|
|
Post by svart on Oct 10, 2013 13:02:41 GMT -6
I actually started something like this over at GroupDIY many years ago but I don't think I finished it.
Rules for modification of gear, part 1: What *really* matters?
1. Find the schematic. You aren't going to get far without knowing what it is that you are modifying.
Analog:
1. Decoupling.
2. Decoupling.
3. Decoupling. I can't stress this enough. It may be the most boring part of modification, but it's the most necessary. Everybody is too engrossed in the fun and excitement of changing opamps and signal path caps but very few actually pay attention to *proper* decoupling. They might change some power supply caps to "better" ones, but rarely do they think about the parasitic effects of long power traces and the inductive and resistive decoupling that happens between the power supply and parts. That inductance and resistance make it hard for the power being created and reserved in the power supply to be supplied in amounts equal to the demand.
Frequently, the parts aren't supplied with the instant current they demand and you have issues with distortion and other non-linearities as the part is starved.
Generally, designs that are made for *mass* consumption are designed to work "good enough". I've been in meetings at companies where the purpose was to dumb down the product and make it work well enough to sell with the cheapest of the cheap parts. This includes most consumer level audio products, regardless of name or lineage. The measurements and specs WILL be measured in the way that suits the marketing the most and you'll never experience that level of performance in the real world. You can bank on that. I've seen it happen and I've stood there aghast.
The first thing that will always be cut from a design, if it was even there to begin with, will be capacitors. Good caps cost a lot more than most anything else besides the big ICs, opamps not included. Opamps are cheap, and in quantity can still be cheaper than caps. In analog audio design, even at the engineering level, proper decoupling is not something that is understood very well. Engineers go by datasheet designs which usually show the bare minimum in design for the rather steady state tests that are done. It's generally expected that the designer using the parts will do their own testing to determine the usefulness of the part in their own design. Besides doing things like sweeping CW tones through a unit, other more rigorous testing is not usually done, unless a distinct problem is found.
Unfortunately, all of this conspires to bring you mediocre performance from otherwise decent and normal parts.
SSL is a good study if you want to see how their parts are implemented. Their consoles are full of the NE553x parts that most people blindly hate and change out, yet are still very misunderstood by a lot of people. Their consoles and devices are about the farthest from the "sound" of the NE553x that most claim to hear and are an excellent example of what can be achieved by designers who understand the needs of the parts they use. In my own studies, the NE553x parts are very sensitive to the minimal amount of decoupling they need, and I've rarely seen products that meet that minimum, and therefor result in a sludgy, slow sound, the "NE553x sound".
Also unfortunately, this misunderstanding and bias towards parts that are misused only complicates issues when the end user decides after reading forum threads, that changing opamps is the *only* fix for the sound they are hearing. Switching from a semi-low power part to the newest whizz-bang-power-hog-super-opamp only compounds the issues. This also manifests in other anomalies and leads the person to continue trying part after part in the search for increased performance. Clearly this is a wild goose chase and a time/money waster, although it's been a guilty pleasure of mine to play around with this as well. What happens is that instead of making a decent opamp perform it's best through making it's supporting circuit work as it should, you are bringing in a faster opamp and bringing it down to nearly the same level of performance and hearing only it's slight difference in tone. Who cares if your super-fast opamp can cleanly operate at 100v/us when all you are allowing it to do is work at 8v/us? So why not improve the circuit so that the hobbled part working at 6v/us can work at 10v/us and save your time and money for other things?
The FIRST thing anyone should do to a device is add local decoupling to the opamps. EVERY opamp should get 100nf from -V and +V to ground. Adding 100nf BETWEEN -V and +V is advised as well. That means you'll be adding 3x 100nf to EVERY opamp.
You should also add 100-220uf on both the -V and +V to ground within a range of 2-4 inches of every opamp. This means that if you have a group of opamps within about 4 inches of each other, add the 2 caps between them all.
