I am, at the moment, somewhere halfway between excitement and fear. I hope you like more shots of wires then. I got a new phone just before starting this build and am constantly fighting to take decent shots with it. The main lens' focal length is so short half the image seems to always be out of focus, and I end up having to use the rubbish macro lens. So, I'm glad you like them anyway.
I like shots of your wires since they are so evenly twisted. You can try to take pictures from further back and then crop
Divert away, I don't mind at all. I've been thinking about the whole gain thing since yours & @Jacob29 posted about it and there certainly seems to be a variety of gain levels to choose from in the First Watt lineup, from the Aleph J and the J2 at 20dB right down to the F4, with none. The Aleph 30 is listed as 20dB balanced and 26dB single ended, but I'm not sure which connection the Aleph J gain figure refers to. The 6dB difference when swapping the Aleph J with the F6 was quite noticeable, but so was the change in sound. I like the F6 a lot, but I'm not sure I'd ever choose it over the J. The SIT-3 is the really interesting one to me, and for someone looking to lower their system gain could be just the ticket. It seems very well liked by reviewers and users, and if you don't look too hard at the distortion percentage the specs seem pretty nice. Great idea, thanks.
Thanks! Yeah, I was looking at that gain table again myself. I’m usually on the bottom step or two of my TVC, and I don’t like that. I am curious about the SIT-3 but my mind doesn’t like the mixing of the SIT and non-SIT output devices. I could go back to a single ended connection to my amp and leave jumper out of the empty XLR, that would give me a few dB less gain. I’m also just super curious to try a different one of his amps, separate from any gain discussion and I’ve been going back and forth on which one it should be.
When I was designing new PCBs for my ANK DAC upgrade I spent a little time unsuccessfully trying to understand the M2x power supply, and the tube regulator it employs, with a view to utilising it in other builds (like a tube preamp for instance). Initially I had just assumed, due to its Audio Note heritage, that it contained within some secret sauce other power supplies lacked - a perception encouraged perhaps by the unique way the schematic was drawn. This is a series regulator (as opposed to shunt), and for anyone wanting to know the difference @Jaytor has provided a clear and simple explanation in post #18 of his excellent thread here: Yet another DIY Preamp - Vacuum State RTP3 Jaytor notes that a shunt regulator is inefficient as it requires the production extra current that will likely be thrown (shunted) away. Well, the series regulator can be inefficient too in its own aspect because the voltage going in must be higher than the regulated voltage coming out, and for the tube variety the differential can easily be into the hundreds of volts. The only other representation of such a tube regulator that I was familiar with coming into this project was from this reprinted Sound Practices article: A One-Tube Regulator Article By Mike Vans Evers Of Sound Practices Magazine and while there were obviously elements in common to the M2x, it was difficult for me to make sense of their similarities & differences at the time. So, from the beginning of this build, as I gathered together its separate pieces, one of the main goals was to finally understand the workings of the tube regulator and whether one based on the 6BM8 (ECL82) would be sufficient to produce not only the required voltage & current for this particular project, but also ones in the future which may be even more demanding. I've made this mistake a few times now, trying to design for universality rather than directly for the problem at hand, and it hasn't been ultimately beneficial to either so this will probably be the last time I approach things in such a way. I decided, only somewhat arbitrarily, that a target of 300V at 100mA would likely prove suitable in most circumstances. The M2x would already be incapable of this, due to its use (pre-regulator circuit) of the 70mA rated 6X5 rectifier tube, so that part of the power supply would have to be re-thought, a little at least. Further research turned up a couple of other relatively in-depth articles: Valve Technology - A Practical Guide Tube Based Voltage Regulators - Part 1 However, much like the Sound Practices article, neither really had satisfying entry-level explanations of their working or the process of component selection & values. So I started to collect the designs I found on the web and re-draw them all into the same basic structure, in the hope of gleaning some insight. This was quite helpful in illustrating how much the building blocks of all the designs were fundamentally the same. I also found it interesting on the other hand that such a variety of component values were all effectively achieving the same or similar results, while showing that the M2x wasn't particularly different in its implementation either. It was a tangentially related search for information on current draw though that produced the key source for my understanding the series regulator - again, at TubeCAD journal. If this project turns out to be a success then I may consider myself obligated to become a John Broskie Patreon member. All roads lead to TubeCAD it seems. This article is by a different John however, so thank you John Atwood for this one: Tube CAD Journal: Tube regulators part 2, July 1999 Here, not only do we have a basic introduction to the series regulator, but some practical guidance for selecting components (plus some killer equations of course). In short, the most simplified version of the 6BM8 (Pentode/Triode) regulator looks like this, with the pentode of a 6BM8 serving as a regulator and the triode as an error amplifier: a. The pentode sections plate is connected to its screen grid to facilitate triode operation (triode strapping). b. The plate resistor for the error amplifier is connected to the regulated output to improve regulation at minimal current draw. c. The cathode of the error amplifier is held above ground by the zener diode as a reference voltage. d. To reduce impedance and AC noise the zener is bypassed with a capacitor. e. The zener is provided operating current through this resistor. Here it is drawn from the regulated output, but can alternatively be sourced pre-regulator. f. A resistor divider sets the output voltage. Often a variable resistor is used for adjustability. g. Unsure - assume for noise reduction. h. A decoupling capacitor across the output lowers the output impedance, and probably smooths as well. The last piece of the puzzle was surfaced by the same search - an AudioXpress article by Terry Bicknell, which outlined a number of tube series regulator designs and, most importantly, tested each for noise and stability at varying voltages & current demands. It was exactly what I was looking for: https://audioxpress.com/assets/upload/files/bicknell2890.pdf It demonstrated that the above 'standard' regulator was not really capable of 300V at 100mA, and although the best performance could be gained by the final unit under test I chose the 3rd circuit in the roundup, due to it having acceptable performance around the operating point of my proposed CCDA build and that it wasn't so dissimilar to the standard version that I didn't understand how it worked. At the time however I hadn't considered the current drawn by the operation of the selected regulator circuit itself, which added an extra 22mA to the 66mA needed for the CCDA circuit. Modifying the circuit a little I was able to shave 6mA off that amount, which doesn't sound like much, but I figure when your dealing with such small values in the first place every little bit counts. Part of this included reducing the zener diode current down to 3mA, below the 5.5mA of the original circuit and well below the datasheets 12mA nominal test current. No ill effects were observed in the prototype build, but I hardly put an oscilloscope over it. The final design: Also, the EL34 in the example circuit was replaced with a dedicated voltage regulator tube, the soviet 6S19P. In all honesty, I suspect this regulator might still end up operating within a range of increased noise, but with the constant current nature of the CCDA circuit, it will hopefully remain largely un-strained.
Thanks for posting all of that! It's fascinating how similar all of those power supply stages are after you normalized the topologies. I've never designed with tubes but I'm starting to think maybe I should come up with a project.
Great post! I really liked your comparison of the various circuits. A 100v zener is going to be pretty noisy, so you want to make sure you use good filtering (high quality film cap for c5). It might be worth considering a second RC stage before the grid of V3.2, although since the grid current is quite low, that might not be necessary or make much difference.
Yes, dirty old zener. I did consider working out a way to replace that with a gas regulator, but board space was too limited for adding another tube in the end. The original circuit had a couple more zeners in the adjustment divider, although I don't know how much of a noise problem they'd create in that position. I assume they were there just as an easy way to drop 150V, but with a B+ of around 300V the current down that leg is over 5mA. Changing the zeners out for resistors and adjusting their values brought that down to 1mA. Similarly, increasing the values of R5 & R6 halved the current in that section to 2mA. Pin 7 of the ECF80 doesn't seem to draw much thankfully. A kind soul on the diyaudio forum actually simulated my tweaked circuit for me, which I was really grateful for, and it all appeared to come out okay.
Soldering the back panel RCA sockets once again had me wrestling with sanity. The ground tabs being a part of the jack body requires more heat than my iron can deliver making it almost impossible to get the solder adhering to the metal. The first run of trying to connect to ground wires resulted in a blobby mess that failed the 'pull test' so I filed off the gold plating down to the copper below and tried again. It still looks terrible but everything seems tightly fixed now. Although the signal wires were only just long enough to reach the solder cups the end result didn't come out too bad. Unfortunately I was to find that the first three out of the four twisted pairs were connected to their output sockets incorrectly - positive to negative and the reverse. I was baffled for a little while. After all I had tested each with the multimeter. However, I did not factor in that (as they are currently unpowered) the output relays were in a closed state, which shorts signal to ground. Measuring continuity between the board ground and the wire end like I did was always going to give a positive result regardless of which wire I tested. I'll have to desolder those pairs from the PCB and swap them around. At least this thing on the other side of the chassis went okay.
Looks nice and clean. I hated those kind of RCA connectors. I switched to using WBT connectors which are high quality and easy to solder to, although a bit pricey.
I had considered the similar style KLE connectors, and kinda wish now I hadn't been so tight-fisted about it! There doesn't appear to be any 'high-quality' knock-offs of these low mass designs like there are for the RCA interconnect plugs, unless I've just missed seeing them.
Work on this has been slow going recently as I've found it a bit difficult to get enthused about the last stretch of awkward close-quarters wiring. There was a lot of tedious maneuvering around the components with a pair of needle-nose pliers. I found a relatively stiff piece of plastic able to be used as a temporary front panel until I can be sure the preamp functions as hoped. With all the mains and transformer wiring completed it will now be quite difficult to make adjustments or fixes without taking it all apart again. The last stage was connecting up the main front panel doodads to the headers in the microcontroller section (minus the LED indicators for now). The IR receiver too is only a short-term bodge up just to make sure it works correctly. I'll have to work out a proper mounting system for this on the actual front panel. Unfortunately the first power-up did not go very well at all. I left the tubes out of the circuit so I could initially ensure the voltages on the transformer secondaries were all correct and concentrate on the low voltage supplies at the front & rear of the PCB. However, none of the input or output controls worked as they should have, and upon further investigation I found the current being drawn from the rear 5V power supply to be extremely high. It was while I was staring at the board, wondering how on earth to troubleshoot my problem, that I spotted this stray washer hugging a capacitor in the output delay circuit. I must have dropped it while assembling the chassis side panels or something, and looks likely to have been shorting the 5V rail directly to ground. Removing the washer improved the power supply performance considerably, but the current draw was still 400mA, which is about 20x too much. This can be alleviated by pulling the two chips in the input selection circuit, so hopefully they were fried and replacing them with new ones sorts the issue permanently. Input 1 was still activate, so I decided it was time to try putting some music into it...
Test runs of the "final product" have been a mixed bag. Let's get the main thing out of the way first - it sounds really fantastic, even straight out of the box. There's a noticeable sense of ease/lack of restraint to the presentation, with the music seeming to just pour out of the speakers. Bass depth and heft seem an improvement over my Aikido, as is the holography and detail in the soundstage. And not the fake detail of a lifted top end, just more revealing of the recording (and even of the noise coming from the source unfortunately). For those that value the indefinable concept of 'musicality' this has it. As predicted, it is a heat generating machine, but I'm still surprised how much some of the adjacent components are being warmed up. The surface of the box capacitor right next to the regulator tube for instance increases in temperature quite notably. In the short time I've been listening though there been issues presented that, if they can't be rectified, will see this project remain a conceptual piece only. Firstly, the output mute circuit works well, but only on startup. As @Davey kind of predicted the circuit doesn't seem to be able to handle the quick discharge of the output capacitors, and a rather large crack is sent through to the speakers. I can't say at this stage whether the capacitor charge is getting through or it's the disconnection of the relay contacts themselves that is the main contributor to the problem. But the purpose of having the circuit there at all is defeated if it doesn't work for both turn-on and turn-off right? The attenuator functions quite nicely, controlled from either the remote or encoder dial, although the first half of the step range is largely useless. I think I'll be able to alter the code and have the software ignore this section. As the attenuator moves through the steps it does ping a little through the speakers, each click amplified a little by the preamp circuit. But it's not a major problem to me - the same thing happens with the Khozmo control in my Aikido. The main issue with the attenuator is that it pops through the speakers quite loudly at certain steps, but only while signal is being applied. I may be able to map out the steps and see if there's a particular relay involved, but my initial guess is a grounding issue. Speaking of which, I have placed a front panel ammeter on the ground rail between the signal circuit and the power supply, to provide a fun indicator showing the amperage drawn by the CCDA. It appears I didn't consider the voltage drop across the meter, which although only about 0.18V then elevates the ground of the entire CCDA section by that much. I wouldn't suspect this to pose any issue to the running of the CCDA, except that the signal ground is also linked to it, and therefore raised above 0V too. I will investigate whether this is a contributor to the attenuator popping problem. The only other things of note at this time are power related. Although all the supplies are running well, 700VCT for the main transformer is about 100V too much, requiring a significant voltage drop across the regulator tube. Things aren't helped in that regard by the 6.3V AC winding on the same tranny in actuality delivering 7V AC, probably due to lack of load. The two over-voltages no doubt share some responsibility for the heat the regulator tube emits. All the other voltages on the board seem good however. A basic first conclusion then would suggest that incorporating anything relay related was a bad idea, and with a standard input selector switch plus a volume potentiometer this project would be a done deal as long as I continued the practice of turning it on/off after/before the power amplifier. I'll certainly be attempting to alleviate the issues above though. It sounds too good not to try.
Well, I say kudos to you! 1. It was a design by you 2. It looks awesome 3. It sounds awesome Good luck on finding those fiddly bugs. You have a lot to be extremely proud of here!
Thanks, I really do appreciate your words of encouragement. It's about 90% there, and hopefully I'll be able to get even more out of it with a little tweakery.
It's great to hear that you are liking the sound. Congrats! I don't recall if you described how your relay muting circuit works. How is this controlled?
There's some chat about it on page 12, from post #284. Designed by Pete Millett, it has a 555 timer which engages the relays 30 seconds after startup. When the power is cut the relays auto-disengage. I have hooked it up in parallel configuration as shown in post #288, but as Davey did predict, a series resistor is probably needed so the discharge through the output caps doesn't swamp the relays.
I've done a little work mapping the functions of the relay attenuator in order to understand what might be making it pop at certain steps. Basically, each of the seven relays can either be in the normally-closed (NC) or normally-open (NO) position. When in the NC position that relay is attenuating via a resistor divider, and in the NO position that same divider is bypassed, and therefore not attenuating. If all the relays are NC then the signal is fully attenuated, and vice versa for NO, which is full volume - the whole attenuator effectively bypassed. At each step different relays are in NC or NO, changing the level of attenuation for 128 steps (0 to 127), and it behaves very much as a binary counter. In fact, it wasn't until I watched an unrelated video on how to count in binary that I came to understand how this attenuator (and those like it) operated. Anyway, here are the troublesome steps that send a pop through my speakers: Step 64 pops when selected after step 63 and the reverse is also true. This makes a bit of sense - all seven of the relays are changing contact positions when moving between these two steps, so the likelihood of noise is highest. The other problem step is 96 (but only when coming from step 95) which has the first six relays all changing at the same time. Strange though, that moving back to step 95 doesn't show the same effect. Another large switch of the contacts occurs also at step 32, but I think there may be too much attenuation at that point to hear anything. Basically each 32 steps, 31/32 to 63/64 to 95/96 is where the most contact switching occurs. Even on my Khozmo step 32 is liable to pop a little. I'm not sure what to do with this info. I was hoping it would be easy to spot a particular relay that was creating noise, but it appears that was wishful thinking.
Yikes! This gets into digital encoding methods. I learned about various encoding strategies in 100 and 200 level engineering. But haven't needed to use that knowledge in over 30 years. I'm a bit rusty on all of this. R-2R DACs have some similar issues with glitching when lots of bits flip at once when moving from one adjacent or nearly adjacent state to another adjacent state. You can avoid that by using a different binary coding scheme. Gray code is one way. With Gray code you only flip one bit (or one relay) when switching between adjacent states.
I think Grey code would be smart (also means you’d have to change all of your relay weights) and that way if there is some not-make-before-break or vice-versa action going on, at least one will have continuity at all times.
