~250kHz pings from neve 1073 type line output stage

[RoadrunnerOZ]

That doesn’t look right at all - the AC has to pass through an electrolytic?? You’re not going to get much power through a series electrolytic. 1000uF for the 24V and 100uF for the 40V - then arrives at a pair of diodes, one of which on a negative cycle would short out the AC if all those flying ends go to ground which appears to be one side of the AC in connected to chassis??

 

[HenryL]

Both the parallel PSU's (24 and 48V) use a voltage doubler at the start of the chain, hence the caps on the input to each psu, this is what enables the doubling action (charge up the first cap on the negative cycle, then the positive cycle gets the tx's push, plus the capacitor's voltage push, so to speak, so the voltage into the reservoir cap is doubled) at a half-wave rectification frequency. It's a useful technique, esp. eg. in low power valve and hybrid units as a way to generate the low current but high voltage for the HT whilst the rest of the circuit, heaters and other stuff can run from a lower voltage ((you can build up a whole ladder of voltage multiplier steps to get to the high voltage you need).

Sorry it was just my own notes - yes all the 'flying tails' go to ground (to the left of where I showed the chassis connection at the bottom). If it was for publication I should have drawn it better

 

[RoadrunnerOZ]

Ok. Got it - just didn’t see it as a multiplier at first. Wonder why they built it that way??

 

[Khron]

Cheaper to get a standard single-secondary-voltage external transformer (obviously)..?

 

[HenryL]

Ok. Got it - just didn’t see it as a multiplier at first. Wonder why they built it that way??

Cheaper to get a standard single-secondary-voltage external transformer (obviously)..?

Yes it's a cost-effective solution.

I must say I was initially a bit surprised that they used voltage-doubling in the 24V supply chain since they then had to shed a load of voltage. The 270R gets far too hot and is cooking everything in the area including nearby caps (I'm going to replace this with a >5w resistor on flying leads bolted to the chassis somewhere). But I recall from my guitar-amp-designing days bumping into some gotchas when trying to combine full-wave bridge rectification with voltage multiplying - I forget the details - and certainly I found the power supply design could be a quite a puzzle to juggle cost, complexity, what's in the cupboard etc..

A benefit of the approach they took here for the 24V supply are that they that get plenty of voltage headroom to allow multiple smoothing and regulation stages to get a nice quiet supply to that class-A amp.

 

[HenryL]

Just another comment on the original problem here of the spiking/ringing - it's easier to see the scenario from the sketch of the circuit rather than my earlier attempt to summarise it in words... Remove the dotted power connections and replace the 'audio board' in the picture with the scope and you basically have the test config which finally showed that the spikes were getting directly onto the output as a result of some kind of parasitic transmission.

I immediately began querying their grounding design - I wouldn't do it like that etc., but then I started to see a few snags... a project for another time perhaps. A first improvement I'd make would be to co-locate the rectification with the power transformer outside the case and split the filtering/dropping resistances across both the V+ and ground paths. But I think the spikes have probably been subdued enough now just by changing the diodes.

 

[Khron]

Could consider also adding some RC snubbers in parallel with each of the diodes.

 

[HenryL]

Some pictures from the gallery for those interested:

Example of one of the spikes originally seen when monitoring the output whilst zobel tweaking (this one is I think on 0.5V/div) :



Example spikes captured and expanded (0.05V/div) :



After replacing 1N4007s with UF4007 in both the 48V and 24V supplies. The improvement was overwhelmingly due to the 24V supply (very little current being drawn from the phantom supply). (0.05V/div)

 

[HenryL]

Could consider also adding some RC snubbers in parallel with each of the diodes.

You could I guess. The advice I recall from the past (maybe not completely accurately) was not to use additional caps if using fast diodes as it tended to make the magnitude of the ring worse again, though that was probably more referring to the use of caps straight across the rectifier diodes rather than more carefully tuned RC snubbers (which is not something I've played with).

 

[Khron]

And/or if you wanna "guild the lily", you could look up some sufficient-current-rating Schottky diodes - recovery should be also softer than the UF-series.

 

[HenryL]

And/or if you wanna "guild the lily", you could look up some sufficient-current-rating Schottky diodes - recovery should be also softer than the UF-series.

Yes that's probably a good way way to go to improve things a little further. I had a bag of UF4007 in the cupboard though...! 4007 are a natural choice for high-voltage amps of course, though not such a factor with this preamp.

 

[RoadrunnerOZ]

I’m wondering also at the choice of 50V electrolytics at the front end of the multiplier - did you see any sign of the parasitic frequency at the junction of the first two diodes?
Also have you scoped the PSU chassis/ground of the box with the box isolated from ground in any way, in other words using the transformer box/mains ground as ground for the scope. It would be interesting to see if the ringing is actually in the ground network of the whole PSU board.

Could consider also adding some RC snubbers in parallel with each of the diodes.

