Hi all
I've been slowly recapping and modding my Soundcraft 1600 console, and have some interesting measurement results I thought I'd share.
I have done all mods in sets of two channel strips, and measured the results with a MOTU 828mk3 interface using Room EQ Wizard.
By far the biggest sound improvement was recapping the power supply. Do this first if you're recapping an old console for sure! Also I would recommend to replace any TO-3 transistor sockets, I had one that was on the way out (weak contact pressure on the pins) and might not have noticed if I hadn't specifically looked at the socket. The sound improvement was noticeably tighter bottom end throughout.
I haven't been able to hear any improvement in sound with any opamp swaps.
Mods done to channel strips are:
* Recapped electros with Panasonic FC (Green traces)
* Recapped with Pana FC, and also recapped film caps with WIMA FKP2 (Blue traces) Some values are not available so used other brand of non-metallized film cap for those
* Recapped with Pana FC, and swapped out input transistors with ZTX951 (Orange traces)
The red traces are an unmodded channel strip with original 80's components.
where can i upload my pictures online
As you can see, recapping gave me back my bottom end, and everything was calibrated to within 0.1dB when making measurements.
The distortion figures are pretty interesting. Going to Panasonic FCs gave a good reduction in distortion over stock, until we get to the top end which is actually a bit worse.
But recapping the old films to new WIMA films gave a massive improvement, which is pretty audible. It actually IS worth recapping those old film caps in this case.
It looks like while the ZTX transistors have lower noise than the 2N4401 stock, they are a bit more distorty in this circuit.
Hope this helps someone! I'll post the actual cap values I used when I get a chance; I collated several old threads from many sources and ended up with some solid recommendations that improved bottom end response, as shown above. (Edit: see post #140)
Cheers
Darren
I've been slowly recapping and modding my Soundcraft 1600 console, and have some interesting measurement results I thought I'd share.
I have done all mods in sets of two channel strips, and measured the results with a MOTU 828mk3 interface using Room EQ Wizard.
By far the biggest sound improvement was recapping the power supply. Do this first if you're recapping an old console for sure! Also I would recommend to replace any TO-3 transistor sockets, I had one that was on the way out (weak contact pressure on the pins) and might not have noticed if I hadn't specifically looked at the socket. The sound improvement was noticeably tighter bottom end throughout.
I haven't been able to hear any improvement in sound with any opamp swaps.
Mods done to channel strips are:
* Recapped electros with Panasonic FC (Green traces)
* Recapped with Pana FC, and also recapped film caps with WIMA FKP2 (Blue traces) Some values are not available so used other brand of non-metallized film cap for those
* Recapped with Pana FC, and swapped out input transistors with ZTX951 (Orange traces)
The red traces are an unmodded channel strip with original 80's components.
where can i upload my pictures online
As you can see, recapping gave me back my bottom end, and everything was calibrated to within 0.1dB when making measurements.
The distortion figures are pretty interesting. Going to Panasonic FCs gave a good reduction in distortion over stock, until we get to the top end which is actually a bit worse.
But recapping the old films to new WIMA films gave a massive improvement, which is pretty audible. It actually IS worth recapping those old film caps in this case.
It looks like while the ZTX transistors have lower noise than the 2N4401 stock, they are a bit more distorty in this circuit.
Hope this helps someone! I'll post the actual cap values I used when I get a chance; I collated several old threads from many sources and ended up with some solid recommendations that improved bottom end response, as shown above. (Edit: see post #140)
Cheers
Darren
Last edited:
Here's distortion plots for a stock subgroup, and a subgroup recapped with Panasonic FCs:
Not as marked a difference, but still an improvement. It also picked up a few dB under 20Hz:
The Soundcraft 800b, 600b and 400b are all pretty similar to the 1600 so I'd expect you would see similar results with those desks.
Not as marked a difference, but still an improvement. It also picked up a few dB under 20Hz:
The Soundcraft 800b, 600b and 400b are all pretty similar to the 1600 so I'd expect you would see similar results with those desks.
You want to look at the green trace (Panasonic FC recap) and the orange trace (same, but also with ZTX Q's).
The ZTX strips had about 6dB more distortion at 10Hz and 20kHz, and maybe 1dB more around the 1kHz mark.
Next two strips I recap, I plan to give the 'trans-amp mod' a go (as well as recapping the electro and film caps) and see what difference that makes in comparison. I have 10x 1600 channels left to recap, and then 10x 400b channels in a sidecar after that, so plenty of room left for experimenting!
Hrm... maybe I need some transformers in there...
