Paralleling ULN2803 outputs for unification of ranks

[John_W._Couvillon]

Gentlemen,

A theatre pipe organ is highly unified.  That means that  one or more octaves of pipes can play with more then one rank.  This allows for much fewer pipes in the same ranges, and allows for combinations of ranks to create new ranks.  In the old days,  with mechanical relays it was not a problem to just tie the relay contacts together and to the pipe magnet so any rank could play the  note. With the ULN2803 driver chips, there seems to be a problem with this approach which causes failure of the outputs. With two outputs tied ogether, for example, to play middle C7 from the diapason or flute, it is not know when the failure occurs;  when one putput is conducting,  when one output is conducting and a second begins to conduct, or when an output turns off while the other is still conducting.  In any event,  The problem really impacts the application of the DOUT configuration with the ULN2803s in place.  Would diodes placed between the ULNs output and the tie point eliminate the problem?

Does anyone have experience with the ULN2984 8 channel, source drivers?

Johnc

[doc]

Would diodes placed between the ULNs output and the tie point eliminate the problem?

Yes. If you put Diodes (1N 4004 or 1N4007 should be the right choice) on every ULN-output , you can connect the ULNs together.

greets

Doc

[John_W._Couvillon]

Hey Doc,

Check the attached schematic.  This is what I am considering.  However, I am concerned.

In this arrangement,  with no inputs to the ULN 2803, the output is high so the voltage at the magnet connection with the diode will be near 12vdc.  with an input to the uln, the output goes to essentially 0vdc, so the full 12vdc drops across the magnet coil.  That means that the diode on the other uln has some small voltage on one side( the voltage across the conducting uln) and whatever voltage the uln ouput pin on the other uln has when not conducting.  Would it be possible to put a pull up resistor on the ULN ouput pin to lift it to say 5vdc when not conducting to be sure that the diode is off.

As a second option for paralleling.  Since the outputs from the Din Shift register (74H595) is TTL level,  would it work to drive one ULN inputs from two shift register output pins. 

Anyone who has an opinion please jump in.

Johnc

[Trevor]

John,

Why not just build a diode matrix relay?  You can have all the unification you want but have only one ULN output for each magnet.  They work great with my midibox.

If I understand correctly (and assuming you meant "Dout" instead of "Din"), I don't see why you wouldn't be able to unify through MIDI.  I think jOrgan has an option that lets you do that.

If I were you, I'd use jOrgan.  It's cheaper and easier.

Trevor

[John_W._Couvillon]

Trevor,

Don't quite follow your comment.  About the ULN, all 8 of the outputs on each chip drive magnets.  Are you suggesting using one 8 channel ULN2803 for each note?  Yes, you are correct,  jOrgan can do all the unification, and will.  Actually though,  This is not unification per se.  For example,  if the 8' octave of the bourdon needs to play as part of the Bourdon and also the tibia, the magnet for each note in the 8' octave must have dual inputs such that if the Bourdon stop is pulled, the 8' octave plays with the Bourdon,  or  if the  8' tibia stop is pulled, it will also play with the tibia.  That means having two ULN2803 outpus driving the same magnet.  With jOrgan, the unification is within the full 8' or 16' rank. If I am wrong, someone please step in here.

Two alternate approaches come to mind:

1. Do the paralleling at the TTL level inbetween the 74HC595 on the dout chip and the ULN2803.  Again,  anyone with hardware savey, please jump in here.

2. Do the paralleling with the .ini file in the midio128 by keying the same outputs from different ranks. In the .Ini file, you specify all the midi data necessary to have output on each pin.  If you repeat the channel no., note, and message type for each rank that a particular octave must play,  it may work.  Any comments.

Johnc

[doc]

Hi Johnc,

In this arrangement,  with no inputs to the ULN 2803, the output is high so the voltage at the magnet connection with the diode will be near 12vdc.  with an input to the uln, the output goes to essentially 0vdc, so the full 12vdc drops across the magnet coil.  That means that the diode on the other uln has some small voltage on one side

ok. And what would be the problem then? Some small voltage wouldn't pull your magnet coil, or?

