I am currently testing on a PCB and keeping the breadboard portotype for reference. I was quite surprised when what worked OK on three different breadboards did not work on the PCB, but just as you said, the stray capacitances and contact resistances helped stabilize the circuit on the breadboard.
A few minutes ago I added a 10R on the emitter of the drivers, with the 82R on the collectors as per schematic. This made absolutely no improvement to the HF oscillations seen at the output but instead reduced the amplitude on the output slightly. Therefore I would say that the 600pF caps is the solution that I will go with for now.
I think I can revert to the darlington scheme on the same PCB just to try things out if it proves to be electrically stable. I have no problems with thermal stability currently as the heatsinks are way too large and am confident they could easily accomodate the darlington pair.
Since I am working off batteries, any extra mV counts, and to be sure a Li-Ion "12V" pack is actually a 9.8V-12.6V pack, so one needs to test going down to 9.8V. 600mV is quite significant on a wave of 8V (almost 8%).
I have not tried it in class A since I must preserve the batteries.
With my little electronics knowledge I am trying to build a device that produces a 200KHz sinewave at +/-50V into a 1K-2K load. The load is a human being in case you were wondering.
So my device works as follows
1) Two, 12V Li-Ion battery packs provide supply power, +/- 12V DC
2) an op-amp wien bridge oscillator produces +/-8V at 200KHz. LT1361, LM4562 or MC33072 works fine. I have settled on the LT1361.
3) an output current stage produces the current needed to drive a 22R load (typical). The output stage is just that, no VAS, no voltage amplification and no global feedback at all.
4) An 1:6 transformer raises the output voltage from +/-8V to +/-50V at 200KHz. I chose the CFP to get any extra mV out of this stage since there are serious regulation issues with the transformer which took me ages to solve. Had to wind about a dozen transformers on RM10, RM12 and RM14 formers and with various cores, and getting awful regulation, before I realised that the single most important factor was the N ratio. I had to lower the ratio from 1:12 to 1:6 to get almost perfect regulation. This also meant that I had to up the primary voltage and therefore also change the battery supplies from +/-6V to +/-12V, resulting in unexpected thermal stability problems with the CFP (especially on the stupid breadboard - not an issue on the PCB). This was another reason I stuck with the CFP.
5) there are additional modules. An LM567 lights an LED when the output frequency is spot on 200KHz. If the LED is off it means something is wrong as there is no other indication that the device is actually working correctly except this LED. An LM3914 monitors the battery voltage with 4 LEDs to give me a visual indication of remaining battery power. The output needs to be switched across two pairs of electrodes once every second, so there is a 555 driving opto-mosfet relays (LH1510AT). In addition there is a 1R sense resistor on the primary transformer side and another tweo stage diff amp and comparator energising another mosfet relay which shorts the output of the op-amp oscillator in case of overloads (imagine electrodes on sweaty or wet skin).
The electrodes are also a small science. You cannot simply use metal on the skin as it will eventually electrolyse and destroy the skin. I am using instead piezoelectric disks made of very high dielectric PMN PT material, with silver electrodes on one side only and amorphous material on the other side. This produces a voltage drop but apparently also prevents skin electrolysis and damage.
In total this has taken me quite some time as I have been hitting on problems on every step of the way and having to learn and experiment with osicllators, output stages, HF transformer design, semi-conducting materials (eg piezoelectrics, tapes etc), charging of Li-Ion batteries, solid state opto-mosfet relays...
I think I have all that is needed now, and the PCB is almost correct, unless of course I revert to the darlington arrangement. The PCB has cost me 80 pounds to make (I got 4 copies) and I was almost dreaming of a machine I could input the gerber files to and out of a drawer pops out a finished and drilled PCB
Imagine the robot in Lost in Space I wonder how much such a machine would cost.
A few minutes ago I added a 10R on the emitter of the drivers, with the 82R on the collectors as per schematic. This made absolutely no improvement to the HF oscillations seen at the output but instead reduced the amplitude on the output slightly. Therefore I would say that the 600pF caps is the solution that I will go with for now.
I think I can revert to the darlington scheme on the same PCB just to try things out if it proves to be electrically stable. I have no problems with thermal stability currently as the heatsinks are way too large and am confident they could easily accomodate the darlington pair.
