Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Current sensor upstream of Bulk of output capacitance in DCDC converter?

Status
Not open for further replies.
T

treez

Guest
Hi,
We have a DCDC module generally giving a 48V, 6.5A output. We wish to clamp the output current to approximately 7A so have a current sensor in the output, which feeds an external error amplifier. (as in the attached schematics)
Our contractor wants to put the current sensor upstream of the bulk of the output capacitance, as in schematic A….but we believe that we should have it the other way round, as in schematic B (ie downstream of the bulk of the output capacitance) …do you agree with us?
 

Attachments

  • Current sense position schematics.pdf
    159.6 KB · Views: 73

If you mean "upstream" as "after the inductor but before
the filter" as opposed to "after the filter but before the
load" I don't see anything but a wire-length's worth of
difference.
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
In control terms, design A can easily achieve stability, stability of design B is uncertain and depends on the load impedance. Design B can under circumstances react faster on load changes.
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Upstream means the sensor sees all the high freq current info into the capacitor - this will always be more stable ...
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
In control terms, design A can easily achieve stability, stability of design B is uncertain and depends on the load impedance. Design B can under circumstances react faster on load changes.
In control terms, design A can easily achieve stability, stability of design B is uncertain and depends on the load impedance. Design B can under circumstances react faster on load changes.

Thanks, I am relieved to hear you both saying this, since “A” is how I did the first prototype PCB.

However, I then modified “A” to be like “B” and the current as seen in the sensor has less disturbance in it. Do you know why?
I wonder if I should also have the sensor upstream of the 220uF electrolytic (C70 in “A”)?

If you mean "upstream" as "after the inductor but before
the filter" as opposed to "after the filter but before the
load" I don't see anything but a wire-length's worth of
difference.
Thanks, The Hall sensor has to be downstream of the 150nH inductor ("L6" or "L2" respectively) otherwise the current seen in the sensor has a 80kHz oscillation in it.

...................*......................................*......................................*......................................*.................
Upstream means the sensor sees all the high freq current info into the capacitor - this will always be more stable ...

Thanks....I am just wondering if it would be even better to have just a 10uF film capacitor at the DCDC module output, then the 150nH inductor, and then the Hall Sensor, and then all the electrolytic capacitance (approx 2mF's worth)?
I initially declined to do it like this because the DCDC module needs 220uF at its output for stability......and so i put this 220uF right at the module's output.

...................*......................................*......................................*......................................*......................
Taking circuit A, there is an LC output filter there with a resonant frequency of 10273Hz (150nH and 1.6mF). The crossover frequency of the DCDC module is said to be 20kHz…….so isn’t this a recipe for instability?.....having said that….i would expect the added external output capacitance to reduce the crossover frequency of the overall power supply……so this stability fear could be wide of the mark…….with Vicor modules, there is no way of measuring the crossover of the overall thing, so we’re stuffed?

...................*......................................*......................................*......................................*......................
I have implemented the external current clamp, because there will be forty of these DCDC modules all operating together….kind of all in parallel with each other….the modules have a droop sharing algorithm (vout reduces by 5% from no load to full load)….however, due to the sheer size of our system, the volt drops from each module to the load will be widely different, and so droop sharing along won’t be enough…and we believe we need the current clamp as shown…(to stop any individual module over-hogging current)
 
Last edited by a moderator:

Obviously, stability is a necessary condition for the design. All other parameters, like more or less "disturbance" in the sensor signal, are meaningless without specifying the exact purpose and intended dynamic behavior of the control circuit. Not sure if you have specified it at all.

You may want to "clamp" the instantaneous output current, but speed of reactions is limited by the bulk capacitors.

I read in the DCM3623 datasheet, that TR input bandwidth is only 30 Hz. At the end, the functional difference between both design variants may be smaller than expected.
 

These 4 scope shots (two from circuit A of top post, and two from circuit B of top post), show how circuit B appears more stable when on constant load.
There is a scope shot of current_sensor_output and SMPS_output_on_AC_coupled for A and B cases.
I made the external current clamp fix the current to 2.53A in both cases, A and B.
The load is 15R.
Vin is 31V
As you can see, in the “A” case, there is a disturbance in the VIOUT signal of the ACS722 current sensor IC that is not there with the B case.
This is not expected. Do you know why “A” appears better than “B”?

- - - Updated - - -

Also, if i take Circuit A , and totally remove the bulk capacitor bank downstream of the current sensor, then the current sensor output "VIOUT" is also in that case, free of disturbance.
I first thought it was because of the LC resonance thing...but even with the L shorted out, the disturbance on VIOUT still happens.

- - - Updated - - -

Ive just disconnected R31 (and R35) and let the module run on its own internal feedback loop (current clamp error amplifier not in the loop any more)……it turns out that once again, having the bulk caps downstream of the sensor makes the VIOUT of the sensor noisy…..having them upstream of the sensor makes the VIOUT of the sensor noise free……so this looks like being a problem of the Hall sensor being upset by the noise from the DCDC module.

- - - Updated - - -

I am realising that this noise is noise coming out of the DCDC module…and going into the cap bank, wherever it may be…the thing is, why is the DCDC module behaving like this? The DCDC module has its own significant internal low ESR capacitor banks, so I am not sure why it is so noisy in operation?....perhaps I need to add more artificial ESR to the output capacitors?....i have no way of calculating the feedback loop Bode plots, as the module datasheet doesn’t give enough info for us to do this.
 

