Douglas Self Line Input

Pretty happy with this little guy so I thought I'd share. 

I used this circuit for a line input on a friend's console I'm recapping/modding. He uses it for mixing only, so no need for a mic preamp.  This circuit works very nicely with a 5532 and a 10K rev log pot.  Very easy to install in our situation, YMMV depending on what the stock circuit was.  Added a couple of (100uF IIRC) caps in front, plus a couple .1uF ceramics from the power rails to ground.

Here's where I got it from:

http://www.douglas-self.com/ampins/balanced/balanced.htm#5

No comparison with the original padded down mic preamp used for line inputs.  Of course, I also recapped the channel we tested, but we'll do the rest of the console in a couple of weeks, that should be enlightening.

Thanks to Mr Self for inventing and sharing. I'm just the messenger.  ;D

thats rather smart, but what about common mode noise rejection? Isn't it based on those input resistors and how well they are balanced?

A1 sees only R3//R1 at its - input, and R2//R4 at its + input, A2 buffer has 0R output resistance

What I wonder is if the pot could become scratchy after a while

That circuit or variants of it have been around for decades. That schematic looks a little sketchy. At min gain A1 is driving 180 ohms to ground, a little low for my taste (and most opamps).  Also any lag(delay) in A2 is added in series with lag(delay) of A1 in series with NF network so the topology affects stability compensation (2x lag). Also for good HF CMR A2 output must be very low impedance. 

I've used a version of this without the variable gain and it is popular with console designers to make a true balanced line input using only two opamps. Most true balanced approaches require 3 opamps and those extra opamps add up tines tens of channels. 

It seems there have been threads about this exact circuit here before.. IIRC Fletcher (Alice mixers, etc) or one of the other brits claimed credit for inventing it.

It might be interesting to see how it behaves with modern uber opamps, I'd still put it on the bench and ring it out before trusting that schematic.

JR

Here's a nearly equal parts count plan as implemented in a commercial product (Studiomaster Trilogy series stereo line input). There are several advantages. You can use pretty much any opamp because it's not fussy and there's no black art done on the feedback loops. And no need to source a rev log pot/fader. Especially for stereo ganging dual 10k logs are found easier. (THD+N is also somewhere around 0.0008% on my test rig, vs. 0.0004% of the plain AD/DA loop)

Kingston said:

Here's a nearly equal parts count plan as implemented in a commercial product (Studiomaster Trilogy series stereo line input). There are several advantages. You can use pretty much any opamp because it's not fussy and there's no black art done on the feedback loops. And no need to source a rev log pot/fader. Especially for stereo ganging dual 10k logs are found easier. (THD+N is also somewhere around 0.0008% on my test rig, vs. 0.0004% of the plain AD/DA loop)

Click to expand...

I've used differential inputs successfully plenty of times but for the record they are not completely balanced for all kinds of input signals. The impedance looking in the + input is well behaved 2x4.7k. The impedance looking in the - input depends on the signal at the + input.  For common mode signals you get better current balance from using equal values in the differential, since - input R is connected to 1/2 the + input voltage due to opamp NF action.

These two approaches are apples and oranges so not really comparable.  FWIW good differential inputs will work for 99.9% of applications. More important to reject out of band noise, and accurately subtract - input from +. 

JR

JohnRoberts said:

These two approaches are apples and oranges so not really comparable.  FWIW good differential inputs will work for 99.9% of applications.

Click to expand...

Apples and oranges? They are both balanced line level input stages with gain trim. I see the same fruit.

To enhance discussion, perhaps it would be a good idea to post that good differential input for 99.9% of applications...

Hey this is great, I only really meant to share this, but I'm all for improving things, especially since the iron isn't going on for another couple of weeks. (only did one input channel, fader buffer, and master). Recap, and "rechip" is that a word? Used some LME49720 or 49860 or something, can't keep those numbers straight. The infamous, it was like a veil was lifted from the eyes/ears comment applies, but like I said, it was also recapped with Elna Silmics in the audio path, and increased-value Panasonic FC's for PS decoupling.

Common mode - For the test channel I hand-matched the resistor pairs as closely as possible, and plan to do the same on the rest of them, but it's also a sort of controlled environment (one guy's mix room, everything fixed in place and working fine for a couple of years already) so I'm not sure CM will be such a problem as in other studios, or live situation of course. The thing basically connects to PT and back with some effects inserted.

As for the scratchy pot, I did add 100uF caps before the 15K resistors which should eliminate any DC coming in to the system, I hadn't considered the offset of the 5532 to be a problem but it's also not set in stone yet what opamp will be used. I just had that and figured it would be the most vanilla opamp I had around, so if it worked, others should (might?). Probably some better choices here, but we need to be power conscious as well as price conscious (it's not my money I'm buying with, so if I go with something a lot more $$ there needs to be a reason).  Maybe a FET input would be better?

Kingston, I've heard only good things about studio master stuff, but I'm intrigued by this design's simplicity, and the fact that I can get it onto the stock PCB with no trace cutting,  just "strategic parts placement" is a huge bonus.  (means it is also fully reversible, although that wouldn't be fun). 