There should also be 1000uf and 100nf on both the -V and +V to ground at the entry point of every board or anywhere power wires come from the power supply to the analog circuit. This should be sufficient for most systems to smooth out any step responses from the power supplies and give the power rails a healthy amount of bulk decoupling.
Most people think this is completely overkill, but I deal with high speed (4GHz) designs with multiple levels of digital analog mixed signals on boards up to 12 layers. I've seen all manners of distortion, ground/current loops, ground bouncing, signal coupling, crosstalk, intermodulation, compression, etc. You name it, I've seen it and had to figure out how to fix it. 9/10 times, if the design wasn't mine, it was poor decoupling or poor layout, which can go hand-in-hand very easily. I mean, if you build your mansion on top of quicksand, does it matter if you buy the finest appliances for your kitchen? Nope. Your foundation is going to fail and the rest means fuckall. Build your foundation on bedrock, and you can build yourself a skyscraper.
So if you have the itch to modify your gear full of IC opamps, DECOUPLE first. You just might be surprised. Unfortunately, most have already made their minds up to change the opamps at this point and very little can be done to change that, but it doesn't make it any less true that decoupling is job #1.
|
|
|
|
Post by svart on Oct 10, 2013 14:32:15 GMT -6
Rules for modification, part 2: keep your mods within reason!
I've seen all kinds of mods, both amateur and professional, on all kinds of gear. Some are simple, some are complex, some are just insanely cool, some insanely bad. I've seen people sell mods that claim a lot and do nothing, and I've seen folks sell mods that could have easily been sold for 2x the price and gotten rich.
In a lot of cases, especially in the DIY world, I've seen folks who expect way too much from their modified gear. They expect low distortion from high distortion parts, or they expect a colored sound from neutral parts, or they expect that modification will give them a sound that just isn't possible. Part of this happens through expectations, part by not being familiar with modification/electronics in general.
You wouldn't drive a Civic and expect Ferrari performance would you? What if you put a turbocharger on your civic? would it perform like a Ferrari? Nope. It'd be faster, but it's still a civic, and always will be, no matter how many parts you change.
So why would you expect your Mackie mixer to sound like an SSL/Neve/API console with a few mods? As a former believer in this situation and now existing as an outsider, I've come to notice that a lot of people do have the expectation that a few simple mods will make their Mackie into a super mixer, even if they don't even realize it. DIY forums have a bad habit of nurturing confirmation bias in people. People want performance increases for very little money so much, that they read these glowing reviews of magic modifications and instantly take that stuff to heart and imagine their wildest dreams coming true. This is never the case, and leads to the perpetual "upgrading" that people end up doing while trying to bring their dreams to fruition.
This leads me to mention some of the worst issues with upgrading.
1. Insufficient power supplies. Most folks focus on the upgrades that they believe will bring them the most performance. This means opamps and signal path(DC blocking) caps. Decoupling and power supply quality almost always suffers, especially as these faster and more power hungry parts are added to the mix. As mentioned before, manufacturers are notoriously cheap. They'll use the bare minimum power supplies that will work. In cases like Behringer designs, they might even use inferior parts knowing that the unit probably will die, leading to more sales as people replace them or have them fixed. All things considered, these supplies are usually not up to the task of supplying addition power to the system.
2. Insufficient layout/trace width/thickness. Besides the power supplies being underrated and the decoupling being poor, another big problem is insufficient traces. Having a huge power supply means nothing if you have thin traces that can't supply the current to the parts. Even large amounts of decoupling can't completely overcome this issue. Manufacturers will go with the thinnest copper they can get away with to save costs. PCB designers might not know better than to use thin traces for all signals on the board, including the power traces. I know of many that I have worked with that didn't and it caused me many a headache trying to figure out why my 5V source was now 4.2V at the receiving end. Thin, resistive traces, that's how. This is also true for GROUND! Having huge power rail traces is great, but not when you have a tiny little ground trace, then they offer you nothing. They and the opamps will be jerking that tiny ground around all over the place any you'll have all kinds of issues. Same goes for traces that are highly mismatched between -V and +V. Also applies to hand wiring. A 30ga wire won't handle the current a 10ga wire will. Why would a tiny little trace be expected to hand the current for a dozen opamps with ease?
