Schottkey diode for smps control.

[dr pepper]

I put together a smps using a tl494 and usual comps for a boost converter.
Page 19 fig 27 shows how to cycle by cycle current limit with the tl494:

https://www.ti.com/lit/an/slva001e/slva001e.pdf

The diagram shows a schottkey connected to the timiag cap, I tried a bat41 and a bat48, neither worked I get the waveform on the cap appearing at the deadtime control (even when no current limit is happening), a 1n4148 works, but this is not the correct diode and it does affect osc freq, posibly due to its capacitance.

Any ideas what diode I should be using here?

 

[ronsimpson]

I get the waveform on the cap appearing at the deadtime control

Are you using a 1,000,000. ohm resistor?

 

[dr pepper]

Yep, I double checked that, I dont have another in the drawers but it does meter out correctly.
I replaced it with a 100k which seemed to make the circuit latch bewteen cycles but not allways un-latch as the osc o/p isnt able to drive against it.
I tried bc547, and 557's and 2n3904/2n3906's, the latter sort of work, the former current limits but doesnt latch and reset between cycles, I checked the pinouts, ok.
Seems that the circuit is particular, or I'm doing something mad.

 

[schmitt trigger]

You may require matching of Q1 and Q3. Not simple to do without a curve tracer, but it can be done.
Something else you may want to try is to add a 1Meg in series with Q1's base. Again, an attempt at matching.

 

[dr pepper]

I have a curve tracer, with variable current and a 3 pin socket, and a tape of bc547's, however that sounds like a pain.
I will try the 1 meg on q1 though (and maybe a 100k as that would equalise base currents as one tranny works with vref on its base and the other with 0.6v from the sense resistor), if that doesnt work then I think I'll try another method.
I'm surprised TI publish such difficult to get to work circuits.

 

[schmitt trigger]

Having known personally the app note-writing business, I can tell you that many in the IC industry have ingrained the fact that transistors are matched...that is a given in IC design.

Now, I know, these are discretes. And my guess is only that, a guess. Another clue to this guess is that note does not list any particular transistor or diode part number. IC designers do this frequently, and let the process engineers figure out the actual device geometry to meet the requirements.

 

[audioguru]

I get 1N5819 Schottky diodes from sunburned solar garden lights that rusted away.

 

[ronsimpson]

You may require matching of Q1 and Q3. Not simple to do without a curve tracer, but it can be done.
Something else you may want to try is to add a 1Meg in series with Q1's base. Again, an attempt at matching.

I do not think the transistors need to be matched.
I think:
Q1 will turn on when current in Rcl causes 0.7 volts.
Q1 on will turn on Q2
Q2 on will turn on Q3 which will keep Q2 on (latch on)

To reset the latch: D1 needs to steel away the current from the base of Q3. The diode needs to have a very low voltage drop. Less than the turn on voltage of Q3!

I get the waveform on the cap appearing at the deadtime control

How can a signal get from (CT) to (Dead Time)?
At first I thought through the 1M resistor.
Now I think it might go into B-Q3 to B-Q2 and to Dead Time. If this is the case try a 1 meg resistor from B-E of Q3.
It may be that the "latch" is almost turned on but not. It is in analog node and is working as a amplifier.

 

[schmitt trigger]

Transistor matching? Maybe not. Or maybe yes.
We don't know, and it is a suggestion that either has to be validated or discarded.
Either way, one will have a little more experience.

 

[dr pepper]

Ron thats exactly what I thought, and its also what the datasheet says the circuit does.

I'll try the 1m accross Q3, in fact all the ideas suggested.

 

[ronsimpson]

For those watching, here is the circuit.

Vref=5V
Switching Circuit is a ramp from 0 to 0.7 volts.
Ct is a sawtooth ramp.
Dead-Time is DC.
All transistors are off. Switching Circuit ramps to 0.7 and turns on Q1, which turns on Q2 and then Q3. This (Q2&Q3) latch on. Q2 pulls DeadTime up to 4.7 volts. When DaedTime is high the SwitchingCircuit turns off. When CT goes low it starves out the base current in Q3 thus resetting the latch.

For some reason the sawtooth on CT is getting to DeadTime. WHY?

Maybe try 1MΩ to 470k or 220k??

 

[schmitt trigger]

geeez.
actual scope waveforms would help a lot

 

[dr pepper]

I moved on to another project for now, I'll get back to this one.