Arbitrary/Function PSU Idea. Usefull or Not?

[Mechatrommer]

I'm in SMPS (offline) learning process, so quite dangerous working with Mains, i need a signal (sine or square) at frequently used SMPS frequency design (30KHz-1MHz?), that signal has to be powerful (at some relative point) to drive the transformer (not the high impedance, ie >= 50ohm virtually no power output from typical AWG/FG), i need to study the SMPS transformer behaviour first before making the PWM controller (probe the secondary signal, measure the primary amperage at open load etc), so i built one small scale (10W) Arb PSU, (or APSU? is there any term?), please vote and share your thought about this. i never heard of it esp larger scale (30V, 10A etc) or i just simply ignorant. and due to my ignorance, IMHO this can be the next thing to the Lab/Bench PSU that we all familiar with.

ps: you can only give one vote, can be changed later. stat only visible to voters. cheers!

[ejeffrey]

The output stage of any push-pull amplifier looks essentially like this, but there are some limitations in your design.  The biggest issue is that you should draw the feedback link on your schematic!  I thought you were running the op-amp without feedback until I saw the text labels. 

The biggest issue with the circuit itself is that when the sign of the current changes from positive to negative or vice-versa, the opamp has to instantly slew ~1.2 volts to turn on the other transistor.  This takes some time, and many opamps don't have the highest slew rates ever.  This will cause crossover distortion in your output signal.  The normal solution to this is to add a bias current to the transistor bases.  This creates a narrow zone near zero volts where both transistors are conducting, so that the opamp can switch between the two transistors more smoothly.  Google for 'class AB amplifier' for more information on how to do this. 

The final problem is that the type of transistors you want to use for this have very low current gain.  Power BJTs often have current gain less than 10.  If you want 10 amps output, you need more than 1 amp from your opamp.  Few devices will be capable of this.  The standard solution to this is to replace the output transistors with a Darlington or Sziklai pairs.  Then you multiply the current gain, and bring your opamp requirements down to a manageable level.

Edited: formatted for readability.

[Short Circuit]

Arbitrary power supplies do exist, but they are only usable for slow changing signals.
http://www.elektroautomatik.de/products/programmable/ps8000.html
http://www.chromaate.com/Product/62000P_series_Programmable_DC_Power_Supply.htm

For your average function generator you can get a power amplifier stage;
http://www.tti-test.com/products-tti/text-pages/gen-wa301.htm
Rigol PA-1011 10 W Power Amplifier

[ejeffrey]

I seem to have misread your schematic.  10 watts, not 10 amps!    1 amp is possible with the transistors you have specified and a relatively beefy opamp, but a pair of transistors is still probably better.

[saturation]

Essentially you want a power amplifier with good wide bandwidth, such as an RF amplifier.  You can then drive the RF amp with the FG to get varying output frequencies, and if the bandwidth can take it, it could output the other waveforms well [ square, triangle etc.,]; any RF amp design can do it, distortion rises with more power output, but a basic design should work for your need then limit output power to get better frequency response.

http://www.nxp.com/acrobat_download/various/SC19_POWER_AMPL_DESIGN_1.pdf

[Mechatrommer]

The biggest issue is that you should draw the feedback link on your schematic!

thank you my master for this advice i was want to be a little bit tidy without wire crossing over.

the opamp has to instantly slew ~1.2 volts to turn on the other transistor.  This takes some time, and many opamps don't have the highest slew rates ever

fully agreed, high end opamp is required for the job.

