Direct output from Neve channel switching unit

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
atavacron said:
Abbey, I wouldn't know how to set up the 1646 for more than 6dB gain. It seems fixed, at least according to the documentation. Is there some trick?
  You need to add another opamp with 4dB gain. That's all.

The Tech Info Sheets are pretty clear on not running an opamp at less than 10dB gain...
  What? Many opamps are unity-gain stable.
 
abbey road d enfer said:
  You need to add another opamp with 4dB gain. That's all.

Too many stages I think.

abbey road d enfer said:
  What? Many opamps are unity-gain stable.

Not the 340 / 440 /640. Unsure why though, just that the 406 was the only one approved as a buffer. These are all non-inverting, and unable to take a diff input for reasons that I’m unable to grok.
 
Most of the early Neve amps are not unity gain stable because, unlike most op amps, they don't take a sledge hammer to stability by adding a pole at a few Hz. This means you don;t get the 20dB/decade rise in distortion that op amps give you. Maybe that is part of why they sound so good.

If yo really want to retain the transformer them use a BA283 with an appropriate pad in front of it,

Cheers

Ian
 
Winston O'Boogie said:
Won't you have to use the inductor too since your trans. are ungapped?  What about a 440 style amp or that BBC specific one?  Someone makes a version you can buy I'm sure.

That’s the issue, they are ungapped and nobody’s even making a PCB for a 340, 440, or 512. The Eisen 512s are really cool but I’d have to rail split. I wish a board and heatsink existed for the original 512 design, it would be the perfect thing. I actually have one that could be reverse engineered but I’m not excited about taking it apart.

There are many solutions involving things that run on bipolar 15-18V, but it would be really nice to leave it single 24V. I’m quite a bit more up to speed on this than when I started the thread. In the land of “affordable,” and mucking about with power supply not originally spec’d, it’s basically the the CAPI BB2521, or the JLM Hybrid.
 
Winston O'Boogie said:
Would it have to be the same footprint of the 512 or 440?  Or would you be happy with, say 8 amps on one board? 
There's also the Neve 640 BBC amp which is just a 5534 and a couple of transistors.

Edit,. I think you said you want this inside each module so 8 amps on one board is out.  Why not just a 440 on a little PCB that is just secured inside and hard wired?

Good questions.

I believe the appropriate placement for these is as close to the 33752 modules as possible. The “direct output” wiring runs from there to another area of the desk, where the LO2567s are clustered without convenient power. So to accommodate both existing  wiring and impedance rules, the ideal solution is a 420mm wide board that mounts up under the channel switching units.

The 640 is not really worth cloning if one is going to the trouble of a custom solution. The 440 would be great. For those who are in the know, the 512 is perfect. A 420mm board would have the opamp layout x 12, with resistors and capacitors according to the BA489 spec onboard. All in one solution.

Then that same layout could work for folks who are not interested in paying $150 a pop for untested vintage 440s on eBay and Reverb. What a rip.
 
BA440

https://drive.google.com/open?id=1LdvKK5EKboJ7RZkHSXE0xIPrwYEUy8zQ

BA512

https://drive.google.com/open?id=12E6kBQZzdEWJMiU26TTVRA0xzoX4-01E

I’m out of town but I can take pics of a 440 and a 512 when I get back. There’s no trace layout in either doc but I could put ‘em in a scanner and do something in Photoshop  to make ‘em usable. I doubt anyone would have a problem with a faithful layout and part substitution only to accommodate an easy Mouser  BOM. I don’t believe the part equivalents are terribly unusual.

I do need to scan the BA489  schematic and call out the correct supporting components for each amp. There are some significant value differences across types.

This is all class A/B, there were no Class A output amps in the broadcast desks AFAIK. The best application example of a 340/440 into an LO2567 is of course the 1081, albeit into 50 ohms parallel series. The common application example of a 340/440 into a LO1173 is the 3405 line amp. This desk has 33415 and 33416 line amps, which output with 440s into the TO129, which is like a LO1173 with a tertiary winding for NFB.

3415: https://drive.google.com/open?id=1rIBRKHPKbknvcBvmehG9iuU2oyn3ZYNA

3405 schematic attached.

With a DIY BA440 solution, a 12-channel board could run the amps inverted at a gain of 2 (+6dB) into the transformers at 200:600. It is my understanding from the Tech Info Sheets that they are not stable below 10dB of gain non-inverting, though I suppose one could clone the pad from the 3405 and just eat the extra 4dB. The 33752 would not mind seeing another 5K-10K trimmer on its unbalanced output, as it already has four aux sends at 4k76 and eight groups at something like 8k each that it feeds off of the same, so if these ran non-inverting, one could basically just clone the output of the 3405 and trim to suit. The trim on the 3415 is in the front half of the amp, by the way, so it’s a little less straight ahead. Klett’s online EDO is the best trafo ref.

