A Summing Amp I Built

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

bjosephs

Well-known member
Joined
May 10, 2021
Messages
104
Location
Massachusetts
So, several years ago I became enamored with the idea of analog summing and feverishly designed and built the devise I will show and describe below. As embarrassing as it is to say I never actually integrated it because I didn’t have much in the way of IO. I had planned to get some more converter channels but just didn’t. I did test it out with a few signals but aside from that the poor thing has sat in a rack for years.

Well stumbling on this forum and seeing other people’s summing projects shared I decided to dig it out and take some pictures. More importantly I pulled the trigger on a 24 out usb interface so I can give it a proper chance and do some mix comparisons.

The topology is simple: 4 virtual ground summing amps, two per channel with each summing the phase and anti phase signals of the balanced pair. Each channel is fed to a transformer for increased cool factor. There’s a DC servo in there that I don’t remember working all that well but we were in the mV range so I left it alone. It’s DC coupled aside from those two features.

I machined that back panel out of an aluminum block on a Bridgeport at work. I clearly took some shorcuts out mounting with the sticky squares and zip ties... probably planned on cleaning up my work after I was sure I made something worth while 😂

Also I borrowed two of the 990s for my mic preamp project and threw in a couple of other discrete OAs from my collection.
 

Attachments

  • 95C3C07A-8465-443F-9B38-6A32CBC62588.jpeg
    95C3C07A-8465-443F-9B38-6A32CBC62588.jpeg
    246.1 KB · Views: 249
  • 55878CAF-B69C-41D5-9EFD-F537E1FC4DC0.jpeg
    55878CAF-B69C-41D5-9EFD-F537E1FC4DC0.jpeg
    212.9 KB · Views: 239
  • 0FE99117-F56F-48FE-97D2-C0FEFD0BCF70.jpeg
    0FE99117-F56F-48FE-97D2-C0FEFD0BCF70.jpeg
    195 KB · Views: 201
  • 28F2F13B-FDB8-4D43-9BC3-A6B18B0519E5.jpeg
    28F2F13B-FDB8-4D43-9BC3-A6B18B0519E5.jpeg
    191 KB · Views: 231
Also wanted to mention that if I were to do the project today I’d probably use PCB mounted DB25s for the inputs. Ironically the interface I bought has DB25 outs and I could have used simpler cabling - just didn’t think I’d ever have that option. I’d also make some of the channels mono or switchable for kick/snare/bass and other center panned stuff.
 
Well I hooked it up today and everything sounds great. I think I did too good of a job because I couldn’t tell the difference between a digitally summed mix and one running through the summing amp when level matched. I would have liked to push the levels a bit harder to see if I could get some heat but the outputs clip the A/D way before then... add an output attenuator to the list of features for upgrades. 😂
 
How many channels are you summing? I'm rather surprised you don't hear a difference. Even without summing the transformers alone should change the sound.
 
Indeed, that’s what I was expecting. They are hefty Cinemag units so maybe I need to hit them harder. I was summing 9 stereo channels (maybe only 7 were playing at a time). Not a complicated mix.
 
Last edited:
I did some mods today. I found out the channel closer to the supply was picking up some stray noise from the power transformer. I think the OLI isolators are the culprit. It was only -80 DB but I decided to relocate the transformer as far away as I could and reposition the bipolar power supply board. It improved a bit but I think I’ll have to try a shield plate next. While I was in there I reterminated the power transformer with some quarter inch spade terminals.

More importantly I decided to try more gain so I pulled the 10K feedback resisters on the summing amps and replaced them with 40K to get 12db of gain. I also rewired my studio to put my monitor controller between the output of the summing amp and my AD converters because the signal is way too hot now... currently working on sourcing some T pad attenuators.
 
You could turn the mains tranny so that the wires are next to the sidepanel. That way the spot with the strongest magnetic field is the furthest away from the rest of the circuit. Additinionally will it help to twist all the wires from tranny to PSU board.

With a 40k resistor in the feedbackloop you´ll pay a noise penalty since you´re amplifying the thermal noise of the resistor with the gain of your summing amp. If you replace the 40k with 2x20k in series then you get 3dB less thermal noise since you have two uncorrelated noise sources instead of one.
 
