|
Post by rocinante on Dec 31, 2017 20:22:38 GMT -6
Oh I know how complex it could be. I'm not saying stick some ne5532 in there and call it done (that would cause oscillation at the very least plus a plethora of other things) but it's definitely a weak point in the mixer. Jeff's ACA summing is just that; it's a summing bus for a mixer just like yours or to be put into an API mixer. It's what I have in the Ghost. It's a replacement for a mixers summing buss and turns things from good to f****** amazing. It's like driving a kia and upgrading to a Mustang.
|
|
|
Post by matt@IAA on Dec 31, 2017 21:00:26 GMT -6
Are you talking about buying his ACA PCB and replacing the built in IC ACA circuitry?
|
|
|
Post by rocinante on Jan 1, 2018 9:07:11 GMT -6
Yep.
|
|
|
Post by matt@IAA on Jan 1, 2018 13:00:07 GMT -6
I’m not sure how I’d do that. I’d have to bypass the circuitry on the card pcb to the bus/rails somehow. That may be beyond my ability here.
|
|
|
Post by allbaldo on Jan 4, 2018 19:12:52 GMT -6
Oh man, I’m happy to see this thread. I have one of these (with super eq) that I’ve always wanted to make into something more useful. Mine has a good many problems, but I enjoy the 6 or 7 channels that do work, so if this experiment is a success, I’ll be doing some tinkering.
|
|
|
Post by matt@IAA on Jan 4, 2018 21:34:27 GMT -6
Oh man, I’m happy to see this thread. I have one of these (with super eq) that I’ve always wanted to make into something more useful. Mine has a good many problems, but I enjoy the 6 or 7 channels that do work, so if this experiment is a success, I’ll be doing some tinkering. I have schematics, layouts, and the manual. If you need anything let me know.
|
|
|
Post by matt@IAA on Jan 4, 2018 21:58:41 GMT -6
I'm thinking about this. During mixdown on this little thing you send the stereo mix to bus 7/8. After the ACA on 7/8 the signal either goes out to the track 7/8 output or to the 2 Track L and Right booster on cards 11 and 12. To replace the signal path on mixdown, I'd have to replace only the ACA on 7/8, which I think is pins 8,9,10 on the U2 4741. Everything else is just a booster. The CAPA ACA card has the two opamps, which would replace 7/8, and then two post-fader boosters, which replace the opamp for pins 2,3,4 on U2 on cards 11/12. ... But how do you bypass one opamp on a socketed IC in a PCB?
|
|
|
Post by matt@IAA on Jan 5, 2018 12:03:43 GMT -6
OK. I'm not sure I want to do the CAPI ACA route. But in case anyone else does, here's my notes on how I think it could be done on this board. The board uses two card-to-card jumpers, J4 (9 rails) and J3 (12 rails). J3 carries solo control bus, solo audio, echo bus, and the 8 tracks. J4 carries phantom supply, common, +V, -V, monitor busses, cue bus, and the channel ACA inputs. Here's the bus ACA portion of the I/O card schematic. Here's the 2 track bus portion from the control room card showing the ACA out from 7/8 going to the master fader, then to the 11/12 booster, then out and to the monitor section. To bypass pins 8,9,10 on U2, I think you could disconnect the J4-9 pins on cards 7 and 8 and send that to the CAPI ACA board inputs. The current bus resistors are already 47k which is what the CAPI wants to see. This ACA main output signal needs to be sent to: - The input of the bus trim pot on cards 7 and 8 (R49). This retains the function of the track 7/8 4741 IC boosters and the track outs. - The ACA out on cards 7 and 8 (connect to the wire in molex connector J6-3). If you to did not want to use the CAPI boosters, the output from the fader would not change (back to the two track boosters on channel 11, 12 - J6-1 on each card) and from there it would just drop back into the normal signal flow and you're done. If you did want to use the CAPI boosters you route the fader output to the L and R CAPI booster inputs instead of back to J6-1 on 11/12. Then, - Remove 47 ohm load resistor R66 on 11/12 (this already exists as R3 on the CAPI Booster on the primary side of the 2503 transformer) - Send the BO Main + out to the output point of R66 (you could do these in 1:2 or 1:1, depending on how you wanted to configure it) for metering and J6-2. - Replace the RCA with a TRS. Connect the existing hot to the +. - Send the BO Main - out to TRS -. Looks like the total cost including transformers for the whole card would be around $230 using GAR2520 kits.
