Allen & Heath System 8 Mk1/2 recap and mods

[musgrave]

as for local decoupling, the 33R/100u/25v, the 33r goes between power and pins 4&8(7). The R is in series with the power source and ic pins and the C is between the R and the IC power pin. it does 2 things. first it keeps the currents on the board returning back to the power. if you have 2 inverting stages of different amps together, there is a tendency for the positive going electrons to leave the amp, and then go thru the negative flowing amp next door rather than back to the supply. oscillations can occur (not usually) but it sets up a tank circuit effect or 'climate' for oscillations a more nuanced word. so this keeps electrons "fresh". fluffy words to paint a concept. you find these resistors on all higher end consoles. then the Cap at the pins is a storage device at the amp. like multiple car batteries the trunk of a hi-end car stereo with thunderous bottom end. for kick's and other transient signals, the current for the waveform comes from the cap which is then resupplied in a less transient time line by the supply. I can see this on my distortion analyzer where I continuously monitor the signal and the distortion component on my scope. channel strips with audio caps only replaced and old power decoupling caps show a second order element from the power being pulled across the module. replace the caps noise is gone. The 4700 on each channel strip is the same medicine for the whole module. the ground plane strapping helps too. ground in audio is like diving in a lake from a dock. lean forward and push off. the doc firmly anchored allows you to dive effortlessly. Dive in from a raft and it's a belly flop for sure. because the raft moves back as you push off. so I stiffen up the (lower) resistance of ground and leave the power traces alone (anemic?) so as I get hot with signal I invite a little 2nd order as you get near clipping for color. the amp will still have transiet oomph from the cap, and the second order will be module wide (the R of the trace drops the current that creates the voltage) and become apart of the signal.

 

[Lasselakken]

Thanks, Musgrave. That made it much more clear to me.
I guess that is the better (and more elaborate) way of making local decoupling compared to just sticking a 0,1 film cap from each power rail pin on the IC and to a nearby ground trace. Or does that method offer other advantages?

 

[Whoops]

I said “based on my experience”. My experience, thank God, is different from yours so it doesn’t need to annoy you so much, what’s worse for me may be the same for you, and that’s perfectly fine.

Don't confuse interest with annoyance.
I'm really interested to know,
"Worse" in what sense?
Can you define what you mean by "slightly worse" and provide technical proof of that?

 

[musgrave]

any kind of capacitance from the power pin to ground will aid in stability. 10 µF to ground 22 to ground 0.1 it’s all good. Sometimes you can get a small drillbit and drill holes through the PCB to catch power or ground on the other side. Also sometimes where I have ground plane on top and bottom, I’ll drill a tiny hole and put a lead through and soldered on both sides just to connect the planes together for more massive ground. Having electron storage out by the opamp‘s helps extend the bottom end. And the series resistance is just doing it up right, but not mandatory whatsoever. 0.1 at the amplifier in a pair of 4700s near the connector where the module power comes in it’s really good to. Springing for a pricey OPA bifet for the EQ helps. EQ seem to love bifets. Nothing wrong with a 5532 or 34 everywhere else. Changing the output resistor from 50 or 100 down to 22 helps sonically to. you just need enough R at the output of the amplifier to swamp out any stray capacitance from the feedback loop.here is a sound workshop module

 

[Newmarket]

I mentioned Z5U because I have a good source of these in axial form, that seemed practical. I’ll get X7R instead.

Is local power rail decoupling by the ICs generally advisable, or is it only necessary if I change to different op-amps?

Do you think I could improve anything by putting in new TL072’s, or would they perform exactly the same as the original ones from ca.1984?

Axial ? I'd recommend using SMT chip parts there. The SMT chip package reduces lead inductance (given that there are no real leads) and so decoupling/bypass is more effective.
Since you are modifying anyway I'd definitely suggest to add in local decoupling where it is 'missing'.
tbh I'm not aware of any 'ageing' effects with op-amps but others may have some experience ? It's not something I've ever factored into my design work.

 

[Lasselakken]

Thanks again, Musgrave. That is great inspiration - I guess I will have to experiment on the actual circuit to find out which approach to go for (and of course the amount of available space on the PCB will factor in too).

