Once the supply was basically built it was time to commission it. It was then that I realised that I had somehow managed not to connect the output pins of all three supplies to their output pads! Passing the wires through the PCB holes and soldering direct to the PCB underneath fixed it and all supplies provided the correct voltages. I then fitted the PCB to an SRS sub-rack and connected the dc outputs to the four module motherboard and tested it with the new Classic mic pre module. This is a 3U Eurocard version of my original mic pre design based on a pair of mu followers. To fit into the standard width module, the 6CG7 tubes of the original design were replaced with 6922 tubes. These also give a little more gain but need to be operated at a higher current to achieve distortion levels as low as the original 6CG7 design. The extra current does mean that this version has a better drive capability than the 6CG7 version (see later).
It was at this point that problems began to surface. I was unable to set the Classic gain to maximum (70dB) without the circuit going into oscillation. At 60dB it did not oscillate but the noise level was far too high and there was a lot of hum. At this stage I did not know if this was the Classic or the power supply at fault. So I built a regular Eurochannel mic pre and tried that. I know this works so when it too oscillated at about 1MHz I knew there was something else wrong. I spent a lot of time adding decoupling, re-routing wires and adding grid stoppers and small caps the the NFB loop but nothing seemed to work. I could not tame this oscillation. I tried removing the input tube. The oscillation was still present at the output but it was smaller and disappeared after a few minutes. Removing the output tubes removed the oscillation from the output. I tried PCC88 tubes instead of the 6922; I even tried the original 6CG7 tubes but they all oscillated. I even tried a couple of boards where only the tube heaters were wired up and I still got the oscillation at the output! Something very odd was going on here.
This went on for most of January without finding the cause. Eventually I decided to put my oscilloscope across the heaters and there was the oscillation, even with the HT disconnected it was there. It was the new power supply heater regulator oscillating. I checked the datasheet of the new 12V regulator the LT1084V and sure enough it needs a much larger output capacitor for stability - at least 150uF. I changed the existing cap for a 220uF and the oscillation disappeared completely.
The Classic could now be turned up to its full 70dB gain and its EIN with the input shorted was -123dBu, a very satisfactory result. Via 2:1 output transformer I loaded the output with 600 ohms. At 1KHz and +4dBu the distortion was just 0.18%. At +20dBu it rose to 1.08%, again very satisfactory for a mu follower. I then made the same tests on the Eurochannel. Its EIN was slightly worse at -121.5dBu but its distortion into a 600 ohm load was much better as expected, It was 0.026% at 0dBu and only 0,25% at +20dBu.
All these issues with the power supply ( squeezing it into the 28HP box, providing adequate heat sinks and the oscillation problem) prompted me to try something I had been meaning to check out for some time - AC heaters. I had an old home made power supply that had a 6.3VAC heater output. I modified the heater wiring on the Eurochannel card so it would work on 6.3V heaters and connected it to this power supply. I was very pleased to discover that the measured EIN was exactly the same as with DC heaters. I then tried various types of 12AX7 input tube on the basis that if heater hum was going to get into the circuit, the first stage would be the most susceptible. I tried WA, WC and EH variants as well as the type with the spiral wound heaters the LPS. All of them performed exactly the same. I also tried changing the output tubes and this also made no difference.
These results meant that AC heaters were a viable option. The main benefit would be a considerable simplification of the power supply along with a consequent reduction in mains transformer size and a useful cost saving. About this time I became aware of a new 19 inch rack mounting enclosure made by Frank Röllen (who also makes my front panels). One unique feature of this new enclosure is that you can fit support rails in a variety of positions and in one configuration you can create a 10 slot 500 series lunch box. My modules are 2.8 inches wide and double width 500 series modules are 3.0 inches wide. This means that with suitable module front panels and a new motherboard you could possibly built a 5 slot lunch box using Frank's enclosure. The question is, is there room to for an internal power supply.
With AC heaters this looked like a definite possibility. There is about 100mm of space in Frank's enclosure between the back of the motherboard and the rear panel. So I designed a 100mm by 100mm power supply PCB; just HT and phantom power and a couple of resistors to make a virtual centre tap for the AC heaters. Here is a picture of the PCB layout:
And here is its schematic:
I used a standard Don-Audio toroidal transformer type TB25048 which has 250V, 50V and 12V secondaries. The 12V secondary is only rated at 2 amps which is only enough for four Eurochannel modules but this was sufficient for my initial test purposes. Here is a picture of a test build I did to see how all the components would fit (the motherboard is a standard 6 module one just used for the test).
