Following this thread with much interest. We have a Neve 53 series board here. Also a "Green Phone" SABC 54 type board with 4 discrete 34127 channels that we would like to add direct outs to. Definitely into buying some boards.
Direct output from Neve channel switching unit
- Thread starter atavacron
- Start date
Help Support GroupDIY Audio Forum:
{Abbey:] I don't see the need for such a big cap in the C2 position [1081=C35]. 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.
Smart. The 3405 uses a 22µF/16V in this position. Image attached. Still can't figure out how to do inline images.
[Winston:] If the cap is at the inverting input junction (1081), the cap is within the loop. Non linearities of it are then equalised.
Good. That's the external IN– point then. Might want to make the pads to accommodate a 150µF, even if we go with 22µF in this application.
[Winston:] I'm not sure I see this as a group project for widespread use though. Maybe I'm wrong?
This wants to be a 12-channel board at 35mm per channel in a v-groove snap-off form factor, which will make for easier mounting in the desk (though each channel does need to have its own mounting option). If I'm going to have 'em made, I want to provide an inexpensive way for people to use just a few channels for projects.
Also I have a 16-into-4 unit that I'm developing that could use four of these channels – rather than mounting BA489s to proto boards with card edge connectors, as was the previous plan. In that application I will need to voltage sum into the inverting input, so I realize we need to have both an IN+ and IN– on the header that also has the 24V and 0V inputs.
I don't think the power wants to be bussed across the boards, as B– impedance rules are so strictly tied to signal path length, and the central B– reference in these desks is the buss channel. Each channel should be considered as part of the module it is wired to.
In my 16-into-4 application I can jump power across. Others can too.
[Winston:] Although I'd still want to use the tantalum 10uF cap on the +ve input
I'd agree that we want to keep the signal path the same within the amp layout itself, save for necessary substitutions (your transistor sub for instance) and that temperature protection with the LEDs, and whichever bias spreader trim technique you feel is easiest at setup time.
[Winston:] How about If I lay out a single channel and post it here or pm it to you
If you want to go minimal at first, four channels on a board would be smart, as I could use the prototypes for the other project I mentioned above, and would have some spares.
[Winston:] Having a place for another gain boost resistor in parallel was just replicating the function that was brought out to the rear connector on Neve amplifiers. A trimmer is fine too if that's best for you.
Pads for either a single fixed Rg or a 3/8" top adjust trimmer please, in the same board position.
[mjrippe:] Definitely into buying some boards.
That's the idea! DM me with channel count and config. I hope a 12-channel snap-off form factor will work for you too.
Smart. The 3405 uses a 22µF/16V in this position. Image attached. Still can't figure out how to do inline images.
[Winston:] If the cap is at the inverting input junction (1081), the cap is within the loop. Non linearities of it are then equalised.
Good. That's the external IN– point then. Might want to make the pads to accommodate a 150µF, even if we go with 22µF in this application.
[Winston:] I'm not sure I see this as a group project for widespread use though. Maybe I'm wrong?
This wants to be a 12-channel board at 35mm per channel in a v-groove snap-off form factor, which will make for easier mounting in the desk (though each channel does need to have its own mounting option). If I'm going to have 'em made, I want to provide an inexpensive way for people to use just a few channels for projects.
Also I have a 16-into-4 unit that I'm developing that could use four of these channels – rather than mounting BA489s to proto boards with card edge connectors, as was the previous plan. In that application I will need to voltage sum into the inverting input, so I realize we need to have both an IN+ and IN– on the header that also has the 24V and 0V inputs.
I don't think the power wants to be bussed across the boards, as B– impedance rules are so strictly tied to signal path length, and the central B– reference in these desks is the buss channel. Each channel should be considered as part of the module it is wired to.
In my 16-into-4 application I can jump power across. Others can too.
[Winston:] Although I'd still want to use the tantalum 10uF cap on the +ve input
I'd agree that we want to keep the signal path the same within the amp layout itself, save for necessary substitutions (your transistor sub for instance) and that temperature protection with the LEDs, and whichever bias spreader trim technique you feel is easiest at setup time.