3. Feedback, impedance circuits. Most opamps need feedback to constrain their gain or frequency response. Some need it to stay stable as well. Most opamps have a SPECIFIC range of feedback or input/output impedances they want to see for optimal performance. Swapping these parts for something completely different might put the new parts outside their optimal ranges and now you'll have noise or worse, instability. One of the worst offenders I see is swapping out low impedance opamps for the super high impedance FET opamps and not scaling their feedback networks. Now you've just created a noise source that may or may not be worse than your original opamp. I've also experienced the SAME model opamps from different manufacturers act completely differently on the same circuit. I exchanged a few TL074 for some ONSEMI MC33079. For the most part they worked fine. I then bought some ST MC33079 due to being on sale and they oscillated like crazy at high gain. Why? Because the circuit was designed for the slower TL074 and lacked frequency shorting caps in the feedback. I thought the MC33079 sounded better in this particular situation but was unstable without feedback caps and therefor unusable. Once the proper amount of cap was installed to squash the oscillation, they sounded absolutely NO different than the TL074 and the subsequent signal nulling test resulted in 99% nullification, meaning there was practically no measurable difference either. Why did I waste money and time on this? Because I read that they would be better. Obviously not the case. I've since gone back and done the same tests with other opamps with very similar results. Others will claim to have proof that their opamp of choice is somehow superior but I always find the results are never posted or they are dubious at best.
In summary, you aren't going to turn a tough flank steak into a ribeye, but you can tenderize it enough to enjoy it. Just don't expect miracles, and don't let someone trying to sell you a product or service tell you that you can achieve ferrari performance on a civic budget. it's not true.
|
|
|
|
Post by tonycamphd on Oct 10, 2013 14:57:02 GMT -6
great stuff svart! very useful as always, thanx
T
|
|
|
|
Post by svart on Oct 10, 2013 15:07:08 GMT -6
It's also worth adding that a lot of tube amplifiers and old timey class A devices get a bit of their "sound" from these same non-linearities/distortions that result from what is now considered poor design practices. "Sagging" power rails and high 2nd order distortions are hallmark for tube amps. The old time designers of high powered radio transmitters would look at amp designs and scoff at the poor design, not really "getting" that the distortion was somewhat intended.
I remember talking with a guy who designed tube guidance systems in old rockets and then worked for radio stations keeping up their old tube transmitters. He asked me what I was working on, and when I showed him the schematics of an old Fender amp, he was dumbfounded at the design, pointing out all the "problems" and noting that the distortion would be really high, especially when the gain was turned up. He couldn't figure out why I started laughing until I told him that the distortion was INTENDED! He just shook his head, like "kids these days and their toys" kind of look.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Oct 14, 2013 4:27:31 GMT -6
Great thread, svart!
Special thanks for rehabilitating the NE553x opamps. I even used them as the upgrade opamp in some cases where decoupling was good with very good results - it really depends on the implementation, but in many cases i would not come to the idea to replace them anyway... because their performance is not a problem at all.
I cannot find a reason to think that a change of these to something else would always gain something. Great audio opamps if implemented right, for decades, and for a reason, quiet and clean. Not only SSL used them successfully, also e.g. TAB(Telefunken) in their last big modular analog consoles...
Decoupling Captain Obvious tipp:
Use the backside of the PCB for placing the additional 100nF film caps. Don't use crowded PCB design as an excuse for not doing it.
And one on modification of mixers/consoles...
Even if you get the PSU and decoupling to be more than sufficiant for an opamp upgrade and calculated the power demand and put up a safety margin on top and everything....
In more than one case where the replace parts totally work as intended i found myself coming back to the original parts despite the better specs of the channel modified.
Simply because the channel lost too much of its character, that i actually liked (otherwise i wouldn't have bought the console...), and the slightly better noise figure did not justify the change in sound. Same can apply to the master section gain stage...
