Golden Age Projects Pre-73 PSU Schematic and Mods

[HenryL]

Here posting notes and conclusions from this other thread ( ~250kHz pings from neve 1073 type line output stage ) in a way that might be more easily findable by anyone looking for stuff relating to the Golden Age Pre-73 and its power supply...

This relates to the Mk1 version of the Pre-73, I don't know what may have changed in the later versions.

I didn't find one online so this is my tracing of the original PSU board circuit and some voltage measurements taken when the circuit was powered up and the PSU under normal load:


The ID's in brackets against various components are as found on the circuit board.

Possibly unclear in the sketch are the 24V and 48V power connectors on the right (CZ10 and CZ11). These have two pin connectors for connecting the power across to the main audio board, but at the audio board end only the +ve pin is connected. All the power supply ground connection is via the chassis. Hence the 'screen only' comments.

The topology of the grounding scheme is as drawn, with the 0v line at the end of both the 24V and 48V supplies connecting to chassis ground at bolt holes + additional strap wires. The 48V ground track extends to the input and output XLR sockets which are also mounted on the PSU board (there is also a TRS output socket I forgot to draw in, but it just piggy-backs on the XLR output).

 

[Khron]

Still baffling as all hell why they went with a voltage doubler also for the 24V DC rail - just literally burning off all that extra power, for really no good reason at all.

24V AC rectified should result in 30-33V DC, while 78xx's need (let's be really generous) 3-4V above the output voltage to most definitely maintain regulation.

"Only" 68dB ripple rejection @ 120Hz - fine, add an RC filter ahead of the 7824, one would worst case have at least 3V to burn off on the way to the regulator.

You'd think they would've at least tried to keep the costs down, anyway...(?)

Whatever those people were smoking (except components) - i want some!..

 

[HenryL]

So the problem that started the other thread (link above) was noise spikes appearing on the output - even with the audio board disconnected. Eventually tracked down to rectification noise, mainly from the 24V supply circuit. A solution to this was to make various modifications to the PSU board:



1) The first mod (*1 in the above scrawl) had been done already before the noise spike problem was detected - it is suggested by Golden Age themselves as a mod to reduce a small noise problem. The solution is to break the chassis ground connection at the phantom supply by breaking the track to the bolt hole and removing the strap wire. EDIT: Don't use this GA mod if you are making these other changes, and if you already have it then undo it and ground the 48V supply again at this point. It's not a good idea to have the rectification noise currents in the phantom gnd path like this, they do break through, and it's prone to some instability, sometimes motorboating.

*2) "7R2" is a 270 ohm filtering/dropping resistor. It is adequately rated for the dissipation in principle but it does get rather hot, the resistor and circuit board in my case were well cooked, it is furthermore pretty closely surrounded by electrolytic capacitors, all very cosy and not helpful. I suggest moving this out to a chassis mounted power resistor on flying leads. I found a spot on the left side of the enclosure between the PSU and audio boards seemed to give reasonable clearance from capacitors with room for internal convection heat to spread out. While interfering with this resistor you can take advantage of the opportunity to place some of its resistance in the ground path to improve the isolation of the rectification stage from the main ground. 270R split into 220R and 47R above.

*3) Replace all the 1N4007 rectifier diodes with faster diodes. UF4007 above. This was the most significant single change. Edit: a small clarification here so as not to accidentally mislead anyone: it's possible to damp rectifier switching noise with resistance without necessarily replacing the diodes with faster ones, if I'd tried that first then it might have been sufficient on its own, but the order in which these changes were tried was to change the diodes first and this took away the big spikes right away, leaving a much lower level and gentler periodic ring, which was then addressed with the series resistances.

*4) Add series resistance to the rectification. Again I elected to split the resistance into both ground and positive paths to help corral the switching currents and keep them out of the ground as far as poss. The 24V supply can stand a little more than the above resistance but beware the phantom 48V has little headroom - I found the above 2R2+2R2 ok with my mics but if you have a slurper then you might find it dragging the voltage down a bit too much.

 

[HenryL]

Still baffling as all hell why they went with a voltage doubler also for the 24V DC rail - just literally burning off all that extra power, for really no good reason at all.

