Author Topic: fast CC/CV power supply take two  (Read 11792 times)

0 Members and 2 Guests are viewing this topic.

Offline iMo

  • Super Contributor
  • ***
  • Posts: 5652
  • Country: gw
Re: fast CC/CV power supply take two
« Reply #25 on: September 30, 2023, 11:00:48 am »
I wonder whether somebody tried a PSU with for example the TDA2020 - that is basically an opamp +/-22V, 3.5A output, 20W, with two pins for controlling the output transistors (two internal "power and current limiting circuits").
That could be a quick way to make a 4 quadrant PSU if CC mode could be added.

Like this PSU playing music :)

Readers discretion is advised..
 
The following users thanked this post: xavier60

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3037
  • Country: au
Re: fast CC/CV power supply take two
« Reply #26 on: September 30, 2023, 02:09:41 pm »
I have combined my other two ideas into one idea for fast CC response limiting.
While in CV mode, Q5 keeps the non-inverting input of the CC opamp switched to ground, shunting away negative feedback from the compensation cap C4.
When the CC threshold is exceeded, the output of the CC opamp only sees the small capacitive loading of C4 so is able to slew down at close to it's full speed.
Q2 senses when the CC loop is in control and turns off Q5 via Q4, allowing C4 to provide feedback for normal CC compensation.
I took a photo of the CC opamp's output at the time the PSU's output was short circuited.

« Last Edit: October 09, 2023, 03:03:51 am by xavier60 »
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline Zoli

  • Frequent Contributor
  • **
  • Posts: 595
  • Country: ca
  • Grumpy old men
Re: fast CC/CV power supply take two
« Reply #27 on: October 03, 2023, 03:59:24 am »
I' playing around with simulations, with a radically different topology(sorry, no details yet). What can I recommend: use faster components; as example 2N5550, 2N5401 as control transistors, 2SCR586J as bypass transistor, OP37 as op-amp; all of them are part of LTspice standard distribution. Now about the pictures: first is is current limit set to 1.5A, the second current limit set to 3A; current sensing resistor 0.1Ω, reference 7V, output divider 360k/100k. Simulation command: .tran 0 4.5m 4.395m startup; load configuration: PULSE(1 2 4.4m 1u 1u 49u).
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3037
  • Country: au
Re: fast CC/CV power supply take two
« Reply #28 on: October 03, 2023, 05:10:49 am »
Do you see an over-shoot across the CS resistor when the output is short circuited?
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline Zoli

  • Frequent Contributor
  • **
  • Posts: 595
  • Country: ca
  • Grumpy old men
Re: fast CC/CV power supply take two
« Reply #29 on: October 03, 2023, 05:32:02 am »
Voltage over the CS resistor, current limit@1.5A; not nice, but does the job.
Edit bonus: output voltage with 330µF/50mΩ capacitor.
« Last Edit: October 03, 2023, 05:40:56 am by Zoli »
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3037
  • Country: au
Re: fast CC/CV power supply take two
« Reply #30 on: October 04, 2023, 12:54:39 pm »
This simpler version also has 3us of short circuit current overshoot.
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline Zoli

  • Frequent Contributor
  • **
  • Posts: 595
  • Country: ca
  • Grumpy old men
Re: fast CC/CV power supply take two
« Reply #31 on: October 04, 2023, 04:44:26 pm »
This simpler version also has 3us of short circuit current overshoot.
Overshoot where? On the CS resistor, or on the load? To understand the importance of the question, see attachments:
-Green: output load
-Blue: current sensing resistor
The first(Load&CS_W_RC(1+4.7).png) has an RC snubber over the output, 1µF+4.7Ω, exactly like in your drawing; current within 10% the target(1.5A) shortly after the 70µs mark.
The second(Load&CS_W_RC(082+22).png) RC snubber is 82nF+22Ω; current within 10% the target(1.5A) around the 15µs mark.
The third(Load&CS_WO_RC.png)has no RC snubber over the output; current within 10% the target(1.5A) around the 12µs mark, but with ringing extending up to the 40µs mark.
Note: the CS resistor is in series with the main supply, exactly like in your drawing; on my previous screenshots were in series with the load.
It takes around 5µs1 for the CC circuit to take over the bypass transistor control, that's the spike around the 11µs mark.
Simulation command: .tran 0 4.5m 4.395m startup ; PULSE(1 2 4.4m 1u 1u 90u).
Schematics: I intend to release it soon, in public domain; since is niche of the niche, not worth trying to monetize IMNHO, even if it would be patentable.
15µs is the time required to discharge the compensation capacitors in the CV section.
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3037
  • Country: au
Re: fast CC/CV power supply take two
« Reply #32 on: October 04, 2023, 08:50:25 pm »
I always measure across the CS resistor. I have also included the Base of Q1.
I personally have no need for super fast CC limiting.
This is the design of the mainly used PSU on my bench, https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg3582664/#msg3582664
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline Zoli