|
|
|
Post by matt@IAA on Jan 5, 2018 13:59:26 GMT -6
Alternately, you could create a completely separate summing bus if you wanted.
I think you could grab the output of the existing 4741 ACAs after the trims on cards 1-8 (either at J6-1 "booster in" or after the booster at output of C14 / input of R66....haven't decided yet) and land those on a breadboard. This would have eight 47k resistors. The 1,3,5,7 resistors would output into the L input of the CAPI; the 2,4,6,8 would go to the Right. Now you just take the output of the CAPI ACA, send it to your master fader and back to 11/12 as above.
This sort of fixes the wonkiness of summing to 7/8 for mixdown - now you have summing through the existing busses, giving you those additional faders. But there's no bus inserts, so it's not that much of a benefit. The one goofy thing is that you don't really have any true stereo busses, so you'd have to be careful about matching left and right. This also prevents bus 7/8 from sounding way different than 1-6.
It is actually adding one additional layer of 4741 processing through the busses. 1st method: line input -> 4741 for gain, EQ, booster -> 2520 for 7/8 ACA -> 2520 for booster. 2nd method: line input -> 4741 for gain, EQ, booster -> 4741 for 1-8 ACA -> 2520 for 2bus ACA -> 2520 for booster).
|
|
|
Post by matt@IAA on Jan 5, 2018 21:50:33 GMT -6
I think I might have talked myself into doing this now. Can anyone help me figure this last bit out? If I go the second route to create a separate summing bus, in order to effectively use them I would take the signals post trim pot. Do I take them before or after the track boosters? Note - on channel 11 there is no ACA and on 12 it is for the solo bus without a trim pot. So effectively everything before the J6-1 input point isn't there.
|
|
|
Post by rocinante on Jan 6, 2018 9:42:04 GMT -6
God damn it, I have seen Jeff and Gary answer this very question before and I'll see if I can look up where I think I saw it. You cerrtainly have a firm grasp on mixer engineering and audio schematics. Also nice job breaking it all down. Everything you need to know about the ACA is probably here: groupdiy.com/index.php?topic=33038.msg402708#msg402708Ygou could just ask at Gdiy. Someone will know.
|
|
|
Post by matt@IAA on Jan 6, 2018 11:56:56 GMT -6
I sent Jeff an email, he said post booster is probably safer.
His documentation says the ACA in his console is fed at -2dBu through a 47k bus resistor. CAPI's docs also say the -2dBu signal comes out of the ACA at -.5dBu, and then the booster brings it up to +4.45 dBu.
However, I already have a pair of Cinemag CMOQ-2S (1:1,2 steel - kinda like a 2503) and those Jensen 11-ECFs (1:1 steel). I think I'll put the Cinemags on the ACA side and the Jensens on the booster side. This ain't "right" because the CAPI card wants a 2623 on the ACA and a 2503 on after the booster, but...I think it'll be fine. I can always swap 'em later. Going from 1:2 to 1:1 on the ouput transformer does mean I'll be -6 dB from the circuit design so for a -2 dBu signal in the booster will be sending out a -1.55 dBu.
As a quick check, the block diagram on an API 1608 has a +4 signal going to the program bus mix input. Since it's already a nominal +4 going into the ACA, there's a 1:1 output transformer on the booster. So I guess I can send a 1.23V / +4dBu signal through a bus, and measure before and after the booster to find out what's what. I think I agree with Jeff - I'll send the post-booster signal to the ACA at +4, which should give me a nominal +4 output even with the 1:1 transformer.