Often enough, if you do not like the sound of your board, you are better off buying one that you do like, instead of trying to make it into something different with time consuming and expensive modifications...
BR,
Martin
|
|
|
|
Post by svart on Oct 14, 2013 7:25:14 GMT -6
In more than one case where the replace parts totally work as intended i found myself coming back to the original parts despite the better specs of the channel modified. Simply because the channel lost too much of its character, that i actually liked (otherwise i wouldn't have bought the console...), and the slightly better noise figure did not justify the change in sound. Same can apply to the master section gain stage... Can't agree more. I spent years chasing the ghost of opamps. Most of the time, the cost wasn't justified as the performance just didn't live up to the hype. The super fast opamps who a lot of professional modders use either imparted a sound that built up over a dozen opamps in series and couldn't be gotten rid of. Once I figured out that I was using a lot of EQ to try to manipulate this sound, I ended up doing the unthinkable (to most), I started putting back the Ne553x and TL074 parts. I couldn't be happier. I still use some MC33078/33079 in my bus modules, mainly due to having a number of quad opamps on those channels but I took out and sold all of the high price super opamps. I think the super fast opamps do allow a more clinical reproduction but they are so dry and sterile. I thought I wanted that, but I realized later that I wanted neutrality over clinical correctness. |
|
|
|
Post by tonycamphd on Oct 14, 2013 9:16:26 GMT -6
hey guys, I don't disagree with the larger point, but Jim modded my master and a bunch of channel strips in my Soundcraft delta 200(which doesn't sound horrible stock), opamps are all upgraded, lots of good film caps, as well as the elimination of most electro caps, and quite a bit of power stabilizing engineering that's way over my head. The mods quite literally render the stock modules as noisy useless toys, the mods are really that good! We blind tested between the stock and modded strips with a few different tests, we cut it short because it was so absurdly obvious, it proved to be a waste of our time.
|
|
|
|
Post by svart on Oct 14, 2013 11:03:53 GMT -6
Well, the delta 200 wasn't known for it's good sound to begin with. It almost exclusively uses TL072 opamps, which are good in unity gain situations but aren't good in high gain applications or applications that require the parts to drive heavy loads.
But of all these things, the power system is notoriously poor. I'm sure the decoupling that Jim added did most of the work in gaining you the new sound.
|
|
|
|
Post by jazznoise on Oct 14, 2013 16:15:33 GMT -6
Great stuff! Realistically is there much to be gained in noise reduction through decoupling? I suppose I'm asking what's the craziest improvement you've seen caused by shunting all that noise to ground!
|
|
|
|
Post by svart on Oct 14, 2013 18:06:06 GMT -6
It depends on how noisy your power rails are. If they're real bad, chances are you'll reduce noise and distortion. On an old mackie mixer, I saw a 15db drop in noise floor just from outrageous decoupling... I think that's the most I've seen.
|
|
|
|
Post by jazznoise on Oct 15, 2013 4:34:55 GMT -6
It depends on how noisy your power rails are. If they're real bad, chances are you'll reduce noise and distortion. On an old mackie mixer, I saw a 15db drop in noise floor just from outrageous decoupling... I think that's the most I've seen. That sh*t is cray cray. I was thinking a 3-6dB would be well worth the additional cost. 15dB..were they allergic to caps in the Mackie R&D department? |
|
|
|
Post by svart on Oct 15, 2013 7:27:05 GMT -6
I think in the mackie, it was a lot of layout issues. Lots of thin traces trying to supply large amounts of opamps, with very sparse decoupling. Traces twisted around parts and around areas and through sections just to avoid routing on another layer (for cost savings) all over the place.
Anyway, I also ran some thick ground wires from some of the farthest sections over to the main chassis ground too, which probably helped too.
Also, when I measure noise floor, I do it with all sections set to unity gain and enabled. Most of the noise I cut out was hum and hiss and it went from somewhere around -45/50db to around -60/65db.
