DIY SCPI programmable dual channel bench PSU 0-50V/3A (now EEZ H24005)

[jleg]

I think what the other guy was trying to say is. If you turn the knob slowly then it moves the numbers slowly kind of like a fine adjustment. If  you turn the knob fast then the numbers move fast like course adjustment. Does that make sence?

...here's 'the other guy'  - meanwhile Martin also explained that on YT, and i think i got it now - dynamic acceleration is already there, and these 2 parameters in settings seem to define the profile of acceleration. Still wondering if these parameters are necessary, though...   Can't wait putting my hands on this great piece open hardware and try myself. And sneak at the code...

[prasimix]

Still wondering if these parameters are necessary, though... 

We just added it to have a more freedom in defining encoder behavior. What is still missing are proper names for that options, so your suggestions are welcome  .

[kolesar007]

Great rotary implementation and I finaly start to build my PS
Will "Velocity control" be proper name ?

[jleg]

"(up/down) acceleration increment"?
"dynamic encoder increment up/down"?
 

[Strada916]

Enc Up Gain
Enc Dwn Gain

Sent from my D6503 using Tapatalk

[0xPIT]

It is called acceleration!

So, any plans to sell & ship a full kit from the EU?
I want to buy, but I'm definitely not going to pay import tax and customs... rather give 20% more to you.

Best

 pit

[jleg]

So, any plans to sell & ship a full kit from the EU?
I want to buy, but I'm definitely not going to pay import tax and customs... rather give 20% more to you.

well, "import tax" afaik is only about 2.4% (naive TARIC research), plus - in my case - adding 19%VAT. In case this would be offered let's say shipped from germany, you would also have to add 19% VAT.
So imo the only thing to bother with (which also made me think twice) really is "to deal with customs" in general, which might be a bit annoying...
And prasimix already mentioned - shipping from Croatia would mean adding even 25% VAT...

[cs.dk]

And prasimix already mentioned - shipping from Croatia would mean adding even 25% VAT...

Unless you provide a VAT number. That works in EU, but not from US. I'll be very happy if they shipped from an EU country to avoid the customs hell. It's time consuming and expensive.

[jleg]

Unless you provide a VAT number. That works in EU, but not from US.

In case you *do have* a VAT id, you may write this off (getting back VAT) in both cases, i'd say...

[cs.dk]

Unless you provide a VAT number. That works in EU, but not from US.

In case you *do have* a VAT id, you may write this off (getting back VAT) in both cases, i'd say...

Yup,
But customs are a PITA to work with, just their handling fee (and time) is ridiculous. 

[voltsandjolts]

Nice, you can try it directly with our Software simulator. Just open a connection on port 5025 (localhost or IP address if simulator is running on other computer) ...

This scpi-raw ethernet remote control app seems to work with the simulator, can someone try it with a real psu please?

Requires python3
On linux you may need to "sudo apt install python3-tk" or "sudo dnf install python3-tkinter" or ...

Edit: v2 Minor update to fix ugly labelframe borderwidth on linux
Edit: Python file moved to https://github.com/kevinee/eez-remote

[prasimix]

Yes, it works, error messaging is not handled but simple setting and monitoring output values looks nice! See attached short video...
Thanks for your effort. For starting on linux one has to use the following command line:

 python3 eez_remote_v1.py

[voltsandjolts]

Thats great, thanks for the video 
Yes, error handling should be added but its a starting point for a simple cross-platform app.

[prasimix]

I'd like to include current measurement in two ranges, e.g. 0-500 mA and 0-5 A on the next Power board that will continue to be compatible with Arduino shield and firmware will take care about it like with all previous Power board revisions. If this works fine it will become a candidate for group buy on the CrowdSupply.
Idea is to use mosfets as range switches that will be well saturated to provide lowest Rdson. The switching between ranges will be accomplished in a "make-before-break" fashion, that both sense resistors will be connected simultaneously for about max. 1-2 ms.
Additionally I'd like to preserve 4-wire (Kelvin) connections that already exists on the previous board but have no experience nor I had a chance to see how other people design PCB for that. All what I found so far is a not so representative picture of ct'lab DCG power supply (Q8-Q11 for 4-ranges) but without detailed PCB view.



In our case everything will be done with SMT components and I created a new Eagle components for mosfet and 2512 resistor with exposed addition "sensing" pin. Since the next higher bias voltage then existing +5V is +48V, it is used for Q13, Q14 saturation that is cut to +13 V with ZD17, ZD18. The existing 8-but I/O expander is already fully occupied and therefore it's replaced with 16-bit version (IC7, MCP23S17-E/SS) to provide two more lines for current ranges selection.



... and here is PCB layout current sensing section:



 I didn't manage to connect other end of ZD17 in a way to not cross sensitive measurement line but at least that is happened under right angle and on the opposite side of the PCB.

Your timely comments are highly welcomed (I'd like to place an order for new PCB very soon).

[Tazz]

Low level physical interfaces are scarce and poor on the PI.
The Cortex M3 of the Due is the perfect balance between pure old school µcontroller like ATMEGA/XMEGA and full blow µconputer like the PI or the BBB more suitable for the present task.
M4 or M7 would have been cool, but need the complete design of a µC board. Discovery board would have been an option but the learning curve/"getting started" with the dev environment is not as simple and accessible as the Arduino env and the STM32Duino is not "plug&play".

This is a just to the point. Arduino IDE is selected for two reasons: it's open source and "plug&play". I'd like to see another more powerful MCU that is directly supported from environment such as Arduino IDE. I was entertained in one moment with idea to squeeze down the TM4C1294 Connected LaunchPad to Arduino Due/Mega form factor and make it pin compatible with Due or simply put it directly on what is currently called Arduino Shield board. That MCU is equipped with many interesting stuff, and together with more capable display such as one with FT80x controller that could be an attractive combination even for more demanding projects. Again, the question is how to make it simple for use, and Lauchpad has Energia IDE but I'm not so sure that has a so dynamic community like Arduino does.