Seeing a further >12 dB mid-band improvement just from swapping polyester film caps out for polypropylene is quite shocking, really. Mind you, it seems that there are a lot of tone stages using them, and the distortion spec for this mixer at 0.01% isn't that awe-inspiring to begin with...
Worse performance with the Zetex devices may be rooted in lower beta and higher Ccb. All hail the mighty 2N4403!
(Quite franky, these transistors with their Rbb' of 17 ohms or thereabouts are amply sufficient for a nominal 200 ohm mic input.)
Going through the schematics, I noticed this and that:
1. Phantom power supply arguably is not as low-noise as it could be - reference zener ZD1 is run a bit on the hot side (5-11 mA), and there's a measly 4.7 µF of parallel capacitance (C9). I'd look for a matching low-noise zener of adequate wattage and would up C9 to whatever 25 V part will still easily fit - comparable designs have seen several hundred µF there.
2. The headphone amplifier (using a Class B buffer with "cheating resistor" to mitigate the worst crossover distortion) isn't the last word in fidelity.
3. The mic inputs have obviously survived for long enough without major damage, but still not seeing any protection whatsoever makes me a bit uneasy.
Worse performance with the Zetex devices may be rooted in lower beta and higher Ccb. All hail the mighty 2N4403!
(Quite franky, these transistors with their Rbb' of 17 ohms or thereabouts are amply sufficient for a nominal 200 ohm mic input.)Going through the schematics, I noticed this and that:
1. Phantom power supply arguably is not as low-noise as it could be - reference zener ZD1 is run a bit on the hot side (5-11 mA), and there's a measly 4.7 µF of parallel capacitance (C9). I'd look for a matching low-noise zener of adequate wattage and would up C9 to whatever 25 V part will still easily fit - comparable designs have seen several hundred µF there.
2. The headphone amplifier (using a Class B buffer with "cheating resistor" to mitigate the worst crossover distortion) isn't the last word in fidelity.
3. The mic inputs have obviously survived for long enough without major damage, but still not seeing any protection whatsoever makes me a bit uneasy.
Well mixers use high signal levels and polyester distortion goes up as the square of the signal voltage IIRC.
Which is what you'd expect from dominant 3rd order distortion, as is typical for nonpolar capacitor nonlinearity. 3rd climbs 18 dB per 6 dB of signal in small-signal terrain, or 12 dB relative to fundamental.
Yeah, high levels are a likely contributing factor, but I was still surprised to see the effect being that dominant given the mix of TL072 and NE5532 opamps. They were treating their TL072s sensibly though - if you give them supplies of close to max, keep their output loading at 20kOhm and up and use them inverting if possible, these really aren't too terribly bad. They may still be accounting for the gentle rise towards the high frequencies that is visible now.
Yeah, high levels are a likely contributing factor, but I was still surprised to see the effect being that dominant given the mix of TL072 and NE5532 opamps. They were treating their TL072s sensibly though - if you give them supplies of close to max, keep their output loading at 20kOhm and up and use them inverting if possible, these really aren't too terribly bad. They may still be accounting for the gentle rise towards the high frequencies that is visible now.
Last edited:
Interesting....what were the original stock transistors? (My Soundcraft 200 used PN4355s) Was there an improvement in noise performance with the ZTX951s?
Hmmmmm. On the ZTX951 or PN4355 datasheets, I don't see a spec for Ccb, which is 8.5pF on the 2N4403. Beta figures are 200 for the ZTX951, ~100 for the 2N4403 and PN4355. Please explain your suggestion that the 2N4403 is superior. How is the collector-to-base capacitance related to distortion?Worse performance with the Zetex devices may be rooted in lower beta and higher Ccb. All hail the mighty 2N4403!
Horowitz/Hill say Cob typ = 5.5 pF for 2N4403 and 74 pF (!) for ZTX951. That's quite a bit of rather nonlinear capacitance right there. Can't imagine that's doing high-frequency performance any good. You don't go overboard with Cob in LTP inputs if you can help it. Aside from absolute Cob, its nonlinearity may also vary.
You are correct in that ZTX951 beta should in fact be higher than 2N4403 beta, up to 10 mA... I temporarily forgot that the 2N4403 has about the worst leakage in any small(ish)-signal transistor. Not sure what the low-frequency difference is all about then.
I wouldn't think it's Early voltage either, the ZTX951 is still given with 120 V and the 2N4403 according to Mark Johnson clocks in at close to half that. All rather odd.