24V AC rectified should result in 30-33V DC, while 78xx's need (let's be really generous) 3-4V above the output voltage to most definitely maintain regulation.

"Only" 68dB ripple rejection @ 120Hz - fine, add an RC filter ahead of the 7824, one would worst case have at least 3V to burn off on the way to the regulator.

You'd think they would've at least tried to keep the costs down, anyway...(?)

Whatever those people were smoking (except components) - i want some!..

My most optimistic guess so far is that they wanted headroom to have both the zener follower smoothing and 7824 stage but I stopped trying to think too hard about it after that

 

[Khron]

My most optimistic guess so far is that they wanted headroom to have both the zener follower smoothing and 7824 stage but I stopped trying to think too hard about it after that

That may well be the case, but that still leaves at least ONE layer of "why???"

 

[HenryL]

Another power-related issue - with the Mk1 at least - is that the power on/off button on the front of the unit does not disconnect the mains power - not surprising perhaps with an external power lump - but it does not disconnect the LV AC input either. The PSU remains powered up when the unit is 'off', it is just the audio board that is disconnected from DC power.

Again this may not be such an issue. The PSU is not having to supply anything much so doesn't consume anything much...

BUT the 48V phantom supply is NOT disconnected by the 'on/off' button. This means all the lights can be off on the front panel, including the 48V light (they run from the 24V supply) the unit is 'off' but if the phantom power button is pressed then the 48V will still be applied to the input !

The board is bolted back into the unit now so I can't give more user-friendly detail but there is a fairly simple modification that can fix this. With a little track breaking and adding of a couple of bits of wire you can repurpose one pole of the on/off switch which is doing something rather pointless and use it to switch the phantom power instead.

All the buttons have indicator LEDs next to them and the design evidently uses some kind of constant current source to the LEDs since all the LEDs are in series with each other and are switched off/on by shorting them out/unshorting them. The shorting is performed by one pole of the switch that they are next to / indicating for. This approach was extended to the on/off switch too, which has a pole dedicated to turning off the 'power on' LED.
This might strike you as a little superfluous as the power is going to go off anyway and therefore the lights will go out anyway....

I disconnected the pwr led from the pwr switch and made it 'permanently on', tapped into the 48V line after 48V switch and routed it via the now available pole of the on/off switch.

 

[Khron]

BUT the 48V phantom supply is NOT disconnected by the 'on/off' button. This means all the lights can be off on the front panel, including the 48V light (they run from the 24V supply) the unit is 'off' but if the phantom power button is pressed then the 48V will still be applied to the input !

Genius!..

Some people really should stay away from any electronic design... Especially people who need ego-stroking THAT badly...
("Look at me, aren't i clever???" while the fruits of their labor are... ummm... sometimes unsuitable for human consumption, as it were )

 

[HenryL]

I couldn't possibly comment.

 

[Mr. Moscode]

Still baffling as all hell why they went with a voltage doubler also for the 24V DC rail - just literally burning off all that extra power, for really no good reason at all.

24V AC rectified should result in 30-33V DC, while 78xx's need (let's be really generous) 3-4V above the output voltage to most definitely maintain regulation.

"Only" 68dB ripple rejection @ 120Hz - fine, add an RC filter ahead of the 7824, one would worst case have at least 3V to burn off on the way to the regulator.

You'd think they would've at least tried to keep the costs down, anyway...(?)

Whatever those people were smoking (except components) - i want some!..

My guess is the 24V xfmer is off the shelf and it was cheaper to do a doubler than get a custom xfmer. And less worries about safety.

 

[Khron]

My guess is the 24V xfmer is off the shelf and it was cheaper to do a doubler than get a custom xfmer. And less worries about safety.

For the 48V rail, sure. But for the 24V rail???

