Balanced output w/ MC33079

[Bo Deadly]

Where does the problem hide that makes it difficult to implement (on existing circuits)?

It's easy to hack an existing unbalanced output to be impedance balanced by simply adding resistors to the output (or frequently only one on pin 3). That ground sensing circuit requires enough parts that you would probably need to make a little PCB to integrate it into an existing device in which case you might as well just go whole-hog and do a THAT1646.

 

[Script]

Space confinement. Oh, yes, I see.

 

[Zen Zeven]

I was planning to use a single TL074 chip for both a insert send and return before I stumbled upon this thread. I would like to be able to insert unbalanced gear when so needed (without any tascam problem).

The insert card I'm designing is a miniature daughter board attachable via headers to a larger board so it can be swapped out if needed in a later stage.

I was for ones stoked to use TL074 mainly for their low power consumption but also for the comfortable price of them. The third reason is that I was planning to swap them out for the pin compatible OPA4134's later when I can afford it but OPA4134's are 5€ a piece and for that you get about nine Tl074's, so that will have to do for now

Would be nice to find a circuit where both Tl074 and OPA4134 could be used in a balanced or unbalanced setup. Could someone add a schematic?

 

[abbey road d enfer]

Would be nice to find a circuit where both Tl074 and OPA4134 could be used in a balanced or unbalanced setup. Could someone add a schematic?

TL074 and OPA4134 are interchangeable without any precaution, except that power rail decoupling must be done adequately.
May I suggest you opt for duals instead of quads. The real estate gain is hampered by the increased difficulty of routing. You have more options with duals. You could use 5532/4560 for the output driver and TL072/OPA2134 for the receiver and buffer.

 

[Zen Zeven]

TL074 and OPA4134 are interchangeable without any precaution, except that power rail decoupling must be done adequately.
May I suggest you opt for duals instead of quads. The real estate gain is hampered by the increased difficulty of routing. You have more options with duals. You could use 5532/4560 for the output driver and TL072/OPA2134 for the receiver and buffer.

For decoupling I was going to add 100n ceramic's for both the positive and negative rails, one for each to ground. This is how I treat 5532's in this same project. How would you suggest otherwise?

I'm using 5532's mainly for the channel strips, like in the balanced line input stage, eq, fader, panpots and also for the summing. I have been calculating the power consumption over and over in different configurations to get my 5A of +/-18V to be enough for as many channels as possible with all the functions that I need. Using only 5532's I could not get the ch count that I wanted. I got stocked when I realised that I could use TL0xx for ch Inserts and aux sends to get the ch consumption down to be acceptable in exchange for some performance, slew rate etc..
I have not used RC4560 (only RC4558) but they seem both affordable and only draw typ 4.3mA and max 5.7mA, which is about the half of 5532 so this is a good suggestion and I might try them for my aux sends as well that comes after these boards. How would you best decouple the power rails for 4560 or is it like with 5532's that I described earlier, 100n for each rail and the other side of the caps are connected to ground?

Quad and dual opa xx34 cost about the same at least where I checked which makes the 2134's to expensive for my project, except in the group and the master sections. Today I drew my first pcb with smd's and I found the routing tolerable around the SO-14 sized TL074, no problem thus far. Was the routing around the quads your only concern or did you consider something else? I'm all ears.

I can order some 4560's and try them out but the question still remains, what kind of balanced circuit could I use that will work with unbalanced signals as well?
(I just realised that this will also be important when I design the aux master sends)
I will add the return receiver schem that goes with the original send schem from this post and that I vas going to use

 

[abbey road d enfer]

For decoupling I was going to add 100n ceramic's for both the positive and negative rails, one for each to ground. This is how I treat 5532's in this same project. How would you suggest otherwise?

No, it's good. Make sure you use ceramic caps and that you have a pair of electrolytics (about 100uF)

Today I drew my first pcb with smd's and I found the routing tolerable around the SO-14 sized TL074, no problem thus far. Was the routing around the quads your only concern or did you consider something else? I'm all ears.

I seldom dab with SMD because my eyes and hands are not so good now.

I can order some 4560's and try them out but the question still remains, what kind of balanced circuit could I use that will work with unbalanced signals as well?