The mains transformer fits on the right hand side wall with the power supply PCB beneath it and the mains inlet above the PCB. I discovered you can mount three VTB2291 transformers on the left hand wall one above the other so for a five slot lunch box you only need to find room for two more output transformers on the rear panel. Here is a close up of the PSU board and the mains transformer:
So far it looked good. Next I designed a 5 slot motherboard with double width 500 series module spacings that could be attached to Frank's support rails:
There were some issues with fixing the PCB to Frank's rails but in the end it turned out the easiest thing to do was to tap the holes in Frank's rails with a 2.5mm thread and use the same 2.5mm machine screws used to fit the motherboard into an SRS sub-rack. Here are a some of pictures of the new motherboard in Frank's case with a couple of modules:
This all looked pretty good. Notice how the right hand module is very close to where the mains transformer will be mounted. I did not realise it at the time but this would prove to be a problem. I carried out some initial tests on the Classic module plugged into the left most slot and the results were fine. EIN was perhaps not quite as good as when I last measured it but it was within a dB.
In the meantime I began building a four slot lunch box for a German customer - SRS sub-rack, four Eurochannels housed in Fischer extruded aluminium cassettes and full lunch box power supply with DC heaters. All went well until I tested each module with all four plugged in. As I tested each one I noticed that the noise got slightly worse as the module I was testing got nearer to the mains transformer and there was a small hum component in it. The hum was not large and only made a couple of dBs difference to the noise floor even when the module gain was set to 70dB. Moving the transformer and testing with an external power supply soon confirmed the hum was caused by stray magnetic field from the mains transformer. The solution is to screen the transformer and at the moment we are waiting for a mu-metal enclosed transformer to be manufactured.
Naturally, this prompted me to test the slot right next to the mains transformer in Frank's enclosure. I used the Classic module and, because it is not in a screened box, I was not surprised to find the noise was about 4dB higher than than in the left most slot. Aaron from Don-Audio kindly sent me some GOSS (grain oriented silicon steel) strip to try screening the transformer. I could not get rid of the hum using this but by placing it close to the transformer I could reduce it by a couple of dBs. Mu-metal is very expensive and therefore not really a viable option for a DIY project. I tried shortening the leads to the transformer and twisting them tightly but this did not improve the hum, so I looked around for a cheaper alternative. I found a steel box 100mm x 100mm x 50mm for £5 on eBay. I fitted the mains transformer into it, made a hole in the lid for the wires to come through and mounted the assembly in Frank's enclosure. I turned on, measured the noise and it was back to normal levels and there was no sign of hum. So, although the leakage is small it has a measurable effect, but it can easily be eliminated by some simple low cost screening. The steel box I built is really too big so Aaron is sending me a GOSS top plate to fit over the transformer. Hopefully this will be as effective as the steel box.
Since it was now clear that AC heaters and an internal power supply would work fine with a little care and the right screening, it seemed appropriate to revisit the 28HP power supply PCB. With no need for a regulated heater supply, the PCB layout was very much simpler. I also took the opportunity to change the large 4 amp SIL bridge rectifiers I had used for the HT and 48V supplies to small one amp DIL types which should be more than adequate for the 100mA or so of output current required. I then realised there was no real need for a 28HP module. The mains transformer could be fitted to an SRS rack side panel just as in Frank's enclosure. If I made the power supply PCB Eurocard size it could possibly just plug in like any other module. If I was smart it could feed HT and phantom power straight to the motherboard and apply the heater elevation voltage to the centre tap of a pair of resistors across the heater pins of the PCB. As the heaters would be AC they could be wired direct from the transformer to the power connector on the motherboard. The only question was how to get the HT and 48V AC inputs to the PCB? I then realised the mic and line inputs on this PCB would obviously be unused. What if I fed the HT and phantom AC supplies into the board through these unused inputs? Here is the PCB layout I came up with:
The HT AC supply comes in via the mic input, goes around the outside of the PCB and exits at the HT pins of the PCB. Inside this loop the phantom AC is fed to the line input pins and its output goes direct to the +48V pins. There was enough room to allow the second two HT smoothing capacitor to be larger diameter types with values up to 470uF for additional smoothing if required. There was also enough room to make provision for a small heatsink to be fitted to the TL783 phantom power regulator. Prototype PCBs are currently being manufactured.
What this means is that it is now possible to build a 5 slot lunch box using the SRS sub-rack and the standard 6 module motherboard. Normal modules occupy the first five positions and the 6th houses the PSU. It should now therefore be possible to build a 5 slot lunch box using either frank's enclosure or a Eurorack. It is an odd number but I am very tempted to standardise on 5 slots for the lunch box. Four Classic mic pres fitted with LCR bus assign toggle switches and a Twin Line Amp for bus amplification would make a simple, self powered, four mic input, stereo all tube mixer in a 3U lunch box.
No comments:
Post a Comment