John, for the 180R, I guess raising it would raise minimum gain, which is not a problem per-se, to counter that could you use a 22-25K rev log pot?  Just have to live with a bit more resistor noise, is that it?  As for lag, a higher speed opamp would be of benefit?  There are going to be compromises (power consumption, cost, stability maybe?).  Not sure about you comment re: the output impedance of A2, is the 4K7 the effective impedance?  I'm not sure how to calculate CMR, are we talking huge degradation? 
 
cheers!

In Self book, value for R6 and R5 are reversed, R6 is 2K (R5 is 180R). So less loading when gain is minimum

Your posted differential will work for 99.9%, while I might select the resistors differently.

======

here is a link to an old discussion of the 2 opamp balanced input.

http://www.groupdiy.com/index.php?topic=12586.0

I found a few other discussions of ted's "superbal" but some of the old links to schematics are broken.
------

http://www.tfpro.com/phpbb/viewtopic.php?p=427) 

here's some comments from Ted about priority of his design.


JR

wow, interesting, stuff. HAdn't seen the thread directed to PRR, lots of good stuff there.  But now I'm not sure which way to go. On the one hand, this one works fine, on the other, the superbal is supposed to be supergood.  Hmm, food for thought, uses the same parts pretty much, so we'll see.  I've got a couple weeks to think about it.  Superbal with 33K input resistors and a 10K rev log pot should work?

http://headfonz.rutgers.edu/SuperBal-Var-TedF.jpg

keefaz said:

In Self book, value for R6 and R5 are reversed, R6 is 2K (R5 is 180R). So less loading when gain is minimum

Click to expand...

I knew I saw that somewhere too, but couldn't find it afterwards. In the end it's the same thing isn't it? The two R's are in parallel so either way it'd end up less than 180R.  I think it would only affect the "taper" of the pot before it gets too close to the end of travel.  OR maybe I'm on crack, I definitely had too much coffee today.  4-5 LARGE bialletti espresso pots. 

No, at minimum gain R6 is in parallel with R5 + Pot resistance (10k), so 1.9K
(with R6 = 2k, R5 = 180R, Pot R = 10k)

If you're not on a tight budget consider the instrumentation amp topology. 3 opamps instead of 2, but everything is well behaved with no unusual asymmetrical feedback paths, odd input impedance, etc. 

Or, use an off the shelf solution like THAT chipsets with well matched precision resistors, etc.

The two opamp solutions are cheaper and nice for large consoles when trying to save space and money, but these days the dedicated chip sets might win on performance.

JR

keefaz said:

No, at minimum gain R6 is in parallel with R5 + Pot resistance (10k), so 1.9K
(with R6 = 2k, R5 = 180R, Pot R = 10k)

Click to expand...

you're right, of course.  I was thinking backwards. (It's the coffee, I swear!)

JohnRoberts said:

If you're not on a tight budget consider the instrumentation amp topology. 3 opamps instead of 2, but everything is well behaved with no unusual asymmetrical feedback paths, odd input impedance, etc. 

Or, use an off the shelf solution like THAT chipsets with well matched precision resistors, etc.

The two opamp solutions are cheaper and nice for large consoles when trying to save space and money, but these days the dedicated chip sets might win on performance.

JR

Click to expand...

Well, somewhat of a budget I guess.  31 more channels to do, plus 7 groups (did one already), and a couple auxes, not to mention the PSU, but that should be OK with just a recap.  I ruled out the THAT chips earlier because of the fixed gain. The variable gain of the Self circuit works very nicely as is, should be similar to the superbal with 33K input resistors.  My friend definitely plans to use the gain knob.  Will look into the 3 amp topology, thanks for pointing it out.

For 32x the superbal works too,,, Like I said it is a fairly well known and well used topology. Just be aware of the stability concerns I raised because of extra delay in NFB path.

That's why we prototype and bench test even simple designs.

JR

Stability of the Self or the Superbal? ( I thought I saw you comment about stability in the other thread).  can't be sure and I'm falling asleep (insomnia last night).  I'll be gone for a day or so, but I'll look more into this for sure. 

btw, on the superbal, besides the inputs, do any other resistors need to be matched?

thanks!

Someone ever had a look at the Rolls RM64 input stage ?

Not really, but it looks like the phase would be inverted after that 3rd opamp?

analogguru said:

Someone ever had a look at the Rolls RM64 input stage ?

Click to expand...

It's clever. Almost as well balanced as a 3-opamp instrumentation amplifier with lesser parts count, but not adequate for unity-gain.

However I'm somewhat surprised at the 10k resistors R104 & R107. Since these are mic inputs, they add 20k to the source impedance, thus raising the noise figure by about 20dB.

analogguru said:

Someone ever had a look at the Rolls RM64 input stage ?

Click to expand...

Not really clear what they're trying to do. The 5% phantom power resistors in front of a clearly line level instrumentation front end looks mismatched (is it mic or line input?). The lack of a differential after the input stage means common mode noise does not get cancelled. Two gain pots seems awkward unless one is supposed to be a channel fader.

Putting a pot wiper across the NF path is nice to reduce closed loop gain when potted down, but wiper bounce or a bad pot connection could leave that opamp without NF (I'd add a high value resistor there JIC).

Several things not to copy there. Design doesn't do anything very well (IMO)

JR