You have a choice.
You can use the standard cross-coupled EBOS - beware it needs to be driven by a low impedance, typically the output of an opamp. And it needs some final adjustment.
You may want to use an impedance balanced stage. Big advantage, it takes only one opamp for the output. So with a dual you can have the balanced receiver and the impedance -balanced output.
You could try a ground-sensing output, which is an extension of teh impedance-balanced thing. I don't think it's really worth it.
Me, I would choose the impedance-balanced output.

 

[Zen Zeven]

You have a choice.
You can use the standard cross-coupled EBOS

Thanks abbey for your input. The EBOS looks like what I'm looking for. It also looks alot like Douglas Selfs Quasi-Floating Balanced Output (schem attached) that I'm familiar with and Heikki in this forum (moro) used a similar balanced output stage for his tube compressor project if I remember correctly. I use it for the master outputs and it was my plan for the ch inserts at first. I didn't want to use NE5532s for the current draw nor three times green nichicon muse per insert, so I started to look at different options and now I'm basically back where I started, but better The EBOS seem perfect for my needs and are without the n/p electrolytics so I don't upset the ch-matamatics. If I use the 4560's I get the power consumption I'm aiming for. I like the sound of the 4558's and I just read in the datasheet of the 4560 that they are basically upgraded versions of the 4558. I have build a spring tank drive and receiving amp with them and also they are all over my nutrom pi-bhase dual phaser clone, so I kind of know them, at least a little. Glad we had this conversation. How ever I don't see any value for C10 or C11 for the EBOS any idea?

you wrote
"beware it needs to be driven by a low impedance, typically the output of an opamp"
Good to know and that sounds reasonable. I intended to take the signal from the EQ output stage that has is an opamp based output and also make the insert switchable pre or post EQ. In pre EQ mode the line input amplifier(schem attached) with low impedance output would be driving the insert. The impedance of the eq output I have to look up and I intend to have different EQ models so they will probably have different output impedance so that is something I will have to solve in the future, I might come back to this topic in a few months when I come to building the parametric equalizers if that is ok

you wrote
"I seldom dab with SMD because my eyes and hands are not so good now"

I know what you mean I have stayed away from smd until now but felt with the little space on the main board for the daughterboard I might give it a try, let's see how it goes SO-14 is still only half the leg pitch from DIP packages so with some flux I hope I'll be fine with my pine64 iron, magnifying glass and some IPA.

 

[abbey road d enfer]

The EBOS looks like what I'm looking for. It also looks alot like Douglas Selfs Quasi-Floating Balanced Output

Of course. D. Self worked at Soundcraft for many years. They're the same circuit.
You must consider the fact that the EBOS requires final adjustment for best performance. The monolithic versions of it are laser-tuned. IMO, there are not enough advantages to overcome the difficulties of implementation.
The main advantage of EBOS is that it gives good headroom specs, but it's moot when driving unbalanced inputs.

I didn't want to use ... three times green nichicon muse per insert, so I started to look at different options and now I'm basically back where I started, but better The EBOS seem perfect for my needs and are without the n/p electrolytics so I don't upset the ch-matamatics.

If the circuit is to be connected to an input that has no DC, there is no obligation using NP caps.

"beware it needs to be driven by a low impedance, typically the output of an opamp"
Good to know and that sounds reasonable. I intended to take the signal from the EQ output stage that has is an opamp based output and also make the insert switchable pre or post EQ. In pre EQ mode the line input amplifier(schem attached) with low impedance output would be driving the insert. The impedance of the eq output I have to look up and I intend to have different EQ models so they will probably have different output impedance so that is something I will have to solve in the future,

In order to avoid any disappointment or regrets, you should have an input buffer. It could even be a differential receiver so you could connect it to any type of source.
You just need one opamp for the line receiver, so the opamp count would be 3 for the line driver + one for the receiver.

 

[Zen Zeven]

No, it's good. Make sure you use ceramic caps and that you have a pair of electrolytics (about 100uF)

Do you mean the electrolytics are in paralell to the ceramics? Also what voltage rating should they have 25V for my dual 18V rails or more like 50V?

You must consider the fact that the EBOS requires final adjustment for best performance. The monolithic versions of it are laser-tuned.