So, the final flow through this thing will be: line input -> 4741 line amp -> 4741 x2 for EQ -> 4741 for channel booster -> 4741 for track bus ACA -> 4741 for track bus booster -> 2520 and transformer for 2bus ACA -> 2520 and transformer for 2bus booster.
|
|
|
Post by matt@IAA on Jan 6, 2018 12:26:58 GMT -6
Also, if anyone else is following along on a 1280b, whether or not you modify it for +4 or not, when it's open is a good time to go ahead and make sure the VU LEDs are adjusted correctly. The manual says the red LED is a quasi-peak indicator, so adjust the LED trim so it barely lights up at 4 dB above nominal out - for +4 dbU, that's +8dBu or 1.95V.
One of my 4558 ICs is bad - that channels first LED is always on and none of the others work, so I need to get that replaced. Fortunately they're like $0.30.
|
|
|
Post by matt@IAA on Jan 7, 2018 14:10:47 GMT -6
I was thinking... you could also do it this way, and create a separate program bus instead of going downstream of the track busses. You just need 24 47k resistors instead of 8. A little shorter signal path (input->eq->fader->booster->pan->summing ACA). This eliminates going through the 4741 on the individual bus ACAs and the post-fader booster 4741 stage. And if I ever want to upgrade the opamps, this would limit the entirety of the mixdown signal to the U1 4741, making future upgrades easier. Hm...and you could send these to the two open busses on J4...that could be pretty clean, although unless I put on a double molex connector or something you'd have to desolder to remove them. Then just add a wire out from the J4 bus to the CAPI ACA input. And the mute is before the bus send but after the direct out, so if I wanted to grab something post fader I can at the direct out, then mute it so it doesn't hit the 2 bus. Practically speaking I can do groupings inside the DAW anyway, so the analog subgroups aren't super critical to have (especially since they don't have insert points). I'm liking this idea more.
|
|
|
Post by jimwilliams on Jan 7, 2018 15:12:27 GMT -6
You will need a small value feedback cap across the sum opamp 47k resistor and the fader recovery opamp. Use 10pf NPO mono ceramic to keep the bandwidth up. That will restrict it to below 100k hz and will cancel the buss resistor's stray capacitance that causes ringing/overshoot and any potential oscillations.
Use a 22.1k feedback resistor for the fader recovery opamp as that will return the gain staging to unity. Then shove the master faders all the way up. Dale CMF55's are a great choice.
Use quality opamps to lower noise and improve sonics. Add local .1 uf mono ceramic caps from pin's 4 and 11 to ground. Recommended opamps include the LME series, the OPA1642 fet input and OPA1602 bipolar rail to rail opamps. Those will increase dynamic range and give greater headroom than the standard types. The 1602's draw 3 ma, the 1642's only 1.5 ma, less than an 072 opamp so you won't need to worry about that weak power supply.
When installed, check the output pin's for any DC offsets. All of those are precision opamps and if the offsets are below about 3~4 mv, short out the el coupling caps.
|
|
|
Post by matt@IAA on Jan 7, 2018 16:08:33 GMT -6
You will need a small value feedback cap across the sum opamp 47k resistor and the fader recovery opamp. Use 10pf NPO mono ceramic to keep the bandwidth up. That will restrict it to below 100k hz and will cancel the buss resistor's stray capacitance that causes ringing/overshoot and any potential oscillations. Use a 22.1k feedback resistor for the fader recovery opamp as that will return the gain staging to unity. Then shove the master faders all the way up. Dale CMF55's are a great choice. Use quality opamps to lower noise and improve sonics. Add local .1 uf mono ceramic caps from pin's 4 and 11 to ground. Recommended opamps include the LME series, the OPA1642 fet input and OPA1602 bipolar rail to rail opamps. Those will increase dynamic range and give greater headroom than the standard types. The 1602's draw 3 ma, the 1642's only 1.5 ma, less than an 072 opamp so you won't need to worry about that weak power supply. When installed, check the output pin's for any DC offsets. All of those are precision opamps and if the offsets are below about 3~4 mv, short out the el coupling caps. Hi Jim, I always really appreciate your posts! Re: bus resistor -- gotcha. How would you install the OPA1642 or 1602 since they don't match the current socket pinout?
