How to speed up this circuit

[uoficowboy]

Hi - can anybody suggest how I might go about speeding up this circuit? My goal is to take a high impedance TTL input and output a lower impedance differential version of the input. The circuit I have attached works just fine in simulation - except that it is slow! Would different transistors make it faster? What else can I adjust to speed it up? I had hoped to have response times around 100ns or so, but instead I'm seeing about an order of magnitude worse than that!

Thanks!

 

[karesz]

Hi,
What about over more speedy transistors pls?
Otherwise you can pick the basis resistors (some capacitances_so at 47 pF_ parallel on the 20KOhms, eventually a Schottky diode parallel on BE(Cathode on the Base), the you can have a small inductance serial with the Collector resistors too...
K.

 

[uoficowboy]

Hi,
What about over more speedy transistors pls?
Otherwise you can pick the basis resistors (some capacitances_so at 47 pF_ parallel on the 20KOhms, eventually a Schottky diode parallel on BE(Cathode on the Base), the you can have a small inductance serial with the Collector resistors too...
K.

Hi - how do I identify faster transistors? What specification do I care about?

What would a parallel Schottky do? Speed up turn off?

Thanks so much!

 

[FvM]

An inductor won't reduce charge storage in saturated switching, which is the basic circuit problem, all other suggestions should work perfectly.

To experience the effect of transistor speed, you may want to try BFR92A and BFT92 in the circuit.

What would a parallel Schottky do? Speed up turn off?

Yes. it prevents from transistor saturation.

 

[biff44]

jeez, what's with the 20K base resistors? That is a good way to slow everything down. Try 220 ohms with maybe 27 pf across each one as a start.

After that, you want to trim stuff so that each trasistor is turned on well, but not way oversaturated. Also, play around with the 1K resistors, they may be too big for a capacitive load.

 

[keith1200rs]

You can probably get down to around 250ns to 300ns by adding Schottky diodes between the collector-base of the two output transistors and dropping the base drives to 4.7k from 20k.

Some speed-up caps across the base resistors (around 22pF) as already mentioned, would then get you down to well under 50ns.

You have the simulation tools so experiment with the values a bit and see what works. Also, be sure to check whether it is the first transistor or the output pair that is limiting the speed to make sure you are adjusting the right stage.

Keith.

 

[keith1200rs]

Oh, and don't bother with BFR92/BFT92 - nice transistors and I use them quite often but you will probably just end up making a GHz oscillator

 

[FvM]

you will probably just end up making a GHz oscillator

Yes, very easy with unsuitable PCB layout. You're right, that these GHz transistors aren't actually needed in the present case. But as transistor
speed was questioned, I think, it should be demonstrated.

 

[uoficowboy]

jeez, what's with the 20K base resistors? That is a good way to slow everything down. Try 220 ohms with maybe 27 pf across each one as a start.

After that, you want to trim stuff so that each trasistor is turned on well, but not way oversaturated. Also, play around with the 1K resistors, they may be too big for a capacitive load.

I tried playing with values before posting and the 20K base resistors seemed to speed it up. I wonder if the value was so high that they were keeping the transistor from saturating? I'm not sure.

Added after 3 minutes:

An inductor won't reduce charge storage in saturated switching, which is the basic circuit problem, all other suggestions should work perfectly.

To experience the effect of transistor speed, you may want to try BFR92A and BFT92 in the circuit.

What would a parallel Schottky do? Speed up turn off?

Yes. it prevents from transistor saturation.

I will check those out. The SPICE models for them seem strange, I'll have to figure out how to use them.

So - they are labeled as 5GHz transistors. Is that how you are identifying them as being fast? Or is there some other parameter that you can look at that will identify a part as being quick?

Thanks!

 

[rfsystem]

From the waveform:

TurnOff of both outputs are slow because they could be driven On by the 20k but not fast enough Off because they go into saturation and accumulate much charge in the base. So for switching them On takes much less charge then to make shure that the output levels are sufficient near to the supply.

The design should be very simple. 3 NPN's, no diodes, as less R than possible, right?


The only solution I could imagine is to make a compromize on the output levels. If you accept a slightly lower high level and a slightly higher low level there is a way to avoid high saturation of all three bipolars.

I have a solution with 3 bipolars, 9 resistors and 2 diodes.

Is there a way to use a web based schematic do draw the circuit?

 

[keith1200rs]

The only solution I could imagine is to make a compromize on the output levels. If you accept a slightly lower high level and a slightly higher low level there is a way to avoid high saturation of all three bipolars.

I have already posted a pdf with modified resistors, two diodes & speed-up caps which achieves <50ns. With a bit more work it could probably be improved further. See above.

Keith.

 

[uoficowboy]

You can probably get down to around 250ns to 300ns by adding Schottky diodes between the collector-base of the two output transistors and dropping the base drives to 4.7k from 20k.

Some speed-up caps across the base resistors (around 22pF) as already mentioned, would then get you down to well under 50ns.

You have the simulation tools so experiment with the values a bit and see what works. Also, be sure to check whether it is the first transistor or the output pair that is limiting the speed to make sure you are adjusting the right stage.

Keith.

Keith - I've taken these suggestions and added them to the model. Now I'm getting it to be crazy fast.

Interestingly enough - the choice of Schottky seems to affect it quite a lot. I first tried it with a BAT54 (as LTSpice has that model built in) and it was significantly slower. Why was the BAT54 circuit so much slower? Looking at datasheets for the respective parts - it looks like the BAT54 has a capacitance of about 10pf and the 1PS70SB82 has a capacitance of about 1pf. Is that the big difference, or is there something else that I am missing?

Thanks so much!

 

[keith1200rs]

Interestingly enough - the choice of Schottky seems to affect it quite a lot. I first tried it with a BAT54 (as LTSpice has that model built in) and it was significantly slower. Why was the BAT54 circuit so much slower? Looking at datasheets for the respective parts - it looks like the BAT54 has a capacitance of about 10pf and the 1PS70SB82 has a capacitance of about 1pf. Is that the big difference, or is there something else that I am missing?

Thanks so much!

As you say, it is probably the capacitance. The easy way to check is to add 9pF across the 1PS70SB82 and see what happens.

By the way, my choice of diode was not very scientific - I just picked one I had models for. I cannot remember which of the Philips 1PS range I used before, but some of them are pretty good for high speed circuits. I think the HSMS-282Y is another I have used. Some Schottkies aren't as good as you might think they should be, depending on the application.

Keith.

Edit:

By the way - the diode across Q1 may not actually have much effect - try it without it.

Keith.

 

[uoficowboy]

As you say, it is probably the capacitance. The easy way to check is to add 9pF across the 1PS70SB82 and see what happens.

By the way, my choice of diode was not very scientific - I just picked one I had models for. I cannot remember which of the Philips 1PS range I used before, but some of them are pretty good for high speed circuits. I think the HSMS-282Y is another I have used. Some Schottkies aren't as good as you might think they should be, depending on the application.

Keith.

Edit:

By the way - the diode across Q1 may not actually have much effect - try it without it.

Keith.

Indeed - it seems to have been the capacitance. Is that the main parameter that one should look at when trying to find a fast Schottky? Or are there others that I should worry about?

One other question: how did you calculate out the values for the 22pf capacitors and the 4.7K base resistors? Simulation?

Thanks!

-Michael

 

[keith1200rs]

Yes, for a high speed circuit the capacitance is the main Schottky parameter to look for, but make sure the forward volts drop doesn't get too high at the current you are using.

If I was doing a complete design I would normally select resistors first by design, probably starting at the load and working back. Then optimise by simulation. The speedup capacitors I would probably just simulate. That doesn't actually mean it's the best way to do it!

Keith