Question to PRR (diferential input)

[chrissugar]

Hi PRR and a happy new year to you.

My question is related to the fact that long time ago (one or two years ago) you said that there are better ways to make a diferential input receiver than the classic one or two opamp diferential input with the same amount of components/effort.
Can you elaborate, please.

chrissugar

 

[PRR]

> better ways to make a diferential input receiver than the classic one or two opamp

I bet I was objecting to the common one opamp diff-input. There are several fine 2-opamp plans.

 

[Samuel Groner]

I remember that you once said that with half a dozend transistors it is possible to make a better diff. input than building an opamp and using it with the standard four resistors.

Perhaps chrissugar is referring to that?

I'd be interested in a solution as well as I thought long enough about it and didn't come up with something satisfying.

Samuel

 

[Brian Roth]

[quote author="PRR"]> better ways to make a diferential input receiver than the classic one or two opamp

I bet I was objecting to the common one opamp diff-input. There are several fine 2-opamp plans.[/quote]

I am a BIG fan of the "inverter feeding an inverter" topology. Two opamps, 5 resistors, and easy to trim the gain.

Bri

 

[Samuel Groner]

balfig12.gif
That one? How do you easily change gain here?

Samuel

 

[Brian Roth]

Nope...along these lines:

http://www.brianroth.com/pix/temp/vu-buffer.pdf

Bri

 

[clintrubber]

I've heard more good things about that topology (but it may have been from the same persons :wink: )

This is as also in the 1176 IIRIC, with added bonus the easy gainsetting as you said.
Thanks for sharing. I've been looking at that 1176-input - the only (minor)drawback I could think of is that when going in unbalanced there'll be two opamps in the path - unless there's some easy polarity-correction further along the way.

I figure a 5532 could live here happily too - with added 5k resistors for the non-inverting inputs for DC-offset.
[overengineering] hmm, and why not make that pot a dual one, to let the non-inv. leg resistance track the resistance the inv.-input sees... for keeping the resistances the same (only a first order compensation it'll be ... poor pot-matching) [/overengineering]

 

[TedF]

I think what you may be looking for is a variable gain version of my old 'superbal' circuit.... The basic 2 opamp circuit has been taken up by numerous manufacturers as a stable and well behaved balanced input with good CMR and no real vices. Last year I spotted a way of making it variable gain without degrading the CMR..... I would publish it here but I'm not clever enough to get an image into this post :shock: ..... but if anyone would like the basic circuit, I will return email it. It's done by adding 3 resistors and a pot. :wink:
Ted Fletcher ([email protected])

ps; and if anyone wants to put my sketch up on the forum, that's fine.

 

[fum]

*snip*

see PRR's pic below

 

[PRR]

> That one?

Thanks for mentioning Doug Self's site.

His page Balanced Line Technology section 5 Line Inputs has a good review of the common schemes.

 

[bcarso]

One issue that is usually not terribly important at audio frequencies with the two-inverter approach is the slightly different speed of the two paths. It is possible to do a pretty good first-order R-C compensation for this, if it be an issue.

The two-inverter approach does have the advantage of low common-mode distortion. And it is practically bulletproof against cockpit errors in the heat of battle.

As far as the use of discretes to make a better diff amp, by many measures today that won't be true. However the combination of judiciously selected dual/other multiple parts and other discretes can be quite good, but hardly easy and simple.

 

[PRR]

I did a little cleaning of TedF's drawing and put the design-notes with it:
{EDIT - refreshed the image}

 

[TomWaterman]

Theres some nice stuff on two amp inputs in the Operational Amps stuff by Walt Jung. You can download the whole chapter that covers the same section available in his current book (the one in the shops) from the Meta thread:

Here Page 38 onwards...makes for good reading.

Tom

 

[PRR]

TedF's vari-gain SuperBal plan is fascinating.

I suspect the pot wants to be reverse-audio. Linear gives:

20.33K = 1.5 = 3.5dB at "0"
10.33K = 1.97 = 5.9dB at "5"
0.33K = 31.3 = 29.9dB at "10"

2.4dB change in one half rotation, 24dB change in the other half rotation: we are wasting half the dial. With 10% reverse-audio
20.33K = 1.5 = 3.5dB at "0"
2.33K = 4.8 = 13.6dB at "5"
0.33K = 31.3 = 29.9dB at "10"

10dB in one half, 16dB in the other half, good. A 20% rev-audio taper would be even better.