Talking about SBCs I see more attractive some open source solution such as BeagleBone instead of RPi.

As a side note, I just received the ST news letter which talk about  Arduino STAR OTTO based on STM32F469. But it's still vapor(hard)ware.

[prasimix]

Yes, in fact that is a second announcement (here is the first one), it's not included into latest Arduino IDE, and price is still not set but there is some "rumor" that it could be up to €80.

[prasimix]

First thanks, everyone for suggestions regarding better naming for encoder params. The new settings page looks like this:



Switch is now always in use, and "Confirmation" and "Selection" mode are merged in one thanks for someone's suggestion. Now you can enable or disable switch Confirmation mode. When enabled you have to confirm new set value with pushing the switch, and when disabled changes will be applied immediately as you turning knob in one or another direction. Moving to next value is performed by switch in round-robin fashion starting from CH1 voltage to CH2 current.

Hopefully new title "Knob acceleration sensitivity" better describe what is all about.

[Kleinstein]

The shunt switching is not working correct as shown. You have to make sure to only measure the voltage over the shunt, and not include the MOSFET_on resistance.

The easy version is having the two shunts in series and use only one MOSFET to bridge out the larger one. The voltage sensing also needs to be switched (e.g. a CMOS switch or two small MOSFETs). Alternatively one could use two of the larger FETs for both current paths and always take the voltage from the upper shunt.

[prasimix]

Thanks Kleinstein, your comments are very useful as usual. I started with thinking about two shunts in series but somehow ends up with proposal shown in one of the recent posts. Please let me know if such configuration makes sense (I'm not sure does one switch like IC3 is enough or not):



EDIT: as Kean notes in post #599 this not going to work: lower leg of HI shunt is shorting LO shunt!

Also if two mosfets are going to use, does something like this make more sense (probably R73, R70 ends shouldn't be connected to mosfet source but POUT- point):

Alternatively one could use two of the larger FETs for both current paths and always take the voltage from the upper shunt.

This part I didn't catch. Could you please elaborate it a little bit more?

EDIT: I'm wondering what is more problematic for accuracy: to include mosfet's Rds(on) or to put something on the sensitive paths between shunts and current monitor IC?

[Strada916]

Why not have both sense resistors in parallel. Have a FET in series with the low ohm. So when in low ohm or high current mode the FET is on negating the high ohm resistor. In low ohm mode you calibrate for the RDSon + low ohm resistor. In high resistor mode the FET is off and you only have the high ohm resistor in circuit. Keep it simple. Kelvin measurements are made at entry and exit nodes of the shunt network.

Sent from my D6503 using Tapatalk

[prasimix]

Why not have both sense resistors in parallel. Have a FET in series with the low ohm. So when in low ohm or high current mode the FET is on negating the high ohm resistor. In low ohm mode you calibrate for the RDSon + low ohm resistor. In high resistor mode the FET is off and you only have the high ohm resistor in circuit. Keep it simple. Kelvin measurements are made at entry and exit nodes of the shunt network.

Sent from my D6503 using Tapatalk

Ok, so you are proposing something like this?



HI range shunt (with lower resistance) cannot negate LO range shunt in total: 10 times smaller current will still flow thru it, but that can be taken into account while calibrating. It's much simpler but I'm not sure how this like any of previously mentioned will behave in practice.

[Strada916]

Why not have both sense resistors in parallel. Have a FET in series with the low ohm. So when in low ohm or high current mode the FET is on negating the high ohm resistor. In low ohm mode you calibrate for the RDSon + low ohm resistor. In high resistor mode the FET is off and you only have the high ohm resistor in circuit. Keep it simple. Kelvin measurements are made at entry and exit nodes of the shunt network.

Sent from my D6503 using Tapatalk

Ok, so you are proposing something like this?



HI range shunt (with lower resistance) cannot negate LO range shunt in total: 10 times smaller current will still flow thru it, but that can be taken into account while calibrating. It's much simpler but I'm not sure how this like any of previously mentioned will behave in practice.

Yes. It's simple and can calibrate in software. Just an idea. The Kelvin connection would be on the high side of the of the FET. Otherwise you are not disconnecting the low ohm. I'll post a pic. Also FET is on 0R02 not 0R2.

Sent from my D6503 using Tapatalk

[Strada916]

Sent from my D6503 using Tapatalk

[artag]

You need to at least have separate resistors from sense resistor to opamp. As drawn, you're shorting out the FET with the positive sense leads. Sense leads on 4-terminal resistors aren't isolated, they're just additional terminals (or PCB traces) that don't carry the through current.

[Kean]

prasimix - in your first series shunt schematic you have the HI shunt kelvin connection shorting out the LO shunt.  You'd need dual analog switches for the top and bottom of both shunts in that arrangement.

Similarly, as Strada916 and artag pointed out, in the later proposed schematic you have the upper shunt kelvin connections effectively bypassing the FET.  And in this arrangement you have to calibrate out the effects of the FET and secondary shunt, which seems a bad idea.

Another suggestion is to have the shunts in series and short the higher value one with a FET under CPU control - as suggested by Kleinstein.  Then have two current monitors, one off each shunt, and an analog switch selecting between the outputs of the two monitor amps before feeding into ADC.  This has the advantage over using the later suggestions in that you don't need to calibrate out the FET RDSon, but it does require another current monitor and a single analog switch.  Actually, if you have a spare ADC channel you don't need the analog switch either...