go smd, then the MMBTA42 is available in SOT-23 package... https://www.mouser.com/datasheet/2/149/MPSA42-196155.pdf (https://www.mouser.com/datasheet/2/149/MPSA42-196155.pdf) You could also use a BAV99 dual series diode, also in SOT23, in place of the two 1N4148s, saving a package.
EDIT: As retiredfeline points out, you can use either approach. In saturated switch mode, you would remove the Re resistors (grounding the emitters) and drive the transistor bases from the logic via series resistors, and of course adding Rc. So both approaches use the same number of parts per segment, if you ignore the single bias network (2 resistors, 2 diodes) for constant current mode.
EDIT1:
Looking at the recent post by pqass with citations from the Burroughs App note 101, that appears to suggest a one transistor approach with a cathode dropping resistor (thought the R values seem low). I believe you are suggesting I am looking at an approach like that to drive from TTL with a cathode resistor, OR a constant current sink that does not need one. And the current sink would not need a voltage dropping resistor.
Beware of cathode driver diagrams (A) and (C) in pqass's citation (and attachment). The 9307 IC is ECL logic, not TTL. You can't directly connect the base of the transistor directly to the output of a TTL or CMOS logic IC without a series current limiting resistor.
Diagram (B) is essentially a saturated switch, and diagram (D) a constant current source - they look different because they are being driven by 7447 BCD to 7 segment, decoder, which has open-collector outputs.
-
Project 1: http://tubetime.us/index.php/2008/10/03/panaplex-wall-clock/ (http://tubetime.us/index.php/2008/10/03/panaplex-wall-clock/), schematic here: http://tubetime.us/wp-content/uploads/2012/04/PFClock.pdf (http://tubetime.us/wp-content/uploads/2012/04/PFClock.pdf). Note the D1 zener + D2,3,4,5 diode "pre-bias".
Project 2: https://github.com/74hc595/Panaplex-Clock (https://github.com/74hc595/Panaplex-Clock) using https://www.ti.com/lit/ds/symlink/tpic6b595.pdf (https://www.ti.com/lit/ds/symlink/tpic6b595.pdf) driver (outputs clamped with built-in 50V zener) and 22K cathode resistor. ie. functionaly the same way Dave's video #950 explains how to use a ULN2003 with 48V zener on the common pin to drive a nixie.
OK, I went back and looked at Dave's video, which I had watched over a year ago when I was considering a nixie clock. I see the reason for "pre-biasing" - two reasons in his use. One is to keep the voltage low on undriven cathodes to prevent glow, the other is to protect the outputs if the ULN2003 were used.
-
I would forget about the ancient clock ICs and use a microcontroller, it's so much more flexible. I've built a bunch of nixie and other clocks and have always used an 8 bit AVR. For high voltage displays there is the good old MPSA42 transistor that works nicely, or there are 8 channel driver ICs, or another good option is HV shift registers, I have used some parts by Supertex that are I think 32 channels in one IC.
-
OK, I went back and looked at Dave's video, which I had watched over a year ago when I was considering a nixie clock. I see the reason for "pre-biasing" - two reasons in his use. One is to keep the voltage low on undriven cathodes to prevent glow, the other is to protect the outputs if the ULN2003 were used.
Sorry to chime in on this one (again). You would need to assure yourself, probably by physical test, that the ULN2003 is capable of reliably switching the cathodes on a Panaplex display. The reason I say this is that, with a 47V zener clamp, it is only capable of swinging about 48V on the output, this is fine for a Nixie, where only one cathode is illuminated at a time. With a Panaplex, you have multiple on and off segments in very close proximity (worst case is displaying '0', where you have all of the segments lit apart from the middle one), providing a strong source of ionisation very close to the off segment.
With only a 48V swing on the cathode driver, this may not be enough to reliably extinguish the middle segment (in the above example), remember that the DS8880 guarantees not to break down until 80V and typically higher, it is not actively clamping to 'off' cathodes. I'm not saying that it won't work with Panaplex, but it might be more sensitive to HT voltage setting for example. Personally, I would want to test it on the bench before committing to a final design / PCB.
Dave's video was based on Nixie.
-
I found a decent explanation and demonstration on how a simple constant current source/sink of the type I am considering works. I apologize to the board for simply not having a clue.
https://www.youtube.com/watch?v=H-W5K9upohM (https://www.youtube.com/watch?v=H-W5K9upohM)
It appears I have about three ways to achieve what I seek. Thank you for your patience.