[Winston:] How about If I lay out a single channel and post it here or pm it to you
If you want to go minimal at first, four channels on a board would be smart, as I could use the prototypes for the other project I mentioned above, and would have some spares.
[Winston:] Having a place for another gain boost resistor in parallel was just replicating the function that was brought out to the rear connector on Neve amplifiers. A trimmer is fine too if that's best for you.
Pads for either a single fixed Rg or a 3/8" top adjust trimmer please, in the same board position.
[mjrippe:] Definitely into buying some boards.
That's the idea! DM me with channel count and config. I hope a 12-channel snap-off form factor will work for you too.
Winston O'Boogie said:
I fully expect that someone will want to load this with 50 ohms. In my 16-into-4 application, I need to drive 75 ohms (API 2623). Does that mean I need to have heat sinks made and we can forget about the current limiting aspects?
Yes they will; they have to sustain the myth.Winston O'Boogie said:
I believe you know Bateman had to make measurements in unrealistic conditions for a well-designed piece of gear. Applying volt level across a coupling capacitor is not commensurate with typical operation. Only notable exception is passive x-overs.
Winston O'Boogie said:Loading 70 ohms would be fine. The output transistors will need heatsinks anyway and it seems folks have made 1081s using individual and easy to buy heatsinks rather than the custom flat piece of aluminium with isolated top hats underneath.
If we can current limit these to avoid death than that still leaves plenty of current to drive 70 or 50 ohms, just means that the amp switches from class A to AB earlier on. That's what it was in the 1081 so.
Oh good.
The aluminum is not so much an issue to be honest, but the isolation biz is finicky. Hopefully appropriate sinks are the inverted top hat type rather than the spider type, as otherwise there’s no way to get the parts to fit close enough to retain the original footprint.
I’ll go look for those threads regarding heat sinks when I have a sec, or maybe someone can post the link here.
Winston would you mind clarifying the output level comment? I’m unfamiliar with the shorthand. I’m aware that the spec on the 440 is max +18.5dBm into 15 ohms (which is what the Neve -6 to -8 buss level is based around for headroom purposes), so of course we want to retain that headroom into 50 ohms.
Thanks for translating! Appreciate it.
That’s the spec on the op amp, 47k input impedance, capable of +18.5dBm into a 15 ohm load. Doesn’t mean they ever used it that way but some engineer must have been proud enough of the 15 ohm aspect to test it as such and make a point of it in docs.
Maybe we could put .250 tabs on the board and include a little 3” Vintage 30 rated for 1.5W.
Winston O'Boogie said:
That’s the spec on the op amp, 47k input impedance, capable of +18.5dBm into a 15 ohm load. Doesn’t mean they ever used it that way but some engineer must have been proud enough of the 15 ohm aspect to test it as such and make a point of it in docs.
Maybe we could put .250 tabs on the board and include a little 3” Vintage 30 rated for 1.5W.
Winston O'Boogie said:
Wow that is really sexy.
I think it’s totally reasonable to expect people to follow Neve cabling rules here, enabling a single layer board. I think no matter how you run the traces from inlet to either the non-inverting or inverting inputs, they’ll be shorter than in the BA489. No worries.
Let’s do Mill-Max 380 series solder cup pins for I/O, .100” pitch. IN+ / IN– / +24V / 0V / OUT+ / OUT–.
The shield for the pot with the slugged wiper will need a position, so having two OV poles wouldn’t hurt.
This labeling will align with the other card I’m ready to have circuited.
In the discussion of the electrolytic capacitors, I defer to your experience, but I would like to keep the total height of the board as low as possible — which is why I spec’d an axial series. With axials and a non-daughtercard all-in-one layout, this should be a very compact build. Might pair well with the tiny Ed Anderson LO1166.
oh whoops, I forgot some practicalities. here we go:
input side:
+24V x 2 (in, thru)
0V x 3 (in, thru, shield back to trimmer if offboard
IN+ (to input of op amp / that first tantalum)
IN– (to junction between NFB cap & Rg/RVg / also conveniently the gain boost position)
any order that makes your layout easiest is fine, there's no law about putting positive before negative.