You are correct in that ZTX951 beta should in fact be higher than 2N4403 beta, up to 10 mA... I temporarily forgot that the 2N4403 has about the worst leakage in any small(ish)-signal transistor. Not sure what the low-frequency difference is all about then.
I wouldn't think it's Early voltage either, the ZTX951 is still given with 120 V and the 2N4403 according to Mark Johnson clocks in at close to half that. All rather odd.
Well, 74 pF isn't really much of a factor in audio-frequency applications----many op-amp circuits use 100 pF in parallel with feedback resistance.
Please explain the avoidance of high Cob.
If anyone is modifying an existing vintage mixing console, I HIGHLY recommend this BRILLIANT power supply modification.
D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg.
This is for LM317/337-regulated supplies, and reduces noise by an astounding 30 db! And it's less than 10 parts, costing a few $$ at most.
D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg.
This is for LM317/337-regulated supplies, and reduces noise by an astounding 30 db! And it's less than 10 parts, costing a few $$ at most.
We are talking voltage-dependent Miller capacitance that quite arguably is effectively outside the feedback loop, and whose effect is only cancelled out by transistor matching, and even then only for even-order distortion components. I believe that the typical degradation in input impedance linearity towards high frequencies that you see in Samuel Groner's opamp measurements is directly related to nonlinear input device capacitance, with the OPA627's superior performance over other JFET input parts (otherwise noted for rather high amounts of nonlinear capacitance) being explained by it effectively using a common-mode bootstrap for its input FETs. With a mere nominal 200 ohm headphone load, however, I don't think that this effect is very relevant here.
More importantly, transistor fT at low currents depends on Ic and device capacitances only. (See eqns. 6-48 and 6-25. At low Ic, the 1/gm terms eventually dominate everything else, and fT ~= gm / (Cje + Cjc) = Ic / Vt * 1/(Cje + Cjc). Cob in turn is the capacitance seen between collector and base in a common base circuit, emitter being assumed open - it is the sum of Cjc and a series circuit of Cce and Cje. Generally Cce is quite small and Cob is dominated by Cjc.)
There's a reason why ZTX951 fT is spec'd at 100 mA - at just 1-2 mA, it would be a lot slower. By contrast, fT for a 2SA1016 (Cob = 2 pF typ) is similar but spec'd at 1 mA. At this current, the ZTX951 could be slower by a factor of 30 (i.e. ~4 MHz only)!
Lower input LTP GBW means that frequency-dependent open-loop gain of the whole composite amplifier is reduced accordingly. Less OLG in the highs, hence more distortion up there.
It is not a coincidence that the top input LTP tranaistors brought up in past threads over in Solid State tend to be very low Cob at still high beta and highish Vceo (which generally means high Early voltage as well). It is a skillset not overly different to that of a good VAS transistor, albeit at a generally lower device current and more emphasis on beta. Sadly, many of these devices have gone the way of the dodo following the extinction of high voltage transistors for CRT applications, including the aforementioned 2SA1016.
IMHO the ZTX951 is way overdressed for a mic input stage. It needs something like 10 mA of device current for the super-low Rbb' of little over an ohm to become relevant at all (as well as gain setting/feedback resistor values to match, of course - this mixer is 47 ohms min), and then delivers voltage noise performance good enough to approach thermal limits for phono MC (~20 ohm) sources. At a beta of 200, that's 50 µA worth of input bias current, or 4 pA/√(Hz) of resulting shot noise alone - for one transistor. In a balanced input stage, just that current noise would already increase noise level with a worst-case 600 ohm dynamic microphone by 3.4 dB over mic thermal noise. That's not ideal any more.
The same exercise with 2N4403 at 1 mA (beta ~= 100) yields a penalty of about 0.9 dB. At the same time, estimated voltage noise yields 1.5 dB more than thermal with a 150 ohm source.
You can still shift the balance down towards lower impedances a bit with 2 mA a transistor, just like the input stage in this Soundcraft mixer does. This gives about 1.7 dB with 600 ohms, so things are just about reversed then. Best coverage of both extremes will be somewhere in between.
For a potential upgrade over the 2N4403, I might try MPS8599 - somewhat lower Rbb' and higher beta, lower Cob (2.9 pF), a bit slower though (must be transition time). Probably not that much in it overall. ZTX718 might be another candidate, which has even lower Rbb', plenty of beta (>300 @ 10 mA), but Cob = 21 pF.
