Greetings!
I've lost loads of sleep researching quite a bit of odds and ends of electronic components, getting the courage up to post a schematic for criticism. I think I'm close to having an acceptable design, but having zero experience with op amps and MOSFETs really puts a damper on confidence. So in an effort to curb rising "magic smoke" levels in the atmosphere, I thought I'd ask for some criticism on a circuit.

My schematic so far:



(Click to enlarge)

Notes:
AWSP40-24 is the 24VDC 40W PSU being used (Forces a big RC time constant...but I can live with it. Might end up using only lower voltages most of the time for battery tabbing)
IC1 - OKI-78SR-5/1.5-W36-C is a DC/DC buck converter outputting 5.0V
IC2A - LM393N is a comparator with an open collector
T1 - 50RIA20 is a thyristor triggered by a footswitch
R2 is overkill in case the short formed by the electrodes during welding is not removed for an extended period of time (but this edge case will generally be avoided)

My biggest concern is whether or not charging of the capacitor will work properly. IC2A (LM393N, comparator) is supposed to compare the voltage set by the voltage divider with that of the capacitor. If it is lower, then it should output 5V to Q1 (IRFP2907ZPBF, N-Channel MOSFET) to charge the capacitor. In the other case, IC2A is supposed to output 0V to Q1 to stop charging.
The idea is that I can charge the capacitor anywhere from 0V to 20V, but using 24V to decrease the RC time constant.
So, does this actually do what I think it does?

Also, please don't be afraid to point out "obvious" things to me - any appearance of experience on my part is purely incidental. For example: All bypass capacitors I believe I need to use (all one of them ) are on the schematic. No fuses/protective diodes are planned to be added (yet?).

 

Is this a "spot welder"?
You are discharging a 1 farad cap into your work--correct?
Single pulse?
Rep rate, even if manual?

For 10V, ε = CV²/2 = 50 Joules
For 20V, ε = 200J
Sounds plausible

What kind of peak current are you expecting?
What is the capacitor spec? ESR? --post link

Your circuit as shown has some problems, but we should be able to help.

First, start with a clean slate and draw the simplified discharge circuit--the SCR triggering has problems because as it starts to turn on, it starves its gate signal as the cathode voltage rises--this may not damage the SCR, but is a poor practice. The gate drive is far too weak--for something like this, you should shoot for a 1A, 5uS gate signal pulse.

I recommend that you ground the cathode of the SCR and place your electrodes between the cap anode and the SCR anode. Is this acceptable with you?

We will worry about the charging issues later--you can manually charge the cap and test the discharge concept first.

 

Is this a "spot welder"?
You are discharging a 1 farad cap into your work--correct?
Single pulse?
Rep rate, even if manual?

Yes. The repeat rate is limited by the capacitor charging rate.
To charge to 20V, I expect it to take ~26 seconds (0.932F).
Also, the $100 CD Battery Tab Welder has had an influence on this project.

What kind of peak current are you expecting?
What is the capacitor spec? ESR? --post link

I assume you mean discharge current - with all the various impedance values scattered all over the place, I have no idea. I do know that similar projects have had success. Here's a gratuitous (2-pulse) example.

I've sourced a Scosche ECAP1. I'm gambling that texaspyro's measurements are accurate - 0.932F, 1.52 mohm @ 20Hz.

I've taken your suggestions and updated the discharge portion of the schematic.

I appreciate your input. I do wish to order all the parts necessary (including for the charge circuit) in one order, though.

Oh, note that I do intend to make the appropriate thyristor/electrode/large value capacitor connections as close as possible, and this will not be reflected in the schematics.

 

OK on the redrawn circuit.
This is what I recommend for a trigger circuit--provides 100nS gate pulse risetime--need to really slam it on--do not recommend a switch as it tends to provide a dirty pulse--the transistor is a programmable unijunction--when the anode voltage exceeds the gate voltage by 0.6V, it triggers and dumps the 0.47uf cap into the SCR gate--will reset by itself. http://www.datasheetcatalog.org/datasheet/on_semiconductor/2N6027-D.PDF

Will think about the charging circuit cutout.

 

Made an attempt at another charging circuit without the need for a DC/DC buck converter. This one uses a new power MOSFET to be compatible with the new gate voltage - and is also cheaper.

One design assumption to note is that this charging circuit will prevent the main discharge thyristor from resetting. To me, this does not appear to be a "big deal" for the following reasons:

• The charging resistor has been chosen to dissipate the extra power indefinitely
• When welding, the area the electrodes make contact with is considered to be "not-safe"
• After welding, the electrodes will typically be removed by hand very quickly (as they are held in place by hand)

Although, I'm still open to a solution that addresses this problem.

New parts used:
PHP29N08T127 (N-Channel Power MOSFET)
2N6027RL1G (SCR)
I think I'm going to use a PS-65-24 (65W PSU) for shorter charge times. I could reduce R2 to 8.9 ohms and get to 20V in 16 seconds, 10V in 4.8 seconds!

I also redrew your discharging circuit.
So, that PUT dumps out the contents of that 0.47uF cap so fast that Q2 will be reset, eh? Does that mean that if I don't take my foot off the switch fast enough, C3 will recharge and Q2 will fire again? I'm probably over-thinking this - the more sane alternative is that the +24V keeps Q2 from resetting, and releasing the switch causes Q2 to reset. Curious choice of words, "will reset itself"

Also, does the comparator need a current-limiting resistor on Vcc+?

Finally, a question about T1 placement: what would happen if I tried to spot weld something that is grounded? It seems to me that this placement is not safe for this.

 

Well, I probably did not have to say that it resets itself as it is obvious--but if you hold the pedal down, it will keep pulsing--howbeit with the energy storage cap already discharged--not a problem.

You do not need the comparator--it makes it unnecessarily complicated.
Is the 24V supply regulated? and short circuit protected? SC current?

On T1 placement, yes it is an issue--that is why I asked if was OK to reposition the SCR--to ground your work, you will have to move the SCR back and drive the SCR gate via a pulse transformer or do something more tricky--will think about it--it is desirable to really hit the SCR gate hard because the SCR has a di/dt dynamic current rating of 200A/uS--this means that in you application, it will be dissipating additional power until the device is fully conducting--there is no way to get a 100nS gate pulse rise time with a pulse transformer, but some kind of isolation will now be required for the floating gate circuit.

This is what I have in mind for the charging circuit--the zener shunt regulator has to be sourced via a current source to attempt to keep it in regulation because there is only 2V head room--the NPN darlington is configured as an emitter follower to provide automatic voltage regulation when the capacitor gets charged--the capacitor charges at the current limit rating of the DC power supply--perhaps a 1Ω series power resistor may be needed to control peak current when it starts--if it pulls the power supply down to a lower voltage, it is OK since it will not need to voltage regulate until the capacitor gets nearly fully charged. What do you think?

When the pedal is pushed, the charging circuit must be turned off--will have to think about that one too.

 

The PS-65-24 is regulated and has short-circuit protection. I'm guessing the current range's upper bound is when short-circuit protection kicks in; 3A. The PSU is rated at 2.7A.

Given the complications involved, I'll just have my work isolated. For its primary purpose (spot welding batteries) this will be no problem.

As for the charging circuit...holy crap I spent 4 hours trying to figure out why everything on the base of the Darlington transistor is necessary when I really just don't understand transistors. I couldn't find any documents on the web to help me with a select problem concerning transistors: When the base to emitter of an NPN transistor is forward biased, but the base voltage is lower than the collector voltage, is the emitter voltage that of the base or the collector?
Perhaps it would help if you just explain why we can't use a simple voltage divider on the base of the Darlington transistor instead of that current source and zener diode.

if it pulls the power supply down to a lower voltage, it is OK since it will not need to voltage regulate until the capacitor gets nearly fully charged

Does this mean that the capacitor won't be charged by a 24V potential? The idea behind using a comparator and MOSFET is that the charging time is sped up since we're not dropping the potential used to charge the capacitor - we just stop charging when the desired voltage is reached.
Also, for the circuit as you've laid it out, I must choose a pot that is rated for at least (22V)^2/(1000ohm)*(1470/1470) ~= 0.25W?