Attachments

  • Ripple at Vout of SMPS on AC coupling_Caps downstream_A.jpg
    Ripple at Vout of SMPS on AC coupling_Caps downstream_A.jpg
    79.5 KB · Views: 71
  • VIOUT pin of Hall sensor_caps downstream_A.jpg
    VIOUT pin of Hall sensor_caps downstream_A.jpg
    84.2 KB · Views: 60
  • Ripple at Vout of power supply on AC coupled_caps upstream of sensor _B.jpg
    Ripple at Vout of power supply on AC coupled_caps upstream of sensor _B.jpg
    80.4 KB · Views: 70
  • VIOUT of current sensor_caps upstream of sensor _B.jpg
    VIOUT of current sensor_caps upstream of sensor _B.jpg
    75.8 KB · Views: 59

yes there is HF ripple of some sort flowing b/n the internal and external caps, upsetting your ( low cost (?)) I sensor - the only way to mitigate this is a choke on the o/p of the module then the I sensor, then the external caps ...
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Thanks, the I sensor is an ACS722 Hall sensor by Allegro.
Will the choke need to be right at the module output?.....what about having a cap (220uF electrolytic) at the output, then the choke, then the I sensor?
 
Yep - sure you can try that ...
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
In one series of tests, we found that the “jinking” module output current problem, appeared to happen after the module had been on for a minute or two and was getting warmer. We also noted that the datasheet says that the module vout actually gets varied as the temperature changes……we then wondered if this “jinking” about is the module kind of digitally trying to vary its vout with temperature, and this effort kind of suffering some kind of “digital resolution error”, resulting in the “jinking” about of the output current?
I say this because the output current can always be regulated to the correct average value….even if it jinks about as in the scope shots shown a few posts back.
There is keen-ness to “get on” and not spend too much time testing and testing. The vout is fine (as long as you don’t look at it on AC coupling at too low V/div)…..Also, the average output current is fine. What would be the opinion on just letting this problem lie and cracking on regardless?
If I try and solve this noise with an inductor right at the module output, I am going to have to do significant surgery on the prototype PCBs to get the inductor in there, and this is going to take time which we are running out of fast.
Can we just live with the “jinking” output current?....the Vicor DCM3623T50M53C2T00 is a totally enclosed module and we have no way of scoping its internals to actually see what is happening. If we can heatsink it so it stays cool enough, then can we just live with the “jinking” output current?
 

Hi,
I was on a role just then, I was using the “A” circuit, and had increased the load from 15R to 11.25R…..was increasing the inductor L6 and it seemed the noise on the Hall sensors VIOUT pin was reducing….then I noticed the noise vanishing completely for long periods (50 secs or so), then little bursts of noise sporadically re-appearing….

…but then I totally shorted out L6, and the initial noise seemed to be slightly worse…but then again, after the module had been on for 2 minutes, the noise disappeared for some 50 seconds…then sporadically returned, with just very minor noise bursts, say 300us every 5 seconds or so.

This isn’t making much sense.

I am actually wondering if the dcdc module is responding to ambient noise in the surrounding lab/office....from electronics devices in the lab that are inflicting noise into the DCM module...after all, the DCM module is tiny, and possibly has no internal shielding of its control circuitry.

I think ill surgically modify the board, and put the inductor right at the module output, and see what happens then....because this "noise problem" is now playing "hide and seek" with me.
 

Hi,
The noisey output current of the DCDC module happens far more just after switch ON, whilst the module is still quite cool…..as its heatsink gets up to 50degC, I notice the output current gets far far less noisey, with only very occasional “blips”.
This seems to suggest component/parameter value change as the module heats up?
I also cut through the output positive power plane, and connected the output inductor right after the module in Cct A….however, this makes no difference to having it just downstream of the first electrolytic (220uF).
 

I also cut through the input power plane of the module, and added a 150nH inductor there (downstream of C68 in the “A” circuit of the top post….this was to no avail…though again, the noisiness of the output current does greatly reduce as the module’s heatsink got above 60degC.

There seems to be only two more things I can try, in order to reduce the noiseyness of the output current…
1…Use a common mode choke at the input.
2…Re-do the board with multiple layers and shield the signal tracks as they get near the dcdc module…ie, shield the signal tracks between “quit” node copper planes ( eg ground and power planes)

Which of these two would you recommend trying first?

Regarding (1) above, we won’t be able to add Y capacitors, since we have no “chassis ground” or “earth ground” to connect the Y capacitors to. Our DCDC module is also not isolated, as we link primary and secondary grounds with a zero ohm link.

The module is a DCM3623T50M53C2T00 by vicorpower.
https://www.vicorpower.com/documents/datasheets/DCM3623x50M53C2yzz_ds.pdf
 

Some of your caps might be a bit iffy - this used to be a prime source of noise once - which was hard to find ...
 
  • Like
Reactions: treez

    T

    Points: 2
    Helpful Answer Positive Rating
Thanks, it does seem odd though, that the noisiness of the output current vastly reduces (but doesnt completely go) when the module heatsink gets to 60degC?
 

Why is the noisy output current of our DCM module not constant?…it can actually disappear for a few seconds…then creeps back.

We note that the output current of the Vicor DCM3623T50M53C2T00 module is noisy. Sometimes, , but not always, we note that this noise gets much less when the DCM module gets warmer (after being ON for a while). Sometimes this noise disappears for a few seconds, but then returns. Usually, just after startup, the noise is more prevalent.

Current on the scope shots is from VIOUT of ACS722 Hall sensor and is {(Volts – 0.33) / 0.264} Amps
 

Attachments

  • Vout_yellow and Iout_bit noisy.jpg
    Vout_yellow and Iout_bit noisy.jpg
    44.8 KB · Views: 50
  • Vout_yellow and Iout_noise free.jpg
    Vout_yellow and Iout_noise free.jpg
    43.5 KB · Views: 58
  • Vout_yellow and Iout_noisy.jpg
    Vout_yellow and Iout_noisy.jpg
    45.4 KB · Views: 54

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top