..bipolar for C7, because actual DC polarity at this point is unpredictable.

In our old Amek, we'd measure DC polarity and mount the capacitor accordingly (the polarity follows a given set of transistors, so only need to be re-measured if a transistor is replaced)..

I have a source for a bi-polar 1000uF cap. As I understand, it would also bo possible to place two polarized caps in series (+ and + connected or - and - connected) to get the same result. I guess I would have to use two 4700uF in series to get around 2200uF of capacitance, right?
I am tempted to try what you described for the Amek, though. Is it just a matter of removing the cap and measure DC on each side to find out which side is more positive or negative?

 

[Newmarket]

The R is in series with the power source and ic pins and the C is between the R and the IC power pin. it does 2 things. first it keeps the currents on the board ...

Apart from filtering/bypass etc aspects of series resistors in the voltage lines there's the safety/protection function of them as fuses.
So that if an op amp went short circuit the resulting current should cause the resistor to fail open circuit and give a degree of protection to the power supply and pcb .
The applicability of this obviously depends on the power arrangements.
For this protection the resistors should be non-flammable (NFR) types.

 

[gyraf]

Just measure DC voltage across the cap already in there..

 

[musgrave]

For power decoupling of the amps or additional storage like the 4700 the caps need to be polarized. yeah non polar will do something, but for maximum efficiency polarized should be used, they will be across a fixed potential like a batterie, so one in a bipolar will be backwards (reversed polarized) and is counterintuitive and sort of just plan wrong, bi-polar is only for coupling. and at the end of the day I think I remember the math as for a 100u at the output of an amp, under clipping at 50hz on =/-16, the drop across the cap is .5 volts AC. the DC is 0v so in a bipolar, you are still dropping .5ac across both caps (making a bipolar) at 50hz. so much for bipolar's invincibility. so this unescapable reverse voltage of .5 volts at the instantaneous peak of a 50 hz wave form (instantaneous as in maybe 45 deg of the 360 degree time period) will cause internal heating of the part and a minor but real derating of the voltage spec. a 5mm-6.5 mm part has an average life of 1000hr spec. an 8-12mm part has 2000 and then you can get into 4000 parts after that thru selection. I recommend a 25v or 35 v polarized part of 8mm or more, with a 4000 hour rating and 105 spec so it is less sensitive to the .5 vac reverse voltage on the kick drum channel. if we were in a listening room with 5.000 dollar amps and speakers most of what I have seen on this thread would be part of the weakest link theory on who sounds best with the best staging and I can hear the rosin on the bow comparisons. but we are trying to get tracks with personality that stand out in a mix and convey the emotion of the music. punch and dynamics are your best friend. put in a big ol cap for coupling so the phase shift at lo freq is best, don't worry about bipolar or lead inductance, don't use parts under 8mm. the lower your ground impedance the tighter the bottom, lower the opamp output and power resistors so more current flows (reduce series current limits - ohms law) and worry about the song. you will sleep better at night. lol again I might be just a fool what the heck do I know...

 

[Lasselakken]

Thanks again everyone for helping out here.
I am about to get ready to shop for components to recap the master section, two output cards (busses) and two input cards.

My plan for the electrolytics on the output card is:

Output PCB	Function	Org. value	New value
C1, C3, C4, C6	coupling	47uF	47-220uF?
C7, C8	Power rail decoupling	100uF	220uF/25v or more
C9, C10, C11	coupling	10uF	22-220uF?
C12	Meter output	47uF	47uF

Should I just go for 220uF for C1, C3, C4, C6, C9, C10 and C11 or is that too high in some places?

I have attached the schematic for reference.

 

[Newmarket]

I think I remember the math as for a 100u at the output of an amp, under clipping at 50hz on =/-16, the drop across the cap is .5 volts AC. the DC is 0v so in a bipolar, you are still dropping .5ac across both caps (making a bipolar) at 50hz. so much for bipolar's invincibility. so this unescapable reverse voltage of .5 volts at the instantaneous peak of a 50 hz wave form (instantaneous as in maybe 45 deg of the 360 degree time period) will cause internal heating of the part and a minor but real derating of the voltage spec. a 5mm-6.5 mm part has an average life of 1000hr spec. an 8-12mm part has 2000 and then you can get into 4000 parts after that thru selection.

Most of your advice seems sensible to me but I don't get your point about bi-polar above.
No one claims that a bipolar cap (or the equivalent "Back to Back" arrangement of polarised caps) somehow has less voltage drop ? But you seem to say that it is a problem. The point is simply that a bi-polar cap' avoids the situation where a cap is reverse biased.
Or am I not reading what you mean correctly ?
As it happens I think it's been reasonably well shown over time (cue jeers !) that a small reverse voltage compared to the voltage rating has little, if any, effect on audio performance and reliability. But you wouldn't choose it as a preference obviously. Depending on the circuit details the polarity might be defined or you can use biasing networks to define a correct dc bias - from memory some AMEK and SSL designs come to mind but don't quote me on that.

re lead Inductance - tbh it's not something I'd worry about either in this application. But the SMT chips are generally convenient to use with thru hole parts - 'bury' one terminal in the solder joint of a pin and take the other end to 0V or relevant voltage rail - eg using 30AWG patch wire. Avoids space issues and/or need to drill board. Reduced lead inductance is a 'free benefit'.

 

[neil.johnson]

Thanks again everyone for helping out here.
I am about to get ready to shop for components to recap the master section, two output cards (busses) and two input cards.

My plan for the electrolytics on the output card is:

Output PCB	Function	Org. value	New value
C1, C3, C4, C6	coupling	47uF	47-220uF?
C7, C8	Power rail decoupling	100uF	220uF/25v or more
C9, C10, C11	coupling	10uF	22-220uF?
C12	Meter output	47uF	47uF

Should I just go for 220uF for C1, C3, C4, C6, C9, C10 and C11 or is that too high in some places?

I have attached the schematic for reference.
View attachment 85680

I would take some time to understand what every capacitor in that circuit is doing, what effects the other components around will have, and then choose capacitor values suitable. Going for bigger or smaller assumes the original designers didn't know what they were doing, or you end up making pointless changes. For example, the power rail decoupling capacitors are 47uF according to your schematic. Coupled with the 22R series resistors puts the -3 dB point at around 154 Hz, compared to most opamps who's PSRR starts dropping around the 300-500 Hz region. In that case increasing it will have little effect.

C11 is just the coupling for the VU meter driver, and that presents a 100k load to the capacitor. That combination of R and C has its -3 dB point at 1.6 Hz. So again, leave it alone.

And so on. For every capacitor go round the circuit either estimating or noting the worst case (lowest) resistor load and then you can calculate the -3 dB point, and then decide based on that which capacitors need enlarging.

Neil

 

[musgrave]