The normal solution to this is to add a bias current to the transistor bases...Google for 'class AB amplifier' for more information on how to do this.

internet is referred, i have here 2 thick books are referred (skipped the math ) 1 is used in academia, 1 is the famous AoE... page 92, figure 2.57 using resistors and diodes to bias the voltage... is done on the breadboard, it almost blow the push pull BJTs for no reason (open/no load, luckily i was there seeing the first smoke), dissasembled and never want to try it back, hence i never analyzed it thoroughly, my suspicion is that BJT's Vbe is less than Vdrop of the diode, resulting "window of both conducting" hence shorting out V+ and V- rail. i believe this kind of thing needs meticulous tuning and parts selection which i'm not intending to get into yet. AoE doesnt give a warn on this, hence i'm warning newbies like me out there, - "dont let the push pull become both conducting". the other biasing method using many resistors and smaller BJTs almost made me fainted, maybe next time. for now, as a newbie i am, i just take for granted the 7V/ns slew rated opamp, for simplicity, it worked (even though not 100% perfect esp under heavy load).

the simplistic idea is there, i'm not get into serious detail of serious APSU just yet such as parallel BJTs/MOSFETs, darlingtoning or such. standalone unit may needs more involved circuitry and more complex than i can comprehend. for now, i will rely on my AWG to give the signal input. anyway, thanks Jeff for the advice, i'll keep in mind and i appreciate it. cheers

[ejeffrey]

AoE doesnt give a warn on this, hence i'm warning newbies like me out there, - "dont let the push pull become both conducting".

Well, the goal is for them to both be conducting at once, but only a small current, say 5% of the operating current.  Otherwise, as you have discovered, they will melt

As mentioned in AoE, just the resistors and diodes are not thermally stable.  You can improve matters somewhat by mounting your diodes to the same heat sink as your power transistors.  The solution here is emitter resistors.  A pair of small resistors between the transistor emitters and the  output will do wonders to stabilize the circuit.  That is the function of R3 and R4 in figure 2.58.  If you use the complete bias circuit of figure 2.58 that is even better.  You can then use the trimpot to adjust the amount of current that flows across both transistors.  This circuit is not really any more complicated.  For your application, Q1 is not needed -- that represents the previous gain stage.  You would just connect your opamp to the base of Q3.  So really, it is only an extra trim pot and the two output resistors compared to the circuit you built.

One more problem with your circuit:  Your spec says 2 MHz, but your selected transistors have a minimum current gain of _3_ at 1 MHz.  I expect you will get more like 100 kHz 3dB bandwidth out of this.

[Mechatrommer]

I seem to have misread your schematic.  10 watts, not 10 amps!    1 amp is possible with the transistors you have specified and a relatively beefy opamp, but a pair of transistors is still probably better.

actually i'm using recycled in old stock C2238 NPN and A968 PNP power BJT, not exact complement, but well, i'm quite sloppy on this one, please excuse me. important part is they have both better hFE (25-35) @ 1A, the 200mA output opamp still can handle them directly (with heatsink of course!). i have better TIP122 and TIP125 darlingtons, but i will keep that for another day. this APSU one is a bit sloppy, as long as it works, its fine, and i feel more secure in case they blow up, i still have better stock

Essentially you want a power amplifier with good wide bandwidth, such as an RF amplifier.  You can then drive the RF amp with the FG to get varying output frequencies, and if the bandwidth can take it, it could output the other waveforms well [ square, triangle etc.,]; any RF amp design can do it, distortion rises with more power output, but a basic design should work for your need then limit output power to get better frequency response.
http://www.nxp.com/acrobat_download/various/SC19_POWER_AMPL_DESIGN_1.pdf

i'm using Texas Instrument's THS3062 for the job (discussed last time in another my opamp thread), its not classified as RF, but this new breed of CFB opamp really rocks (hence the 500ohm in the line of feedback there). NONE opamps in my stock i've tried ever comes close to the performance of this 7V/ns opamp, you name it 324, 741, 4558 they are all far behind. you got to see it to believe it, but one thing is that you seldomly see an opamp will smoke so easily like this 3062, already burnt a pair last time. thanks for the link.