output side:
OUT+
OUT– (all the way thru from 0V please, just want to make sure I said it)
Coupla dumb notes. There's a 3µ3 AC coupling cap on the input of the 440 in the 33415. I thought the tantalum takes care of that and thus I am curious why it's there. Also a further 'lytic thought: I am totally not attached to what they used originally. I think the most robust, longest lasting 105° caps with the best specs for each circuit position, which are as small as possible and still reasonably affordable, should be the ones chosen.
input side:
+24V x 2 (in, thru)
0V x 3 (in, thru, shield back to trimmer if offboard
IN+ (to input of op amp / that first tantalum)
IN– (to junction between NFB cap & Rg/RVg / also conveniently the gain boost position)
any order that makes your layout easiest is fine, there's no law about putting positive before negative.
output side:
OUT+
OUT– (all the way thru from 0V please, just want to make sure I said it)
Coupla dumb notes. There's a 3µ3 AC coupling cap on the input of the 440 in the 33415. I thought the tantalum takes care of that and thus I am curious why it's there. Also a further 'lytic thought: I am totally not attached to what they used originally. I think the most robust, longest lasting 105° caps with the best specs for each circuit position, which are as small as possible and still reasonably affordable, should be the ones chosen.
Some application notes for folks following this thread about 0V and summing. You folks who have been doing this a lot longer than me are welcome to make any corrections necessary. Anyone who wants in on the prototype or production run of these (which will be short), please hit me up.
On the 5316 in question, the fader goes between the front half of any given line amp and the back half. This card is very much the back half of a module once you add an output transformer. The fader ground is drawn from the frame, immediately behind the fader on the ISEP rail.
That fader chassis ground is used to shield both the signal input *and* the signal output, but neither connect to the 0V (B–) of the source or the load. In other words, it is entirely possible that people using this new 440 card could telescope the shield from their signal source using its 0V, and begin 0V anew at the card, which is tied to the power supply. If the card follows a Neve module with an unbalanced output, you draw 0V from the module. If the card follows a Neve buss, you draw 0V from the buss channel. If the source is a step-down line input transformer, the other end of its winding needs to go back to 0V, and that 0V point may need to be the one on the card (like a 3416).
The card can add 20dB of gain maximum, so assuming you have 10dB to 4dB of free gain from your output transformer, your input level to the card must be between -26dBu to -20dBu minimum, respectively, to come up to +4dBu line level at the secondary.
If you want to do resistive/voltage summing into the card (a virtual earth sum in this case), you use the inverting input, take the non-inverting input to 0V, and there's a set of calculations to do for the Rg/RVg and Rf positions. I think the Rg in this case needs to be at least 5K, with Rf proportionately higher. This virtual earth sum is the Neve-recommended choice for voltage sums of less than eight channels (due to the sub-5K1 summing feed resistors that would be involved in a sum into a step up transformer and initial op amp, prior to the output amp). It is also the recommended way to sum stereo into mono, using the front halves of two amps and the back half of one amp. If crosstalk is a non-issue, some rather low value feed resistors are possible. I have also been told that the summed signals "disappear" into a virtual earth sum set up in this fashion, so it may be that this is a real weapon against crosstalk in certain situations.
Chassis grounding is to be avoided at the card, unless for some reason your 0V *is* your chassis, in which case you may wish to look at a different power supply or do some careful testing.
If the "fader" is the offboarded input trimmer, you draw the shield from the card and not the source, you use no more than a 5K log or 10K linear with 9K1 slug between wiper and bottom (shield), and you keep the path from "fader" to card to no more than 1 meter. Otherwise you're gonna need a fader buffer.
And you know that this is specifically for 24V single supplies, because that is why the thread exists. If the board ran bipolar, I'd have the Eisen BA512s and a bunch of CAPI Inv-ACAs on order.
On the 5316 in question, the fader goes between the front half of any given line amp and the back half. This card is very much the back half of a module once you add an output transformer. The fader ground is drawn from the frame, immediately behind the fader on the ISEP rail.
That fader chassis ground is used to shield both the signal input *and* the signal output, but neither connect to the 0V (B–) of the source or the load. In other words, it is entirely possible that people using this new 440 card could telescope the shield from their signal source using its 0V, and begin 0V anew at the card, which is tied to the power supply. If the card follows a Neve module with an unbalanced output, you draw 0V from the module. If the card follows a Neve buss, you draw 0V from the buss channel. If the source is a step-down line input transformer, the other end of its winding needs to go back to 0V, and that 0V point may need to be the one on the card (like a 3416).
