Function Generator output stage

[stickjam]

I had a really cool dsPIC-based synthesized function generator cobbled together on perf board.  It did a great job--you could set it on 440.0 and you tune a guitar dead on.  Note my use of the past tense since its recent not-so-brief encounter with a 325V B+ rail.  The tube supply fuse held on tight and the genny ended up as an acrid clump of plastic, metal and silicon.  I should have taken a picture of it reposing in the snow bank I threw it in to cool off.  I transported it there by its AC power cord--the only thing not glowing.  Picture the scene in the movie Ghostbusters and you're there; "Class Five free roaming vapor coming through!"  

I can keep a sense of humor about it since there was no real property damage once things aired out, I still have the source code, and I've always wanted to add features like outputs to trigger a scope and more accurate (read digitally-controlled) output level and DC offset controls.

So, my issue for Drawing Board discussion is this... Other than maybe fast-acting fuses P) what design considerations should be taken to produce a reasonably robust, commercial-grade output stage for a 0.1-20,000Hz function generator?  Capacitive/inductive load considerations?  What should the generator "look" like to any circuit location one may desire to inject a signal?   How and to what extent should it tolerate DC/AC that might be fed back from the circuit/device under test (eg. bias, phantom power?)

I'd also like to add TRS and XLR-M outputs to enable connection to devices using normal audio patch cords. (phantom blocking comes to mind here)  The "Mark I" could output up to around +/-10V PP out the balanced banana posts. (I used a DRV134 chip in the original model.  There may be a better choice.)  Am I correct in thinking it prudent to have a lower upper-limit on the TRS and XLRs?  If so, what should the limit be?

Finally, does anyone have any well-designed commercial audio generator schematics from which I might "borrow" output stage features?


Thanks

-Bob

 

[stickjam]

Just remembered the BAMA library, so as a starting point here's a sample function generator output stage I found (from the B&K 3010)  Any improvements or alternatives to this design? A fuse (0.1A fast blow) inserted to the right of R75 comes to mind, immediately. ;D

In my unit, I'd replace the amplitude and DC offset controls with a couple multiplying DACs (eg. Maxim MAX532) to provide digital control.

 

[JohnRoberts]

You can add a decent buffer to pretty much any opamp, the right side of that schematic...

Instead of the pnp pull up transistor, add another resistor to plus supply like the 2.7k to minus supply, and drive the middle of the two diodes with the standard opamp output.

You can add some simple current limiting with two more diodes, backwards across the driver base-emitter junctions.

enjoy, but I can't guarantee these will survive 300+V either...

JR

 

[Steck]

I'm going to jump in on a related yet tangential thought...  

I have a heathkit oscillator that I'd like to use for measuring transformer impedance.

Its output stage is pretty similar to what's here, save there's an attenuator setup that ends up raising the output impedance into the 1.2K + realm to get the signal down from +20 to +4 into 600 ohm load, and an even higher impedance for a lower amplitude.  

I'm thinking an additional buffer *after* all the level-setting should get the output impedance back down into something where I can measure some transformers with an amplitude somewhere lower than stun.  I'd like to be able to deal with older transformers that have 50 ohm nominal impedances....   Or am I overthinking this?

 

[JohnRoberts]

If you have a meter with good sensitivity, you can use a buildout dropping resistor in series with the load.

JR

 

[stickjam]

You can add a decent buffer to pretty much any opamp, the right side of that schematic...

Instead of the pnp pull up transistor, add another resistor to plus supply like the 2.7k to minus supply, and drive the middle of the two diodes with the standard opamp output.

You can add some simple current limiting with two more diodes, backwards across the driver base-emitter junctions.

enjoy, but I can't guarantee these will survive 300+V either...

JR

It'll probably be a lot cheaper to fix than 300V directly into a $25 DAC chip.  ;D

Would adding a fuse between R75 and output, then a couple back-to-back zeners from the r75/fuse node to ground be helpful or harmful?

 

[jdbakker]

This Linear Tech appnote shows a few useful output stages. There is another one, but I can't find it ATM. Protection diodes from the outputs to the rails plus zeners on the supply lines should help protect you from B+ incidents.

If you need low-Z output, short-circuit survivability and unconditional stability and don't mind a tad higher distortion and dissipation, you can always brute force it with any old power amp followed by a resistive pad.

JDB.

 

[JohnRoberts]

+1 to diode clamps back to the rails, and a flame proof resistor in series with the output that will get sacrificed in case of such a drastic fault voltage/current.

A fuse in series with the output could introduce distortion at low frequency, if the current flowing is enough to heat the fuse.

JR

 

[PRR]

> Would adding a fuse

Transistors fail to protect fuses. It just won't work.

Your best bet is to find a technician who knows better than to connect tone to a B+ rail.

A further step for less drastic abuse is to use the H-P 200AB output stage: two small 6V6 tubes into a 20K:150+150 transformer. The output floats, so you could in fact inject 440.0Hz into your B+ rail. The output is low DC resistance, and has a good chance of hanging on until a fuse blows (but not all B+ lines have fuses). Yeah, the bandwidth and THD isn't modern-spec, but what's the bandwidth/THD of your PIC after a snow-bath?

Use different tools for different jobs. Wire a socketed TL072 on a 9V batt for wild probing and experimentation. You don't need sines, you hardly need three approximate frequencies, to find out if signal passes or not, and if the audio path is roughly flat and clean. Less to go wrong and less to cry about with a battery tri-wave whiner. Then save the "good" oscillator for final testing, when you KNOW what you are doing and where you are sticking that lead.

> 50 ohm nominal impedances....  Or am I overthinking this?

Yes.

On all but the top amplitude, the output of the transistor Heath is 600 ohms. Put 550 and 50 in series across the output. Feed your 50-ohm gear here. The voltage is about 20dB down, the impedance is near 50 ohms. For government work, you could design a pad with exact 600:50 ports and minimum or round-number loss; for ad-hoc general small audio iron tests, just read the voltage at the transformer and pretend that 47.9==50 for all practical purpose. (580+52, plus the 600 inside, is mighty close to 50.0 ohms.)

> thinking an additional buffer *after* all the level-setting

I did that for loudspeaker tests. The 600 ohm one-click down amplitude into 8 ohms was mighty low for a standard VTVM, under 0.05V. I added a 383 or 2002 or something 12V car-amp chip, pumped from the 6VAC lamp-winding, to get a solid Volt into 4-8 ohms, or a good tenth-Volt through a 50 ohm resistor for mapping the impedance. I'm not sure it was worth the effort.

Also, I did that when I had NO spare power amps, I'd have to pull an amp from a classroom. Amp prices are way-down, you find stuff in the trash, a one-side-dead stereo is still a fine bench booster. And measuring iron, you are not likely to demand 0.001% THD or 100KHz bandwidth. Put an old Sansui after the Heath, optionally load with Radio Shed's 10 ohm 10W brick, add 50 ohms in series, feed your iron. Watch the level!! Even a modest-power loudspeaker amp can put many Watts in 50 ohms, which is usually a 100mW world at most.

 

[Steck]

PRR & JohnR : thanks!!

 

[stickjam]

> Would adding a fuse

Transistors fail to protect fuses. It just won't work.

Your best bet is to find a technician who knows better than to connect tone to a B+ rail.

LOL!  ...or knows better than using a crappy alligator clip to put tone to a grid socket pin nub on the foil side, then walk out of the room to get another scope probe only to have the clip let go and land on the B+ trace.  There's a nice mark where the clip started to weld itself to the copper.  :


Thanks guys.  Good info.  I knew you could talk some sense into me.  Now if I can keep the feature bloat to a minimum on the software side, I'll be all set. 

-Bob