Since I am working off batteries, any extra mV counts, and to be sure a Li-Ion "12V" pack is actually a 9.8V-12.6V pack, so one needs to test going down to 9.8V. 600mV is quite significant on a wave of 8V (almost 8%).
I have not tried it in class A since I must preserve the batteries.
With my little electronics knowledge I am trying to build a device that produces a 200KHz sinewave at +/-50V into a 1K-2K load. The load is a human being in case you were wondering.
So my device works as follows
1) Two, 12V Li-Ion battery packs provide supply power, +/- 12V DC
2) an op-amp wien bridge oscillator produces +/-8V at 200KHz. LT1361, LM4562 or MC33072 works fine. I have settled on the LT1361.
3) an output current stage produces the current needed to drive a 22R load (typical). The output stage is just that, no VAS, no voltage amplification and no global feedback at all.
4) An 1:6 transformer raises the output voltage from +/-8V to +/-50V at 200KHz. I chose the CFP to get any extra mV out of this stage since there are serious regulation issues with the transformer which took me ages to solve. Had to wind about a dozen transformers on RM10, RM12 and RM14 formers and with various cores, and getting awful regulation, before I realised that the single most important factor was the N ratio. I had to lower the ratio from 1:12 to 1:6 to get almost perfect regulation. This also meant that I had to up the primary voltage and therefore also change the battery supplies from +/-6V to +/-12V, resulting in unexpected thermal stability problems with the CFP (especially on the stupid breadboard - not an issue on the PCB). This was another reason I stuck with the CFP.
5) there are additional modules. An LM567 lights an LED when the output frequency is spot on 200KHz. If the LED is off it means something is wrong as there is no other indication that the device is actually working correctly except this LED. An LM3914 monitors the battery voltage with 4 LEDs to give me a visual indication of remaining battery power. The output needs to be switched across two pairs of electrodes once every second, so there is a 555 driving opto-mosfet relays (LH1510AT). In addition there is a 1R sense resistor on the primary transformer side and another tweo stage diff amp and comparator energising another mosfet relay which shorts the output of the op-amp oscillator in case of overloads (imagine electrodes on sweaty or wet skin).
The electrodes are also a small science. You cannot simply use metal on the skin as it will eventually electrolyse and destroy the skin. I am using instead piezoelectric disks made of very high dielectric PMN PT material, with silver electrodes on one side only and amorphous material on the other side. This produces a voltage drop but apparently also prevents skin electrolysis and damage.
In total this has taken me quite some time as I have been hitting on problems on every step of the way and having to learn and experiment with osicllators, output stages, HF transformer design, semi-conducting materials (eg piezoelectrics, tapes etc), charging of Li-Ion batteries, solid state opto-mosfet relays...
I think I have all that is needed now, and the PCB is almost correct, unless of course I revert to the darlington arrangement. The PCB has cost me 80 pounds to make (I got 4 copies) and I was almost dreaming of a machine I could input the gerber files to and out of a drawer pops out a finished and drilled PCB
Imagine the robot in Lost in Space I wonder how much such a machine would cost.I said the CFP only made sense in exceptional cases.
By the standards of DIYAudio your application is clearly an exception and CFP seems a reasonable choice. You can probably stabilize the CFP in a more heavy handed way than would be acceptable for audio.
What is it for BTW? Your solution of all the problems by yourself sounds like a university project.
I mentioned class A only as another exception that interested me, obviously not suitable for your battery powered equipment.
Battery powered for earth isolation is used in medical, or is it for portability?
Some of the PCB companies do remarkably low prices on prototypes, I expect they must have very automated solutions not dissimilar to what you want.
Best wishes
David
By the standards of DIYAudio your application is clearly an exception
and CFP seems a reasonable choice. You can probably stabilize the CFP in a more heavy handed way than would be acceptable for audio.What is it for BTW? Your solution of all the problems by yourself sounds like a university project.
I mentioned class A only as another exception that interested me, obviously not suitable for your battery powered equipment.
Battery powered for earth isolation is used in medical, or is it for portability?
Some of the PCB companies do remarkably low prices on prototypes, I expect they must have very automated solutions not dissimilar to what you want.
Best wishes
David
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.