  • Frequent Contributor
  • **
  • Posts: 595
  • Country: ca
  • Grumpy old men
Re: fast CC/CV power supply take two
« Reply #33 on: October 05, 2023, 12:59:34 am »
I've posted my design:
https://www.eevblog.com/forum/projects/voltage-regulator-in-current-feedback-topology/
Pay a visit for some reading and comment.
TIA,
Zoli
 

Offline luudee

  • Frequent Contributor
  • **
  • Posts: 283
  • Country: th
Re: fast CC/CV power supply take two
« Reply #34 on: October 05, 2023, 02:47:21 am »

I've been following this thread with great interest.

Why not use an approach similar to high speed SERDES. They face the same
problem, trying to turn the output on or off really fast.

To avoid over and under shoots, the use a three stage driver:
1. Pre Driver
2. Main Driver
3. Post Driver

All three are very fast, the Pre and Post are weak drivers.

I wonder if you guys can adopt the same principle to this "fast PSU".

Good Luck,
rudi
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: fast CC/CV power supply take two
« Reply #35 on: October 05, 2023, 03:22:15 am »
That's rather cryptic.  Could you provide a little more information or background on that?  Unqualified, "pre" "main" and "post" could be anything: any ratings, any connection.  It's also not clear if it would be a specifically digital solution, or at all amenable to analog function.  At the least, it doesn't sound like it needs to be reactive -- high speed digital signals transmit into controlled impedance and that's it, if the impedance is lumpy your signal is lumpy, end of story; whereas this circuit is intended to respond in a very specific way to different load conditions.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline luudee

  • Frequent Contributor
  • **
  • Posts: 283
  • Country: th
Re: fast CC/CV power supply take two
« Reply #36 on: October 05, 2023, 04:02:18 am »
That's rather cryptic.  Could you provide a little more information or background on that?  Unqualified, "pre" "main" and "post" could be anything: any ratings, any connection.  It's also not clear if it would be a specifically digital solution, or at all amenable to analog function.  At the least, it doesn't sound like it needs to be reactive -- high speed digital signals transmit into controlled impedance and that's it, if the impedance is lumpy your signal is lumpy, end of story; whereas this circuit is intended to respond in a very specific way to different load conditions.

Tim

Hi Tim,

yeah sorry, I know it is a bit cryptic, lol.

Basically you have 3 drivers, first one and last one are "weak".

To turn on a signal (drive to logical '1'), you first engage the pre-driver, then the
main driver. The idea is that the pre-driver will charge up the line and prevent
overshoot.

To turn off the signal, you disable the main driver and enable the post-driver,
before disabling it as well. The idea here is to avoid undershoots.

It's a little bit of both digital and analog, but then at 56Ghz, nothing is digital nor
analog any longer, lol

Best Regards,
rudi
 

Offline exeTopic starter

  • Supporter
  • ****
  • Posts: 2665
  • Country: nl
  • self-educated hobbyist
Re: fast CC/CV power supply take two
« Reply #37 on: November 26, 2024, 09:45:36 am »
Hello everyone!

Some time ago I abandoned this circuit because I could not figure out how to do accurate current sensing. Last night it landed my mind that I can just do a normal series sensing (attaching pic). I didn't want to do this in the beginning because that would mean a second current source loop and yet another shunt resistor (the first one sources current to shunt regulator). But I think this may actually work.  Here is the idea.

1) The "main" current source set the "hard" limit on total current supply. It's fast, but not accurate. In case of overcurrent it engages first.
2) The second current loop uses own shunt resistor to do precise adjustment of current. It controls the shunt regulator (not shown on the pic).

Having two shunts is suboptimal, but I learned that I can drop like just 0.1-0.2V on each and still have a good speed and precision. If I keep the shunt on low side, I can accurately sense sub-mv drop.

I have to go now, and I'll experiment more with the circuit later. Just wanted to share my idea with you guys).
 

Offline exeTopic starter

  • Supporter
  • ****
  • Posts: 2665
  • Country: nl
  • self-educated hobbyist
Re: fast CC/CV power supply take two
« Reply #38 on: January 25, 2025, 09:19:07 am »
Hello everyone!

I decided to finish this project. It was quite a bumpy ride. A lot ideas I tried and fail. I think one of the most challenging aspects was to define the specs. I think I finally set my mind. So, it's current source with precise voltage clamp. I've already built a TIA amplifier for low-current measurements and, while it's a cool small device, I really miss the adjustable voltage source for it. I kinda need the same functionality, but with voltage/current sources.