|
|
|
Post by matt@IAA on Jan 8, 2018 9:05:59 GMT -6
Whew. SOIC to DIP adapter plus OPA1644 or 1604s get pricey quick, even for only the 12 channels. Reading a bit on slew rate and I came across this thread: groupdiy.com/index.php?topic=12926.msg149433#msg149433If a general rule of thumb is that slew rate needs to be minimum .5 V/us per volt of peak output to prevent audible distortion, you can make this table: dBu | -2 | 0 | +4 | +10 | +20 | +24 | V rms | 0.62 | 0.77 | 1.23 | 2.45 | 7.75 | 12.3 | V peak | 0.87 | 1.10 | 1.74 | 3.45 | 10.9 | 17.4 | min V/us | 0.44 | 0.55 | 0.87 | 1.73 | 5.48 | 8.68 |
The Harris 4741s have a slew rate of 1.6 V/us. For reference, an API 2520 has a slew rate of 3 V/us (meaning rule of thumb says it can handle peaks of ~15 dBu without audible distortion). Based on the Walt Jung criteria, the board should probably work fine at +4 without any audible slew rate related issues, but it lowers your peak headroom before distortion to about 6 dB over nominal (~+10 dBu). Food for thought.
|
|
|
Post by svart on Jan 8, 2018 11:01:10 GMT -6
So I'm just now getting to reading this, and haven't read it all..
A few words of warning, and hopefully enlightenment..
Be wary of the summing amps. The original amp is virtual earth (inverting), which while having the detriment of amplifying the ground noise (best option would be to decouple the opamp as good as possible and strap a lot of extra grounds to the board for decreased ground impedance), the main benefit is that it does not change gain when more or fewer channels are hooked to the summing bus.
Take that for what it's worth, and I have no idea how the CAPI summing amp system works, but I would refrain from changing it from the virtual earth style.
Also, while Jim and others are advocates for ridiculously over-able opamps, they tend to end up needing feedback caps to tame oscillations. This is in effect reducing GBW down to levels corresponding to the original low-slew opamps, while introducing possible issues with phase shifting unless the opamp open-loop bandwidth is at least 10 times the filter value.
This isn't much of an issue with most fast opamps, but begs the question of why would you spend the extra cash on a high speed device, only to neuter it back to the region of operation that the original opamp already inhabits, without the external parasitic components??
It's the typical buying a ferrari to only get groceries cliche'...
In my own studies, I've found that there is typically a middle ground, where you can place a similar spec'd part with a single arbitrarily better spec (slew, etc) while not seriously needing to consider other attributes/issues.
These being 4 channel opamps makes this harder than it needs to be though.
However, another few things to consider is that people use the term "slew" as a singular metric of operation, but do not consider that "slew" does not operate the same over a range of gain.
An opamp such as the 4741 might *only* have 1.6V/us slewrate, but at what gain and what output voltage? 1.6V/us might be perfectly fine at 0.5Vpp, but might be very poor at 10Vpp.. So we need to know a lot more about your application and operating conditions than just "put these crazy expensive fast opamps in there and be done".. Because a few tweaks to feedback and output levels might be enough to turn a turd of an opamp into a performer..
|
|
|
Post by matt@IAA on Jan 8, 2018 11:15:05 GMT -6
Thanks Svart. The CAPI ACA is an inverting style also, so it wouldn’t be schematically changing. I mentioned before, I’m not really interested in changing the ICs out right now. The CAPI thing just piqued my interest. I’m still not sure if I want to add the CAPI as a 9/10 “always on” stereo bus from the channel trims, or do it after the 8 existing busses (12 into 8 into 2). Obviously the 12 into 2 is cleaner, shorter signal path, etc. But, I need to think about metering, since this board only has meters on - you guessed it - the 8 busses and the 2 mix. That being said, I could hook the meter bridge to the direct outs and get at least 8 channels of metering... and then the onboard LEDs can be the busses. That wouldn't be terrible.
