Do I need i/o buffers for this circuit?

Hi,

I'm pulling a spring tank out of an old mixer and building a rackmount unit for it. I plan to clone the drive and recovery amps from the mixer schematic. It's pretty simple, just a dual op amp circuit. I'm wondering if I need to add input and output buffers as well.

Attached is a schematic of the drive and recovery circuits. I simulated them in LTSpice and they seem to be providing enough gain on both sides and the frequency response seems right for driving a tank.

Thouhts?

There's no need for additional amplifiers, but you may want to add a few passives to the driver circuits you have to perform the function that a buffer would. On the device input (the input to the driver stage) you should probably add an input coupling cap as well as an RFI shunt capacitor. It's always good to roll off excessive bandwidth and protect against random DC sent to the device. So, add a 50-100pF shunt cap across the input jack positive and ground, then add something like a 50Ω series resistor from the input jack to a 10µF or higher series coupling cap. Or, you could make the 10KΩ input shunt resistor higher and use a smaller coupling cap, or make the coupling cap smaller to roll off more LF (not a bad idea with a spring).

On the output side, you should add a 50-100pF capacitor across the output jack positive and ground, and then connect the op amp to the output jack positive through a 50Ω series resistor. The purpose of the 50Ω resistor is to isolate the op amp from cable capacitance, and from the capacitance of the 100pF RFI filter cap. Without it, the 5532 is likely to oscillate from the shunt capacitance of the output cabling.

Most shielded audio cable has about 30-40pF of capacitance per foot, so try shunting the output driver stage with 1000pF (around 25 feet of cable) and then run a .tran analysis in LTspice and see if it's stable. You'll probably find that without the 50Ω resistor, the 5532 will ring, or actually oscillate, and with the 50Ω resistor, you'll have no problems.

Thank you. I don't need a coupling cap between the op amp and the output jack?

Now that I'm running .tran analysis instead of .ac , I'm noticing that both these circuits run well into clipping in the mid-range. I wonder if this is part of the design, or a misprint on the schematic or what.

By increasing the values of R38 to 390ohm and R102 to 910, I can eliminate the distortion, but it decreases the gain and flattens out the frequency response especially in the makeup circuit.

Guess I'll play with it on the breadboard.

samgraysound said:

Thank you. I don't need a coupling cap between the op amp and the output jack?

Click to expand...

It's not absolutely necessary, since offset at the output will be only a few millivolts, but it is good practice.

Now that I'm running .tran analysis instead of .ac , I'm noticing that both these circuits run well into clipping in the mid-range. I wonder if this is part of the design, or a misprint on the schematic or what. 

Click to expand...

You definitely need to adjust the input level. Typically, I would say the sensitivity is about 100mV for nominal drive, so you need an input pot. Just as well, you probably will need to boost the output; an output pot would be a nice addition. I would add a balanced line receiver and a balanced line driver to round it up.

By increasing the values of R38 to 390ohm and R102 to 910, I can eliminate the distortion, but it decreases the gain and flattens out the frequency response especially in the makeup circuit.

Click to expand...

It also changes the frequency response of the drive circuit, which is a constant-current source only as long as R38 is much smaller than the actual drive coil. Use an attenuator to set the drive level.

samgraysound said:

Thank you. I don't need a coupling cap between the op amp and the output jack?

Click to expand...

You can certainly add one, and it's not a bad idea. The typical difficulty is that it'll have to be huge, since you don't know how low the load impedance will be. But, if you ever drive a transformer input, it's really good to make sure you're not stuffing DC into the primary winding of the next device. So, yes! Something like a 10-47µF high quality electrolytic (Panasonic FR series) will work fine. Too large, and the leakage of the cap will defeat the purpose, and too small and you'll suffer reduced LF response, which might not be important if the reverb output is 100% wet.

A polarized electrolytic is not absolutely correct, but in practice, a small amount of reverse bias will not cause problems. You want the voltage across the coupling cap to be low anyway, so a polarized aluminum electrolytic is not such a horrible answer, especially for a spring reverb tank return. If you want to get fancy, a 10µF metallized polypropylene will be fantastic, will provide full LF response into a 10kΩ load, and will never wear out.

Now that I'm running .tran analysis instead of .ac , I'm noticing that both these circuits run well into clipping in the mid-range. I wonder if this is part of the design, or a misprint on the schematic or what.

Click to expand...

.tran analysis is always a great idea. While .ac does something useful, it does not show you everything that can go wrong, and can easily fool you into thinking that a circuit is stable when it is not. You need to do both, and make sure that the circuit survives a .tran analysis before you extract info from an .ac analysis.

As for your clipping, make sure you're testing something valid. Your circuit has a drive section, a reverb tank, and a reverb tank preamp / output section. I don't know what you're doing to model the reverb tank, but to determine clipping, I'd imagine you connected all three modules together and did some sort of analysis by driving the system and watching what happens at the output of the whole chain.

The problem is that modeling the reverb tank is going to be difficult since its gain and detailed behaviors are not published, much less available as a pspice subcircuit. So, you could model the reverb tank as a simple amplifier with a specific gain (i.e. loss), and then plug that between the two driving and receiving circuits.

If the reverb tank is not specified closely, you can measure it easily. Drive the tank with an oscillator and measure its output voltage. Because spring reverbs are high-Q resonators, it's wise to measure the gain/loss at a handful of frequencies in the speech band (around 1kHz), and average the gains.

Glad to see you're making progress, and report back with what you find!

*slaps forehead*

I have them as separate schematics and I was just modeling them separately with 1v input signals. Obviously the makeup circuit is expecting a much lower input!

The drive circuit can take a max input of 285mv. The makeup one can take a max input of 120mv.

I'm not going to mess with the makeup one. I'd like the drive circuit to be able to take a +4 line level signal. How can I implement this? I don't want to add a variable input control as I will be sending signal with a mixer aux send and don't want an extra control in the way.

Could I just add a L or T pad to the input?

samgraysound said:

I'm not going to mess with the makeup one. I'd like the drive circuit to be able to take a +4 line level signal. How can I implement this? I don't want to add a variable input control as I will be sending signal with a mixer aux send and don't want an extra control in the way.

Click to expand...

However, a good old potentiometer is quite handy; you just set it and forget it. If one day, you think you've gone too far in either direction, it's easy to readjust. It could be a trimmer, if it makes you more omfortable.

Hmm a trimmer. I could do that. I guess it's just a L-pad aka voltage divider.

Sam

So how much headroom over +4 should I design for in the reverb driving amp? 18db?

Sam

samgraysound said:

So how much headroom over +4 should I design for in the reverb driving amp? 18db?

Sam

Click to expand...

Since you have means to control the send level, you should not need the same headroom as you would from a raw source. I have found 16dB headroom to work quite well. That's typically what you get using opamps powered from +/-15V rails (+20dBu peak).

I need some help troubleshooting.

I set this circuit up on strip board and have been testing it. The output is not nearly as much as my model in LTSpice. With a 250mv input I should be seeing a 22v output, I am seeing a 12.5v output. If I drive it into clipping it maxes out at 13v.

I'm a bit bewildered. I double checked all my component values. I rebuilt the circuit with all new components on a breadboard. Even simplifying the circuit to just resistors, my headroom is still capped at 13v.

The power supply is providing a solid +15v and -15v and  is massively overbuilt for the circuit because I pulled it from the mixer. I'm sending -15v to pin 4 and +15v to  pin8 of the 5532.

What impedance are you driving with your test breadboard?  And how does it compare to what you were driving in LT Spice?

Set it up with doubled 9v batteries for each rail and headroom was still capped at 16v. I feel like this has to be human error.

ruairioflaherty said:

What impedance are you driving with your test breadboard?  And how does it compare to what you were driving in LT Spice?

Click to expand...

driving both open for now.

I can literally set it up like this (attachment) and still have headroom capped at 13v

> doubled 9v batteries for each rail

36V total supply means <36V peak to peak.

36Vpp is 12.8V Sine RMS. Near your 12.5V and 13V numbers.

I say you got more than you should expect. (The chip will lose 1V-2V each side, I'd pencil 32Vpp or 11.3V Sine RMS.)

If SPICE says different, it is probably right in its terms but you are not speaking the same language. SPICE is not audio centric and does not always assume "Sine signals". Mine I have to tell it the Peak voltage. Some SPICE opamp models are just utterly bogus and pull output signal from thin air (not from the supply like real chips do).

PRR said:

> doubled 9v batteries for each rail

36V total supply means <36V peak to peak.

36Vpp is 12.8V Sine RMS. Near your 12.5V and 13V numbers.

I say you got more than you should expect. (The chip will lose 1V-2V each side, I'd pencil 32Vpp or 11.3V Sine RMS.)

If SPICE says different, it is probably right in its terms but you are not speaking the same language. SPICE is not audio centric and does not always assume "Sine signals". Mine I have to tell it the Peak voltage. Some SPICE opamp models are just utterly bogus and pull output signal from thin air (not from the supply like real chips do).

Click to expand...

Yes this is it thank you. I was seeing pp in spice and RMS on my multimeter. So everything was working how it should.

Okay. Circuit is working as it should but now I have 60 hz hum.

Moved things around trying to figure it out and discovered that the tank itself is picking up the hum from being too close to the power xfmr. When I remove it from the chassis the hum goes away.

Everything is packed in the chassis tight, I don't really have any room to move the tank farther away. Do I have any other options?

Could I shield the wires for the pickup transducer, or the transducer itself? Could I put shielding around the transformer. I suppose last resort is sticking the transformer in a external box.

Here's a gut shot