Finally, must the charging circuit be turned off during discharge using your circuit? Is the only reason being that the charging circuit prevents T1 from resetting? I'm tempted to just throw on 75W worth of 8 ohm resistors on the charging circuit. What with the PUT rapid-firing the SCR, I don't see any immediately obvious simple solutions. Plus, it is not very likely that the electrodes will be maintaining the circuit for very long anyway - in order to prevent welding the electrodes to your work, they should be removed pretty quickly.

I do appreciate your help, though. I've learned quite a bit (and lost even more sleep) since I started the thread

 

The power supply current limit is probably in the order of 4A or so--this will charge /recharge C1 in about 6S.

The work may now be grounded as I figured out how to "float" the trigger circuit--really quite simple because R10 provides the return path for the trigger charging circuit--when it fires, the voltage across R10 increases to the output voltage, but it is too late to stop C2 from discharging--D1 prevents it from backfeeding to Q1 (D1 nay not be necessary).

OK you win on the voltage divider--good point--it is now quite simplified--I had made it unnecessarily complicated because I was thinking that the 24V supply was not voltage regulated. On understanding how the transistor works, I do not understand your question so I will offer a simple pointer that will help visualize how it works. Just remember that the emitter voltage will always be one junction drop below the base voltage or two junction drops in the case of a darlington (1.2V)--so if you hold the base at a certain potential (22V in this case), the emitter voltage will be 20.8V. When the power supply is in current limit, charging the capacitor at the max rate, all voltages will be lower, but will be increasing as C1 charges--when C1 gets to its set voltage, everything gets into regulation and the emitter voltage of Q3 can increase no further and the power supply regulates again at 24V.

The capacitor is not charged by a potential, but by a current--the current flows due to a difference in voltage potential--now when this current is lower than the power supply short circuit capacity, the power supply will output its rated voltage, but this happens at the trailing end of the charge cycle.

Your original circuit actually performs much in the same way, but you need the comparator to tell the transistor when to turn off as it is used only as a switch--the 16Ω resistor prevents the power supply from going into current limit, but adds significant charge time as it is now subject to the RC time constant (16S) and will take about 4 time constants (64S) to recharge. Charging at a constant current (power supply current limit) is much faster <10S.

The foot switch now turns off the charging circuit--this is necessary to prevent the SCR from latching on--The SCR turns off only when the current is below its holding current--your original circuit has potentially the same problem. R3 is 10W because it must support 24V during this time--it is rated for the switch continuously closed state.

D2 keeps C1 from backfeeding Q3 (may not be necessary). D3 helps provide a path for resonant welding current to return to the capacitor--without it, the cap may charge reverse significantly due to the resonant effect of the cap and lead inductance (also may not be necessary).

I think this circuit will work well--if not, I will breadboard a scaled down version so I can scope out any problems.

 

Looks awesome, thanks for the help! Transistors finally make sense, too.
R10 does look like it is unnecessary. Keep in mind that when electrodes are placed on the work, those electrodes are almost shorted together (with a small amount of resistance). This effectively does the same thing as R10.

Just one thing left that I'm struggling with - what keeps the charging circuit from drawing more than 4 amps from the power supply? It seems to me that we are exploiting the power supply's hiccup mode overload protection to charge the capacitor.

 

Never heard of "hiccup mode" before, so I checked it out. It provides perhaps 200% current for 1S to get the load up and running.

The PS-65-24 does not have a stated current limit other than a 73W max power rating. It does indicate that it has hiccup mode limiting, and yes, this will not work in this application. Mean-well otherwise makes good stuff and inexpensive--I have used them before.

The only stated overcurrent feature in the ASWP40-24 power supply is current limit which is 120% or 2.04A. Yes, we are exploiting the power supply's current limit feature and that may be perhaps an unusual app for a DC power supply, but there is no good reason why it cannot provide 120% current for 10S. Thermal considerations are generally in terms of 1 to 5 minutes. It does state: "Avoid short circuit >30 sec." You are well within this time period. Note that the current and short circuit ratings must hold up for the maximum ambient temp spec of 60°C--you will be running 25°C--tons of margin.

