Keyboard DI/Preamp/Balancer

[RFSiesta]

Hi,
This is my second thread on groupDIY, trying to build a 10-ch dedicated Keyboard buffer with balanced outputs.
Our studio is two-level, two control rooms and all rooms will be interconnected. Cable length is an issue, and while I'm used to record keyboards directly through line inputs or sometimes even to microphone preamps for transformer colouring (only devices with low enough output Z), we decided we need an impedance converter with 0, 6 or 12 dB gain (maybe switchable, maybe w/ locut as well) and an electronically balanced output for being able to drive cable lengths in excess of 15m without loss.
The device(s) will be included in our patchbay system(s), so rather small footprint is important. I want to have it all on one euro-PCB (160*100mm) including supply & in/out screw clamps and an external PSU for +-15V.
I'm planning a JFET input stage, still not sure if IC or discrete, zener protection diodes on inputs and 4004 on outputs, and a balancing line driver like the THAT1646.
Everything looked neat using ONLY the line driver, except the input impedance is 5kHz, which is a bit low, especially for some older instruments. There are other products on the market, like the SSM2142 with Z=10k, but maybe it's better to add a simple impedance converter to the input to increase impedance to 20-100k or something and get higher gain as well.

Since almost all of the circuits I gOoOrgled are designed for hi Z instrument outputs rather than actual keyboards, I would like to know if there's a typical approach to this without the use of transformers (space and cost issue).

I'm trying a simple version with a opa134 in an impedance converter configuration, according to it's datasheet (with feedback v. divider set for +6dB), since I truly like the relaxed sound of these opamps in the return section of my trident 80, where we exchanged the TL071 for OPAs in a 100%fb impedance converter. 134/137 seem to be known to be hassle-free regarding RF issues, so this seems like a simple solution for good quality. TL071 sound OK anywhere in our desk, but in full feedback they actually *suck* IMO...

I will post schematics as soon as possible!

Richard

 

[RFSiesta]

OK - some SPICE simulations later
seems like it's no big issue to combine the Bo Hansen DI schematic with a THAT1646 for a unity gain impedance converter, except:
1) the circuit sees a 4-5kHz input Z instead of a much higher one through the 1:10 (EDIT- 5:1, sorry) transformer in the DI box.
2) no attenuation like a DI, but that's what I wanted : close to line level at the outputs.
The Hansen DI seems to be so utterly stable that most changes in values don't make a big difference in SPICE, except smaller capacitor values change lofrequ. response.
Am I missing anything important?

 

[Andy Peters]

1) the circuit sees a 4-5kHz input Z

Just for clarity's sake. impedance is given in ohms, not Hz.

-a

 

[RFSiesta]

yea sorry, didn't read it a second time. I meant kΩ of course.

 

[Chris_V]

If you want something transparent just put an opamp in non inverting configuration in front of the 1646, you will already have the needed power supply and it will be the simplest solution.

 

[Whoops]

Why the Bo hansen DI circuit?

this approach seems really what you need:

"I'm planning a JFET input stage, still not sure if IC or discrete, zener protection diodes on inputs and 4004 on outputs, and a balancing line driver like the THAT1646."

 

[RFSiesta]

Why the Bo hansen DI circuit?

this approach seems really what you need:

"I'm planning a JFET input stage, still not sure if IC or discrete, zener protection diodes on inputs and 4004 on outputs, and a balancing line driver like the THAT1646."

I was intrigued by the symmetry in the circuit. If I don't misunderstand, the two transistors work in a way that inherently compensates for nonlinearities/drift... Don't know if I'm completely wrong with this.
I don't need extremely high Z on the inputs, but maybe the jfet still is a better option.
But what are the benefits of using an opamp? Isn't it sufficient to take a jfet in, say, source follower configuration, just for impedance conversion?
One more option would be to include a hi Z switch to be able to use the board as guitar/bass input as well.

Btw.: in SPICE, the b.h. DI works flawless at 15V as well, bias voltage can be corrected by only changing the divider network.

Thank you for the inputs - I am merely a studio owner and sound tech, not an engineer
I'm OK with drawing pcbs, tried some basic opamp based designs and PSU stuff. I'm not at all experienced with transistor circuits...
Have a nice weekend!
Richard

 

[Chris_V]

But what are the benefits of using an opamp? Isn't it sufficient to take a jfet in, say, source follower configuration, just for impedance conversion?

A simple JFET follower is far from being as linear as a decent opamp follower especially driven with a "line level".
You can add gain very simply with an opamp so you can drive your long lines with an higher level than the one given by the synths.
I see no benefits of using an jfet, Bo Hansen DI (which is designed to be phantom powered) or other discrete circuit, unless you want to add a special color to your synth sound.

