Group Actions
Engagement
  • Author Author: Jan Cumps
  • Date Created: Date Created
  • Views 6005 views
  • Likes 7 likes
  • Comments 29 comments
Related
Recommended

Hercules LaunchPad and GaN FETs - Side Note B: Look at the PCB

Jan Cumps

I'm designing a BoosterPack to evaluate GaN devices with the help of a microcontroller.

I've received PCBs from Seeed. Let's have a look ...

image

 

This is my first Seeed order (a kind element14 community member gave me a rebate coupon that I happily used).

Board specifications:

 

  • PCB Dimension - 10cm Max*10cm Max
  • Layer - 4
  • PCB Thickness - 1.6mm
  • PCB Qty. - 5
  • PCB Color - Green
  • Surface Finish - ENIG
  • Copper Weight - 1oz.
  • Panelized PCBs - 1
  • Expedited Option - NO

 

The price without shipment for this board is 65.12€. Shipping was 7.37€. The coupon got me 9.15€ off ($ 10). Totals to 63.97€

image

This design requires 2oZ copper layers. My budget can't carry the cost of such a board - so I will not be able to draw full current out of my design.

This is a compromise, one of the many I had to make. I tried to avoid compromises in the design, but did make them when ordering.

Getting a board that has enough copper on it will not require a design change. It's just ticking a different option when ordering and coughing up the monies.

I'd rather do that after I validated that I didn't make mistakes in the PCB layout.

image

image

Just like previous designs I did with OSHPark, the Seeed boards look very good at first inspection.

This design is 4 layers, so I have no mechanism to look at the two inner layers. The top and bottom layer look very nice, so that's a good sign.

 

In the images below, the artist renditions on the left side are generated by the Seeed portal.

The photos on the right side are the actual thing. Click on them to get a detailed view.

 

imageimage
imageimage

 

 

Populate Order

This is the first time I'll populate a two-sided board. I'll use hot air for the smd components.This is how I plan to approach it:

  1. solder the GaN IC and the inductor. These two have the pins under the device. I'll have to massively bombard them with heat to transfer the energy underneath where the solder paste is.
    I'll use the pre-heater that shabaz gave  me to get the whole board to a close-to-reflow temperature.
  2. then all the smd components on the bottom side. If they fll of later, that's not a big deal, because the are only few.
    I'll also use the pre-heater but on lower temperature, because its heat will be radiating on the inductor and GaN chip.
  3. Then I'll do the remaining top smd components.
    The pre-heater again on a temperature low enough to not desolder the bottom components.
  4. the trough-hole components by hand

 

It is my first time. I'd be happy to take your advice on the preferred order to solder the components.

The tools I have at hand:

  • hot air
  • soldering iron
  • pre-heater

 

 

edit: I've attached the waveform of one High input pulse as hi01.wfm.

If you have a Rigol DS1052E you can upload that to your scope and examine the ringing with cursors.

image

 

 

Related Blogs
Hercules LaunchPad and GaN FETs - Part 1: Control Big Power with a Flimsy Mouse Scroll Wheel
Hercules LaunchPad and GaN FETs - Part 2: Make a BoosterPack
Hercules LaunchPad and GaN FETs - Part 3a: BoosterPack Layout - Reference Design
Hercules LaunchPad and GaN FETs - Part 3b: BoosterPack Layout - my version
Hercules LaunchPad and GaN FETs - Side Note A: BoosterPack Layout - Custom KiCad Parts
Hercules LaunchPad and GaN FETs - Side Note B: Look at the PCB
Rotary Encoders - Part 1: Electronics
Checking Out GaN Half-Bridge Power Stage: Texas Instruments LMG5200 - Part 1: Preview
Rotary Encoders - Part 4: Capturing Input on a Texas Instruments Hercules LaunchPad with eQEP
Vintage Turntable repair: Can I fix a Perpetuum Ebner from 1958 - part 4 - Hercules LaunchPad Enhanced PWM try-out
Attachments:
hi01.zip
Parents

  • The board is built up, not tested yet.

    Reflow worked ok, except for 0402 and some 0603s.

    C22(0402) toombstoned, I had to rework C15, C21 and LED D2 too.

    I'm happy that the (somewhat) complex footprint of quadrature encoder - with one oval and two round mounting holes fitted perfectly.

     

    The two screw terminals however, P2 and P3, are a complete goof-up. The holes are too small, the distance between them is wrong and the total size of the footprints doesn't match the component.

    I have to check how that happened, because I made this footprint myself from the data sheet. (wrong part ordered or delivered, wrong drawing selected in the sheet, inch/cm switch, I don't know yet)

     

    image

     

    image

  • in reply to Jan Cumps

    Two steps forward one step back at the moment.

    When I apply bias voltage to the GaN device, power supply goes trough its knees. Even if this is supposed to be a 3mA maximum. The led indicator in parallel is taking more current normally than the whole bias part.

    The LMG5200 is getting real hot too.

    I first used a DMM to check if there are shorts (I did that just after mounting the GaN and inductor too). No.

     

    So I have now desoldered the GaN device and will check all of the signals going there. I may have damaged the device.

    Hope not. Because I only have once IC left and they aren't cheap.

     

    image

     

    On the other side, this is a chance to go and probe the circuit without fear of damaging the GaN device...

  • in reply to jc2048

    Jon Clift wrote:

     

    ..."Focus on ringing. It's acceptable, I think."

     

    With respect, I might disagree. Unless you have a scope with a high bandwidth and an active probe with very low capacitance, ringing is always worse than you see on the screen.

     

    ....

    In comparison with my usual captures I mean image.

    This was my capture of the original board:

    image

     

    I was happy that DAB wasn't close by when I posted that. He'd have hit me with several big stones I think image

     

    Jon Clift wrote:

     

     

    Question is, do the peaks get anywhere near the input switching voltage? Are they CMOS levels on the inputs or TTL? Hopefully they're CMOS. (I can't look at the datasheet because I can't see one on the product page - maybe you have to register or something.)

     

     

    The datasheet is available as PDF only. No web version for this product:

    http://www.ti.com/lit/gpn/lmg5200

    We're looking at the signals that will go into LI and HI input.

    image

    image

     

    Next time e14 runs a high frequency oscilloscope road test, I *must* be selected.

     

    edit: the LMG5200 isn't currenly populated on the board. The signals are open-ended during the measurements today.

  • in reply to jc2048

    This is a simplified schema of the input circuit. The PWM signals enter left.

    The two 10puff caps are on the PCB, as close as possible to the GaN input pins.

    The 200K represent the nominal value of the internal pull-down resistor on the GaN die.

    image

     

    I measured before the 10puffs, without the GaN IC mounted. I'll redo the exercise once I have the LMG5200 back in place.

  • in reply to Jan Cumps

    I don't understand what I was doing with the datasheet. I've been back to the product page and there is a link to the datasheet. How strange.

     

    I see what you mean with the previous capture. That's the ground lead (and possibly, in some degree, the probe itself which isn't really intended for looking at these kind of edge rates). So that really leaves us up in the air. My confident assertion that the ringing is always worse than you see isn't necessarily true (you are testing the test equipment at the same time as the board).

     

    The input levels are TTL, but with hysteresis. I would be concerned with the ringing you show, given that the oscilloscope may be under-reporting. The peak of one of the ringing cycles has to get to 2V. If it does, do you get shoot-through of the FETs or is there logic on the chip that prevents that?

     

    The 200k isn't a terminator, it's just a light pull-down to ensure the input remains in the inactive state if nothing is connected to it. The stripline across the board will have an impedance of something like 40-100 ohms depending on where the ground underneath is and the width of the track. For end termination the resistance needs to match, but that assumes that the driving end can actually drive it which might not be the case if it's a weedy processor pin. If the processor output has FETs with a channel resistance of 20 or 25 ohms, you'd be better off with the series termination at the sending end - basically you add resistance to bring the channel resistance up to the line impedance, so if the channel resistance is 20 ohms and the stripline is 50, you'd have an external 33 ohm to match.

     

    What is the width of the track and the height above the ground plane?

  • in reply to Jan Cumps

    If you're reworking the board layout, you might want to give more clearance between these tracks, particularly if it might be made with 2oz copper rather than 1oz.

     

    Minimum track and gap is often 0.25mm (10 thou) for 2oz copper.

     

    image

  • in reply to jc2048

    I'll check your remarks, Jon.

    For the time being, a scope capture of the high side input, measured after the cap, with the IC in place (just soldered a new one in place).

     

    image

     

    I'll redo this capture with the trigger point a little more to the left to capture the additional ringing after falling edge, then use cursors to measure the peaks of the rings...

    The ringing stops right at the edge of the screen. The last rising edge you see is also the last significant ring.

     

    I've uploaded the waveform in RIGOL DS1052E wfm format. If you have access to that type of oscilloscope, it's possible to load the file via USB and perform measurements on the scope.

Comment
  • in reply to jc2048

    I'll check your remarks, Jon.

    For the time being, a scope capture of the high side input, measured after the cap, with the IC in place (just soldered a new one in place).

     

    image

     

    I'll redo this capture with the trigger point a little more to the left to capture the additional ringing after falling edge, then use cursors to measure the peaks of the rings...

    The ringing stops right at the edge of the screen. The last rising edge you see is also the last significant ring.

     

    I've uploaded the waveform in RIGOL DS1052E wfm format. If you have access to that type of oscilloscope, it's possible to load the file via USB and perform measurements on the scope.

Children
No Data