Vref is cap C2 sense+ right? I have two possibilities: lytic 47uF/63V Elna Cerfine & Mundorf 4.7uF ZN poplypropylene film & foil. Also I have the same possibilities for C3.
Wich will put first the lytic or the film?
you will fit both.
Try one for a week. Swap them and try for a week, swap again and then decide which you prefer.
The ones I posted as symmetric 1.2 in page# 190 can do for +/-24V. Those are the ones that Bill initially tested for supplying his OP AMPS. There is a Q2 2N5459 JFET which is 25V VDS max that will set a 30V absolute max Vin given that it sits on 1.85V and there is a Vgs and R1 drop above it. Respect that Vin max. That Q2 must have 7mA or more IDSS for the cascode to work efficiently for a 7-9mA 2SK170BL Q9 under it with the given R4 (220R). If you want to feel free from Vin restrictions or choosing JFETS, you can do the BJT cascode left side. See post #2432 for negative example.
Making an additional positive one like those for 5V, only needs R13 1k trimmer to be 5k trimmer so it can go that low.
R1 is changing current setting, by 0.6V/R1=CCS current. Steadily predictable because that 0.6V is Q7's Vbe. Does not vary significantly per part or batch as with Leds Vf.
As for Vref and output caps subjective preferences, as Andrew said.
Good luck!
I finally had some time handy to do the PCB design I wanted to do for the version 1.2.
First I drawed the schematic again for having an easier time to do the PCB-layout. Then I thought of using a PCB-version of a SMA-connector for the remote sensing and added that one.
I have included options to replace the variable resistor with a network f fixed resistors for lower noise and the pcb allows for the use of a zobel network or a electrolytic cap as output compensation.
The two IRFP9240 are meant to be soldered from the bottom and directly screwed onto a big heatsink - it is for experimenting with lower values for R1 and correspondingly higher heat dissipation.
I usually only design smaller uC-Boards and so I have no experience whatsoever in designing analogue PCBs. Therefore I would be very happy about feedback on how the design could be improved.
Blue is top and red is bottom copper-layer. Aside from two very short sigals, only R-Sense-Ground and general Ground is routed on the bottom (red) layer. The final PCB will have a groundplane on the bottom-side. It is just omitted to show the ground pathes.
Talking about ground-planes, do you advice to add a second ground-plane to the top copper-layer, too?
First I drawed the schematic again for having an easier time to do the PCB-layout. Then I thought of using a PCB-version of a SMA-connector for the remote sensing and added that one.
I have included options to replace the variable resistor with a network f fixed resistors for lower noise and the pcb allows for the use of a zobel network or a electrolytic cap as output compensation.
The two IRFP9240 are meant to be soldered from the bottom and directly screwed onto a big heatsink - it is for experimenting with lower values for R1 and correspondingly higher heat dissipation.
I usually only design smaller uC-Boards and so I have no experience whatsoever in designing analogue PCBs. Therefore I would be very happy about feedback on how the design could be improved.
Blue is top and red is bottom copper-layer. Aside from two very short sigals, only R-Sense-Ground and general Ground is routed on the bottom (red) layer. The final PCB will have a groundplane on the bottom-side. It is just omitted to show the ground pathes.
Talking about ground-planes, do you advice to add a second ground-plane to the top copper-layer, too?
Attachments
Last edited:
I would use the second layer and mounting under, so to bring the 120R gate stoppers and base stopper pads in touch with the semiconductor pads they address. Especially the BC560's base stopper is very far as it is now. C3 and/or C4 output caps can (think pin2 C3) and must solder nearer to the output Mosfet. SMA connector must go to the other side so to be near the components, long free runs negate using a coax Kelvin connection (sense) to a point. You could mix TO-220 Mosfet pads too, along TO-247s?
V3 is attached. Board size is 68x72mm.
I corrected the trimmer and rearranged some components for overall shorter traces. I added the option to use a electrolytic cap for C2, too.
There is one major design decision, I'm unsure with. Looking at the base of the BC546B there is a 1K and a 27 ohms resistor connected (central portion of the PCB).
I have two options here - put the pins connected to the base as close as possible to the base and increase the length of the other traces or leave it as is in the attached layout which resluts in slightly less overall trace-length.
I don't know whether the proximity of the two resistors to the base of the BC546B is as crucial as with the base stoppers on the FETs.
Additionally I was thinking whether it would be good to have another electrolytic capacitor with bypass on the Vin side. Space would be available on the pcb. Which values would you choose?
I corrected the trimmer and rearranged some components for overall shorter traces. I added the option to use a electrolytic cap for C2, too.
There is one major design decision, I'm unsure with. Looking at the base of the BC546B there is a 1K and a 27 ohms resistor connected (central portion of the PCB).
I have two options here - put the pins connected to the base as close as possible to the base and increase the length of the other traces or leave it as is in the attached layout which resluts in slightly less overall trace-length.
I don't know whether the proximity of the two resistors to the base of the BC546B is as crucial as with the base stoppers on the FETs.
Additionally I was thinking whether it would be good to have another electrolytic capacitor with bypass on the Vin side. Space would be available on the pcb. Which values would you choose?
Attachments
If the minimum order are several pcbs & you don't need all, If you are interested I can buy some so togheter we can share the cost to make professional pcbs?
The 27R has proximity priority to the base of BC546B. The 1k should still go nearer than now. Give them 90deg rotation opposite to each other from now parallel, will be easier.
Its a good idea to have a Vin decoupler since you got the space. When rather long cables come from rectification it may help. 47uF//0.01uF.
Also make the red GND trace double width from C4 to 1 OUT.
P.S. Those C2,C3 4u7 MKP look a bit fatter than usual, or is it my idea?
You would have to know it works fine on a prototype first. Are you going to make its negative mirror too?
When the design for the positive version is ready, I will do the negative and a bipolar version, too.
Regarding doubling the pcb trace's width, I have done so for the next iteration of the design (which I will post shortly - just some polishing and exporting), but I was actually thinking of using a ground-plane covering the complete bottom-side of the PCB which will "significantly" increase all ground "traces".
When showing the design with groundplane it is just hard to see a "real" ground path.
Or would you advise against having a ground-plane?
You got eagle eyes!
They are indeed "fatter" because I opted for having space for up to 10uF, but all the MKP10 ones are available in 37,5mm pitch. To allow for that modification, it is actually 5mm wider...
I am not a PCB expert but a ground-plane is something non current flow defined and creates a capacitor with almost everything (I guess).
Next Iteration of the layout:
- swapped the resistors around to be very close to the base of BC546B
- added 47uF and 0.01uF decoupling
- increased the width of pcb-traces
Anything else that could use some tweaking on the latest PCB?
I'm definately not so deep into the circuit that I would know which signal pathes should be kept as short as possible and which ones could be longer...
I used MKP10 for 0.01uF and Silmic II for 47uF - better choices?!
- swapped the resistors around to be very close to the base of BC546B
- added 47uF and 0.01uF decoupling
- increased the width of pcb-traces
Anything else that could use some tweaking on the latest PCB?
I'm definately not so deep into the circuit that I would know which signal pathes should be kept as short as possible and which ones could be longer...
I used MKP10 for 0.01uF and Silmic II for 47uF - better choices?!
Attachments
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.