Tuned with the RV2 trimpot or what kind of tuning are we talking about? I have DMM's and an digital oscilloscope it's an RIGOL DS1054Z and REW on my mac. Can it be set with these tools. What would the process look like for adjusting this topology?

IMO, there are not enough advantages to overcome the difficulties of implementation.
The main advantage of EBOS is that it gives good headroom specs, but it's moot when driving unbalanced inputs.

In order to avoid any disappointment or regrets, you should have an input buffer.

..you still recommended this circuit so I assume that with the buffer added it should be a viable solution for my insert cards

so the opamp count would be 3 for the line driver + one for the receiver.

yes this sounds good using two dual opamps then. I have re-thinked the smd approach and will try to make the insert cards bigger instead. Now with two dip8's, three 5mm cans and about fifteen through hole resistors and the headers, this will be a challenge but will simply have to make more room some how from somewhere.

How hot will the 4560's get with 18V supply voltage on them, the 5532 are warm but not to hot even though thats their reputation

 

[Zen Zeven]

EBOS requires final adjustment for best performance

now I remember, when I build my proto master outputs with the quasi-floating circuit I matched the resistors really close and then I tuned the trimpot so that I had the same resistance on both sides. Would be good to update the procedure if possible now that I own a scope

 

[Newmarket]

Of course. D. Self worked at Soundcraft for many years. They're the same circuit.

Yes. And Dave Dearden of DDA also ex-Soundcraft. Similar EBOS module on DDA desks and also Ground Sending outputs in some cases for other Sends.

 

[abbey road d enfer]

Do you mean the electrolytics are in paralell to the ceramics?

You only need one pair of lytics; they don't need to be mounted vey close to the opamps.

Also what voltage rating should they have 25V for my dual 18V rails or more like 50V?

25V should be OK, though I now use 35V, since they are the same size, due to technology improvement.

Tuned with the RV2 trimpot or what kind of tuning are we talking about? I have DMM's and an digital oscilloscope it's an RIGOL DS1054Z and REW on my mac. Can it be set with these tools. What would the process look like for adjusting this topology?

The common method consists in joining the two legs via 10k resistors and nulling the resultant voltage, however it is not the best. The correct method consists in measuring CMRR, which involves the use of a good AC analyzer and a test rig. The method is described in the AES standards. I think it can be found on the Jensen site.

..you still recommended this circuit so I assume that with the buffer added it should be a viable solution for my insert cards

Yes and no. Don't neglect the caveats about stability and tuning.
My fave is still the impedance-balanced circuit.

How hot will the 4560's get with 18V supply voltage on them, the 5532 are warm but not to hot even though thats their reputation

The 4560 dissipates about half of the 5532, so it gets moderately warm.

 

[Zen Zeven]

The common method consists in joining the two legs via 10k resistors and nulling the resultant voltage, however it is not the best. The correct method consists in measuring CMRR, which involves the use of a good AC analyzer and a test rig. The method is described in the AES standards. I think it can be found on the Jensen site.

Yes and no. Don't neglect the caveats about stability and tuning.

oh no not the CMRR test lol. I have read about it before and now I can recall a bit and it's a bit of a hastle isn't it. Reminds me that my current setup needs to be tuned at some point, not my favorite thought at all

My fave is still the impedance-balanced circuit.

I must admit that I confused the EBOS and the impedance-balanced circuit as the same earlier while reading your first comment about it. Could you show a schem so I can skip the confusion part

or is it something like these (attached file)

b) has the unbalanced output could just skip the relay

 

[JohnRoberts]

"Impedance balanced" is very simply using a passive resistor connected to ground (actually 0V), instead of an active negative signal leg. If the source impedance of both + and - outputs are the same, the output is balanced.

JR

 

[abbey road d enfer]

or is it something like these (attached file)

No. These are two versions of the dreaded "Tascam problem".
See attachments for ground sensing and impedance balanced outputs.

 

[Zen Zeven]

You can use the standard cross-coupled EBOS - beware it needs to be driven by a low impedance, typically the output of an opamp. And it needs some final adjustment.
You may want to use an impedance balanced stage. Big advantage, it takes only one opamp for the output. So with a dual you can have the balanced receiver and the impedance -balanced output.