Would it be possible to put a pull up resistor on the ULN ouput pin to lift it to say 5vdc when not conducting to be sure that the diode is off.

Possible (I never tried). Depending on your coil-data it should be a 2W type for a pullup.

As a second option for paralleling.  Since the outputs from the Din Shift register (74H595) is TTL level,  would it work to drive one ULN inputs from two shift register output pins.

Anyone who has an opinion please jump in.

Also possible. You should use diodes (1N4148) in that case, too (Diodes turned around then).

greets

Doc

[Trevor]

Okay, I see what you're doing now.  You're doing borrowing/shared bass.

I'm trying to say that you should use one 8 channel ULN for 8 notes (one channel/pipe).  What you need to do can easily be done with a diode matrix if jOrgan can't do it.

[xpa]

Using the ULN2803 outputs in parallel is not be a problem and requires no external diodes, BUT make absolutely sure that the +12 volt supply for the coils is tied back to pin 10 on the chip and the negative side of the 12 volt supply is connected to the negative side of the 5 volt logic supply. If the grounds of the two supplies are not tied closely to the ULN2803's ground and pin 10, the chip will literally blow it's top. I've had the chip tops pinging around the chamber.

By closely tied, I mean the driver board containing the ULN2803s and the 12 volt brute supply should be located close together (within a few feet) and 18 gauge or better wire should tie the grounds of the 12 volt brute supply and 5 volt logic supply. Do not rely on power cord grounds. It's a good idea to plug the two supplies into the same power strip. Especially in old buildings (theatres) you can find a significant voltage between the ground connection on outlets on different circuits. I've measured ten volts or more.

I had carefully tied the 12 volt supply tied to pin 10 and ASSUMED that the 110v power plug grounded them together. Result: blown chips. I have had no problems since I connected the negative sides of both power supplies together.

George

[John_W._Couvillon]

XPA, Trevor:

Problem comes with paralleling between adjacent uln2803's. Not within one uln2803.  I agree with your grounding suggestion as that is how my setup is wired, but that has not helped solve the problem

Trevor,  so you are suggesting to dedicate one 8 output uln2803 per note. So you could drive the note from up to 8 different sources. You could also do one note with 4 outputs, 2 notes per chip, or even 2 outputs, 4 notes per chip depending on how many different ranks would use the same notes.  That would keep the paralleling within the uln2803. I must agree,  using a diode matrix would be more economical, using all 8 outputs for one note.

Johnc

[Shum]

Hi Johnc,

If you have not had a look at the date sheet for the ULN2803 you can find it at this site. www.st.com/stonline/books/pdf/docs/1536.pdf.    It mentioned that the output of each driver can be connected in parallel. I have included the attached diagram to show an equivalent circuit. Please see Fig.1. When both inputs are switched off, the voltage at point “a” will be at +12V. When input 1 is switched on current will flow from the +12V rail through the relay, point “a” and T1 (conducting) to the 0V rail. The voltage at point “a” will fall almost to 0V (the small voltage equals to the voltage drop across T1). No current will flow through T2 as it is not switched on or conducting. If it is now switched on, both T1 and T2 will share the current through the relay (half goes through T1 half goes through T2). Figure 2 shows an equivalent circuit. Based on the data sheet for the ULN, it is my opinion that there should not be a problem connecting the output of the drives across individual chip, provided the +12V rail and the 0V rail on each ULN board are connected solidly. This was already mentioned by XPA. Based on the circuit, theoretically there should not be a problem. Not sure of the problem you are facing.

Shum

2803_1.JPG

[jimhenry]

Consider the inputs when the problem occurs, one is at ground and one is at +5V. I think you get a low resistance path from +5V to ground from the off input to the on output. If that is the problem then either diodes on the output or resistors on the inputs might be the solution.

[John_W._Couvillon]

Shum,

Thanks for the detailed explanation.