-
No need for apologies. If everyone knew everything there would be no reason for questions. :)
-
About the use of Ra and limiting cathode current to 200uA (in your original post):
I don't think you need Ra since it's not mentioned at all in the Burroughs appnote (as a protective measure). Also, the appnote endorses multiplexing so a higher cathode current of 1-2.5mA is recommended (see "Biasing" paragraph) given the 60Hz refresh interval per digit (see "Timing"). A lower current is prudent in a non-multiplexed design but how low, you'll have to experiment (brightness vs. longevity).
About off-bias levels for anodes and cathodes:
See "Biasing" paragraph and Figure 9 (CR19, CR20) in the appnote where off anodes and cathodes are recommended to be biased to 90-100V. This seems to throw a wrench into using a high-integration design with ULN2003 or TPIC6B595 + (cathode) DIP resistors. Maybe you can omit it entirely. The appnote mentions that this may be necessary for displays with multiple digits/tube. If you are using single digit/tube this may not be an issue for you.
I would also recommend you bread-board your choice first (driver chip+cathode resistors OR discrete transistor+emitter resistor current sink, with/without 48/90/100V cathode off-bias, with/without the 90/100V anode off-bias). If you take Project 2 at face value (link in my previous post), TPIC6B595+22K supposedly works. Although, 22K gives 9mA which is too high in my opinion. Also, rather than the appnote anode drivers, I'd use the MPSA42+92 anode driver shown either Project 1 or 2 (link in my last post) schematics.
About the Fairchild 9307 being an ECL part:
See page 8-34 of http://www.rsp-italy.it/Electronics/Databooks/Fairchild/_contents/Fairchild%20TTL%20data%20book%20-%209000%20series.pdf (http://www.rsp-italy.it/Electronics/Databooks/Fairchild/_contents/Fairchild%20TTL%20data%20book%20-%209000%20series.pdf)
It is a TTL part whose output is an NPN with a 2K pull-up (equivalent circuit diagram and Vout vs Iout graph on page 8-37) so internally there is a base resistor. But, I think the base current is limited by Re; according to the falstad.com simulator anyway. See attached. Play with it here: http://www.falstad.com/circuit/circuitjs.html?ctz=CQAgjCAMB0l3BWcMBMcUHYMGZIA4UA2ATmIxAUgoqoQFMBaMMAKACURcAWETFXrlT5QRg6lQnQELAC4guXPLwz8EYfsKoQmxBLHgouYSGGzY8kbFGiEwebGEIEEXDFzQ5kVACZ0AZgCGAK4ANjIsAE4U6rx4Smr82Aj8WhaQLADm0YnJ2Zx4PBLseWAYhCUxEvJCcCKS0tgYtDGl5Qm8COUQRQAOFGWx8QNJ-N1QLADu-eUocXmzSulTCMO57SPjjVTmPMIKSsJj6VHcyokFZyKsQA (http://www.falstad.com/circuit/circuitjs.html?ctz=CQAgjCAMB0l3BWcMBMcUHYMGZIA4UA2ATmIxAUgoqoQFMBaMMAKACURcAWETFXrlT5QRg6lQnQELAC4guXPLwz8EYfsKoQmxBLHgouYSGGzY8kbFGiEwebGEIEEXDFzQ5kVACZ0AZgCGAK4ANjIsAE4U6rx4Smr82Aj8WhaQLADm0YnJ2Zx4PBLseWAYhCUxEvJCcCKS0tgYtDGl5Qm8COUQRQAOFGWx8QNJ-N1QLADu-eUocXmzSulTCMO57SPjjVTmPMIKSsJj6VHcyokFZyKsQA)
I think the same current sink action would work whether it's the 7447 (open collector) or a standard TTL/CMOS (push-pull) output. The former is disconnecting the emitter entirely vs the latter raising the emitter voltage to the base voltage; both shutoff the transistor (Figure 8 (d) in the appnote).
-
I breadboarded the MPSA42, with an 11k emitter resistor, and the bias that Chris showed me of 3.1V. It was instructive. I have to leave town for 2 days tomorrow at 5am, but I used a Vcc of 49V (to temporarily simulate the HV). In varying Re from 25k to 161k, the load current changed from .225 to .223mA as Vc changed from 43.4 to 13.1V.
So the theory works. I bet most readers here are shocked of my ignorance of basic active analog electronics. Again, I appreciate the input from everyone. I am leaning towards this current source method, and the rest of the method in the schematic Chris supplied. I apologize for punching out, but I really need to eat dinner, figure out if I need to order a part for my mower before I leave, and go to bed.
Don
-
In varying Re from 25k to 161k, the load current changed from .225 to .223mA as Vc changed from 43.4 to 13.1V.
That doesn't look right; such a large Re range without changing the current. Confirm when you get back.