I read through your comment again and I had missed that there was three different topologies in there. I somehow thought that you meant that I needed a driver/buffer before the ebos

So with a dual you can have the balanced receiver and the impedance -balanced output.

in my confusion I read this as needing a pair of ic's so 4 opamps, 1x for the receiver, 1x for the driving stage and two for the ebos. Don't ask me how but I was clearly wrong.

No. These are two versions of the dreaded "Tascam problem".
See attachments for ground sensing and impedance balanced outputs.

Now with the impedance balanced schem added I read the comment again and facepalmed when I saw what you really meant. One 4560 per insert section/card couldn't get better; low power consumption, low cost, low heat dissipation and less space on the pcb. I have most of the components at home for the impedance balanced circuit, just have to build a monster out of three resistors to get the 36k and the biggest caps I have is 220u but will prob be fine for a test board. Thank you ones more for your time and patience. Here is a mockup with the receiving circuit I intend to try out to get the insert section to be used with both bal and unbal signals

 

[abbey road d enfer]

"You can use the standard cross-coupled EBOS - beware it needs to be driven by a low impedance, typically the output of an opamp. And it needs some final adjustment.
You may want to use an impedance balanced stage. Big advantage, it takes only one opamp for the output. So with a dual you can have the balanced receiver and the impedance -balanced output. "
I read through your comment again and I had missed that there was three different topologies in there. I somehow thought that you meant that I needed a driver/buffer before the ebos

No confusion. There were actually several possibilities, and the EBOS, needing to be driven for a very low Z, you'd better implement a buffer because you don't know what source you may have to use.

in my confusion I read this as needing a pair of ic's so 4 opamps, 1x for the receiver, 1x for the driving stage and two for the ebos. Don't ask me how but I was clearly wrong.

Just to make it clear.
A complete bal/debal with EBOS takes a total of 4 opamps, when the impedance-balanced and ground-sensing take only 2.

Now with the impedance balanced schem added I read the comment again and facepalmed when I saw what you really meant. One 4560 per insert section/card couldn't get better; low power consumption, low cost, low heat dissipation and less space on the pcb.

I have most of the components at home for the impedance balanced circuit, just have to build a monster out of three resistors to get the 36k and the biggest caps I have is 220u but will prob be fine for a test board.

The values there are for a specific circuit with a definite gain. For unity gain, it's much simpler.

 

[RogerAF]

https://www.analog.com/media/en/technical-documentation/data-sheets/SSM2142.pdfThis is 3 opamps and 10 30K precision resistors (and a few others). It can drive balanced and/or unbalanced outputs. The secret is the crossed Sense lines which are normally tied to the outputs, but can be used with super long cables, in which case you would use quad core and tie them at the far end. The other secret is the precision matched resistors. TI makes the DRV134 which is still active, but out of stock: https://www.ti.com/lit/ds/symlink/drv134.pdf
And of course there's the THAT 1646. They are all basically the same chip. I was using the SSM2142 in custom gear I used to make for recording studios and sound companies in the 90s. They compared very well against transformer outputs. Nobody ever complained. So you need a precision resistor chip and 2 dual opamps and copy the design in the SSM2142 data sheet. Like Picasso said "...great artists steal."

 

[abbey road d enfer]

This is 3 opamps and 10 30K precision resistors

Precision is the magic token here. The degree of precision achieved by laser-trimming cannot even be approached by a DIYer, even with trimmers.
The limitations of SSM2142 and DRV134 are well-known (An Improved Balanced Floating Output Driver IC – THAT Corporation), although perfectly acceptable in most situations.
The THAT 1646 use a different topology that fix these issues.

 

[RogerAF]

Precision is the magic token here. The degree of precision achieved by laser-trimming cannot even be approached by a DIYer, even with trimmers.

Yes, however I was thinking about array networks. Vishay has thin-film 0.125W and 0.25W arrays at 0.01% tolerance.

The limitations of SSM2142 and DRV134 are well-known (An Improved Balanced Floating Output Driver IC – THAT Corporation), although perfectly acceptable in most situations.
The THAT 1646 use a different topology that fix these issues.

Now THAT I didn't know. ;-) They weren't around when I started back in the 80s.