 

[RFSiesta]

Ok, you got me convinced.
I guess I thought about "discrete" because I wanted to learn how it is done and be able to experiment
And I kind of disregarded the level difference between line and instrument out, because many keyboards are somewhere inbetween in reality. That said, most nowadays products ARE almost line level...
I actually wanted to avoid coloration, as we have enough couloured vintage stuff - though I like the idea of "cleanish with musicality" that I sometimes miss from typical opamp devices. But then again, I found both to be the case with the INA217 instrumentation amplifier even in the most simple setup.
Ok, the BH DI is designed not only for phantom power, but for an output transformer, too - and I don't have months for balancing to get it to sound good while dramatically changing important variables...

So that's what it will be for now:
An OPA134 as +6dB noninverting amplifier with in/out protection and maybe an adjustable voltage divider in the feedback for gain adjustment, followed by a 1646 balancing line driver.
thank you!

Soon be back with eagle layout

 

[Whoops]

So that's what it will be for now:
An OPA134 as +6dB noninverting amplifier with in/out protection and maybe an adjustable voltage divider in the feedback for gain adjustment, followed by a 1646 balancing line driver.
thank you!

That sounds a pretty good plan for your needs

 

[JohnRoberts]

If you want something transparent just put an opamp in non inverting configuration in front of the 1646, you will already have the needed power supply and it will be the simplest solution.

You might as well make the input buffer differential to reference the keyboard ground to the bump box ground (I mean audio 0V).

Most modern keyboards use op amp outputs so can drive 10K easily. Only the old Rhodes needed a proper preamp or DI. I don't know all the THAT chips by heart but IIRC they like being driven from low Z, so the op amp input buffer makes sense. .

If your studio is passing audio between different levels and rooms with different mains power drops I hope none of your gear has pin 1 problems. 

I guess you'll find out.

JR

 

[RFSiesta]

You might as well make the input buffer differential to reference the keyboard ground to the bump box ground (I mean audio 0V).

This might be a good idea! We have been discussing issues with synths that are on battery power, like the Korg monotribe or even laptops without the psu connected. We need a reference to ground for these, but that way will get into trouble connecting it over the patchbay system either with missing ground reference for battery devices or potential ground loops coming from gnd differences on power connected keyboards. Don't remember exactly, but there was no perfect solution for both - except a transformer in every line plus an optional ground lift button. I thought to adress this by grounding from the balancer input instead of the line/mic amplifier in the other control room, but that only solves part of the problem.

Most modern keyboards use op amp outputs so can drive 10K easily. Only the old Rhodes needed a proper preamp or DI. I don't know all the THAT chips by heart but IIRC they like being driven from low Z, so the op amp input buffer makes sense. .

THAT1646 has a typical input Z of 5k, according to the datasheet. And I've also read that it does prefer low source impedance anyways. We wanted to use the buffers for some vintage gear like tape echo or germanium stuff like the Telefunken echo mixer (for spring reverb as well, so maybe even switchable input Z for 600Ω termination may be appropriate. The Telefunken is humming btw., but we haven't refurbished it yet - it's from the 50s and I assume it could be because it's never even been recapped...
As I've heard there's other old synths that can have limited output current, but we haven't time to measure everything we see when people bring their old keys for sessions. AND we have a Fender Rhodes, but we would still use regular transformer driven DI like our Radial for this.
Things maybe get too complicated if we seek a solution for everything in one box...

If your studio is passing audio between different levels and rooms with different mains power drops I hope none of your gear has pin 1 problems. 

I guess you'll find out.

I've already build the Jensen "Hummer" and tried on some stuff, but didn't find anything yet. (knock knock) But chances are we have such gear or will have someone bringing in a box with the pin1 problem. Would a transformer help this?
Btw - for splitting stuff to the different rooms, we have a block of 10 Neumann V442 splitters and 10 nice clones, hardwired to our central (lemo triax^^) patchbay. And 5 spare channels as mobile utilities. They sound very, very musical to my ears. We wanted 16 of their channels to split sub-monitor signals that come from our monitor submixing thingy to be able to do separate monitor-/roughmixes in any room... So there's only four channels left plus the 5 channel box.
Should I maybe even turn it all around and use the neumanns for keyboards and my diy crap for the monitoring?
Then again, the V442 DOES colour the signal, I know that it's not always what I want to hear - f.ex with the AKAI mpc. It tends to flatten transients a bit, especially with line level signals I think. (Would make sense)

Hmm.. Much more enlightened and at the same time even more confused 
R

 

[RFSiesta]

Good morning!
After a night of dreaming weird stuff about opamps and balanced cables, this seems to be a working solution (sorry for the tiny sketch). A differential input with one opamp, including a switchable cold/gnd connection and two gain options. The gain topology is proposed by Douglas Self and enables me to choose between -6 and +6 in the first stage, adding to a 0 or +12dB gain together with the +6 of the THAT line driver.