I think if you have a mixer that is noisy, you should take a look at a spectrum analyzer on the output and determine what kind of noise it is. You should see peaks at intervals of 60hz for hum (look for peaks at 60,120,180,240, etc) or general hiss. The hiss could be made up of all kinds of noise, from Pink (1/f) to white or shot noise from the electronics themselves when they use noisy opamps and poor quality resistors, or just use the wrong feedback values for the opamps.
|
|
|
|
Post by jazznoise on Oct 15, 2013 8:02:47 GMT -6
Well I'm in Ireland. So I look for 50,100, 150 etc.! But good advice. I suppose the thin traces also lead to alot of parasitic capacitance, since they'd be quiet close? Running a ground bar in a mixer is one I've heard of before. Nothing says f-u to ground loops like a thick piece of high quality copper.
|
|
|
|
Post by svart on Oct 15, 2013 9:20:21 GMT -6
Well I'm in Ireland. So I look for 50,100, 150 etc.! But good advice. I suppose the thin traces also lead to alot of parasitic capacitance, since they'd be quiet close? Running a ground bar in a mixer is one I've heard of before. Nothing says f-u to ground loops like a thick piece of high quality copper. Ah, I see. Yes, any harmonics of the primary line frequency that your country uses is where to start. Parasitic cap isn't terrible on power traces (unless of course it's so terrible that it kills step response of a power supply, but that's an entirely different discussion and not really applicable here). Series resistance/impedance on these excessively long traces is probably the primary problem in audio designs, with noise immunity being a close second. As I mention in the earlier posts, if the opamps draw a lot of current, either by slewing a very large signal or multiple opamps synchronously slewing the same signal, then the first source for instantaneous current would be the local decoupling and bulk capacitors. The next source would be the power supply itself, with it's current attempting to be drawn through those thin traces. Obviously this will lead to current starvation of the opamps with distortion and frequency response issues showing up in severe cases. A lot of folks don't experience this problem in a severe enough form to understand why their audio sounds muffled and harsh, they just think it's the "opamp sound" and thus reading glowing reviews about changing opamps and setting off to run on the upgrade treadmill. Those newer opamps can be both faster and more(or less in some cases) efficient but overcome some of the issues of decoupling by having raw speed. I suppose it's kinda like drowning out the noisy muffler in your car by turning up the radio. In any case, I feel like it's my duty to draw attention to the basic issues in these designs before ever suggesting that one do any opamp swaps. |
|
|
|
Post by tonycamphd on Oct 15, 2013 12:35:09 GMT -6
Great stuff! Realistically is there much to be gained in noise reduction through decoupling? I suppose I'm asking what's the craziest improvement you've seen caused by shunting all that noise to ground! I have an opinion on this, dynamic range improvements, and a few db of noise reduction adds up over lots of equipment, and a bunch of channels, with proper gain staging, and the proper use of compression a low noise floor, wide dynamic range= POWER!!!!!! in your mixes. I love POWER!!!! moooohhahahaha! |
|
|
|
Post by svart on Oct 15, 2013 12:59:23 GMT -6
Great stuff! Realistically is there much to be gained in noise reduction through decoupling? I suppose I'm asking what's the craziest improvement you've seen caused by shunting all that noise to ground! I have an opinion on this, dynamic range improvements, and a few db of noise reduction adds up over lots of equipment, and a bunch of channels, with proper gain staging, and the proper use of compression a low noise floor, wide dynamic range= POWER!!!!!! in your mixes. I love POWER!!!! moooohhahahaha! I can't disagree but I've always found my definition of power as being well designed attack and decay of sounds. A super fast transient doesn't have "power", but a rounded transient with a longer decay will be perceived as harder hitting, thus the popularity of the SSL bus compressor and it's ability to slowly round a transient to a hard thump rather than a harsh crack. I think getting the noise floor down allows the ear to hear more of the signal without being veiled. This leads to clarity and less confusion while the brain is processing what it perceives. It's well known that the brain will "fill in" signals that it expects but doesn't hear, so clinical clarity is good in the fact