I am down with leaving the meter alone particularly on the out put side. The thing in manufacturing is you need to make money so things are usually done against a price point. that's where the values come from. if you are making something and you add 1 cent, that add's against the total bill of materials; the values of c cost more as you get larger, the cost of r's stay the same for different values in the same package. pick your C on cost, pick your R on math. this forces you into higher values and higher impedances of R for less pricy smaller C. on the out put R values of amps things are made to be safe and minimize return and calls to the dealers or tech staff about stray radio stations in the output. these are the things the engineering manager is enforcing upon the employees. no one where I have worked did the math on the freq point of the power decoupling, its already DC. its what value will keep it stable, and how few parts do we need to pay people to solder, yes there are machines, wave solder, but its the point Im making. smaller BOMs cost less, not just in materials but in process from the warehouse to the assembly floor. end users want performance and spend on how easy money comes and where they perceive the return is greatest for the task they have at hand. go 220/25v and down to 6.5 v if size matters. leave the 47 on the meter output as its a rectifier and the smoothing against the input of the meter device will determine the ballistics.
for bipolar and my math,Caps work on charge like when you rub a ballon of your sweater and it sticks. So as you present a charge say the positive flow of an amp to a cap, the other side will attract the opposite charge and a potential or voltage occurs. no electrons actually cross, we probably all know this, but just to start the story. An electrolytic has a positive charged juice inside to amplify this effect, there's the need for polarization of the device, at a frequency and voltage there will be an instantaneous impedance based on the moment of phase, which causes a voltage drop or a part of the overall voltage across its terminals. Robin Porter and I were in the Neve LA office talking about cap failure in the VR's and he was maintaining the cap had 0 volts, so we broke out a calculator and voila' at 50 hz at max it was .5 vac. so in a bipolar as the amp charges the first cap surface, the second surface attracts the opposite charge. this intern leaves the next surface (1st of the second internal device) charged the opposite that then leaves the last surface connected to the other terminal of the opposite charge. add the math and you see what I mean about just use a single cap. true non polar's don't have the positive juice inside. I used to use bipolar when I recapped consoles, now I just use better parts.
on the V3 the mix buss used 2200 coupling, the VR's 470, I forget what SSL4000 use but they add an inductance to swamp out the stray C of the mix buss. I use a 4R7 resistor, it swamps the mix buss C and I put 120pf around the mix amp fb to keep it stable. as you get neat the -3db point you also have phase shift so the bigger the C the lower the -3db point pushing the phase shift down with it. you see a 10 uh on the op of VR'r and 4K, this is to push the phase back in the other direction. On a mix at Conway once on the SSL K, the engineer called me in to show me the kick was out of phase looking at the waveform in protools. the kick was a composite of the kick plus an oscillator at some low freq. So I grabbed the test gear and no, no phase flip from PT, but going thru the channel at 20hz there was a 50 degree shift at the insert...
The main thing is get tracks that talk to you, that have personality and tell you a story. manufactures have to make weird decisions and honestly some guys in the engineering dept are just weird and have never been in a studio... Im not the smartest, that guy Abby teaches me so much, he's the bomb! but I have been in rooms and labs with the most gifted producers and audio electronics designers and I have learned a lot. the main is 'how does it sound'. I wish you the best on your recap. you can see I don't post much but I read and learn all the time. lol god speed my friend this is where I slip back into the shadows.

 

[musgrave]

ok one more, newmarket you are right on the money, no disrespect from me. and I have used smt,s too, so easy one side on the trace one side on the plane done! thanks for taking the time with others, where I have not. peace out to you too!

 

[Lasselakken]

Thanks again - this is very educating.
I will definitely consider going for SMT for the local bypass caps even though it seems a little scary for someone who has never done anything but through-hole.

I would take some time to understand what every capacitor in that circuit is doing, what effects the other components around will have, and then choose capacitor values suitable.
(...)
For every capacitor go round the circuit either estimating or noting the worst case (lowest) resistor load and then you can calculate the -3 dB point, and then decide based on that which capacitors need enlarging.

You're absolutely right - I am trying to achieve that understanding.

Here are some questions about caps in the input module, that I'm still struggling with:

Leave the phantom blocking ccaps at 47uF. Increasing them could create other problems. 47uF is plenty.

How come you recommend not going higher than 47uF for the phantom blocking caps (C3 and C4)?

I would make the following changes:

- all 10uF/16V change to 22uF/25V except C11 - 100uF/25V
- C3, C4 - 47uF/63V
(...)
- C13, 14 - 220uF/25V, not higher
(...)
P.S:
You can use the same value for capacitors C11, 13, 14, 29 of 220uF/25V, that way you will reduce the total cost because you will get a better discount for larger purchase.

Why not make C13 and C14 (power rail decoupling) larger? It seems that others pretty much make it as big as possible? Is it because anything larger than 220uF might not really make a difference, or do you recon it will actually make something worse?

If C11 and 29 can be 220uF, couldn't C23 and C27 be 220uF as well?

C1 and C2 at the line input are 10uF - would it cause trouble to raise them to 220uF as well?
C8 and C9 around the transistors are also 10uF - could I benefit anything from raising them to 220uF or would that cause trouble?

I have attached the schematic for the input module again here for reference.

 

[neil.johnson]

Thanks again - this is very educating.
I will definitely consider going for SMT for the local bypass caps even though it seems a little scary for someone who has never done anything but through-hole.


You're absolutely right - I am trying to achieve that understanding.

Here are some questions about caps in the input module, that I'm still struggling with:

How come you recommend not going higher than 47uF for the phantom blocking caps (C3 and C4)?