I am satisfied that paralleling outputs on the ULN2803A is acceptable when done within the 8 outputs of a single chip, with careful attention to grounding as stated on other replys. However,  I am not so sure that paralleling outputs on different chips is equally ok. This condition is more the problem for me then within a single ship.

My pipe organ output relay consists of midipox cores and digital output drivers, driving magnets. Each core supports 128 driver outputs, hence 128 pipes or roughly 2 ranks of pipes.  Some of my ranks are incomplete, i.e. the open diapason stops at tenor c,  C3 so I would like to play the 8' octave of the Bourdon when the open diapason plays.  That means that the 12 pipe magnets of the 8' Bourdon are driven from the open diapason drivers when the open diapason plays,  and are driven from the 16' Bourdon drivers when the 16' bourdon plays.  So the output drivers of two different DOUT cards  are tied together to the magnet coils.  Not only are the outputs on different uln2803's, the uln2803's are on different DOUT cards, connected to different Cores.  There are 3 cores functioning as mididecoders, and each core supports 4 DOUT cards.

A little more info.  I use only one 12vdc source for all magnet supplies, and the grounds of all 5vdc ps's are tied

together and to the 12vdc- of the magnet PS.

Johnc

[Shum]

Hi Johnc,

You mentioned that you are comfortable with connecting the ULN in parallel as long as it is on the same chip. As I see it, there should not be any circulating current between different chips driven by different Douts/Cores. This is provided that the 12V and the 0V rails are solidly wired up. Pin 10 of the ULN is a common line where all the flywheel diodes are connected to, and this pin on every ULN must be connected to the 12V rail solidly via a short connection as possible.

When switching on or off a magnetic circuit (magnet/relay) voltage transients are produced. The transients if not suppressed will cause failure of the ULN.  Especially when switching off the magnet, the electro magnetic energy that is stored in the coil will be released as ‘Back EMF’. This EMF will cause a current to flow through the coil and the diode and dissipates as heat. The more magnets that are connected in parallel the higher is the current that flows through a single ULN (one ULN controls two or more magnets). If the connections were not done appropriately it will lead to eventual failure of the ULN.   

BTW have you actually paralleled two or more magnets/relays to one ULN of the same chip with success and failed when making parallel connections across adjacent chip?.

Instead of connecting the ULN in parallel, would it not be a better way to do it on the DOUT side, please see the attached schematic. The 4066 is a switch, which when switched on will link for example in this case, the C notes of different footage. At the bottom of the schematic is a typical keying configuration of an analogue electric organ. As my knowledge on the pipe organ is limited, I was trying to figure how the pipes are connected similarly but I could not find the answers.

After some more thinking an idea came to mind. The second schematic (TOL - thinking out loud) shows how a 16' and 8' pipes can be selected individually or together.

Regards

Shum

ULN.JPG

TOL.JPG

[John_W._Couvillon]

Shum,

Thanks for the input.  Sorry I am so slow responding.

actually,  there is a third solution which I think is the best.

Ver. 3.3 of jOrgan which is a Shareware, Virtual Organ program,  will accomplish the unification for you in software, whereby, you only need one ULN driver per magnet. In addition you can have a full multilevel memory, and combination action on top of that. Not only that,  You can  have one 8' octave of pipes linked to any number of ranks, all done with configuration of the software.

thanks again,

Johnc

[John_W._Couvillon]

Hello Forum,

Having scratched, dug, solicited advice on this subject, a very simple solution is availlable.

the latest version of jOrgan, version 3.3  includes a very flexiible message system for defining I/O on the organ, and between jOrgan and remote midi decoders for driving pipes. It is now a simple task to accomplish in software what we were trying to do in paralleling the outputs of the uln2803. each ULN driver is dedicated to one magnet.

Look up jOrgan its now a fully functional organ relay, including driving SAMS (new), multilevel memory, combination action, and fluidsynth.

Thanks to all who responded with advice and input.  jOrgan coupled with midibox technology is awsome!

Johnc

[Shum]

Hi Johnc,

I am glad to note that you found a solution to your problem. I agree that software implementation is more flexible than hardward option. Good luck with your project!

Shum