I don’t know if they are unity gain stable when inverted, but regardless, it feels like a stretch to ask them to do unity gain inverted just to be able to run the trafo parallel/series for its 10dB of free gain — also it would really be a lot of current draw. But they certainly will drive 50 ohms all day!

It really would make my whole dang Summer to see an appropriate amp birthed as a result of this topic, and I’ll bet a lot of other folks would dig it too. I am supremely lucky to have a spare power supply with this desk...
 

Attachments

  • C624ABAB-18E9-4C97-900C-9FB512AF0693.jpeg
    C624ABAB-18E9-4C97-900C-9FB512AF0693.jpeg
    241.7 KB · Views: 50
I'm gonna attempt to have an intelligent conversation about the BA440 vs the BA512 and potential directions for a reproduction, but I'll be missing some key terminology due to not knowing as much about how opamps work.  (What's CCS?) For reference, this console's setup level is +4dBu. Everything on the patchbay is that...except for these -2dBu weirdos.

The BA512 is spec'd a bit more carefully than the 440, 7 to 17dB for lowest distortion (I would assume to please the client), whereas the BA440 is spec'd generally up to 20dB. The BA512's .1uF output load spec is interesting, I wonder what the capacitance of 1 meter of Canford FST going into the series primary of an LO2567 would be. Probably way lower.

Do I need to do a listen test? I kinda feel like taking the last opamp of the broadcast series and improving it slightly for DIY is probably the best call. And with people nowadays having plenty of BA283/LO1166 options, I'm gonna guess that the BA512 would be considered a slightly more hifi / snappier transient starting point for line amps and low-channel-count summing amps and other output sections. Might also appeal to people who would otherwise default to ICs, but do need to drive a transformer.

On the other hand, doesn't the BA512 split the rail, and isn't that what I'm trying to avoid?

If either amp will let me do 7dB of gain non-inverting, I would run it that way with a little input trim, or just ignore the extra 1dB.  I do wonder if the BA440 is as stable as the BA512 at that low level. The Tech Info Sheets list the "not less than 10dB gain" rule for all opamps other than the BA406 in a very prominent list of Don'ts, but this was before the BA512 came along (with its diff input as you mention).

I think a trimmer in place of R16 would be a good call...a side adjust 3/8" or smaller, so you could get to it without taking off the heatsink. 2K 25-turn I presume, if the goal is always in the neighborhood of 1K. Or are you saying wholesale import the 5K arrangement from the 340? Seems more reliable to just change one resistor than two at a time.

Why not just integrate the supporting parts from the 1081 or 3405 into the board itself? It would make everyone's life easier (except for folks who need a drop in replacement). In this arrangement, I'd put decoupling and smoothing below or above the core amp layout, to keep the total width less than the broadcast module size of 35mm. 24V, 0V, IN+, IN-, OUT could all just be on a .100" header at one edge, or separate the power and the output by .100 off the inputs.

Last but not least, maybe there's a way to lay this thing out so that some beefy clip-on heatsinks could be used (see for example how the 33752 mimics the BA640 but without the top plate). Having to have a custom plate milled just adds cost...they do get hot but not crazy hot.


 
Righto. Well I don’t know the BA512 well but I do know the 3405 and 33415 schematics, so here is an idea specifically for the BA440, attached.

Having the pads for the trimmable pad, the feedback resistor/cap, and the gain resistor labeled generically would be smart. I believe this would work with any step up transformer. The primary shunt might want to go on the board though.

The more I am thinking about this, the more I am thinking BA440. It’s better supported in Neve literature and is more native to the systems people would be aping (including this desk, which is all 3114 / 3415 / 3416). The one BA512 I have came from a 33422, which is from that weird era where Marinair was ditched for Belclere but the module size remained 35mm.
 

Attachments

  • A3C0C520-C9DB-4B6C-9A39-0FD44C44C8C8.jpeg
    A3C0C520-C9DB-4B6C-9A39-0FD44C44C8C8.jpeg
    473.1 KB · Views: 47
Winston O'Boogie said:
Won't you have to use the inductor too since your trans. are ungapped?  What about a 440 style amp or that BBC specific one?  Someone makes a version you can buy I'm sure.

Yes you would which seems rather wasteful of iron especially as the inductor need to be gapped anyway. Simpler to spring for the gapped Carnhill transformers and be done with.

Cheers

ian
 
atavacron said:
Righto. Well I don’t know the BA512 well but I do know the 3405 and 33415 schematics, so here is an idea specifically for the BA440, attached.

You don't need the 150 ohms across the transformer primary do you?

Cheers

Ian
 
I like where this is going.

I had been wondering about that primary resistor, and why it was different in different schematics. Great to know!

All the outputs and inserts on this patchbay have 1K2 across the normals, which are out of circuit when feeding anything in addition to their normal loads in the next row.

What is the purpose of the 47p cap? It is a 470uF when it comes from the tertiary of the TO129. Would it need to change based on the value of the NFB resistor?

The safety value with parallel pads for R16 sounds good, as does the LED trick. The series resistor on the output would be more Neve, the inductor and resistor more Hardy I guess.