Another option might be 2N5087s, assuming you get ones with an Rbb' of 40 ohms like those that Horowitz/Hill tested (it's an old jellybean part, and e.g. the On Semi datasheet noise curves suggest nothing below 200 ohms). If you do, noise performance would still be adequate, but your beta would be up to ~200 and Cob down to ~2 pF; paralleling two each may also be worth considering to get the noise performance back while the rest is still somewhat better.
On the whole, the stock parts are like 99% there. Noise performance should still be about as good as most anything in practical use, just distortion at higher gains is not likely to distinguish itself very much from a basic modern-day Behringer mixer - the preamp topology is no great shakes by modern standards.
Re: power supply tweaking, there are two different power supply board schematics for different boards. On the second one, the +/-17 V supplies are using LM338s with ADJ pin cap plus some extra RC filtering elsewhere - I would imagine they're fairly quiet as-is, probably diminishing returns territory already. The first one features discrete +/-18 V regulators, with much the same 11 V zeners and 4.7 µF parallel caps in the +48V supply that I had already commented on earlier.
Last edited:
Whew! I'm not sure that I dig what you're rappin', here, brudda....Hmmmm.....What I am trying to do is improve my Soundcraft 200 mixing console---its microphone input is a differential pair of PN4355s feeding a TL072 opamp, and it's NOISY! I will replace the TL072s with John Curl's new OPA1656; but I'd like to improve the transistors as well. So your suggestion is to try the 2N4403, a MPS8599, or a ZTX718----Is that correct?
If you looked at Elvee's magic power supply denoiser postings, you'll see that it is quite an improvement for very little effort and parts;
D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg.
and therefore, I think quite worthwhile.
Soundcraft: 200 : Free Download, Borrow, and Streaming : Internet Archive
Some notes from the designer of the succeeding model (200B) here.
Are all the mic inputs equally noisy? Can the noise be influenced with the input EQ controls? Does it vary with gain setting? What is the microphone used for testing? With the same sound source, level and mic position, how much noisier than others is the input at near maximum gain?
I would suggest making sure that:
1. all the small electrolytics in the power supply section have been replaced (also see previous notes on C10 and ZD1 - btw, there is a drafting error in the schematic, the connection between the two is accidentally omitted)
2. at least one channel input board has been recapped (note: best upgrade C7 to 220-330µ - also rec'd for C6 in the 1600)
3. Pad and Line input switches as well as gain pot on said input board are squeaky clean (same also recommended for EQ pots)
Proceed from there.
I don't know much about PN4355s, but they would have to be badly degraded for excessive noise to be a problem. Still, trying 2N4403s would be worth a shot, along with reducing R5/7 to 7k5 like they are in later models. I wouldn't be too surprised if it's not stable though. PN4355 Cob is around 20 pF, so quite a bit slower. ZTX718 might be a better candidate.
In a composite amplifier like this, replacing the opamp with a faster one can backfire badly, so be warned. Unlike in later models, the 200 does not (crudely) drain open-loop gain via a capacitor across the inputs. Not sure why they don't have any decent compensation for the input LTP.
I did find that the original 200 is reputed to be somewhat noisy.
I also found a star grounding mod which may be worth pursuing; it is supposed to bring down hum by as much as 12 dB.
Some notes from the designer of the succeeding model (200B) here.
Are all the mic inputs equally noisy? Can the noise be influenced with the input EQ controls? Does it vary with gain setting? What is the microphone used for testing? With the same sound source, level and mic position, how much noisier than others is the input at near maximum gain?
I would suggest making sure that:
1. all the small electrolytics in the power supply section have been replaced (also see previous notes on C10 and ZD1 - btw, there is a drafting error in the schematic, the connection between the two is accidentally omitted)
2. at least one channel input board has been recapped (note: best upgrade C7 to 220-330µ - also rec'd for C6 in the 1600)
3. Pad and Line input switches as well as gain pot on said input board are squeaky clean (same also recommended for EQ pots)
Proceed from there.
I don't know much about PN4355s, but they would have to be badly degraded for excessive noise to be a problem. Still, trying 2N4403s would be worth a shot, along with reducing R5/7 to 7k5 like they are in later models. I wouldn't be too surprised if it's not stable though. PN4355 Cob is around 20 pF, so quite a bit slower. ZTX718 might be a better candidate.
In a composite amplifier like this, replacing the opamp with a faster one can backfire badly, so be warned. Unlike in later models, the 200 does not (crudely) drain open-loop gain via a capacitor across the inputs. Not sure why they don't have any decent compensation for the input LTP.
I did find that the original 200 is reputed to be somewhat noisy.
I also found a star grounding mod which may be worth pursuing; it is supposed to bring down hum by as much as 12 dB.