|
|
|
Post by svart on Jan 8, 2018 12:07:51 GMT -6
Since the LME quad opamps are EOL for some time now, you might look into the MC33079 for the quad opamps, as well as the MC33078 (dual) for any BJT input swaps. Their slew is 7V/us, and the GBW isn't terribly over-spec'd, but they still might need small feedback caps in some cases since the audio bus impedance is relatively high. They have a lower input impedance, but isn't so low to cause me to worry about loading.
I did use them in a phase shifting EQ, and they tended to oscillate in some situations when the phase margin wasn't enough. A small feedback cap fixed it, but in the end I went higher impedance and used the TL074 in place and it sounded fine without the cap. You still might be able to use the TL074 in these spots since the impedance is pretty high and doesn't look like loading is much of an issue. The TL0 parts don't have much GBW and drive strength, so I'd limit them to low gains and high impedance loading.
However, at the summing points, a higher slew part is a better option. I'm not sure what that is, as I've personally not found the burr brown sound to my liking, so I'd not use the OPA stuff if I had a better choice.
|
|
|
Post by matt@IAA on Jan 8, 2018 13:11:38 GMT -6
Since the LME quad opamps are EOL for some time now, you might look into the MC33079 for the quad opamps, as well as the MC33078 (dual) for any BJT input swaps. Their slew is 7V/us, and the GBW isn't terribly over-spec'd, but they still might need small feedback caps in some cases since the audio bus impedance is relatively high. They have a lower input impedance, but isn't so low to cause me to worry about loading. I did use them in a phase shifting EQ, and they tended to oscillate in some situations when the phase margin wasn't enough. A small feedback cap fixed it, but in the end I went higher impedance and used the TL074 in place and it sounded fine without the cap. You still might be able to use the TL074 in these spots since the impedance is pretty high and doesn't look like loading is much of an issue. The TL0 parts don't have much GBW and drive strength, so I'd limit them to low gains and high impedance loading. However, at the summing points, a higher slew part is a better option. I'm not sure what that is, as I've personally not found the burr brown sound to my liking, so I'd not use the OPA stuff if I had a better choice. Thanks, I'll look into those. They look very promising, with better channel separation and power consumption than the LME49740. I'm leaning towards going from 12:2 so in the mod all of the summing to "tape" would be through the new CAPI ACA, using GAR2520s. It sidesteps the 4741s in the groups completely for the 2 track mixdown.
|
|
|
Post by jimwilliams on Jan 9, 2018 14:49:10 GMT -6
Hi Jim, I always really appreciate your posts! Re: bus resistor -- gotcha. How would you install the OPA1642 or 1602 since they don't match the current socket pinout? 1602's for the sum stages, 1642's in other places if your PSU is stock as it's a POS. Use the Brown Dog dual SOIC to 14 pin DIP adaptors. Open loop gain specs are where summing opamps succeed or fail. The only discrete opamp I know of with good OLG specs at 10k hz is the old Jensen 990, about 85~90 db. The 5534 is 60 db in comparison, 30 db less feedback/distortion correction. API's are around 40 db like the 072 and 741. The better IC's do around 70~80 db OLG at 10k hz. Having a couple more db's of headroom is another benefit of those parts, less clipping throughout the desk and an increased dynamic range with the lower noise floor. No one I know likes hiss. You will also want to ad the small feedback caps across the feedback resistors even if you keep the original opamps. Phase margin, ringing and overshoot is why. You will get better square wave geometry if you do that plus added stability. I'd consider the lack of them a design error. SW tended to use very large resistor values (47k buss resistors) that created excessive thermal noise and reduced crosstalk due to resistor capacitive coupling.