The values shown are convenient. Are they good for the real world? One hard-case is when someone hands you a way-hot output. Modern outputs are push-pull opamps, can deliver 20V before clipping. Of course levels "should" be 1V to 10V, but stuff happens. With the common unity-gain input, a 20V push-pull signal will clip a 10V max opamp: inputs should have a minimum gain less than 0.5. The values TedF shows above give a minimum gain of 1.5. He also tells us the solution: raise the input resistors as needed. Making them 33K will give minimum gain less than 0.5, so it will swallow any modern output even if the idiot on the other end is flashing clip-clip-clip.

w/ 10%RA and 31K input resistors
20.33K = 1.5 = -6.5dB at "0"
2.33K = 4.8 = 3.6dB at "5"
0.33K = 31.3 = 19.9dB at "10"

Zero dB comes up around "3" on a 0-10 scale, and will be a fairly uncritical setting on a standard 2-segment "taper" pot. When that idiot feeds you 20V levels, you can pull them down to 9V and not clip in your input (if his output is clipping, that's not your problem: you can capture his audio without added injury). If you are calibrated for +4dBu, and take a -10dBV feed, it pots-up about "6" on the knob. If someone hands you an under-recorded cassette, and your only tape player is a 2-AA WalkMan outputting 0.1V, you can pot-up to nominal level.

20K rev-audio is not the most common value. If it were 10K, and keeping TedF's general scaling, then the input resistors could be 15K-17K, the pot-end resistor say 150 or 220 ohms, and all the others 5K or 4K7. It happens that 5K rev-audio pots are commonly used in mike preamps; that leads to ~8K input resistors, ~100 pot-end, and 2K7 everywhere else.

 

[gnd]

PRR, TedF, and all:
I made vari-gain superbal, as input for mixer. I have an oscillation that I cannot get rid of, and would really need some help. Can you please help me troubleshoot it?

R1 and R12 are there for output biasing from -15V.

I tried different values for compensation C's, and found this:
1. C6 needs to be added over R5 and R6.
2. Values for C2 and C6 need to be 15p or 22p in this circuit with these R values.
3. I get oscillation on pin 6 of IC, when signal is present, and cannot get rif of it.
4. C1 is connected to oscillation, but doesn't remove it.

This oscillation is not seen on output, it is just on pin 6. It is 500mVpp, and it is some 650-700kHz (freq depending on what is plugged in on input).

If I put in C1, some 15p or 22p, oscillation becomes dumped, but starts to become visible on output signal instead. So trying to tame down oscillation, I make output signal worse.

Otherwise signal looks nice, 30kHz square looks nice through it, no oscillations in signal itself, smooth rise/fall ramp arround 20% of halfperiod.

Can such oscillation of 700MHz exist within circuit? If not, how do I get rid of it?

thnx
gnd

 

[JohnRoberts]

Some small things to suggest, but posisitve feedback is obvious concern.

Lag/phase shift due to additional delay of invertor IC1B can be problematic. While you don't have enough invertors in series to make a full "bubba" oscillator we need to account for the extra lag. From observation Ted's version is lacking your C6, while that alone may not be the way to stabilize. I have used the non variable version of this many times and it can be quite stable, so focus on everything in IC1B loop.

JR

PS: Small things like, good PS decoupling (to ground), perhaps small C to ground at IN+ and IN- if testing unterminated. I kind of like making R10 and R11 pairs of resistors in series with a small cap between them. This LPFs the input before it hits any active circuitry for good RFI and rejection of above band material..

 

[TedF]

Hmmmm. I have to suggest that adding C6 is asking for trouble!
In commercial circuits I use MC33078 as the amps; they seem more friendly than the 5532, and the extra drive capability is not needed. I have come across this instability and cured it by adding a bit of a 'pole' in IC1A by directly slugging the pos input with a bit of C. I have to admit that it was not very scientific, but it fixed the problem, which was on a particular PC layout with some long tracking around the input; and I put it down to my careless layout.
Regarding the pot value..... I use a dual pot; the second section alters the gain of subsequent amplifiers and the combined gain structure gives me the law I want right across the rotation of the pot.

 

[JohnRoberts]

What Ted said,,, he is the man on this topology which has been widely copied for decades because it works well and uses less parts than 3 opamp intrumentation topology (a cost and PS concern across tens of console input channels).

I think I recall some benefit from unconventional capacitive lead in IC1B loop but was reluctant to mention as I suspect it is dependant on specific opamp package input capacitance, value of feedback network, layout, etc.

:thumb: I second the use of popular bi-fet opamps. The drive capability and slightly lower noise voltage of 553x series is of little benefit here. Bifet is slightly quicker, not that its an issue.

JR

PS: Hi Ted: cheers, glad to see you still in the game...

 

[bcarso]

The dual opamp has obvious appeal, but were you to use two 5534s you could tailor each one's response with the external compensation pins.

 

[JohnRoberts]

[quote author="bcarso"]The dual opamp has obvious appeal, but were you to use two 5534s you could tailor each one's response with the external compensation pins.[/quote]
If one is really adventurous the 5534 run decompensated could exhibit less lag/delay in the positive feedback loop. I have seem them stabilized with very small feedback caps to define stable 3x closed loop gain at HF working against the pFs of input pin package capacitance to ground.

This is pretty tricky design stuff and I would be inclined to go with Ted's advice (bifet).

JR