Author Topic: Class D FET driver questions (long...)  (Read 341 times)

0 Members and 1 Guest are viewing this topic.

Offline Chris Wilson

  • Frequent Contributor
  • **
  • Posts: 942
  • Country: gb
  • Race car engineer, dog lover, hoarder.
Class D FET driver questions (long...)
« on: January 04, 2018, 12:26:05 am »
I have run a Class D push pull FET based PA on LF (137kHz) for a long time, with no issues. It's exciter is a cheap kit from QRP Labs called a U3S. The amp is from an American hams' design and has worked fine. In a can't leave well alone moment I decided that as the designer of the U3S exciter had issued a new firmware update for its Atmega 328 processor that allowed toggling the Si5351A synthesizer to output two 180 degree out of phase signals, from CLK0 and CLK1 pins, these could be used to delete an IC (Flip Flop 74F74 that divides by two and inverts signal) in the amp and to drive the FET driver IC directly.

Rather than spend an hour typing it all out I am taking the liberty of copy pasting what advice and questions I have received and posted on the UK LF forum, so you can see what's been tried. Basically I have changed from the IR2110 FET driver that needs 5V and 12V supplies to an MCP1404 driver IC that uses a single 12V supply, so also i lose a 5V voltage regulator too. I have it working (ran all night last night) but to do so I have to meticulously adjust the voltage to this driver IC to around 9V. Much less and the gate and drain waveforms of the PA FET's go to hell in a hand cart, and the same, but more so, with a near 12V supply.

Here's where we have gotten so far! Thanks for reading, it's a bit long winded...

I use my U3S QRP Labs exciter to drive a PA on 137kHz, and it works
fine. Hans, the designer, made changes allowing a 180 degree out of
phase pair of signals to come direct  from the Si5351A synthesizer chip with
the intention of this being able to drive a push pull Class D FET
driver chip directly without a X2 signal being divided and split by a
preceding IC, please see original and simplified circuits below. I
tried this but it appears the circa 2.2V outputs of the Si5351A chip
are not sufficient to drive the inputs of the IR2110 FET driver.
Any ideas please?     Thanks.
Original  PA  circuit (works fine with "normal" U3S driving the 74F74
 direct with CLK0 output):

Modded  amp  circuit (simplified in the hope of building with one less

IR2110FET driver specs:

74F74 IC specs:

U3S exciter schematic:

Seems an odd choice of driver chip. A high/low side driver used in a non-bootstrapped mode to drive two low-side FETs
How about one of the normal FET driver chips - picking out one I've used in the past, the MCP14E 3/4/5 family. Logic level input, spec. Logic '0' max 1.3V, Logic '1' min 2.4V. So bias half way and your 2.2V swing takes it into the valid range.
They three types are inverting, non inverting, and one of each in a package

OK,   this   seems   to   continue  the  simplify  quest  rather  than
transformers,  plus I can delete the 5V regulator. I think I have some
MCP1404  driver  chips,  if  I have I'll knock something together and report
back, thanks Andy.


Hello Andy,

MCP1404 IC fitted, as soon as power to the PA FETS is applied (even
6V) one of the gate square waves either disappears or becomes
extremely random. Swapping inputs to the driver IC makes no
difference, it's one pair of PA FET's dependent. FET's work fine when
returned to "as designed" status! Driving the driver IC through the
1nF caps and 15k resistors to ground. Outputs from CLK0 and CLK1 are
180 out of phase and seem correct. Any ideas please? Thank you!

So am I ok using another pair of 15K to pins 2 and 4 up to pins 6 (12V
DC  +)?  Thanks,  this  is  all  rather academic, given the latest U3S
firmware  allows  direct output of two signals from the Si synthesizer
180 degrees out of phase I just fancied seeing if I could reduce the
component   count of an amp.


When I said bias half way, meant half way between the two logic threshold limits. Ie mid way between 1.3V (upper limit of guaranteed '0' and 2.4V (lower limit of guaranteed '1') so bias at 1.8V
Not half the supply volts.
Study the data sheets and calculate. Don't just play with components willy-nilly


Hello Andy, OK,  understood,  I did some calculations then checked them against an
on-line  calculator and get figures of 68K and 12K for the divider. If I
retained  the existing 15K to ground I might be able to use 82k as the
other  half?  Or  are  these  values too high for needed current flow?
Thanks. Chris

An "Online calculator" Whatever is that. Do you mean the basic equations for a potential divider?
Vout = Vin . (Rb / (Rt + Rb)
There is no leakage current into a CMOS input so those will do as well as any other pairings. Check the chip data sheet, it gives you all the values you need


Hello Markus / Andy / all,

Please see

I  added  2  off 86k resistors as resistive dividers as shown. Voltages
were as expected 1.87 and 1.86V on pins 2 and 4 but both driver output
pins  sat high at circa 12V . Removing the added resistors but running
the  MCP1404  on  9V  all  works fine with gate and drain waveforms as
shown here:

But on much less than 9v, say here at 6.8V the waveforms deteriorate and output is noisy:

and with the supply at near 12v things totally fall apart with the traces rapidly degenerating into hash around 11.7V :