|
|
|
Post by matt@IAA on Jan 9, 2018 16:02:42 GMT -6
Interesting about the OLG. The Harris 4741s have a typical sepc of around 50 dB at 10k. Also interesting to note that the CAPI ACA, which is a copy of the API ACA circuit, actually attenuates with the ACA op amp (47k bus resistor and a 28k feedback resistor) and gains up with the output transformer. The post fader booster on the API design is unity on the op amp. Lower gain on the op amp means the OLG doesn't affect it as much... wonder if that's why they did it? The 1280b has no transformer and the gainstaging is a bit different - approach, -1 dB on the ACA and the booster is (56k/22k) = 8.11 dB. For a total of 6.8 dB. I believe the OL gain figures from the chart means that at 20 Hz you have 8.11 dB but at 10k you're down to 8.05 dB, and 20k to 7.99 dB. I don't know enough to know if that matters. Either way, I'm not summing through 'em so it's not a worry. As for the caps across the resistors, older API designs had them even if the SW did not. I suspect it wasn't a design error as much as a decision, probably to save cost. I haven't seen the higher end schematics from SW but it wouldn't surprise me if they had them. I always learn something when you post John. Thanks.
|
|
|
Post by svart on Jan 9, 2018 16:34:32 GMT -6
Interesting about the OLG. The Harris 4741s have a typical sepc of around 50 dB at 10k. Also interesting to note that the CAPI ACA, which is a copy of the API ACA circuit, actually attenuates with the ACA op amp (47k bus resistor and a 28k feedback resistor) and gains up with the output transformer. The post fader booster on the API design is unity on the op amp. Lower gain on the op amp means the OLG doesn't affect it as much... wonder if that's why they did it? The 1280b has no transformer and the gainstaging is a bit different - approach, unity on the ACA and the booster is (56k/22k) = 8.11 dB. I believe the OL gain figures from the chart means that at 20 Hz you have 8.11 dB but at 10k you're down to 8.05 dB, and 20k to 7.99 dB. I don't know enough to know if that matters. Either way, I'm not summing through 'em so it's not a worry. As for the caps across the resistors, older API designs had them even if the SW did not. I suspect it wasn't a design error as much as a decision, probably to save cost. I haven't seen the higher end schematics from SW but it wouldn't surprise me if they had them. I always learn something when you post John. Thanks. More likely is that they found it to have issues driving the transformer at full gain. Reducing the gain decreases the requirement for higher slew rates for high current outputs.
|
|
|
Post by matt@IAA on Jan 16, 2018 23:06:14 GMT -6
Ok, finally had a chance to work on this a little. Got the ACA board populated except for the 47k feedback resistors I miscounted and the transformers. This was a no big deal build, not bad at all. The PCB is really, really nicely made. Did first ever 2520 DOA, then realized you have to buy Mill-Max pins separately. Ha! And, finished my little 47k bus resistor card and installed wires from all of the channels to lead to it. Sadly, I thought I might be able to figure a way to cut the CAPI ACA card in half but there's really no good way to do it, since both channels / stages are powered by the same input. And it's just too nice to hack up. So, since its too big to fit in the case, I'm gonna have to put it in a box and go out of the chassis to it. That's ok, since there's a port marked "expander" that's blank. I'm thinking I can use a DB-15: 1. PS + 2. PS Common (ha just kidding, this thing doesn't have a power supply ground) 3. PS - 4. SB L In 5. SB R In 6. To fader L 7. To fader R 8. From fader L 9. From fader R 10. SB L out + 11. SB L out - 12. SB R out + 13. SB R out - 14. Audio common 15. Chassis ground To do list: - Make three more gar2520s - Finish the ACA card - Land the wires from each channel L, R to the 12 pairs of 47k resistors - Mount the bus board to the chassis - Rig up the DB15 in the case - Run SB out, to and from fader, and final balanced SB outs to the DB15 connection - Make a box for the ACA-Bo and mount it - ?? - Mix stuff
|
|