that the brain can perceive the signal as it was intended to be. However, harmonics from various sources are still an expected and popular source for adding "color", thus the popularity of tube and transformer containing products. These are notoriously "noisy" but the spectrum is psycho-acoustically pleasing due to the inherent harmonic content. These harmonics mix and combine to create a "sheen" (as some call it) that is sought after and expected. Also, we should look at other designs like the API 2520 opamp which is reportedly somewhere around 2-3V/us and generally combined with transformers that test out to be even slower, meaning that they can't even reproduce the whole human auditory spectrum, yet are coveted by most who use them.. So I have to ask, why would certain designs like the API type preamps, which CLINICALLY test out *worse* than pretty much any audio opamp on the market, have much more appeal than the cheap IC opamps which are almost universally disregarded as trash? Hype, misunderstanding and poor usage on the IC opamp's part. It's gotten more than it's share of bad raps due to poor designs since it's inception. Even the worst hated IC opamps on the market technically have less noise than the API2520, but you'll be hard pressed to find an expensive 2520 in cheap, commodity, designs that would likewise give it a bad name when used poorly like you see day in and day out with the IC opamp. I've said it before and was dismissed quickly without reason, the IC opamp is very similar in design to the discrete opamp when you look at the schematic of the internal structure vs. the parts used to make a discrete opamp. There is no magic in either one of them, just basic designs. Another case in point.. www.jensen-transformers.com/as/as017.pdfI would bet that Jensen/NE5534 preamp would sound stellar when compared to anything. Why? because it's well thought out. the impedances are matched fairly well, the compensation is right, the feedback is in the correct range and has frequency compensated. Most importantly, it states that "adequate power supply bypassing is necessary, use ground plane techniques". This is the basis for low noise and stability that all opamps rely on to work correctly. |
|
|
|
Post by tonycamphd on Oct 15, 2013 13:10:08 GMT -6
^ nicely stated, and cottons to my understanding, i personally make music with a lot of rhythmic "silence" in it, the empty space being truly empty, really adds to the impact/power, not as important for wall of sound crush stuff i suppose? As far as the api 2520 thing, i don't get it either, but I absolutely LOVE what they do! It's part of the voodoo i believe exists for whatever reason in tubes, transformers, discrete opamps, these are the 3 things that defy logic to me?, worse specs, better sound for certain things?
magic and music, love it.
|
|
|
|
Post by jazznoise on Oct 15, 2013 14:05:18 GMT -6
Great stuff! Realistically is there much to be gained in noise reduction through decoupling? I suppose I'm asking what's the craziest improvement you've seen caused by shunting all that noise to ground! I have an opinion on this, dynamic range improvements, and a few db of noise reduction adds up over lots of equipment, and a bunch of channels, with proper gain staging, and the proper use of compression a low noise floor, wide dynamic range= POWER!!!!!! in your mixes. I love POWER!!!! moooohhahahaha! Well, to be fair, the rule with nose is it's a 1/root of N improvement, right? That is that taking 10 volts of noise and adding 1 volt of noise gives you 10.1V of noise? Very interesting! That makes a lot of sense but I'd never thought about it that way. Well I learned something new today. Sort of off topic, but I was finishing lunch today and got a call from my old college department. They're asking if I want to do a research masters. You'd never guess what my research topic would be centered around, now would you? |
|
|
|
Post by tonycamphd on Oct 15, 2013 14:08:59 GMT -6
Well, to be fair, the rule with nose is it's a 1/root of N improvement, right? That is that taking 10 volts of noise and adding 1 volt of noise gives you 10.1V of noise? Jazz, i'm sorry, i don't understand the Portuguese language....  |
|
|
|
Post by jazznoise on Oct 15, 2013 14:58:06 GMT -6
Well, to be fair, the rule with nose is it's a 1/root of N improvement, right? That is that taking 10 volts of noise and adding 1 volt of noise gives you 10.1V of noise? Jazz, i'm sorry, i don't understand the Portuguese language.... My mistake. It's Vtotanoise = Square root of (Va^2 + Vb^2 ...+Vz^2). Sum of the squares, as long as the noise is totally decorrelated. |
|
