ringing with high speed switching

[jimmylovesni]

hello all.

Im working on reconstructing a circuit that someone else made. Im by no means an expert. Im having a little trouble understanding a ringing effect Im seeing.

So I have one mosfet driver driving 4 other mosfet drivers at approximately 2MHtz. These chips have 40MHtz band limit (EL7158) and Im pretty sure they are not the problem. I have the all 5 chips in a line staring with the one driving the 4 others. The output of this first chip is directly connected to the input of each of the others via wire wrap (size30). As I progressively get further away from the chip driving the others, a ringing becomes more pronounced on the inputs to the chips and this is amplified in the outputs that they produce. I rebuilt the circuit twice with the same effect.
Id like to understand what is going on. Is it some type of inductive effect that takes place at higher frequcines in the longer length wire. Is there a way to fix it? I tried disconnecting the wires and using thicker size 22 solid wire. Maybe stranded? I dunno.

Ill attach some photos of the 5 pin drivers (front and back) as well as the inputs to the pin driver (same node), as well as the driver outputs

 

[VEBill]

If the square wave is at 2MHz, then the ringing is at about 20 MHz (roughly). At these frequencies (especially with square edges), you're getting into "RF" design. You might have to use better construction techniques AND add some series resistive loading to dampen the ringing. But the resistor has to be the optimum value so to not impact the switching speed.

Standby for others with more experience in this exact area.

You might want to upload the schematic of the section in question.

 

[jimmylovesni]

i think ive got a handle on it now.
its not perfect.

but there were really two problems
if i scoped the ground, there was some ringing on it and it should have been ground potential.
On the supply voltage was a ringing that appears on the high end of the pulse
So i had to put 2 more different value caps on each chip to decrease the ringing, i already had 2 and assumed it was not the problem but it was.

then also adding a lot more ground wires in for redundancy and reducing also helped.

 

[Comcokid]

Proto board, thin wires, no ground or power planes, IC sockets, no small value/ultra short lead bypass capacitors, thru-hole components. It's hard to tell you where to start.

OK, additonal grounds and bypass capacitors are a start. Keep leads short - solder them point-to-point on the bottom of the board (none of this through-the-hole-rightangle-bend-make-it-look-neat stuff). Add short leads to SMT capacitors if you have them. Keep the electrolyitc and bigger capacitors - since these are drivers, they supply the bulk energy for the pulses, but later since they have more ESR than ceramic caps.

For grounds go for low inductance. Copper foil or tape cut to 1/8 to 1/4" width. If you don't have that, use solder braid.

Also, you can get distorted signals with the scope depending upon where the scope ground lead is connected, and how long it is.

Finally, if there is still some ringing, then try dampening it by using a series resistor in the range of 10 to 33 Ohms. Your connection distance is short, so you should not need terminations on both ends. Put the resistor on the output of the first driver with the lead as short as possible.

 

[electricuwe]

With todays free and easy-to-use PCB-Cad Software an the PCB pooling providers offering cheap and fast prototype PCB services there is no justification for using such prototype set-ups like you do.

 

[DHambley]

The wire-wrapping adds hundreds of nano-Henries of inductance between stages, on both the signals and ground returns. If you are stuck with with wire-wrap as the only construction method, Comcokid is on the right track. Add series inductance and add resistance to ground at each chip input to reduce the reflection as well as small capacitors to each Vcc to gnd. the input impedance of most drivers is low enough so that you might not need a driver just to drive the drivers.

 

[jimmylovesni]

so im pretty much done with this project last week... just felt like checking in. its not that bad and works up to 1-3.5Mhtz although FAR from ideal. Even At this high frequency, one row (driver and mosfets) has a phase shift which is pretty out of whack with the others so am just not using it for now since it switching at the wrong time will just generate huge amounts of heat. This timing issue is probably due to some type of inductance or something. I retried redoing the wires but it keeps coming back and probably need to just gut that whole area (pins) etc and then rebuild it.

Its true that its not ideal way to make a prototype board. Ie the soic to dip adapters and wirewrap sockets, something i knew ahead of time as very bad a high frequency. I do appreciate your input though, its just something it was hard to get around with the tools I have at hand. These matched high frequency (3mhtz) high voltage (100V) matched mosfet pairs are the only chips that fit this application and only come in this package. They are unique chips although a ixys is making a new dip chip for higher power (200V) matched arrays which is soon to come out. The wirewrap looks pretty bad too but it gets the job done although messy... Its a good step before using pcb express. Im not sure about the wire wrapping tool adding hundreds of uH as i can measure inductance and get 1-5uH on each connection. Probably has to do with the core being the same dielectric as the wire.

There is one problem i really dont understand. With the 1Mhtz (1072nF)transducer, the waves are very powerful, rippling through the water bath but they get much smaller after about 5 seconds. If I shut the power off and then back on, the waves stay at low power. During this time, If i scope the power output, there is no change in frequency, amplitude, or phase. If I let the device/transducer rest untouched for a while and then start it up again, it goes for about 1- 5 seconds again and then returns to the small output. The same type of transducer in 2Mhtz (1582nF) does not have this problem and I can let it run perpetually (hours) with very high output. I did notice there seems to be a capacitance change on the transducer when it chokes and that the capactice is very different when it is in water (DI or not DI). What is going on?

I attached a few photos of the transducer and the circuit

 

[jimmylovesni]

attachment

 

[DPolak]

Hello.
Looking at photos - bad prototype.
This EL7158 driver chip has rise/fall time in 10ns range.
Piece of wire =10nH for every inch, so few inches of thin wire may give you 50nH inductance.
Few Amps (say 5A), and 50nH, and 10nseconds will give you lots of ringing (now I'm too lazy to calculate).

just remember - when you go into nanoseconds and nanohenries field - there is no more ZERO ohm/Henries wire. Every piece of copper should be viewed as resistance and inductance.

advice:
for every 'hi speed' connection use wide, flat piece of copper (low inductance). Short as possible (low inductance).
2MHz is not very high frequency, but what makes is difficult is rise/fall time of your driver.

Capacitors - ceramic (or film) with very low ESR and low ESL. Sometimes you will have to connect few caps, not to get higher C, but to lower ESR.
For switching few Amps (and more) - forget about tantalum or aluminum.

Scope - connect probe (and GND!!!) at the point you want to check. CAN NOT connect probe at the chip output pin, and GND at supply wire from main PS.

thx