The card can add 20dB of gain maximum, so assuming you have 10dB to 4dB of free gain from your output transformer, your input level to the card must be between -26dBu to -20dBu minimum, respectively, to come up to +4dBu line level at the secondary.
If you want to do resistive/voltage summing into the card (a virtual earth sum in this case), you use the inverting input, take the non-inverting input to 0V, and there's a set of calculations to do for the Rg/RVg and Rf positions. I think the Rg in this case needs to be at least 5K, with Rf proportionately higher. This virtual earth sum is the Neve-recommended choice for voltage sums of less than eight channels (due to the sub-5K1 summing feed resistors that would be involved in a sum into a step up transformer and initial op amp, prior to the output amp). It is also the recommended way to sum stereo into mono, using the front halves of two amps and the back half of one amp. If crosstalk is a non-issue, some rather low value feed resistors are possible. I have also been told that the summed signals "disappear" into a virtual earth sum set up in this fashion, so it may be that this is a real weapon against crosstalk in certain situations.
Chassis grounding is to be avoided at the card, unless for some reason your 0V *is* your chassis, in which case you may wish to look at a different power supply or do some careful testing.
If the "fader" is the offboarded input trimmer, you draw the shield from the card and not the source, you use no more than a 5K log or 10K linear with 9K1 slug between wiper and bottom (shield), and you keep the path from "fader" to card to no more than 1 meter. Otherwise you're gonna need a fader buffer.
And you know that this is specifically for 24V single supplies, because that is why the thread exists. If the board ran bipolar, I'd have the Eisen BA512s and a bunch of CAPI Inv-ACAs on order.
Winston O'Boogie said:
Very nice! Here are my initial notes, I'll spare you another list until V2:
1. RV1 & RV2 as 3/8" vertical top adjust trimmers
2. Shift R14 & TR5 left to enable TR7 to move down and get a heat sink that doesn't overflow the board
3. Ensure room for .125" mount holes at corners, watch proximity of traces if metal pan-head 4-40 screws are used
4. Adjust placement and distance of jumper above TR4 to enable standard size 0Ω jumper
5. Choose 0Ω jumper to match size of preferred resistor series
6. Preferred resistor series are the brown mil spec Dale/Vishays, sizing up is ok (if applicable) as we want to enable bulk pricing: https://www.mouser.com/datasheet/2/427/cmfmil-223788.pdf
7. Nix the two outputs, just go through the series output resistor and I'll write in a 0Ω jumper as an option in the build doc.
8. Combine H2 & H3 as the Mill-Max solder cup headers are quite spendy, and I'm going to be going for bulk pricing on the 6-way version (on this board and another one I'm making).
9. For the output connector, now that the direct trace from the AC coupling cap is gone, Run B– up under the series output resistor. Thus:
Pin 1: 24V (in)
Pin 2: 24V (thru)
Pin 3: 0V (in)
Pin 4: 0V (thru)
Pin 5: 0V (shield))
Pin 6: OUT
9. What 'lytics and tant are drawn?
10. I haven't really dug into the op amp layout itself, but can do so on v2.
11. I understand why you put the power I/O at the output side. If there is a way to bring the raw 24V trace to the input side, it would be advantageous in a number of circumstances. Let's have a think on that one. If 24V could come to the input side, I would just make it another 6-way header, with the additional 2 pins labeled 24V IN & 0V IN.
These are Bourns 3362 series as I know them, do you have any preference or alternative you trust?
Ah, I see...1/4W would have to be sufficient to get a smaller top adjust footprint. I haven't used BI/TT Model 64 or Bourns PV37W, but those seem to be the inexpensive Mouser options. Bourns 3339 or PV12P are pricy but the 5/16" round format is interesting.
I was considering the 'Flying Nun' heat sinks
Mouser and Digikey both stock the Aavid 5783 but it's ~3/4" off the board altogether. The 5782 is ~1/2" but non-stock. Hmm.