Not sure why this is so....Have I misunderstood the bias at 1.8V Andy
suggested??  By e-mail from Andy:
"When I said bias half way, meant half way between the two logic
threshold limits. Ie mid way between 1.3V (upper limit of guaranteed
'0' and 2.4V (lower limit of guaranteed '1') so bias at 1.8V
Not half the supply volts.
Study the data sheets and calculate. Don't just play with components

Thanks. Any /all pointers to what is happening welcome!
I am wondering if the input capacitance of the two pairs of FQA34N20L FET's (worst case 7800pF, best case of 6000pF for a pair) is pushing this driver which is rated to deliver into 5600pF in 34nS

If I used a pair of TC4452 driver IC's (each 8 pin device has only one input and output) these are capable of supplying 22,000pF in 42nS so have a lot more oomph. Am I looking in the right direction or way off track please??
Best regards,

                 Chris Wilson.

Offline b_force

  • Super Contributor
  • ***
  • Posts: 1122
  • Country: 00
    • One World Concepts
Re: Class D FET driver questions (long...)
« Reply #1 on: January 04, 2018, 01:19:46 am »
What is the question exactly?

Also, all these components are a bit old, nowadays there are a lot newer, cheaper and easier to use components.
137kHz is also very low.
"If you can't explain it simply (or at all), you don't understand it well enough." A. Einstein

Offline diyaudio

  • Frequent Contributor
  • **
  • Posts: 637
  • Country: za
Re: Class D FET driver questions (long...)
« Reply #2 on: January 04, 2018, 01:27:04 am »
o boy, what a nebulous long read.  :)

Firstly what is your problem ?  Short term answer get newer chips from Ti and use a reference design, if you want to get abstract and use IR gate drivers and self oscillating loops well then buckle up you have lots to learn. You may want to look at diyaudio's class-d section.

Good Luck.

Offline capt bullshot

  • Super Contributor
  • ***
  • Posts: 1131
  • Country: de
    • Mostly useless stuff, but nice to have:
Re: Class D FET driver questions (long...)
« Reply #3 on: January 04, 2018, 01:55:18 am »
In general, it's a bad idea to use AC coupling capacitors with the inputs of the driver chip.
Next, the IR2110 is weaker than the MCP1404, so drives the MOSFETs somewhat slower.
One more, the IR2110 has asymmetric propagation delays (turning OFF faster than ON), this provides some dead time in the original circuit.
Last: you can use your 12V supply for the IR2110 logic side, this is clearly stated in the datasheet, so no need for the 7805.

If your exciter doesn't provide enough level to drive the IR2110 input, you'll have to use some kind of level translator chip, like a 74LCX08, not AC coupling. AC coupling is "forbidden" with this kind of drivers. Treat them as "logic" ICs (like the 74xx series or whatever), not analog stuff that may work fine with AC coupling. These drivers must be driven by proper "logic" levels, like the output of a '04 inverter (or the '74 Flipflop) operating at 3.3V or 5V.

Provide a solid ground plane for the MOSFETs sources and place the driver chip on this same solid ground plane, do not use wiring to connect the driver outputs to the gate and source. Keep the ground connection from MOSFET sources to the driver chip as short and solid as possible. You'll also need a proper ground plane from the logic chip driving the MOSFET driver to the MOSFET drivers. These drivers are quite fast and can pick up all kind of ground noise at their inputs, resulting in oscillation. The IR2110 is way much less sensitive here than the other one, so for a non-optimum layout, the IR2110 would be the better choice anyway. Stay away from more modern parts, as they often are faster and more sensitive to noise and ground shifting. They need real good layouts, decoupling and so on, and won't work reliably on a poor prototype built. Don't drive MOSFETs faster than required for your purpose.

Driving the MOSFETs too fast (by using a beefier driver chip) can cause all kinds of unwanted feedback and oscillation. Real good ground plane design helps you out of here, or just use larger gate resistors to drive the MOSFETs somewhat slower. Your description of 12V supply voltage causing havoc points into this direction.

Having not enough dead time can cause cross conduction with the MOSFETs, resulting in large unwanted currents circulating in your circuit through the ouput transformer. These also can cause all kinds of distorted waveforms until full destruction of the power stage. If the driver chip (the MCP1404) doesn't provide dead time on it's own, you'll have to provide some circuitry. Two 180° shifted waveforms aren't enough here, there must be some delay from the falling edge of one phase to the rising edge of the other phase, in both directions. The delay is required in the ballpark of 10ns ... 100ns.

« Last Edit: January 04, 2018, 02:06:42 am by capt bullshot »
Safety devices hinder evolution

Offline Chris Wilson

  • Frequent Contributor
  • **
  • Posts: 942
  • Country: gb
  • Race car engineer, dog lover, hoarder.
Re: Class D FET driver questions (long...)
« Reply #4 on: January 11, 2018, 10:21:37 am »
Some very very useful replies, thanks very much indeed. Captain Bullshot especially has given me much to consider, I have much to learn about what initially seemed simple amplifiers! Is *anything* simple in RF electronics? :)
Best regards,

                 Chris Wilson.
The following users thanked this post: capt bullshot

Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo