Differential phase splitter with level shift capability

[Reactance]

Originally Posted by Workhorse

Go for LM361 which has same specs 20nS delay and complementary outputs but also works on +/-15V like an opamp frontend and +5V for logic interface

Thanks you too workhorse, its sightly more expensive but worthwhile supporting the +/- 15V voltage selection. looking at the datasheet and the nice thing is my local supplier has stock of these

[Reactance]

Some progress:

Moving along swiftly... and reading other material on the forum going back two years ago (and lots of patience) ... vie made it until this point with no smoke or component death...

The following is the from the MGD IR2110 (Notice the overshot on the yellow trace its a bad probe , i swopped it around and it didn't show up need get a better probe)

No dead-time has been employed (for now) .. vie grounded PIN VS to (-V 21) to give symmetrical output from the driver just to debug the switching nodes and observe switching behavior.

IR2011 Outputs are switching cleanly between 800mv to 13.8V with no noise artifacts ( taking into consideration its an old bread board )

Using a for 15V regulator i wonder if should step it up to something like an 18V regulator.. will it give better switching results as the gate is charged more effectively.

[ChocoHolic]

Originally Posted by Reactance

The following is the from the MGD IR2110 (Notice the overshot on the yellow trace its a bad probe , i swopped it around and it didn't show up need get a better probe)

My interpretation would be different.
The signal without ringing is looking by far to slow to be the correct output of an unloaded IR2110. With the shown time scale the sloping times of the IR should be a perfect real vertical step, invisible rise time. In your screen shot we can guess 100-200ns, but the IR typically slopes unloaded within 20ns. The round edges of the trace make me guess that the concerned probe is pretty slow, or not adjusted correctly.
The trace with ringing could be correct. It is rising fast, fitting to IR. It shows ringing, fitting to breadboards.
Another option could be that both wave forms are not fully correct, if both probes are not perfectly adjusted. Note, fast probes sometimes have two adjustments.

Originally Posted by Reactance

IR2011 Outputs are switching cleanly between 800mv to 13.8V with no noise artifacts ( taking into consideration its an old bread board )

800mV is also strange. It should be close to zero.
Offset? Do you get really zero, when shorting the probe?

Originally Posted by Reactance

Using a for 15V regulator i wonder if should step it up to something like an 18V regulator.. will it give better switching results as the gate is charged more effectively.

No, the higher voltage will not bring much advantage. Note when you put a higher driving voltage you may reach the ON-plateauvoltage slightly faster, but in turn it will need longer to turn OFF. You can slightly influence the effective dead time by this, but won't getter a better quality of the gate drive signals. The gate drive signal quality will depend more on the PCB layout and MosFet properties.
Last but not least, unnecessary high drive voltages will result in unnecessary gate drive losses.


P.S.
Basically your signals do look reasonable.

[Reactance]

Originally Posted by ChocoHolic

My interpretation would be different.
The signal without ringing is looking by far to slow to be the correct output of an unloaded IR2110. With the shown time scale the sloping times of the IR should be a perfect real vertical step, invisible rise time. In your screen shot we can guess 100-200ns, but the IR typically slopes unloaded within 20ns. The round edges of the trace make me guess that the concerned probe is pretty slow, or not adjusted correctly.
The trace with ringing could be correct. It is rising fast, fitting to IR. It shows ringing, fitting to breadboards.
Another option could be that both wave forms are not fully correct, if both probes are not perfectly adjusted. Note, fast probes sometimes have two adjustments.




800mV is also strange. It should be close to zero.
Offset? Do you get really zero, when shorting the probe?



No, the higher voltage will not bring much advantage. Note when you put a higher driving voltage you may reach the ON-plateauvoltage slightly faster, but in turn it will need longer to turn OFF. You can slightly influence the effective dead time by this, but won't getter a better quality of the gate drive signals. The gate drive signal quality will depend more on the PCB layout and MosFet properties.
Last but not least, unnecessary high drive voltages will result in unnecessary gate drive losses.


P.S.
Basically your signals do look reasonable.

Probe number 2 (yellow) is definitely faulty it cannot seem to read fast signals, the internal from the scope calibration (1Khz @ 5V) the yellow probe reads fine however, after probing faster switching nodes from the modulator (250Khz) it shows ringing that's not present i have proved it by using the probe 1 (red) and being able to measure the faster 250khz signal node.. im only using dual trance mode to view symmetrical output from the driver and at the moment as im tying dead-time as promised previously i really need to get a reliable probe else i will be chasing ghosts as Eva put it..

I'm sure the doesn't exists from the breadboard the (red) probe reads a clean switching events.. see below



The signals from the IR2110 are starting to look great (and professional ) at the moment vie re-calibrated my scope and probes, the results show a clean switch between
LOW0.0mV HIGH 13.40V


As for the gate drive voltage increase idea, i will stay with the existing 13+ volts feeding the mosfets, I'm just about introduce the dead-time, its set to 130ns (I guess this is to high) but, for now testing its okay i will reduce dead-time to 50ns or so later as well as switching frequency at a later stage.I don't feel the need to switch the modulator this fast ie: (255Khz) reducing this as well to 150Khz maybe later..