RC Snubbers are commonly used in DC/DC converter networks as an effective way of getting rid of HF noise. As the RC values, generally R = 2Ω and C = 470 pF can be used to start with, but actual measurements are performed to determine optimal values.
From what I remember you calculate the minimum value for the resistor in the RC snubber by dividing the voltage across the switching device by the maximum current rating then calculate the capacitance for the snubber by multiplying the switching frequency (in this case the parasitic frequency) by the squared value of the voltage measurement acquired. Take the inverse of this number to give the capacitance.
C = 1 / (V^2)*F. (In Farads)
The capacitance should be low enough to not affect the multiplier.

 

[Admiral-dk]

Adding a 1-10 Ohm Resistor in series with the Fuse - will normally completely remove the problem - as it dampens the Reverse-Recovery Spike. This is obviously not always a great solution, as you loose Voltage and Power after Rectification - but in this case, you already have a way too high Voltage -> No problem (as far as I can see).
I have used this solution in many Tube Amps, with success + if you increase the resistor to 47 Ohms 10-11W -> you have a GZ34 equivalent ...!

Per

 

[HenryL]

I’m wondering also at the choice of 50V electrolytics at the front end of the multiplier - did you see any sign of the parasitic frequency at the junction of the first two diodes?
Also have you scoped the PSU chassis/ground of the box with the box isolated from ground in any way, in other words using the transformer box/mains ground as ground for the scope. It would be interesting to see if the ringing is actually in the ground network of the whole PSU board.

RC Snubbers are commonly used in DC/DC converter networks as an effective way of getting rid of HF noise. As the RC values, generally R = 2Ω and C = 470 pF can be used to start with, but actual measurements are performed to determine optimal values.
From what I remember you calculate the minimum value for the resistor in the RC snubber by dividing the voltage across the switching device by the maximum current rating then calculate the capacitance for the snubber by multiplying the switching frequency (in this case the parasitic frequency) by the squared value of the voltage measurement acquired. Take the inverse of this number to give the capacitance.
C = 1 / (V^2)*F. (In Farads)
The capacitance should be low enough to not affect the multiplier.

I don't see a problem with the first caps being 50V types. The maximum they get charged to is about 34V.

I'm afraid I didn't bother scoping at the junction of the two diodes. My line of thought was that once I had the strong theory of what the cause was I took the "it's easy to just replace them and see if it fixes it" approach. I saw the problem mostly go away so I stopped doing much further testing! Including looking further at what was happening on the circuit ground network.

I do have some (not properly thought-through) doubts about the overall grounding scheme as I mentioned but the ringing was everywhere - you could wave the scope probe near the LV AC power cord, or the first cap in the voltage doubler, and get a trace. Since the scope was getting a clear differential picture from the line output socket across pins 2&3 without any galvanic connection to the circuit ground or anything else (just two parallel decoupling capacitors from pins 2&3 to circuit ground) then finding the spike just about anywhere would not be a surprise, so I stopped looking for it

(I didn't do scoping in differential mode - the scope signal was always referenced to mains earth using one of the probes with ground clip. There was no other mains earth connection than the scope's own.)

One observation perhaps is that the audio circuit was pretty well-behaved and did not pick up the ring much at all - if anything the output TX damped it a little when connected to the OP sockets. When I monitored the 'BA283' circuit just before the final output drive amp section there was little detectable ring signal appearing except at the highest gain settings. It seems to get imprinted more strongly on the output directly by some parasitic influences, which was a little surprising.

Re RC snubber info: Not sure I need to do more at the moment to improve things, but interesting. Or should I ?!

cheers

 

[HenryL]

Adding a 1-10 Ohm Resistor in series with the Fuse - will normally completely remove the problem - as it dampens the Reverse-Recovery Spike. This is obviously not always a great solution, as you loose Voltage and Power after Rectification - but in this case, you already have a way too high Voltage -> No problem (as far as I can see).
I have used this solution in many Tube Amps, with success + if you increase the resistor to 47 Ohms 10-11W -> you have a GZ34 equivalent ...!

Per

Dohhh!

I'm embarrassed to admit that I didn't consider this approach here, even though (a few years back into the amnesia cloud...) I've designed more than one valve amp using exactly that principle to avoid spikey rectification!

Although I was happy enough with the spike improvement already, I was thinking of splitting the 270R filter/dropper and putting some of it in the ground path but could also easily add some at the start too. As you say there is voltage headroom to play with, especially in the 24V supply.

Excellent suggestion

(+ indeed the impedance of tube rectifiers does bring some advantages)

 

[RoadrunnerOZ]

If the problem is resolved without the use of a snubber I’d not bother. At this stage it’s more a matter of what’s causing the ring - you say it’s 250KHz? Reduced by change of diode type but at the same frequency?
Is the AC feed to the preamp from the transformer case grounded at the transformer end as it is in the box? Or is it isolated at the TF case and only grounded in the preamp box.

I don't see a problem with the first caps being 50V types. The maximum they get charged to is about 34V.