Larger capacitors can supply higher fault currents in case of unintended shorts:
https://www.aes.org/e-lib/browse.cfm?elib=15251

Why not make C13 and C14 (power rail decoupling) larger? It seems that others pretty much make it as big as possible? Is it because anything larger than 220uF might not really make a difference, or do you recon it will actually make something worse?

You're putting more start-up load on the power supply. Yes there's the 22R series resistor, but if you do this for all boards in the mixer that's still a lot more initial load than the power supply may be able to handle.

If C11 and 29 can be 220uF, couldn't C23 and C27 be 220uF as well?

Who said they could be so big? Again, calculate the worst case load and then calculate the capacitor value for the desired -3 dB point. Let's say worst case load is 10k, and put the -3 dB point at 2 Hz, then C = 8 uF. So you might consider going up to 22 uF, but 220 uF would be rather excessive.

C1 and C2 at the line input are 10uF - would it cause trouble to raise them to 220uF as well?

Again, work it out - they feed into 22k resistors.

C8 and C9 around the transistors are also 10uF - could I benefit anything from raising them to 220uF or would that cause trouble?

Did you buy a truck load of 220 uF capacitors? I'm seeing a common pattern here....

I have attached the schematic for the input module again here for reference.

Thanks. I know it quite well...
https://www.njohnson.co.uk/index.php?menu=2&submenu=4&subsubmenu=5
Neil

 

[TwentyTrees]

...
Thanks. I know it quite well...
https://www.njohnson.co.uk/index.php?menu=2&submenu=4&subsubmenu=5

Neil - just to say, your site was an absolute godsend when I picked up an unloved System 8 Mk2 (16>8) a few years ago for the princely sum of £50! Thanks for sharing your knowledge. I loved that desk, as limited as it was, and if it wasn't so damn huge I'd still have it today (had to downsize a couple of years ago unfortunately). Got one of the best acoustic guitar sounds of my life with those preamps and a pair of Oktava MK012s in XY.



Ok, sidebar over. OP - good luck with your project, the System 8 is well worth the love!

 

[Lasselakken]

Again, calculate the worst case load and then calculate the capacitor value for the desired -3 dB point. Let's say worst case load is 10k, and put the -3 dB point at 2 Hz, then C = 8 uF. So you might consider going up to 22 uF, but 220 uF would be rather excessive.

Thanks, Neil. I see your point here. And thanks for the link - what an inspiring site!

What I haven’t quite understood yet, is how to estimate the load on a coupling cap. What is the load on C8 or C9 in the input circuit? Or C11?

 

[neil.johnson]

Thanks, Neil. I see your point here. And thanks for the link - what an inspiring site!

What I haven’t quite understood yet, is how to estimate the load on a coupling cap. What is the load on C8 or C9 in the input circuit?

C8 and C9 are in series with 22k.

Or C11?

Bit trickier as there are two cases. The simpler one is with something plugged into the insert jack. Reasonable to assume 10k, needing 8 uF.

The more complex case is when there's nothing inserted, then you have to follow the path through the EQ and the pots on the way out. I think the worst case is with EQ out: you have the two AUX pots and their bus feeds, plus the main fader (no wiper load), and the PFL bus feed, and the peak LED driver (annoying half-wave rectifier, but assume it's 22k) -- all those in parallel. And you've got the 47 uF return DC block capacitor as well. I reckon you have about 3k. The 100k bleed has marginal effect, same as the 22R opamp stabiliser. So for 2 Hz point you want a total capacitance of at least 27 uF. That is split between the 47 uF input and the output capacitor. So you might need 63 uF, or stick with 47 uF for simplicity while still close to the 2 Hz target (2.25 Hz).

Neil

 

[neil.johnson]

Neil - just to say, your site was an absolute godsend when I picked up an unloved System 8 Mk2 (16>8) a few years ago for the princely sum of £50! Thanks for sharing your knowledge. I loved that desk, as limited as it was, and if it wasn't so damn huge I'd still have it today (had to downsize a couple of years ago unfortunately). Got one of the best acoustic guitar sounds of my life with those preamps and a pair of Oktava MK012s in XY.

View attachment 85690

Ok, sidebar over. OP - good luck with your project, the System 8 is well worth the love!

16:8 -- very nice!!

Neil