I'm starting to get nervous, as its almost time to introduce the mosfets vie got a pair of IRF640 laying on my desk (yes for now i will use them and upgrade to something better at a later stage)

[Reactance]

Just looked at some entries on the forum regarding dead-time was really dead simple to follow using RCD approach after the XOR gates set to 140ns

This is the results after introducing dead-time (Note: ignore the yellow probe overshot i will replace this probe) (I know its awfully high)

[ChocoHolic]

Yup, here the red one is looking blameless and sloping speed would also fit to the IR driver.
Definitely nice!

[Reactance]

Originally Posted by ChocoHolic

Yup, here the red one is looking blameless and sloping speed would also fit to the IR driver.
Definitely nice!

I feel sleepy excitement tells me to wire up the mosfets now but common sense tells me i cant concentrate now, i feel vie achieved enough today will look at the mosfets and LP filter over the next few days i know the real fun and headaches only starts when the filter and signal drive is applied to the circuit...

Thanks for the assistance thus far Chocolate !

[Eva]

Originally Posted by elevator

It's a question of experience, familiarity and mindset. When asked for a solution to a problem, one tends to use the tools one is most familiar with, regardless of objective fit.
Case in point, your own post! : You prefix your ideas with
PIC 1F628A Protection
That's putting the cart in front of the horse; instead of just stating "firmware", or "uC" you impose a specific architecture and a specific model right from the beginning. Presumably because you happen to be familiar with the PIC1F628A. By analogy, analog DYIers (and the majority ef engineers) instinctively map a problem onto the set of tools and means they have experience with. While most of the time this turns out perfectly ok, sometimes the solution is convoluted, complicated and limited eg in expandability.
Cheers,
E

Choose a PIC with ADC and you will be able to avoid some analog protection circuitry and make the protections much more flexible. It will be a good chance to start learning low level DSP programming (doing simple IIR and FIR filters in assembler or C, ie: DC protection requires a LPF, why using a big cap when you can do it in software? FilterShop is a very valuable tool). For example you can use 16F684 or 16F690. They are similar to 16F628, which is similar to 16F84. If you want something more advanced, go for the new 16F1823/4/5/9, with 32Mhz internal clocks and extended instruction set, which are likely to replace all the other ones I mentioned in the future.

btw: I'm working in hybrid digital/analog PFC now, with analog current loop and a novel digital voltage loop. Next step will be fully-digital single-micro PFC, specially designed for single stage audio SMPS (excellent transient response without tons of bulk capacitance or very poor P.F.). In a world of finite resources, part counts and the cost of each part must be kept low, while getting as much as possible out of each part and the the topology. This requires developing complex control algorithms. The ideal electronics designer of 21st century must be a skilled programmer too. Too much specialization makes you weak and easy to exploit by unfair employers (people willing to eran 10 times more money than you with 10 times less skills, no mercy on them! I'm with 15M people)

[Reactance]

Originally Posted by Eva

Choose a PIC with ADC and you will be able to avoid some analog protection circuitry and make the protections much more flexible. It will be a good chance to start learning low level DSP programming (doing simple IIR and FIR filters in assembler or C, ie: DC protection requires a LPF, why using a big cap when you can do it in software? FilterShop is a very valuable tool). For example you can use 16F684 or 16F690. They are similar to 16F628, which is similar to 16F84. If you want something more advanced, go for the new 16F1823/4/5/9, with 32Mhz internal clocks and extended instruction set, which are likely to replace all the other ones I mentioned in the future.

btw: I'm working in hybrid digital/analog PFC now, with analog current loop and a novel digital voltage loop. Next step will be fully-digital single-micro PFC, specially designed for single stage audio SMPS (excellent transient response without tons of bulk capacitance or very poor P.F.). In a world of finite resources, part counts and the cost of each part must be kept low, while getting as much as possible out of each part and the the topology. This requires developing complex control algorithms. The ideal electronics designer of 21st century must be a skilled programmer too. Too much specialization makes you weak and easy to exploit by unfair employers (people willing to eran 10 times more money than you with 10 times less skills, no mercy on them! I'm with 15M people)

I was thinking along same lines regarding PIC based controlled protection schemes, its really fun and interesting employing software protection schemes (but hard asat the same time its rather a steep learning curve especially in the DSP arena where few dare go )

The idea of the IIR and FIR filter with ADC is great however, can one do low pass filter DSP using an 8-bit uC using the ones you mentioned above ??

Ive looked at the filter coefficients and ALL of them use double precision, 8-Bit MCU memory cant store word lengths this wide can it?? confused face

FilterShop looks great really amazing tool looking at that now...

[Eva]

There are many shortcuts for making simple digital filters with 8 bit micros, using fixed point and replacing the multiplications by shift and add/sub operations. This is a simple 6dB/oct IIR LPF. It's called a "recursive filter" when done that way:

LPF_ACCU-=LPF_ACCU>>S
LPF_ACCU+=NEW_VALUE
LPF_OUTPUT=LPF_ACCU>>S

And this little basic program tells you filter cutoff depending on S and FS (sampling freq.)
10 FS=1000:S=8
20 SC=(-1/LOG(1-1/2^S)):TC=SC/FS:FC=1/(2*3.1415*TC)
30 PRINT "Fs=";FS;" S=";S;" Sc=";SC;" Tc=";TC;" FC=";FC

Another application of this code is to get extra precision out of the ADC by oversampling and then LPF. 2^n oversampling adds n/2 bits of precision at the output of the LPF.

Keep this info for later. You still have a lot of work to do to get class D working.