So, desired specs:
- Voltage range: 0.1v-12.5V (ideally down to zero)
- Current: 10nA - 50mA+
- Overshoot should be below 0.3V.
- It should limit current within 2us at currents above 50mA.
- The accuracy of set current is not important, I'm happy if it's within 1%.
- Accuracy of current readback is important. I want to get a dynamic range of 100db from a single shunt. I believe this is possible, already achieved on this project: [1].
- Voltage readback: within +-2mV absolute accuracy, but really not that important.

I'm attaching the circuit I want to build. The two "tricks" I used are the following.
- Put a small capacitor across measurement shunt. I can't believe I didn't do that before. It greatly helps with stabilizing voltage regulation, without, I believe, compromising other parameters too much. Even 1nF cap helps a lot
- The voltage is clamped with a bjt in emitter-follower configuration (Q2 on schematic). It has to be set slightly above the set voltage so it doesn't normally conduct, but only dumps current on overshoot event.

The clamping with a bjt is kinda "meh". It has to clamps as close to the output voltage as possible, but also not to affect accuracy of set current. That's not possible to achieve for all ranges of currents. So, in the real circuit I make the clamping voltage configurable, the user can select tradeoff between overshoot voltage and how precise the output current is set (NB this doesn't affect current reading precision).

There are other issues with this circuit. One is stability. I think without clamping with Q2 it starts oscillating for some values of output capacitance. So, compensation might need to be adjusted, I left some provisions for that on the pcb.

Also, implementing "floating" reference voltage V28 adds some circuitry. In the real circuit I implement it with instrumentation amplifier (ad8227) with buffered reference pin. What I don't like is that most inamps cannot work with the voltage headroom I have. AD8227 is pretty much the only device I could find that would fit the circuit.

To be continued...

[1] Measures from <0.1pA to 32nA+ on single 100M shunt: https://www.eevblog.com/forum/projects/yet-another-picoammeter/msg5498422
 

Offline exeTopic starter

  • Supporter
  • ****
  • Posts: 2665
  • Country: nl
  • self-educated hobbyist
Re: fast CC/CV power supply take two
« Reply #39 on: January 26, 2025, 09:12:19 am »
About AD8034. I chose it because I need a fast rail-to-rail part. And there is a big catch about this specific part. While AD advertises it as "1pA jfet-input R-R amplifier", look what happens when you approach the positive rail. The input bias jumps from a few pA to over 30uA, a factor 30_000_000. I wouldn't call this a rail-to-rail amplifier in my book.
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 15240
  • Country: de
Re: fast CC/CV power supply take two
« Reply #40 on: January 26, 2025, 05:03:29 pm »
It is a bit odd conbination of having a JFET and BJT mixex input stage, but at least they clearly show it in the data sheet.
It is quite common that RR amplifiers have a cross over region somewhat some 1 V from the supply limit, where they behave not that well. For demanding used this input cross over region may not be acceptable. So RR input is often to be taken with a grain of salt.

There are a few exceptions though (e.g. OPA388) that have an internal charge pump and hinger supply
 to get away with a single input stage. For low speed may AZ amplifier can also be OK, as the switched part is often working over the whole range.
 
The following users thanked this post: exe

Offline exeTopic starter

  • Supporter
  • ****
  • Posts: 2665
  • Country: nl
  • self-educated hobbyist
Re: fast CC/CV power supply take two
« Reply #41 on: January 27, 2025, 01:06:37 pm »
Folks, I think I found a critical flaw in the design. Kinda.

So, when circuit switch to CC mode, the shunt transistor gets reverse-biased, Vbe voltage. That's because it's an emitter follower output, and when output drops, voltage on the base goes above emitter voltage. So, Q2 and Q7 might get damaged.

That's a fatal flaw. Let me know if you have ideas how to fix this. May be there are special bjts that can handle higher Vbe voltage, or some schematic improvements?

Anyway, I did some quick-n-dirty experiments with four transistors I had at hand. Here are rough break-down voltages of base-emittor junction:

- MPSA92 foschan -- 11.4V
- MPSA92 semtech -- 11.4V
- KSB772YS (I believe onsemi) -- 9.4V
- Tip2955 onsemi -- 12.4V

So... I could use TIP2955 to clamp the source current... but it's huge... Also, only rated for 7V Veb, but I'm fine to push it up to, say 11V in my home lab power supply)

PS found 2N2946A in metal can that can survive 32V. And it's only 15bucks/piece, yeah)
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7532
  • Country: pl
Re: fast CC/CV power supply take two
« Reply #42 on: January 27, 2025, 03:20:04 pm »
You mean that some opamp reverse biases the BE junction of an emitter follower?

I suppose you could add a diode in series with the base. Or clamp base voltage with another emitter follower hanging off the emitter - opamps don't source much current.
 
The following users thanked this post: exe

Offline exeTopic starter

  • Supporter
  • ****
  • Posts: 2665
  • Country: nl
  • self-educated hobbyist
Re: fast CC/CV power supply take two
« Reply #43 on: January 28, 2025, 10:23:07 am »
Thanks for suggestions. I think they still require some research. With diode in series with base I see two potential problems:
- it should be a very low-leakage diode, less than the leakage of b-e junction, or it won't divide voltage properly. It's the same issue as with a string of diodes connected to high voltage.
- opamp won't be able to "push" current into the base. This can potentially reduce performance due to storage charge, I think.
But this could be viable, I'll simulate the circuit.

Clamping voltage with emitter follower... Hm, interesting, I'll try that too. On schematic, U3 provides a low-impedance "copy" output voltage. It can also clamp it.
 

Offline Zero999

  • Super Contributor
  • ***
  • Posts: 20533
  • Country: gb
  • 0999
Re: fast CC/CV power supply take two
« Reply #44 on: January 29, 2025, 10:02:16 am »
I've just found a controller IC which might be suit this application, the MP26085.
https://www.monolithicpower.cn/cn/documentview/productdocument/index/version/2/document_type/Datasheet/lang/en/sku/MP26085/

It's only rated to 20V, but you should be able to add a transistor to the output, forming a cascode with the internal transistors.

« Last Edit: January 29, 2025, 01:25:01 pm by Zero999 »
 
The following users thanked this post: exe

Offline PCB.Wiz

  • Super Contributor
  • ***
  • Posts: 2145
  • Country: au
Re: fast CC/CV power supply take two
« Reply #45 on: January 30, 2025, 12:19:52 am »
That's a fatal flaw. Let me know if you have ideas how to fix this.
May be there are special bjts that can handle higher Vbe voltage, or some schematic improvements?

You should be able to easily clamp the opamp drive with either a diode or an emitter follower, or you could use a PMOS as the clamp device.

There are BJT designed for audio muting that have higher VBEO, but I've only seen those in NPN.
They also typically have high gain in both directions.

eg  2SD2114, 2SC3326, 2SD2351

 
 
The following users thanked this post: exe

Offline exeTopic starter

  • Supporter
  • ****
  • Posts: 2665
  • Country: nl
  • self-educated hobbyist
Re: fast CC/CV power supply take two
« Reply #46 on: January 30, 2025, 09:38:42 pm »
Yeah, I also found muting bjts and I have 2SC3326.

BTW, I measured beta of BC549C (onsemi). So, initial beta was 601 (according to transistor tester). After about two minutes of passing ~10mA of reverse base current iit dropped to 477. Then after a few minutes of 40mA it dropped to 453. Heating with hotair at ~250C (don't really know how accurate is the temperature) for about 20 seconds, it started to smell plastic). Anyway, after heat treatment beta recovered to 547, and Vbe raise by 10mV, to 653mV. I let the transistor to cool down before taking measurements.

I decided to further abuse bjt, and passed 40mA for about an hour. So, beta right now 438. But there is a slight temperature variation in the room, so all those numbers are not very accurate. Like, heating the bjt with my fingers raised beta to 467, for example.

I didn't do leakage measurements due to a busy schedule. But I plan to do it, at least a quick one.

There was a tread on the forum about B-E breakdown, but I wasn't able to find it, so posting my findings here.
 

Offline exeTopic starter

  • Supporter
  • ****
  • Posts: 2665
  • Country: nl
  • self-educated hobbyist
Re: fast CC/CV power supply take two
« Reply #47 on: February 08, 2025, 07:19:51 pm »
I did a few more measurements. I measured leakage current of B-E junction of new bc549c vs the one I already tortured in the previous tests.

It's hard for me to measure down to picamps, so take the results with a grain of salt.

- abused bjt (device #1): ~5-7pA leakage current, measured by this device: https://www.eevblog.com/forum/projects/yet-another-picoammeter/msg5707001/#msg5707001 . There is 4-5pA offset, and device showed 10pA when connected in series with 3.3V cr2032 battery. The other device I used for measurements is nanoranger (https://www.eevblog.com/forum/testgear/nanoranger-8-5-decade-ammeter-micro-review/msg5421981/), it showed 26pA average reading, and it seems it has 19pA offset reading with leads disconnected. So, 26pA-19pA=7pA.

- brand new bc549 (device #2): nothing. I can't reliably read its leakage current at 3.3V reverse bias.

- same new bjt, after 5 mins of reverse 10mA current (Veb 9V): beta dropped from 540 to 440. Reverse current: 6pA.

- Collector-emitter leakage current, at 3.3V bias, base is left unconnected: below what I can reliably measure for both devices.

PS temperature in the lab is 19C.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf