Active DI Design

[seanm]

Long story short, I have switched to going straight to the board at one jam. Since the mixers idea of Hi-Z is 50K, I am using a DI

I was not happy with any of the active DIs I have. I had an SSM2142 that I bought by mistake when I meant to get an SSM2141    So I thought I would try making my own DI. I have breadboarded the DI and the results so far are very promising!



As you can see, the SSM2142 makes for a very small parts count at a cost. I think I paid $8 CDN for the part.

I do have a couple of questions though. Do I need a cap between the emitter and the input to the SSM2142? It seems to work well without it, but I am worried that it may not be a good idea. I don't like caps in the audio path though if I can help it.

Also, I tried to follow the instructions at GEOFX for the voltage divider and came up with 1.8k for the resistors. That seemed awfully small, so I used 10k.

If anybody sees any other mistakes or gotchas, please let me know.

I will be using a passive bass and the +6dB of gain from the SSM2142 helps. Active instruments might need a pad.

I first used an opamp for the buffer and if anybody is interested, I have the schematic for that too. The transistor version works almost as well and draws much less power.

[alextheian-alex]

I think you'll need to bias the base of your input bufer... try 1/2 of your Vcc rail tied to the end of your ground reference resistor.  If you are using this for bass, you can have some huge peaks... up to 7 or 9 volts or so, so keep the rail voltage high to avoid clipping when transients swing up to the rail.  Also, your emitter resistor is pretty high... not a huge deal, but drop it down to give yourself some more current... try 10k or so to start with.  It's good practice to put a small stopper resistor on the input of your BJT as well so that you don't oscillate.

And yes, as far as i know you'll need a DC blocking cap between the emitter and the line driver's input, as well as a ground reference resistor.  and it is good practice to have balancing resistors on the output of the driver, 1 per phase to ground, equal to the input resistance of what you'll be drivein.  don't know if it is crucial or not... maybe take a peek at the data sheet.

hope that helps

[seanm]

I think you'll need to bias the base of your input bufer... try 1/2 of your Vcc rail tied to the end of your ground reference resistor.

R1 goes to Vref, which is 1/2 of Vcc. Won't that be good enough?

If you are using this for bass, you can have some huge peaks... up to 7 or 9 volts or so, so keep the rail voltage high to avoid clipping when transients swing up to the rail.

Agreed. I am not sure how I am going to power this yet. One of the problems is that I would have to bring an extension cord to reach a power outlet So I may use batteries. Otherwise, I will probably use a 24V supply. I tested with the 24V supply (which provides closer to 30V under this load) and a weak 9V battery.

Also, your emitter resistor is pretty high... not a huge deal, but drop it down to give yourself some more current... try 10k or so to start with.  It's good practice to put a small stopper resistor on the input of your BJT as well so that you don't oscillate.

I wondered about the emitter resistor too. However, compared to the SSM2142, it draws basically no power. I will try adding the stopper resistor.

And yes, as far as i know you'll need a DC blocking cap between the emitter and the line driver's input, as well as a ground reference resistor.

From the datasheet: "The input of the SSM2142 should be driven directly by an operational amplifier or buffer offering low source impedance and low noise."

It was the "driven directly" that prompted me to drop the coupling cap. And it does work without it......

and it is good practice to have balancing resistors on the output of the driver, 1 per phase to ground, equal to the input resistance of what you'll be drivein.  don't know if it is crucial or not... maybe take a peek at the data sheet.

I believe the SSM2142 already has them. It is meant to be connected directly to the XLR.

hope that helps

It does. This is the kind of feedback I was looking for 

[brett]

Hi
A couple of suggestions.
If the IC is anything like ordinary op-amps, you've got a big DC offset at the input.  If it works without a cap, maybe it has a cap built in?
To be sure, I'd add a 0.1uF cap between the emitter and pin 4. 
Also, try reducing the 100k resistor on the MPSA18.  Something like 10k is the usual.  It'll use a tiny tiny bit more power.
Put a 2.2 to 22uF cap in parallel with R4.  You'll see that in many biasing networks.  It reduces noise and "stiffens" Vbias.
cheers

[David]

Agreed. I am not sure how I am going to power this yet. One of the problems is that I would have to bring an extension cord to reach a power outlet So I may use batteries. Otherwise, I will probably use a 24V supply. I tested with the 24V supply (which provides closer to 30V under this load) and a weak 9V battery.

Hey, Sean!  Could you run it from phantom power off your mixer?

[seanm]

Hi
A couple of suggestions.
If the IC is anything like ordinary op-amps, you've got a big DC offset at the input.  If it works without a cap, maybe it has a cap built in?
To be sure, I'd add a 0.1uF cap between the emitter and pin 4.
Also, try reducing the 100k resistor on the MPSA18.  Something like 10k is the usual.  It'll use a tiny tiny bit more power.
Put a 2.2 to 22uF cap in parallel with R4.  You'll see that in many biasing networks.  It reduces noise and "stiffens" Vbias.
cheers

Great suggestions! Thanks.

Hey, Sean!  Could you run it from phantom power off your mixer?

I would love to! But the mixer's owner does not like turning on phantom power. However, the mixer allows blocks of four channels to have the phantom power enabled/disabled via jumpers. I may be able to convince him to allow one four channel block to be turned on.

If so, you may see a change to the design

[alextheian-alex]

I think you'll need to bias the base of your input bufer... try 1/2 of your Vcc rail tied to the end of your ground reference resistor.

R1 goes to Vref, which is 1/2 of Vcc. Won't that be good enough?

If you are using this for bass, you can have some huge peaks... up to 7 or 9 volts or so, so keep the rail voltage high to avoid clipping when transients swing up to the rail.

Agreed. I am not sure how I am going to power this yet. One of the problems is that I would have to bring an extension cord to reach a power outlet So I may use batteries. Otherwise, I will probably use a 24V supply. I tested with the 24V supply (which provides closer to 30V under this load) and a weak 9V battery.

Also, your emitter resistor is pretty high... not a huge deal, but drop it down to give yourself some more current... try 10k or so to start with.  It's good practice to put a small stopper resistor on the input of your BJT as well so that you don't oscillate.

I wondered about the emitter resistor too. However, compared to the SSM2142, it draws basically no power. I will try adding the stopper resistor.

And yes, as far as i know you'll need a DC blocking cap between the emitter and the line driver's input, as well as a ground reference resistor.

From the datasheet: "The input of the SSM2142 should be driven directly by an operational amplifier or buffer offering low source impedance and low noise."

It was the "driven directly" that prompted me to drop the coupling cap. And it does work without it......

and it is good practice to have balancing resistors on the output of the driver, 1 per phase to ground, equal to the input resistance of what you'll be drivein.  don't know if it is crucial or not... maybe take a peek at the data sheet.

I believe the SSM2142 already has them. It is meant to be connected directly to the XLR.

hope that helps

It does. This is the kind of feedback I was looking for 

yes, R1 to Vref @1/2 Vcc will work fine.  As far as the DC blocking cap, you don't *need* it, but you may have DC offset if you don't have it, which *may* lead to assymetrical clipping... I dunno, I'd have to see the insides of the chip to know for sure, and they don't post a complete schematic, so just use your judgement--you're probably fine without it.

Also like Brett said, bypass the lower leg of the voltage divider with a large capacitor or you may get noise and sag.  You could also raise the values to 100k/100k or more... there is a bit of math involved to determine the "best" values, but it does not really matter either way as long as you don't suck down too much current (not likely).

AND as David said, you could tap off the phantom power pretty easily to buy a free 48v supply.  I doubt that your circuit draws more than a dozen mA.

[seanm]

Also like Brett said, bypass the lower leg of the voltage divider with a large capacitor or you may get noise and sag.  You could also raise the values to 100k/100k or more... there is a bit of math involved to determine the "best" values, but it does not really matter either way as long as you don't suck down too much current (not likely).

AND as David said, you could tap off the phantom power pretty easily to buy a free 48v supply.  I doubt that your circuit draws more than a dozen mA.

I was using 100k/100k. It was after reading the GEOFX article that I lowered them. What is the advantage of 100k over 10k?

With the transistor and the original design, it draws about 5-6mA (don't have my notes in front of me). Which makes sense, the SSM2142 draws about 5mA. With the opamp based version it draws closer to 10mA.

[alextheian-alex]

The impedance of a voltage divider will represent a current sink to the power supply.  it is really no big deal as long as you don't go too low or too high.

[brett]

Hi
You can think of the Vref or Vbias network as a place that sets a "pseudo ground" for op-amp circuits that use a dual supply op amp (ie +/- supply) when using a single supply (ie + only).  As such, it needs to behave like a true ground and not vary.  Preferably, it has virtually no resistance to "true" ground for any signal (or any AC) that arrives at Vref (hence the large capacitor bypassing R4 to ground). 

These factors mean two simple things:
It is far easier to make a mistake by choosing too much resistance than by choosing too little resistance (2 x 10k is a good choice for 99% of low-power op-amp circuits).
It is better to choose a large bypass cap than a small one, or none at all (10 to 22uF is a good, safe choice).

There are detailed discussions that can be had about equalising bias currents and other things, but most guitar op-amp circuits don't need to explicitly consider those factors.
cheers

[Ardric]

The SSM2142 idles at around 5-6ma with no signal.  It's rated at sourcing 70ma into a shorted output.  Many mic inputs are 600 ohms impedence.  You'll need to rethink your power supply budget.  IMHO there's little chance of powering this thing from a few ma of mixer phantom power.  It'll be way too hungry.

As it's drawn now, plugging it into an input with phantom power enabled will cause the SSM2142 to emit fragrant smoke.  You'll need to either block DC on the outputs with great big series caps, or come up with some clever floating supply arrangement.

There should be a DC blocking capacitor on the input side too, unless you're going to carefully manage the DC offset.  They want the input driven from a low impedence because they're mixing feedback with it internally.  Look at the diagram on the first page of the datasheet to confirm.  So the lower you make that buffer emitter resistor, the better.  You'll quickly reach a point where you'll get better performance and current draw by using an opamp as a buffer instead.

[seanm]

The SSM2142 idles at around 5-6ma with no signal.  It's rated at sourcing 70ma into a shorted output.  Many mic inputs are 600 ohms impedence.  You'll need to rethink your power supply budget.  IMHO there's little chance of powering this thing from a few ma of mixer phantom power.  It'll be way too hungry.

I added some resistors to the outputs to get the phantom power. My little Behringer mixer puts out 24V under no load but it dropped to almost 4 volts when connected to the DI. All I got was a small square wave generator Now the Behringer is probably a bad example because I can run a Fishman DI from it, although it does cause the low battery LED to stay on.

There should be a DC blocking capacitor on the input side too, unless you're going to carefully manage the DC offset.  They want the input driven from a low impedence because they're mixing feedback with it internally.  Look at the diagram on the first page of the datasheet to confirm.  So the lower you make that buffer emitter resistor, the better.  You'll quickly reach a point where you'll get better performance and current draw by using an opamp as a buffer instead.

I do have a 1uF blocking resistor on the breadboard. I should update the schematic.

Using an opamp is a real trade off. My tests have shown increased headroom at 9V with the opamp, but increased current too.

[/quote]