This solves a couple of problems at once:
1) ground lift is now possible
2) I can use the same patch for balanced OR unbalanced equipment if gnd is lifted
3) I can use balanced cables for unbalanced signal, using the "cold" wire for signal gnd and the screen connected to only one side (supposed to increase shield efficiancy)

Still a question:
What happens if I intend to even lower input Z for vintage impedance matching, like with a 600Ω shunt? (cold stays connected to ground, of course) Still not sure if there's any advantage in this... seems some old tube outputs don't like to see Z>1k or so.

cheers, R

 

[JohnRoberts]

Your gain switch will diminish the common mode rejection of the differential amp.

While local audio 0V and chassis ground will ultimately get connected, the details of how and where matters.

JR

 

[RFSiesta]

Your gain switch will diminish the common mode rejection of the differential amp.

While local audio 0V and chassis ground will ultimately get connected, the details of how and where matters.

JR

This is not the case, according to douglas self: http://www.douglas-self.com/ampins/balanced/balanced.htm
The principle is simple, and explained - better by the author than I possibly could with my mediocre english - in the text around figure 10...

I will put some thought into the place where audio common and ground join. I guess sinking currents from the cable screen should not be done over pcb ground, even if it carries signal. Some voltage from the interference would drop over the ground plane and its connections, making the reference ground "dirty" for the reference pin of the line driver. Will it be a good idea to connect where ground enters the chassis? If understood correctly, this is somehow common practice.

 

[JohnRoberts]

Your gain switch will diminish the common mode rejection of the differential amp.

While local audio 0V and chassis ground will ultimately get connected, the details of how and where matters.

JR

This is not the case, according to douglas self: http://www.douglas-self.com/ampins/balanced/balanced.htm
The principle is simple, and explained - better by the author than I possibly could with my mediocre english - in the text around figure 10...

I will put some thought into the place where audio common and ground join. I guess sinking currents from the cable screen should not be done over pcb ground, even if it carries signal. Some voltage from the interference would drop over the ground plane and its connections, making the reference ground "dirty" for the reference pin of the line driver. Will it be a good idea to connect where ground enters the chassis? If understood correctly, this is somehow common practice.

Well I guess I can't say I never read anything from Douglas Self.  (I had seen the figure 11, 12, and 13 used in designs before and the attenuated feedback to make gain, but only in the middle of an audio path not in the very input stage).

That looks like it will deliver adequate CM rejection with both inputs driven from low Z voltage sources.  For your application it should be fine (differential input).

I've always used two op amps or more to make balanced inputs, but don't think you need to get that fancy for a reasonably short path..

JR

 

[RFSiesta]

That looks like it will deliver adequate CM rejection with both inputs driven from low Z voltage sources.  For your application it should be fine (differential input).

I've always used two op amps or more to make balanced inputs, but don't think you need to get that fancy for a reasonably short path..

Tough choice. I think it would only be a problem with a combination of hi output impedance AND balanced lines, and I don't expect that scenario in a keyboard "buffer" amp.
With balanced keys I actually wouldn't need the device..
3pin input is merely to stay flexible with the patchbay - it will be in the "for rental"-control Room B, so MUST be bulletproof for even the most unexperienced user.

Maybe I should go further and really use two opamps. Cost is not really a factor with 10*2€ or so even for hi quality chips.

Btw, how would an instrumentation amplifier like the INA217 perform in this "almost" unity gain? I happen to have a couple of them lying around for my mic pre project

Pro: ANY input Z is possible, gain goes up to 70 dB - so this could be used even as guitar DI, I guess.
Contra: Gain <6dB would need attenuation...

 

[RFSiesta]

I'm building a layout and prototype at the moment, for the one-chip differential input as shown, followed by the balanced line driver. Star grounding, jumper for two different gains. three ground planes/stripes for input, output and circuit ground. 5 on one pcb, two cards in one housing, back to back plus a shield plate inbetween. External PSU over screened cable. Hope my Reichelt order is coming soon, still missing couple of parts...

 

[JohnRoberts]

The sources (single legged keyboards) and path length (modest distance on input side) suggest that a single op amp differential input should be adequate. 

The simple non-inverting differential will sink the same current into the + input as returns through the - lead, so there is (should be) zero net current flowing between the two chassis grounds (a good thing).

Perhaps build one and see how it works?

JR

 

[RFSiesta]

Yes! Got a first layout. I'll do some testing with one or two channels on the board before completely populationg it.
Thank you very much for the help and insights - very curious how it turns out