Still if there’s spiking after those caps the voltage could exceed 50V. I’m wondering what you would see across those front end caps with an ungrounded scope.
The addition of a front end resistor is certainly a good idea - low value would give not much voltage drop. I’d try a separate one for each supply.

 

[Tekay]

This is more or less the PSU that we used for Vintagedesign, just need a single 24V AC.

 

[HenryL]

If the problem is resolved without the use of a snubber I’d not bother. At this stage it’s more a matter of what’s causing the ring - you say it’s 250KHz? Reduced by change of diode type but at the same frequency?
Is the AC feed to the preamp from the transformer case grounded at the transformer end as it is in the box? Or is it isolated at the TF case and only grounded in the preamp box.

Still if there’s spiking after those caps the voltage could exceed 50V. I’m wondering what you would see across those front end caps with an ungrounded scope.
The addition of a front end resistor is certainly a good idea - low value would give not much voltage drop. I’d try a separate one for each supply.

Well, the spike and ring is much reduced. From the scope traces I posted above for example I'm not sure how I'd reliably quantify the reduction between pics 2 and 3. I'd estimate it's overall about a factor of ten, but the initial spike seems less consistent. The ring is still there in pic 3, for a slight ripple or two, and you can see it follows the pic 2 trace with a period of around 2 divs, so at 2 microsecs/div then ~250kHz before and after the diode change. (In the other thread discussing input tx damping this kind of ringing magnitude didn't seem to worry most people! Here it is uncorrelated with the audio signal of course. )

Another reason I had for stopping at this point (before Per in Denmark jogged a few more braincells back to life) was that the kind of disturbance visible in the last pic (3) occurred with the 24v psu branch disconnected from the ac supply completely, just the 48v connected. Since upgrading the diodes in the 24V supply produced behaviour about as good as with no 24V supply connected at all then I felt that particular aspect was fairly well dealt with.
Finally I also upgraded the diodes in the 48V supply but it made little if any visible difference.

The AC feed is floating - not grounded at the TX and just connects to the signal ground through the chassis at a single point - except for those 1nF decoupling caps. (1nF presenting about 600ohms to a 250kHz signal)

ok I see where you're coming from now re the spiking on top of the peak ac voltage at the cap. I don't think there is going to be anything significant there, now at least, but I'll take another look. I can't (at least won't) unground the scope as such but could do a differential probing across the input caps perhaps. I'm generally trying to avoid using both channels on my old scope as one channel is a bit flaky, though it can be cajoled into stable operation sometimes (worn out/corroded internal switch contacts I suspect, but not very easy to get at).

Re front end resistors "I’d try a separate one for each supply" - agreed.

 

[HenryL]

@Tekay
hi - this is the kind of circuit I might well use myself if I was building something or replacing the PSU board completely.

( I'm not sure if you are responding to a side comment I made earlier somewhere about running into some unexpected complications when combining full wave rectification with half-wave / parallel voltage doubler as you do in the circuit you posted - if so then: yes... I mean: I know it can be done like this - I just have a nervous twitch on this subject because I goofed once trying to combine some rectification circuits - luckily I discovered my mistake before building it..)

cheers

 

[RoadrunnerOZ]

It would make for a more stable and less noisy supply to have a bridge for the 24V - the 48V is fine as the draw is typically less than 30mA. I’m just not sure how much the ringing will return under a heavy audio load where the preamp is really working with your current setup.
The original Neve modules were in a console with power supplies that had massive filter caps as well as a rack of caps in the console at the power distribution rails.
I recently replaced all the DC power in a Neve console - this uses a separate supply set for each side of the console with 6 separate supplies for each side. I used medical grade, low noise supplies - ripple better than 0.6% at 30A max load - the +16V supplies only drawing about 12A max. Everything star grounded in a 3RU rack drawer with the power supply fans removed and the supplies stood on their sides with a draught fan with less than 7dBa of noise set into the front of the drawer. After setting up, the console was dead silent and so was the room - previously with the old analog supplies there was noise and with all channels open you could hear it quite plainly. Plus the big old supplies weighed a ton, had quite audible transformer hum and got really hot.

Bottom line if it were me, I would chuck the old DC supply completely and replace the 24V and 48V supplies with good quality switch modes, remote them from the preamp unit (maybe they’d fit in the transformer box), run shielded DC cabling to the box and add filter caps at the preamp end matching the tail end caps on the original board. You wouldn’t need the cooling fan on the supplies - if you used a 100W supply for the 24V the load would be minimal and it’d stay cool. As long as they’re hard mounted to the box and well vented you can just remove the inbuilt fans as they’re tiny and noisy. The lower wattage 48V supplies don’t have an inbuilt fan anyway. If needed you can install a fan like the Noctua NF-S12B Redux 700 in the box - so quiet you can’t hear if it’s on and only 120mm.
Photos of the supply drawer 1st AC wiring, 2nd with DC multicore: