Power Supply Build

The custom toroid mains transformer has arrived so I can complete the build of the power supply. This toroid had dual primaries so it can be configured for either 115V or 230V mains and it has secondaries of 250VAC at 150mA for the HT and 50VAC at 100mA for the phantom power supply. It has a GOSS band (Grain Oriented Silicon Steel ) for reducing the radiated magnetic field and an electrostatic screen between the primary and secondary windings for minimum noise. This transformer is made by Terry at Canterbury Windings Ltd. They carry a wide range of standard toroids and will also wind any toroid to your specifications at very reasonable prices.

I also received the remaining components for the two power supply boards. The estimated total HT draw is 90 mA (6 Eurocards at 10mA each plus 30mA for the headphones amp) so to obtain 10V drops across each of the three dropper resistors in the HT supply PCB, each resistor needs to be 110 ohms. See the HT Power Supply Design document for details. The first of the three resistors is a 5 watt type.

I also made some changes to the heater elevation circuit. Normally I use a potential divider consisting 220K and 75K resistors to produce a nominal 75V heater elevation voltage from a 300V HT supply. However, many tube manufacturers specify a maximum resistance value between heater and cathode of rather less than 75K. To date this has never been a problem but I thought this was a good opportunity to update the design. One problem with using smaller value resistors s that the current they draw increases and so they dissipate more heat. As many tube data sheets mention a heater cathode resistance of around 20K, I decided to change the 75K resistor to 22K. With 75 volts across it, this will dissipate 255mW so a 0.5W type is required. The 220K I decided to change to a pair of 33K resistors in series. Each one of these will dissipate nearly 400mW of heat so 1W types were used. The heater elevation voltage is now exactly one quarter of the HT voltage. An unplanned side benefit of the smaller value resistors is that the HT supply does discharge more quickly when turned off.

Lastly I needed to set the SOT resistor in the phantom power supply. There is a nominal 1.25V dropped across the 120 ohm resistor between the output and the ADJ pin of the TL783 regulator which means for a 48V output voltage, the SOT resistor from the ADJ pin to ground needs to be about 4.6K. I made this from a 2.2K and a 2.4K resistor connected in series. The measured output is 49.8V which is due to the tolerances in the resistors and the chip's on board reference. Fortunately the phantom power spec is 44V to 52V so we are comfortably within that. In future I might update the PCB to replace the SOT resistor with a 10 turn pot. However, one of the reasons for using a resistor is that it dissipates 500mW of heat so a 1W type is needed. I used two 0.5W types in series.

Here is a picture of the completed power supply. The white cable is the mains input which will come from a filtered, switched and fused IEC connector mounted at the rear of the mixer.

AUX Send and Return Panel Layouts and the Crosstalk Issue

It is time to do the front panel layouts for the AUX send and return controls. The AUX return controls are fitted to a 3U panel that is attached to the main L & R bus amplifier. There are two reasons for this. First, the AUX returns are fed to the L & R buses and this is an easy connection to make inside the L & R bus amplifier. The other reason is the L & R bus amplifier is a Twin Line Amp  (TLA) board. This has provision for a pair of input transformers (not needed by the bus amplifier) which can be used for the AUX return input transformers. Two AUX return inputs are provided which could be connected to either two mono or a single stereo source. Since they could be mono it makes sense to include a pan pot on each return. If the two are used with a stereo source then they can simply be panned hard right and left. None of this is a problem from the font panel layout point of view but it could lead to a worsening of crosstalk. The reason for this is that a simple passive input has a source impedance that is determined by the position of its level control. If we use a 10K pot for the AUX return level control then the worst case source impedance it presents to the pan pot circuit is 10/4 K = 2.5K. The reason this can affect crosstalk is shown in the sketch below.

The circuit at the top left shows an AUX return channel. The input goes through a 10K:10K transformer and is fed to the level pot. The wiper of the level pot feeds both halves of the pan pot which in turn feed the left and right buses via 100K resistors. Each pan pot is 20K and is slugged with a 10K resistor. The circuit below it shows its equivalent circuit for crosstalk calculation purposes. The transformer and pot are replaced by a 2.5K resistor to ground. Each pan pot is replaced by a 5K/10K pot divider representing the pan pot at the mid position.  Assume we have 0dBu on the bus at the right of the 100K resistor. By the time this reaches the pan pot is is reduced by the pan pot 10K to ground by about 20dB. This signal is then reduced by the pan pot 5K acting with the 2.5K of the fader in parallel with the other pan pot leading to a further loss of about 10dB. The the left hand pan pot itself drops the level by about 3dB resulting in a total loss of 33dB.

To get the crosstak we have to add the bus loss. We have four channels in this mixer so there are three more 100K resistors between the bus and ground plus the 10K master pot for the bus. The total parallel resistance of all these is close to 7.7K so the bus loss is 107.7/7.7 which is about 23dB. Hence the crosstalk is 23dB + 33dB = 56dB. What does this mean? If we send a 0dBu signal from a channel and pan it fully right, and we have the AUX return set so the fader presents exactly 2.5K source impedance and the AUX pan is centred, then that 0dBu signal will appear on the left bus at -56dBu. Note that this is the absolute worst case. If the AUX return pot is at any other position the crosstalk will be better and it will also be better an any other pan position.

In 99.99% of cases this will not cause a problem as the stereo separation between channels in a mix is rarely better than about 30dB. The sole reason for the increased crosstalk is the source impedance of the AUX return level pot. In normal channels the pan pot is fed from the output of a TLA type amplifier which has an output impedance of about 150 ohms. This reduces the crosstalk by a further 24dB to about -80dB. An alternative therefore is to feed the AUX returns through a TLA amplifier which would give us a very low source impedance and we could also have some gain in hand on the return level pot, but this does increase cost. Note also that crosstalk improves if you have more channels because the bus loss increases. If there were 16 channels instead of four, the crosstalk would improve by about 5dB. You could achieve this artificially by slugging the bus to increase the bus loss but we have already partially done this with the 10K master bus pot.

The circuit at the top right shows the equivalent cicuit if the pan pots are 50K instead of 20K. I expected this to give worse crosstalk but it doesn't. The loss from the bus to the pan pot is less (14dB instead of 20dB) because the pan pot has a higher value resistance but the loss from the pan pot to the fader is higher (16dB instead of 10dB) for exactly the same reason. The result is the loss from the bus to the other pan pot is exactly the same which is a very interesting result. It appears that cross talk is independent of the value of the pan pot.

An interesting aside is that when I was at Neve in the 70s, the nominal bus send level was about -8dBu. The reason for this was they wanted to maintain a 26dB headroom and the 24volt rail powering the class A amplifiers meant their maximum output was about +20dBu. At that time, AUX returns were fed in through a 10K:600 transformer which dropped the level by 12dB from a nominal +4dBu to the magic -8dBu. The other benefit is that you can now use a 600 ohm level pot and the worst case source impedance of a 600 ohm pot is 600/4 or 150 ohms. What a coincidence!!

So, having gone through all that I decided pan pots are OK of the AUX returns. The sketch below shows the probable layout of the controls for both AUX send and return. The AUX send front panel is attached to another TLA which acts as the bus amplifier for the AUX sends.

I have kept the send and return level pots at the same height. There is plenty of room on the panels for these controls. All you need to do is allow 10mm top and bottom for the support rails of the sub-rack and 10mm on the left for the fixing of the PCB to the front panel. For the sake of symmetry I also allow 10mm on the right hand side. I plan to use 20mm diameter knobs for the level pots and 15mm diameter ones for the pan pots. The next step is to input these to front panel designer so I can get them made.

To Crimp or Not to Crimp....

When building a mixer, there is a need for many audio connections to be made between the active modules of the mixer and other components. For example, all the mic and line inputs need to be connected from rear panel mounted XLR connectors to their individual channel modules which in turn need to be connected to and from channel faders. The current design of the EZTubeMixer uses a motherboard into which the channel modules are plugged. This motherboard provides all the common connections between channel modules such as power and audio buses which saves a lot of tedious error prone wiring. However, connections like mic and line inputs, faders and direct out still have to be hand wired on a per channel basis to the back of the motherboard.

There are two major problems with this. The first is that access to the motherboard is not always  easy so wiring to the motherboard in-situ is not always possible. To overcome this the motherboard can be loomed before it is fitted, with the free end of the loom being wired to XLRs and faders once the motherboard is in place. That's fine but often the XLRs and faders are not in a place that is easily accessible for soldering. The second problem, is that once all this wiring is complete, it is now very hard to access it to make modifications or correct errors.

The answer, of course, is to use connectors for these signals on the rear of the motherboard. XLRs can then be wired to flying leads with connectors attached . The whole seemly than then be offered up and plugged into the motherboard. Similarly a complete fader assembly can be built with flying leads that simply plug into the motherboard. Soldering can be done of the bench where access is not a problem and the assemblies can easily be disconnected for fault finding or to make modifications.

The only question remaining is what connectors to use to fit on the rear of the motherboard and on the ends of the flying leads? We need a PCB connector to fit onto the motherboard and a mating half that can take a screened lead. It looks like the only viable solution is to use 0.1 inch pitch crimp connectors. The Molex KK range is typical. It has two and three pin headers that can be soldered directly to the motherboard and crimp contact free receptacles (sockets) that plug into them. They can be polarised and they are not expensive. The only problem is the free sockets use crimp contacts and I have never quite got on with crimp connections. It would be nice if there were solder connection versions of the free sockets but I have not been able to find any.

As crimps can potentially solve so many mixer construction problems I thought I ought to give them a another chance. So I bought myself a proper hand crimp tool, some two and three way Molex KK range housings and a packet of crimp contacts. I sat down and tried to come up with a reliable way of connecting a twin screened mic cable to a three way 0.1 inch pitch free crimp socket. Crimping screened cables had been a problem in the past for me because the screen ended up being a much greater diameter than the two signal wires. Fortunately Van Damme does the 'Install' range of twin screened cables that has a foil sheath and a multi-strand drain wire that is little bigger than the two signal wires. The picture below shows the results of my initial efforts in crimping this cable to a three way Molex.

The unsheathed connection was my third attempt. As it has no insulation, it is hard to judge the correct depth to insert the drain wire into the crimp and it is all too easy to push it in too far. This not only leads to a bent connection because one wire is shorter than the others (as you can see in the picture) but sometimes can prevent the crimp contact engaging properly in the plastic housing. I was also unsure how much insulation to strip off the signal leads. Too much and you get the same problem as with the drain wire, too little and you get no electrical connection. That's why it took three goes to get it to work. The green sheathed one was my fourth attempt. It is still bent because I have not got the drain wire length right yet, but signal leads went in fine now knew how much insulation to strip off.

The two brown sheathed connections were my final two attempts. By now I have just about got the hang of it. The connections are straight because the drain wire length is right.

The thing with the blue handles is the crimp tool. It works well but working with crimps is fiddly and you never seem to have enough hands. I found the best way was to put the crimp contact in the tool and start the ratchet action. This holds the crimp in the tool so you can now hold the tool in one hand and with the other offer up the wire (which in the meantime I was holding in my teeth).

It is still a bit of a clumsy process. Holding the fairly heavy crimp tool in one hand and offering up the wire with the other is not easy. I think I might try clamping one handle of the crimp tool in a vice to hold it steady. The I only have one wobbling hand to worry about.

Overall I think with a bit more practice I could become competent enough to make reliable connections with it. Thanks to Holger  ( http://analogaud.io/aa/de.html )  whose motherboard design using Molex KK connectors got me thinking about this again. His two module motherboard is a really neat idea. I might just have to do my own version.

Power Supply Assembly

In my larger mixer designs I use an external power supply as it avoids any possible interference of mains transformers with sensitive microphone input circuits and their transformers.  For this smaller mixer I wanted to try building the power supply inside the mixer as much as anything to see if it could be done without compromising performance. From past experience I generally avoid steel enclosures because they can easily conduct interference magnetically from mains transformers to input transformers. Unfortunately the Rackz enclosure is entirely steel so magnetically conducted interference remains a possibility. The best way to minimise it is to increase the distance between mains and input transformers.

The microphone transformers are towards the rear of the case as are the input connectors and the output transformers so mounting the power supply at the back of the case seems like a bad idea. It turns out there is not really enough room there anyway. The only other space with sufficient volume is right at the front of the case.  This would be directly below the channel and master faders but these will be connected using screened cable so they should not be susceptible to interference. So I decided to fit the power supply at the front. Since the SMPSU heater experiments were inconclusive I have decided to use the big old International Power linear 12V 5.1A PSU for the heater supply. This has a rather large transformer so I decided this had better be fitted on the right hand side at the front, as far away as possible from the microphone transformers. Initially I tried fixing it directly to the floor of the Rackz case but this proved very awkward as access to mark drilling points is restricted by both the case and the power supply itself. Even if it could be done that way, access for wiring would also be limited. All in all not an easy solution. What I really needed was a neat way to build the entire power supply as a separate assembly.

After some thought it occurred to me that the 3U panel at the bottom front of the mixer is unused (see Sub-Rack Build post). The entire power supply, heaters, HT and phantom, could be built and wired together as a complete and fully functional assembly on this panel. I could even fit a mains on/off switch there too. After laying the components on the panel it became clear they would just fit so I drilled the holes and fitted the parts I have to hand:

The International Power heater supply is on the right and just below it will fit the mains on/off switch (this will be at the top of the panel when it is fitted to the Rackz case). In the centre is the HT350 PCB for the HT supply and on the far left is the phantom power supply PCB. Between these two PCBs will fit the custom toroid transformer for the HT and phantom supplies. Mains will entier via a fused IEC connector at the rear and be routed along the right hand side of the case to the mains switch from where it will be wired to the two transformers. Heater, HT and phantom supplies will all exit to the left and be routed to the rear of the backplane. The picture below shows the power supply panel fitted into the case. You can see it is quite close to the bottom of the sub-rack:

You might think having these two large transformers would make the mixer front heavy. However, there are six quite large output transformers to be mounted on the rear input/output panel which I expect will largely balance them out.

Inrush Revisited

A while back I reported the problems I was having with a SMPSU I planned using to power the heaters and how it would not power up with a full compliment of boards. I suspected this was due to inrush current which occurs because the heater resistance is much lower when cold.

I have now investigated this further. First I purchased an International Power linear 12 volt power supply rated at 5 amps. Another EZTubeMixer builder had reported that even this power supply would not power up with six boards (check out Pierre's blog here):

http://theworldin35mm.com/espressojazzdiy/category/diy/).

I suspected this might be due to a fold back current limit circuit in the power supply. The manufacturer does not supply a schematic so the first thing I did was take mine apart to try to trace out the current limit part of the circuit. The main regulator element of this power supply turns out to be the ancient LM723 which was introduced in the 1970s!!! Together with three equally ancient 2N3055 power transistors wired as darlington series  pass transistors this forms a linear regulator capable of at least 5 amps output current. So why would it not power up six boards? Further investigation showed it includes a current limit pot which, when somewhere in the middle of its range, implements a foldback current limit, but if turned fully clockwise should revert to a regular current limit. So in theory, turning this pot clockwise should fix things.

First I tried it on a six board load with the pot untouched. It did power up but very slowly. It was at least 15 seconds before you could see the glow from the 6922 tubes. Of course, once the tubes have been powered up, if you turn off and then turn it on again, it powers up straight away because the heaters are still warm. I found I had to leave it for at least 15 minutes before you could repeat the test. Next I hooked up a DVM, set to its 10 amp range, in series with the supply to the mixer. On powering up, the current started out at about 0.8 amps then slowly rose to about 3 amps when it suddenly peaked at over 6 amps and then settled slowly down to about 2.75 amps as the tubes lit up. I repeated this several times at 15 minute intervals with the same result. This seems to clearly demonstrate the foldback current limit operating but allowing enough current through to to slowly heat the tubes until a point is reached when the foldback limit shuts off and the regular current limit takes over.

The odd thing was that my supply worked on 6 boards but it did not work for Pierre. This could be differences in the tubes and also the tolerances in the power supply. Either way it seemed marginal. So next I tried with the pot fully clockwise and I was very pleased to discover that the foldback current limit was no longer operational. Instead the current rose rapidly to over 6 amps then slowly fell back as the tubes lit up. The time from switch on to seeing the heaters glow was less than 5 seconds. I told Pierre who tried this with the same result. So it looks like rotating the pot clockwise does turn the supply into a regular current limit and allow you to power 6 boards.

The international Power 12V, 5 amp supply is large, heavy and expensive so I wondered if I could come up with something smaller, lighter and cheaper. After looking at a range of regulators I finally realised that the LM317 regulator in its TO220 package, has the same pin out as the TL783 I use in my phantom power supply PCB. Using this chip, changing a couple of resistor values and changing the smoothing capacitor would allow the same PCB to be built as a 12V heater supply capable of supplying one and perhaps 2 amps of current. Quite handy for a couple of tube mic pres or a tube gain make up in a passive summer, but not enough for this mixer. To provide in excess of 3 amps really needs the TO3 version of the LM317 and a bigger heat sink. Perhaps this part with is chunky heat sink could be mounted on the rear of the PCB and wired by flying leads to the PCB? The TO220 heat sink I use on the PCB has two mounting holes. Not surprisingly, these are not the same spacing as a TO3. However, by chance, the heat sink is off centre and if you drill a hole a few mm past the off centre fixing hole you can fit a pair of pillars to the rear of the PCB that mate with a TO3 and it sits almost central to the PCB.

So I built one. I had a spare 12V 5A toroid transformer, not ideal but good enough for an experiment, I built the PCB and mounted the TO3 version of the LM317 regulator  and its heat sink from the rear of the PCB, like this:

Then I connected it up to the mixer via an ammeter as before and switched it on. And it worked! The meter flipped up to just over 6 amps then slowly settled down to just under 3 amps as the tubes lit up. I ran it for half an hour and the heat sink got quite warm but no warmer than the International Power supply did. I will probably use the International Power supply for this mixer because it works and I would need to get a higher voltage toroid to ensure the LM317 has enough operating voltage but at least I know an LM317 can be used.

The PCB will need to be updated though for two reasons. First, the IN5400 series rectifiers it uses are not really up to supplying more than 3 amps continuously so I plan to modify the board to accommodate readily available 4 amp and 6 amp bridge rectifiers. The other problem is the power tracks were never designed for 4 amps; they were designed for 100mA of phantom power current so they need to be replaced by some copper areas. The good thing is the new PCB can still be used for phantom power as well as heater supplies.

REDD EQ Added

I have just completed tweaking and testing a new EQ based on the EQ used in the famous EMI REDD desks. I have duplicated the standard per channel +- 10dB range stepped bass and treble controls (both 'classic' and 'pop' versions) and also included the brilliance boost/cut control from the EMI RS127 plug-in module which provided boost/cut at frequencies of 2.7KHz, 3.5KHz and 10KHz in addition to the 4.7KHz of the 'pop' plug-in. This all fits on a small PCB just 100mm by 100mm.

The three level control switches are mounted directly onto the PCB and the brilliance frequency select switch is connected by a short length of ribbon cable. The schematic is a little more complex than some of the other EQ circuits I have designed because the RS127 appears to have the same Q in both cut and boost. This necessitates using separate inductors for brilliance boost and cut as shown in the protootype schematic below:

I am going to fit this EQ to one of the channels of the demonstrator mixer. I plan to have two channels with Sowter input transformers and Pultec EQ. As well as being suitable for recording, these two channels can also be used for mastering as they have identical transformers and EQ. The Cinemag input transformer will be paired with the Helios 69 EQ and the Jensen with the new REDD EQ. Prototype test results can be found here:

Test Graphs

I also made a short demo recording showing the REDD EQ in action on a pre-recorded track. You can hear the result here:

Just Bob

The left channel is the track with EQ applied and the right channel is me commentating on the EQ settings at each point in the track. The track was recorded by by good friend Bob Wright:

The Song Factory

Building and Testing

I have been busy soldering the last couple of weeks. All six amplifier boards are complete apart from a small number of components that have not yet arrived and I have two EQ boards yet to build. As I have a full complement of amplifier boards I thought this would be a good opportunity to check that the little 12V 3A SMPSU, that I got to power the heaters, was going to work. I hooked up the PSU output to the motherboard, plugged in one board and turned on the power supply. The tubes lit up. So far so good. So I plugged in a second board and turned on again. Again the tubes lit up. So I carried on adding boards. This is how far I got:

The SMPSU refused to power up more than four boards which is a nominal consumption of just over 2 amps. Four it will do, five it will not. It seems to go into some sort of short circuit shut down mode because with five boards in the voltage across the heaters is zero volts. It seems to be completely unharmed by this as you can go back to four boards, turn it on again and it is fine.

I suspect this is an inrush current problem. Cold heaters have a much lower  resistance than when they are powered up and hot, just like any incandescent bulb. It appears the SMPSU can cope with the inrush current from four boards but no more. I shall next try a SMPSU with a much higher reating, say 6 amps to see if that works.