I would go with the ASWP40-24. It can charge your 0.93F cap to 20V in 9.3S. Use the simple relationship I*T = C*V where I is in amps, T in sec, C in farads and V in volts.

On the circuit, I added a safety bleeder and an LED--also changed R1 & R2 to provide enough base current to make it work for a β of 10 on Q1.

 

On the ASWP datasheet, it says short-circuit protection available. I have no idea what that means, nor can I find additional information on the manufacturer's website or anything to indicate that separate models are produced. Is "Over Current Protection: ... Auto recovery" commonly interpreted as having the current limiting capability you mention?

I'm really anxious about charging the cap with minimal resistance.

 

Over Current Protection:
At Load>120% 115/230Vac, Auto recovery
Over Voltage Protection: Zener clamping at Vo>110%, Non auto recovery
Short Circuit Protection: Short circuit protection available.
Avoid short circuit >30 sec.

My gut feel is that this is simply current limit as stated at 120% current.
Since it states to avoid short circuit >30 sec, I believe that the current does not change when short circuited.
Note that this is not exactly a bolted fault in which the terminal voltage is zero--as the capacitor charges it is not shorted, but simply overloaded in current limit.

What you can do is to purchase the power supply and actually measure its short circuit current and also determine if it has any non-stated fold-back current limiting.

Another solution is to add power resistors in the collector circuit of the darlington transistor with the resistance calculated for full load current--14.4Ω for the 40W or 9.6Ω for the 60W power supply. Since the power dissipation is momentary, no safety factor is necessary, so use the actual power rating. This will increase charge time to about 40sec or 30sec respectively for the two power supplies.

Another solution would be to add current limit to the darlington transistor--see new thumnail. It stated power rating is 65W or 40W @ Tc=75°C. The heat sink must be less than (75°C-25°C) /40W or less than 1.25°C/W for continuous power dissipation @ 24V--my guess is that 2.5°/W would be fine for your intermittent, variable voltage application. Note that 75°C will really feel hot to the touch and if you do not like it, simply blow a little air on it. Check out this heat sink:
http://search.digikey.com/us/en/products/FA-T220-64E/FA-T220-64E-ND/2416494
or http://search.digikey.com/us/en/products/530002B02500G/HS380-ND/1216384

How the current limit works--when the voltage across R11 exceeds 0.6V, Q4 turns on and starves the base drive for Q3 thus taking it out of saturation--very simple and widely used technique.

 

If there is interest, I'll post updates about this project. I think the electrical problems are pretty much solved (Thanks again, jimkeith!) (Perhaps the title of the thread could be updated to reflect this?)

I went ahead and ordered parts, including parts for your current limiting scheme considering the cost of shipping. I drew up some schematics for the chassis for a machine shop to fabricate when I realized that the connector I opted to use may or may not close a circuit to ground if mounted to a grounded metal chassis: NMJ2HF-S. I'll find out Friday if "OPEN CRKT" means the first pin is connected to the mounting panel or not. If so, I'll just throw a BJT from Radioshack in the circuit to control the firing signal. Speaking of the firing circuit, I got a foot switch intended for tattoo machines for only $6 shipped. It has a handy 1/4" mono plug already attached, so that's one problem neatly sorted. For those interested in building a foot switch, I've read somewhere that mounting a momentary switch on a doorstop works great. Now, both of these solutions are dual-use devices - so choose the one that your roommate/tattoo artist friend will be least-likely to steal. If, however, your roommate is a tattoo artist, you'll have to come up with your own solution

Would not recommend FreeCAD for drawing schematics yet, by the way. Major pain to use - had to use Inkscape to render dimensions manually. I'm not sure if there are any solid free alternatives, but I don't plan to be drawing a lot of schematics...so this works fine.

I'll be having a chat with the personnel at the machine shop to see what the best (read: cheapest) way to ventilate everything will be; I'm hoping they can make slits and bend metal out. As a last resort, I might just go at it myself with a drill to make a lot of holes.