If you were to forgo the 3.1V vref, attach all bases to Vcc (5V) and just power the Re bottom leg from your MCU pin/logic gate, then your calculation would be: with Vol=0.35V, then Re=(Vcc-Vbe-Vol)/Ic=(5-0.65-0.35)/0.000223=18K
If the driving MCU pin/logic gate goes H (off position), then Voh must be no lower than Vbe (0.65V) lower than Vcc (5V) or 4.35V or be an open collector/drain.
-
...
I don't think you need Ra since it's not mentioned at all in the Burroughs appnote (as a protective measure). Also, the appnote endorses multiplexing so a higher cathode current of 1-2.5mA is recommended (see "Biasing" paragraph) given the 60Hz refresh interval per digit (see "Timing"). A lower current is prudent in a non-multiplexed design but how low, you'll have to experiment (brightness vs. longevity).
Agreed, I think the Ra is intended to protect the likes of the DD700 and DS8880 with their fairly fragile breakdown voltage and current capability. It's hard to imagine a 300V 0.5A MPSA42 or equivalent running into any difficulties. For the SP35x that I am used to, 300uA is the typical non-multiplexed segment current, although, in practice, you can go much lower, with better character definition (less 'bloom'), 130uA being the minimum spec for full segment discharge coverage.
About off-bias levels for anodes and cathodes:
See "Biasing" paragraph and Figure 9 (CR19, CR20) in the appnote where off anodes and cathodes are recommended to be biased to 90-100V. This seems to throw a wrench into using a high-integration design with ULN2003 or TPIC6B595 + (cathode) DIP resistors. Maybe you can omit it entirely. The appnote mentions that this may be necessary for displays with multiple digits/tube. If you are using single digit/tube this may not be an issue for you.
No, I can't see how the ULN2003 can work in practice. Taking the SP35x again as an example, its typical spec Anode voltage of 180V, subtracting 50V for the ULN2003 (clamped at it's maximum permissible voltage), you are left with 130V. According to the same spec, the typical anode-cathode discharge voltage of the SP35x is 135V. It is hard to see how, with the edges of the glow discharge from adjacent lit segments slightly overlapping its ends, an off segment wouldn't attempt to light (albeit dimly). It might light ok, but would it extinguish again? People seem to have got it to work, but to my mind, it is too close to call. Maybe reducing the anode voltage gives some more headroom, or maybe it only works with multiplexing - with blanking periods.
About the Fairchild 9307 being an ECL part:
See page 8-34 of http://www.rsp-italy.it/Electronics/Databooks/Fairchild/_contents/Fairchild%20TTL%20data%20book%20-%209000%20series.pdf (http://www.rsp-italy.it/Electronics/Databooks/Fairchild/_contents/Fairchild%20TTL%20data%20book%20-%209000%20series.pdf)
It is a TTL part whose output is an NPN with a 2K pull-up (equivalent circuit diagram and Vout vs Iout graph on page 8-37) so internally there is a base resistor.
Ah, ok. I saw 93xx and the configuration and assumed ECL. Open collector with pullup would indeed explain it. Thanks.
I think the same current sink action would work whether it's the 7447 (open collector) or a standard TTL/CMOS (push-pull) output. The former is disconnecting the emitter entirely vs the latter raising the emitter voltage to the base voltage; both shutoff the transistor (Figure 8 (d) in the appnote).
Yes, it would (in fact the logic signals on my Datron diagram are driven by a standard push-pull 7400 NAND). On this point, I think the two diode + resistors base bias network is there to adjust the transistor emitter thresholds to TTL level (the 7400 outputs that I mentioned are also feeding other TTL inputs). I didn't comment on this previously because Don specified TTL in the thread title. If using 74HC CMOS outputs, I think it would be reasonable to connect the transistor bases directly to 5V - with the RE emitter resistors suitably scaled to pass the required segment currents at 4.3V. The Datron bias network would also work with 5V CMOS of course (the reverse Vbe being minimal).
-
Hello,
I was just wondering where you got to with your project.
Did you complete it, how did it go?
I have been bitten very seriously by the Panaplex bug!
I have mostly used Microchips HV55xx drivers to drive displays such as SP-35X and derivatives.
The time has now come for me to attempt multiplexing for panaplex displays that have multiple digits in the same display - so they will ONLY work in a multiplexed mode.
- Richard
-
Hello,
I was just wondering where you got to with your project.
Did you complete it, how did it go?
I have been bitten very seriously by the Panaplex bug!
I have mostly used Microchips HV55xx drivers to drive displays such as SP-35X and derivatives.
The time has now come for me to attempt multiplexing for panaplex displays that have multiple digits in the same display - so they will ONLY work in a multiplexed mode.
- Richard
I suggest you have a look at the service manuals for old test equipment/calculators that use Panaplex displays, e.g. the Datron 1061. The Solartron7075 might give you design concepts, but it uses unobtanium ICs.