I sized according to PRP resistors and Dale are the same size I think.
PRP are cool, either way. Are the audio ones all that? Haven't checked pricing. Dales are $0.25ish single.
Here is what Cyril said regarding tantalum caps of 10uF and above: "Measured at 0.3 volts with and without DC bias, my Tantalum capacitor stocks produced at least ten times more distortion than found with low cost polar Aluminium electrolytics. I decided to exclude Tantalum bead capacitors from further tests."
Might be interesting to spec a UKA in the doc as an alternate for tweakers, I'm guessing the tant is dipped and the 'lytic would be similar size. We could spec the Kemet body type so the leads line up (there are like six body types).
This is a good opportunity to mention that I myself am more interested in representing the signal well than precise traditionalism – to wit, the alternative before you stepped in was bypassing the transformers altogether and going with a THAT 1512 into a THAT 1646 (which would be really sad). Having a low- or no-distortion AC coupling option would be smart, even if in practical use a listening test would only reveal very minor differences.
Also on the part swap note, BC441 and BC461 are in stock at Mouser, they just cost almost double what the 2Ns you spec'd do. I haven't compared the values. Same pinout?
Ah, I see...1/4W would have to be sufficient to get a smaller top adjust footprint. I haven't used BI/TT Model 64 or Bourns PV37W, but those seem to be the inexpensive Mouser options. Bourns 3339 or PV12P are pricy but the 5/16" round format is interesting.
I was considering the 'Flying Nun' heat sinks
Mouser and Digikey both stock the Aavid 5783 but it's ~3/4" off the board altogether. The 5782 is ~1/2" but non-stock. Hmm.
I sized according to PRP resistors and Dale are the same size I think.
PRP are cool, either way. Are the audio ones all that? Haven't checked pricing. Dales are $0.25ish single.
Here is what Cyril said regarding tantalum caps of 10uF and above: "Measured at 0.3 volts with and without DC bias, my Tantalum capacitor stocks produced at least ten times more distortion than found with low cost polar Aluminium electrolytics. I decided to exclude Tantalum bead capacitors from further tests."
Might be interesting to spec a UKA in the doc as an alternate for tweakers, I'm guessing the tant is dipped and the 'lytic would be similar size. We could spec the Kemet body type so the leads line up (there are like six body types).
This is a good opportunity to mention that I myself am more interested in representing the signal well than precise traditionalism – to wit, the alternative before you stepped in was bypassing the transformers altogether and going with a THAT 1512 into a THAT 1646 (which would be really sad). Having a low- or no-distortion AC coupling option would be smart, even if in practical use a listening test would only reveal very minor differences.
Also on the part swap note, BC441 and BC461 are in stock at Mouser, they just cost almost double what the 2Ns you spec'd do. I haven't compared the values. Same pinout?
ruffrecords
Well-known member
I downloaded and tried that app. Do you use a stylus? I have great trouble seeing where the trace is going when I use my finger.
Cheers
Ian
Cheers
Ian
ruffrecords
Well-known member
Winston O'Boogie said:Haha, yep if takes a bit of getting used to. No I don't use a stylus, just my fat podgy index digit!
I was on a smallish Lenovo tablet and just really zoomed in to get the traces where I wanted. I set the raster size to 25 thou which got the components on grid and it snapped the traces to where I wanted for the most part.
I don't think it'll be my new primary layout app but it was nice to just sit on the sofa painting away with my finger.
Cheers Ian
OK, I will give it another go once it gets a little cooler!!
Cheers
Ian
ruffrecords
Well-known member
Winston O'Boogie said:
We spent much of today in the air conditioned car and at Sainsbury's which is also air conditioned. We even went out for tea this evening to Morrisons for the same reason. It was till over 30 degrees C when we got home around 7pm. For fun we switched the car temperature gauge from Centigrade to Fahrenheit 'cause us old folks don't understand these new fangled units. According to the car, the outside temperature reached 100 degrees which by anbody's standards is bloody hot.
In the last half hour we have had some rain with some pleasant thunder and lightening which has definitely cooled things down about here in North Norfolk. After last year's summer we toyed with the idea of having A/C fitted. This year we have decided we definitely will.
Cheers
Ian