 

[jokeramik]

So the problem that started the other thread (link above) was noise spikes appearing on the output - even with the audio board disconnected. Eventually tracked down to rectification noise, mainly from the 24V supply circuit. A solution to this was to make various modifications to the PSU board:

View attachment 112356

1) The first mod (*1 in the above scrawl) had been done already before the noise spike problem was detected - it is suggested by Golden Age themselves as a mod to reduce a small noise problem. The solution is to break the chassis ground connection at the phantom supply by breaking the track to the bolt hole and removing the strap wire. EDIT: Don't use this GA mod if you are making these other changes, and if you already have it then undo it and ground the 48V supply again at this point. It's not a good idea to have the rectification noise currents in the phantom gnd path like this, they do break through, and it's prone to some instability, sometimes motorboating.

*2) "7R2" is a 270 ohm filtering/dropping resistor. It is adequately rated for the dissipation in principle but it does get rather hot, the resistor and circuit board in my case were well cooked, it is furthermore pretty closely surrounded by electrolytic capacitors, all very cosy and not helpful. I suggest moving this out to a chassis mounted power resistor on flying leads. I found a spot on the left side of the enclosure between the PSU and audio boards seemed to give reasonable clearance from capacitors with room for internal convection heat to spread out. While interfering with this resistor you can take advantage of the opportunity to place some of its resistance in the ground path to improve the isolation of the rectification stage from the main ground. 270R split into 220R and 47R above.

*3) Replace all the 1N4007 rectifier diodes with faster diodes. UF4007 above. This was the most significant single change. Edit: a small clarification here so as not to accidentally mislead anyone: it's possible to damp rectifier switching noise with resistance without necessarily replacing the diodes with faster ones, if I'd tried that first then it might have been sufficient on its own, but the order in which these changes were tried was to change the diodes first and this took away the big spikes right away, leaving a much lower level and gentler periodic ring, which was then addressed with the series resistances.

*4) Add series resistance to the rectification. Again I elected to split the resistance into both ground and positive paths to help corral the switching currents and keep them out of the ground as far as poss. The 24V supply can stand a little more than the above resistance but beware the phantom 48V has little headroom - I found the above 2R2+2R2 ok with my mics but if you have a slurper then you might find it dragging the voltage down a bit too much.

Were you able to measure the noise floor?

Why did you make the mass so high impedance? That's not a good idea.

Best regards!

 

[Cqwet Dbdfte]

Strange design, McGyver ran out of parts?
Voltage multipliers is not a go-to solution unless you are making a bug zapper, but for low noise...
Relying on a random wall-wart for AC power is suspect, usually low grade parts, bad isolation, no electrostatic shield etc. Measure the leakage current from secondary to neutral, it should be very few microamps.
High end equipment should use premium parts.
I can only see remoting the transformer to get rid of magnetic fields as a possible reason for the wall-wart, but then the whole thing could be remoted. A shielded toroidal may produce a minimum of stray fields.
HF noise from light dimmers and LED bulbs are ubiquitous, differential line filters should be strongly considered.
Electrolytic caps are the usual suspects in old equipment, use only top grade parts.

 

[Voyager10]

There are a lot of good reasons (aside from audio performance) why you might want to buy in everything to do with mains power in a ready-made, pre-approved block. I can understand the wall-wart but it defeats me how they then make such a dog's dinner over turning 24V AC into 24V DC.

 

[Tekay]

Here's the PSU we used for all Vintagedesign products, uses an internal or external 24v AC transformer.

 

[Cqwet Dbdfte]

https://www.edn.com/simple-circuits-reduce-regulator-noise-floor/The sub 10 Hz noise dominates in many regulator chips, a well filtered voltage reference and a low noise OP amp may be needed to drive a pass device, to get better LF performance.

 

[drtechno]

I have one that needs to be modded. But I'm going to do it different using a 48V supply, and trim back the circuitry in the 48V section and alter the 24Vsection so it could handle the input. Another way I can deal with it is alter it so all of it runs off 48V.

 

[drtechno]

Here's the PSU we used for all Vintagedesign products, uses an internal or external 24v AC transformer.

That isn't too bad, but I think using with a 55VDC supply that is used for POE switches and ditch using 24AC transformer all together 317 would just have to be the 60V max type (LM317HV) But just a simple voltage divider could be just used for the 48V if switching to a 55VDC regulated supply, So that would cut it down just to one regulator.