You could turn the mains tranny so that the wires are next to the sidepanel. That way the spot with the strongest magnetic field is the furthest away from the rest of the circuit. Additinionally will it help to twist all the wires from tranny to PSU board.
The position of the wires has nothing to do with the magnetic field, which depends only on the orientation of the core.
The wires induce some electrostatic field, which is of lesser efect, except when running very close to signal traces, which would be serious malpractice.
With a 40k resistor in the feedbackloop you´ll pay a noise penalty since you´re amplifying the thermal noise of the resistor with the gain of your summing amp. If you replace the 40k with 2x20k in series then you get 3dB less thermal noise since you have two uncorrelated noise sources instead of one.
This is absolutely silly. The noise of a 20k resistor is 0.707 that of a 40k. When you combine quadratically it results in exactly the same noise as a 40k.
What's more, the bus impedance is dominant; noise-wise, it is in parallels with the NFB resistor. Whatever the NFB resistance, the noise impedance is almost constant.
 
The magnetic strayfield of a torroidal is worst at the position where the connection wires leave the core because the overlapping of the windings is less perfect. This is directional and if magnetic interference is a problem then turning the transformer can help. And looking at the first picture I see wires a few cm away from the tranny, hence el field interference might as well be possible.

Regarding noise you´re right, my bad....
 
It’s interesting. I’ve had situations where rotating a toroidal transformer helps with noise quite a bit but in this case it didn’t seem to influence what I was picking up on the left channel. I’ll get some pictures later but I ended up sliding the toroid as far into the left-hand corner as I could go and then spinning the power supply PCB 45° to make it all fit. I’ll also get some screenshots from my spectrum analyzer plug-in in my DAW guess that’s where I noticed the noise in the first place.

I’m starting to consider this a rough draft, as even with the extra gain it’s all so transparent that I want to try a different design to see if I get more of what I had imagined the summing would do. Maybe passive with tube makeup gain next time.
 
I could swap in anything with a 2520 footprint but I've never known op amp swaps to be dramatic changes as long as everything is working. Maybe lower loop gain would increase THD at high frequencies... anyway to try without dropping $200+? 😬
 
Maybe try one channel first and build or buy the op amps knowing you will need them in the future for a mic pre or mixer project?
I might be insane, and given the time and effort required, its probably more expensive than just buying a potted DOA, but I do like populating op amp kits.
This one is cheap and has some flavor, but it is not super robust. RED-25 Discrete Opamp Kit
But you could see if 990’s are too accurate for your purpose for $50...
What would you expect then?
990c is a relatively clean DOA from my limited experience with clones, and vs 1731 and 2520 clones at identical gain settings in identical circuits. Honestly, my introduction to audio tinkering was with replacing 4556 ICs with OPA2228, with a complete lack of understanding of the datasheets or the circuits. But I was happy with the results and wanted to know more.
I know you’re trying to open up discussion, and I don’t want to sound like I know more than I do.
Maybe I should make it clear: I abandoned my EE degree after my Electricity and Magnetism class convinced me that I don’t have the math chops, and none of my professors could help me understand why the output transistors in my solid state amp blew up. In the 15 years since then, I’ve been comfortable with the notion that the datasheets are written by marketing, and you have to use the device before you know how it sounds. My opinions are no doubt skewed because I use microphones and speakers for work, and those specs are intentionally hard to compare.
I doubt the OP wants to get into the weeds of rolling op amps, just was wondering if he had some laying around to do an A/B test.
A null test will likely reveal differences, but thats not exactly the question at hand. Better is depending on your source material and expectations.
Alot of I statements in this post, just want to share without being a pain in the butt.
 
Last edited:
It’s a very reasonable starting point for sure, easiest variable to manipulate. I just think this topology is likely to be on the cleaner side of things and that side is cleaner than I had expected. I like clean but I also want to bang out a couple other topologies while I’m still in the return window for my new interface. 😅
 
I feel ya! Im frugal to the max, unless we’re talking lunch.
You must have done a really good job of component selection, layout, PS bypass and construction if your main issue is that its too clean.
Do you have any multitrack sessions you could try that max out your channel count? Might get into the yellow that way
 
Most of my sessions are pretty simple, I might be able to break out the drum tracks a bit. More on that later.

And ya, it was 10 years ago now, but I tend to get very far into the details when I pick a project. The inclusion of the isolators was from Jensen and Hardy datasheets on summing amps, there are high frequency bypass caps hidden under the 990s on the bottom of the board. The regulator circuits are all exactly to to data sheet recommendations. All metal film 1% resistors. I also always give myself options for connecting and breaking grounds for lowest noise. It’s a simple design and I took a low risk path.
 
Last edited:
With the gain increased (and some attenuation on the output to match) I definitely hear something different between digital and analog summing. I won't say what yet because I have some files linked below if people want to listen and see what they hear. I did my best to level match them in pairs. I also named them discretely to prevent bias. Not sure how badly the file conversion on Soundcloud will muck it up but that's what I have right now linking to them.

Without Mixbus Compressor:
RED
WHITE

With MixBus Compressor:
YELLOW
BLUE
 

Latest posts

Back
Top