Making that output really flexible is a good call. Maybe someone’s got eight transformers in a box wired in to their patchbay or whatever, and they want to run the output of these amps to hi/lo and be able to patch to different transformers or something. That would require a robust approach. The instructions would read to jumper that if everything was in the same box.

Here is a version with labeling for board layout, attached.

 

Attachments

  • B328976D-A664-43F9-8D99-2DFBC90FF544.jpeg
    B328976D-A664-43F9-8D99-2DFBC90FF544.jpeg
    1.8 MB · Views: 38
I have no experience with laying out traces. I can draw size, positions, rough pad placement, I/O, silkscreen. Illustrator. DXF. Same for a schematic.

What software are you using? Maybe I will learn Eagle someday.

Uh, yeah of course you’d get a free board out of it. I’m in California. I’ll do a quick dimensioned drawing for a starting point.
 
Lemme get a clearer pic of the messy drawing?

Here’s my version with a layout, 35x70 is tight but possible. The top left and bottom right are mounting holes, the pad stands up next to the trimmer, and the caps are BC 138.

Transformers are remotely mounted (1-1.5m of FST) with plenty of space around them - I don’t think they’re close enough to interact. I will ensure leads are telescoped.
 

Attachments

  • 869568D0-F6A3-49DC-A8C6-D3614A9CFC39.jpeg
    869568D0-F6A3-49DC-A8C6-D3614A9CFC39.jpeg
    480.3 KB · Views: 47
I like that you have a solid idea for my specific application and want to see it through. Thank you so much. There are some things to bear in mind to minimize revisions.

While a perfect accommodation of my application is welcome, I want to be clear that the supporting circuitry has to work for other applications, as a standalone gain stage. This needs to be usable with not only the LO2567/LO1166 and modern knockoffs, but also the Carnhill and Sowter LO1173 equivalents, and even the API 2503 and 2623 and their equivalents. Basically any step-up between 4dB and 10dB gain whose secondary is set up for 600Ω. As I mentioned above, the values of components that have a user option associated should be indicated by their function and not their value (these being Rfext., Rg or RVg, the series output resistor, and Rp1/Rp2/RVp should they be a good choice)

The 150µF on the inverting input is a size advantage over the 470µF, so I'm glad that's possible. No BA489 version uses a 1000µF smoothing cap, and because of its size (12.5x30 as opposed to 10x25), I'd like to understand if it is completely necessary in the 1081 version. Also I am curious how running series/series will change the output choices as opposed to the 1081 parallel/series.

Does it change the operation of the amp to bring back NFB into the inverting input at the node between the fixed gain resistor and the blocking cap (1081), versus bringin it back in to the node between the blocking cap and the inverting input? Regardless of values, this won't be a user-selectable option.

I don't think we need a separate set of gain boost pads in parallel with Rg, because its value will be set by the user prior to assembly. It may however be smart to give Rg additional pads for a trimmer (RVg), which potentially could negate the need for a pad. A trimmer in this position would need a range fix in series.

I see the advantages of a slugged rotary fader. Are you thinking of putting this onboard as a trimmer, and providing hookup points for folks to offboard it if they wish (i.e. a pin at the non-inverting input as opposed to just a solder pad)?

Any difference in function within the 10Ω-47Ω range of the series output resistor, or is that just a way of saying "anything within this range will do"?

Maybe more detailed thoughts than necessary in advance of any actual layout, but it'll be nice to be on the same page. Not trying to look a gift horse in the mouth. Again, I really appreciate the interest.
 
I don't see the need for such a big cap in the C2 position. With a typical value a a few kiloohms for Rg, the -3dB point is <0.1Hz. Considering the LF response is dominated by the xfmr, I would set this cap at about 10-22uF. Will save some space, at the benefit of the output or decouping cap.
You don't want the output stage striving to drive the xfmr into something it can't handle anyway.
 
Winston O'Boogie said:
Although I'd still want to use the tantalum 10uF cap on the +ve input to retain that particularly Nevesque distortion 😀
Considering the very low level across this cap, I very much doubt there's any resulting distortion.

I'm also thinking that since the amp is now not driving 50 ohms but, at the lowest, 200 ohms, the feedback network can be reduced in impedance to lower noise.
Make sense or you think I'm off base?  It was good enough for Rupert but that doesn't mean we can't better it
Since the input transistor runs at about 200uA, the OSI is likely to be about 1-2kohms. Indeed, decreasing the NFB elements by a factor 2 could reduce noise, although so tinily it may not be perceptible. Anything below that will meet with the law of diminishing returns.

Edit:. Also, I don't think these spots on the 1081 were the old Philips axials, I think they were better types that were radial.  Which would suit me as the Philips BC series have quite high ESR.  I'd use a better radial and save board space myself.
Just use the Philips for B+ filtering etc.
Current 'lytics are so much better than they were then, almost any decent type will be an improvement.
 

Latest posts

Back
Top