it seems it tends to wander to technical smorgasbord side of the design rather than its "sociology", the thing is the circuit above is "quick tested" and working as i've expected, however under heavy load, the output will distort so badly, but let be realistic with this 10W circuit. i'm yet to make the real testing, driving and analyzing the smps transformer, but i will test the low wattage working condition first, hence the low spec of 1A... soon. picture below is the finished "no so tidy" deadbug to save time. it gets its input signal from Hantek DDS3x25 (with my 10x25 amplification in between).

anyway, Short Circuit indicated it does exist (german site i cant understand) so its good to know, i wonder how they are priced, seems like a serious equipments. however i have to agree on low frequency capability, that there will be less usefulness for high frequency, i cant imagine what, except for simulating mains frequency, can be made as an audio amplifier (if the input port is provided) for testing DIYers winded speakers, smps etc, maybe above 2-5MHz will be pointless, up to tens and hundreds of MHz, i believe it will invite EMI law enforcer to come home, not to mentioned the annoyed hams (none around here safer for me)

For your average function generator you can get a power amplifier stage;
http://www.tti-test.com/products-tti/text-pages/gen-wa301.htm

i believe the wa301 is signal or voltage amplifier, not power amplifier since there's no mentioning about the wattage. and ... "can drive up to 300mA peak into a low impedance or short circuit" is not that tempting for power amplifier (normal FG spec).

for the http://www.signaltestinc.com/Rigol-PA1011-10W-Power-Amplifier-For-DG3000-p/pa1011.htm, i believe thats what i'm looking for, but working as an addon for DG3000 series (not standalone), 10W at $300+ is quite a fortune IMHO. i just built a 10W 2MHz (5MHz actually, up to 30-50MHz with very bad heat dissipation and signal attenuation due to the opamp heat) out from junkyard for free.

[DaveW]

I built one recently, although single quadrant and somewhat slower....
http://wattcircuit.com/2012/01/28/100/

For the 2 quadrant you're looking at, it'll keep your op-amp slew rates down tremendously if you use a class AB output stage instead, where the transistor not being used is biased onto the edge of conduction at all times. With that, it should be a standard op-amp that can be used.
You'll need a fast output transistor, and you may well get better results with a FET at those frequencies. For example for a BJT, the transition frequency for a high power BJT like the 3055 is around 3MHz, but this is the point where the current gain drops to 1-not a lot of use in this application!
Also, you can parallel FETs to share the current and the thermal load, without needing the balancing emitter resistors that BJTs need

[amspire]

There is a big difference between a wideband power amplifier and a power supply.

Amplifiers are expected to be stable into resistive loads - perhaps conductive loads, but almost never pure capacitive loads.

Power supplies should be stable into a pure capacitive load. 

There definitely are supplies like this. The Fluke precision programmable series (like the 4265) can do +/- 65V at 1A up to a frequency of about 10KHz and a 0.01% settling time of about 100uS.  They are sometimes available at bargain prices, as there basically is no front panel controls. They have both an internal 0.01% accurate DAC, and the option of an external in that can take an AC reference.

But to make a completely stable 4 quadrant wideband supply can be hard work - it can often descend into some very heavy duty frequency compensation design work that can take more time then the rest of the power supply.

But thanks to Op-amps, we have become addicted to negative feedback circuits, and it is the feedback circuit from the input to the output of the amplifier that is at the heart of stability issues. 

It is easy to forget that you can design amplifiers with no feedback at all and they are stable with all loads.

So for example, if you had a standard wideband opamp amplifier circuit that had the 2MHz bandwidth driving a power amplifier built from a transistor emitter follower stage driving a MOSFET source follower stage, you would get something slightly less then unity gain in the power amplifier, but it could have a great bandwidth and total stability.  It would have a DC offset, but that is where you could have a low frequency DC offset loop that restores the correct average DC offset levels.

To get the 1A currents, this circuit would definitely be wasteful of power -  you may have a few amps flowing continuously through the MOSFET - but you can get a good result while avoiding the scary job of stabilizing a wideband feedback circuit for all possible combinations of output load, voltage and current.

Now just before anyone complains, yes, an emitter/source follower is actually using the emitter resistor as a local feedback, but the point is as it's gain is always less then one at all frequencies and so it is unconditionally stable with all combinations of passive loads.

Richard.

[Mechatrommer]

I built one recently, although single quadrant and somewhat slower....
http://wattcircuit.com/2012/01/28/100/

nice to know someone in the same boat. in fact i built this "1 quadrant" amp earlier, but soon after i finished soldering, the idea popped up to build the 2 quadrant amp (i should read about whats this quadrant all about), so basically, i wasted sometime during the learning (1 quadrant amp), not to mention the wasted copper clad. and your captured below is very familiar, thats the opamp slew rate limitation i experienced earlier (already leave a comment on this in your blog, GBW vs SR limitation)

thanks dave and richard for the advice, i need to polish more on my knowledge.

[Short Circuit]

anyway, Short Circuit indicated it does exist (german site i cant understand) so its good to know, i wonder how they are priced, seems like a serious equipments. however i have to agree on low frequency capability, that there will be less usefulness for high frequency, i cant imagine what, except for simulating mains frequency, can be made as an audio amplifier (if the input port is provided) for testing DIYers winded speakers, smps etc, maybe above 2-5MHz will be pointless, up to tens and hundreds of MHz, i believe it will invite EMI law enforcer to come home, not to mentioned the annoyed hams (none around here safer for me)

For your average function generator you can get a power amplifier stage;
http://www.tti-test.com/products-tti/text-pages/gen-wa301.htm

i believe the wa301 is signal or voltage amplifier, not power amplifier since there's no mentioning about the wattage. and ... "can drive up to 300mA peak into a low impedance or short circuit" is not that tempting for power amplifier (normal FG spec).

for the http://www.signaltestinc.com/Rigol-PA1011-10W-Power-Amplifier-For-DG3000-p/pa1011.htm, i believe thats what i'm looking for, but working as an addon for DG3000 series (not standalone), 10W at $300+ is quite a fortune IMHO. i just built a 10W 2MHz (5MHz actually, up to 30-50MHz with very bad heat dissipation and signal attenuation due to the opamp heat) out from junkyard for free.

Such power supplies cost roughly 1 euro per watt, example the EA PSI8080-60 (80V 60Amp 1.5kW) can be found for €1700 ex VAT.
As fo the speed of such solution; the datasheet mentiones a high-speed option with reduced output capacitance. This option enabled step response 0-100% & vv in less than 1msec.

For more speed & power, check these out; http://www.ar-amps.com/html/12100_rf_amplifier.asp Most of these have 5-6 digit pricetags though 

[erupter]

Since you want to learn SMPS I strongly suggest you ditch the AB class amplifier scheme opting for a full H bridge one.
It's much easier to operate since, unless you're doing things VERY wrong, the commands to turn on/off the two sides are separate (AB class has only one, being designed for Analog duties).

Better still with a logic port the two commands can be linked as in
Side1 = NOT Side2

Although it would be much better off with a micro.

[Mechatrommer]

thanks for the advice. example link? please...

[erupter]

http://lmgtfy.com/?q=full+h+bridge+schematics

[erupter]

I built one recently, although single quadrant and somewhat slower....
http://wattcircuit.com/2012/01/28/100/

nice to know someone in the same boat. in fact i built this "1 quadrant" amp earlier, but soon after i finished soldering, the idea popped up to build the 2 quadrant amp (i should read about whats this quadrant all about), so basically, i wasted sometime during the learning (1 quadrant amp), not to mention the wasted copper clad. and your captured below is very familiar, thats the opamp slew rate limitation i experienced earlier (already leave a comment on this in your blog, GBW vs SR limitation)

thanks dave and richard for the advice, i need to polish more on my knowledge.

I don't get this: if it were a slew rate limitation it would have shown on both directions, rising as well as falling.
But it showed only on rising...
Quadrant is simple: X axis is input, Y axis is output.
The fact that you can go negative on one or both axis defines how many quadrants you can reach.


To better clarify some of the principles involved, you should read material regarding
-analog pwm
-power inverters
-switching mode
-space vector modulation (modulation of the power vector)
-6 step modulation

[Mechatrommer]

i just read about 4 quadrant being able to source and drain inductive load (motor) CW drive, CW brake, CCW drive, CCW brake. so i'm guessing 4 quadrant operation is being able to go + to -ve rail, and also being able to absorb from inductive energy, 2 quadrant only able to supply +Ve and -ve rail, 1 quadrant only +ve rail, correct?

peculiar you cant see the signal is slew rate limited on the falling edge, look closely in the middle of the falling (where the input/sine signal at the max |slew rate|. i believe its harder to see because the sharp edge rising and immediately slow slew rating during the fall make it non symetric shape signal.

edit: this is exactly what i got when using 4558 as the feedback controller, iirc the max f based on "no distortion" slew rate is 167KHz. even though the GBW is rated at MHz figure.

[erupter]

i just read about 4 quadrant being able to source and drain inductive load (motor) CW drive, CW brake, CCW drive, CCW brake. so i'm guessing 4 quadrant operation is being able to go + to -ve rail, and also being able to absorb from inductive energy, 2 quadrant only able to supply +Ve and -ve rail, 1 quadrant only +ve rail, correct?

From what I can pull out of my memory cap yes.
Keep in mind that it's not a limitation on inductive loads: if you have a 4quad you can drain power from the load.
This is used in regenerative breaking, but it requires PM motors, which are different from usual DC motors.

peculiar you cant see the signal is slew rate limited on the falling edge, look closely in the middle of the falling (where the input/sine signal at the max |slew rate|. i believe its harder to see because the sharp edge rising and immediately slow slew rating during the fall make it non symetric shape signal.

I still can't... In fact I expect that if it's the slewrate that's limiting it, it would be more or less symmetrical.
Instead you can see there is both the linear ramping and a delay. I think there are more than one thing at work here.

edit: this is exactly what i got when using 4558 as the feedback controller, iirc the max f based on "no distortion" slew rate is 167KHz. even though the GBW is rated at MHz figure.

You're watching a 4.3V swing at 220KHz.
That makes (I could be wrong) 2.76 V/us, while you're only given 1 V/us.

If you want to test the GBW of the opamp, try with a voltage swing of 0.5 V.

[Mechatrommer]

I still can't... In fact I expect that if it's the slewrate that's limiting it, it would be more or less symmetrical.

you need a bit of imagination here. its not stated in the site what the yellow and blue trace really is. i'm guessing the yellow probe is opamp noninverting input (Vref) and blue is opamp (driver) direct output (not actual power amp/bjt output). there's about 0.3V offset which i'm guessing the power transistor Vbe drop. here again the stable output signal picture from the site, not exceeding slew rate limit...



and here's the (edited) slew rate is exceeded... notice the purple region, compare the yellow and blue traces there, yellow is more steep, blue is not and is constant (straight line), the red arrows showing the transition between slew rate ok and slew rate ko. the orange arrow showing how the bottom arrow/region is more bulged than the top indicating opamp not following during the fall (slew rate limit).

[Mechatrommer]

You're watching a 4.3V swing at 220KHz.
That makes (I could be wrong) 2.76 V/us, while you're only given 1 V/us.

maybe my mistake from the memory, maybe it was a different opamp i was using different setup. but the point is there. lets say a 10MHz unity GBW and 1V/us slewrate opamp. at unity gain, at the limit frequency of (edit) 100MHz 10MHz, it only capable of following 16mVpeak signal perfectly. but if your actual signal is 10Vpeak (20Vpp), the opamp will only be following the signal perfectly at 16KHz, beyond that frequency at 10Vpeak, the opamp output will get distorted (edit) 16mVp less than unity gain signal... Slew rate (SL) = 2*pi*f*Vpeak