• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Finally building my MkIIIs, a pictorial thread

Hi Jxdking. I got these amps with the original circuitry already out of them. They had some horribly messy home made pcb driver boards in them. So that's why I had to replace original circuitry-It already wasn't there. As for why I chose the tubes4hifi driver boards, they make my life easier in designing a new amp around KT88 outputs.

As for the gain, I'm not really sure what you mean by "either one is too hot for me". Gain requirements are specific of each amplifier, and I don't see how a gain of 40, or 69, could be inherently too hot or cold for one's taste? But please, I'd love to hear why you think so in this case. I don't have the exact gain decided yet. But I know there's more than I need, which means I can tune the feedback circuitry, of a circuit I already know works well in this application, using KT-88 like I am 🙏

-Loren

Here is the close loop gain that I was talking about. "Hot", I mean the gain is too large. Usually, the overall gain should be around 20, so that the amp will not pick up some noise easily.
Regarding feedback point, just use the 8 Ohm tap, unless you own some 16 Ohm speakers.

PS: Before tinkering the gain and feedback, you'd better build the kit as intended way first. Make sure it does work as designed. Then you can mod the gain as you need. Watch for any instability issues. With sqarewave, overshoot is fine, but you should avoid any ringing.
1729216073325.png
 
Last edited:
Yes that is the plan. I'm making some small changes at first, but nothing to change its operation and barely changing its operating points. Years ago I was in a horrific accident and lost a lot of memory. I've only been back into valve audio, seriously, for a little over a year now, and am constantly learning and re-learning. Sometimes I worry on the forum about coming off like the guy who just wants to mod things. I do like to sit down and meditate on soldering. But only modding something if I understand, only if it's something I can do the math for. I like learning, and am so grateful for this resource and the ability to receive constructive criticism🙏

Speaking of which, your comment...Yes I understood where the feedback components were and their function, sorry for that confusion. What I don't understand is "normal" gain? When you say normal is 20, are you talking about 20db? I've been referring to voltage gain. But normal for what, this circuit? I just don't understand, you must be talking about a specific circuit with specific operating points.

Loren
 
  • Like
Reactions: huggygood
What I don't understand is "normal" gain? When you say normal is 20, are you talking about 20db? I've been referring to voltage gain. But normal for what, this circuit? I just don't understand, you must be talking about a specific circuit with specific operating points.
Sorry, “normal” means the majority tube amps have the voltage gain about 20. 20 voltage gain is 23dB. The original MK III has about 22 voltage gain with the 16 Ohm tap.
 
Oh now it makes sense. 1.5v input to 23v out, 23.6db. Speaking of the 8 ohm tap, of course. Of course the amp up to the OPT has much more, but voltage is dialed back down after the transformer. I still get confused about db sometimes. I think about power doubling every 3db, I always remember that one speaker sensitivity considerations...

Ok so thank you for that. I'm still having a little trouble with the math. Here's a pic of the calculator I've been using, and the equation I'm trying to figure out.
IMG_0560.png

db=20 log(Vo/Vin). So I headed over to some other sites to find examples and figure out how to solve these,
IMG_0559.jpeg

So, reversed it is 20*log(Vo/Vin)=db
Which means log(22v-1.5v)(20)=db
So log14.6*20
And I come up with 1.1643...x 20= 23.287db

I'm sure I'll be using the calculator to find the common log of the voltage gain for a bit, but that's ok. I wanted to know the math, and now I do. I knew of the logarithmic scale, and basically how it functioned, but reading up on it I did not realize there were different logs (bases). Other parts of the history were super interesting too, like where the name decibel came from, and some of the historical uses the equations held.

Thanks again Jxdking. I've been busy lately but when I get back to the amp I've still got a bit of work dressing up the chassis with components. Next I want to learn more about poles, because I know the introduction of yet another one (I'm used to single coupling cap amps) is something I need to understand.

Loren
 
Hello again all. Work has been really busy this time of year (construction, before rain hits) but I got to sit down for a few hours last night. I've got more on the chassis now, binding posts, choke and fuse, rcas, octal sockets. I'm trying to figure out where I'll put everything, before I put it there. So, it was time to open up PSUD2 and see what and how many things I still need room for.


Image 10-22-24 at 10.31 PM.jpeg

First I wanted to see how the supply faired wired as stock, with extra choke filtration for the drivers and input stage. The ripple was satisfactory, but I wanted a little better, and a little better transient response (I'm using a stepped load to see how it reacts, I know not perfect, but helpful).

Image 10-22-24 at 10.34 PM.jpeg

So I increased the second cap. I've always wondered why Dynaco made the first cap larger, don't see that often. I figured less stress on the choke, but when I modeled it, it was such a slight difference it still made no sense. Anyhow this is where I fell. First cap is a 30uf film 600v>stock c-354 choke 1.1H>68uf 550v (2 sections of 105degree can cap in parallel). This gets me down around .01% ripple, better than I need. The 7H>34uf (3rd section of can)>~6k>47uf then makes ripple less than I could read.

To finalize results, I needed to pick a rectifier and get down around 500v raw B+. I plan on hexfreds and using the rectifier as a slow start/dropping device. Also, we'll see, I may very well have enough voltage on the 5v winding to run a damper diode (that's why you see 6de4). But I got to thinking, I'm not 100% on how much voltage the device would drop in this application? With the 6de4, it's probably real close to what psud would model. But if I went with a full wave 5v rectifier, then I would have both sections in parallel. So would this halve the resistance and voltage drop, approximately? I looked all over for this particular question, but could only find references to parallel when people were needing more ma. and were so doubling up on full waves.

Thanks all, I'll finish and post the PS soon as I figure this out, bias supply too.

Loren
 
Last edited:
Hi Loren,
A little late with this response but I used SS full wave rectification in my Mk2's with a 6CG3 damper diode downstream in series to give a slow start. Works well heating it with the spare 5V winding and as I remember gives about 10-15VDC drop. I didn't read back through the thread to see if I'm repeating myself.
John
 
I don't think you are...I'm going a similar route, only with 6DE4. But since I didn't get any response on the dc voltage drop questions in my last post, I think I will test a few things in there, at the nominal 150ma draw, and include some results. I've been making slow progress, one chassis is all done but for a couple lug strip placements, and the other is on the way. I struggled with deciding to cut into the chassis, but I'm glad I did because I rather favor having an ice socket like this. Makes it much easier for me to change in/out components of my system here. Just haven't been back to the thread because I want to have some of that data in hand. I will be soon though to show the build before I begin wiring.

Loren
IMG_0617.png
 
I just use a sheet metal nibbler, a handheld one. I must admit it's a little tough, but probably wouldn't be if I invested in a good quality pair of them. I'm sure there are longer handles and other means of more leverage for the better nibblers. If you have to do it on a painted chassis, just tape around your hole so the face of the tool doesn't scratch the paint, and make as many drill bit pilot holes as you like. By the time I get the nibblers in there, it's really just for the corners that the drill bits leave. Then last I get the files, so I don't take too much because there's no do-overs on power sockets.

Loren
 
Progress has been slow, but last night I finished one of the amps. Sound check after I post this, but the voltages are where they're expected to be and amps are stable&mechanically mute. So let's see, I left off with adding the IEC outlet. Huggygood, I procured a 6.5"x6.5" throat coping saw, like that. Of course after completing both chassis sockets haha, but they turned out alright.
IMG_0758.jpeg

I'm re-using the original fuse block, a dpdt switch I had, and so it was time for power supply. I ordered a 34uf550vdcx3 can from Hayseed Hamfest, (I like the look, and supporting domestic makers), and have plenty of my own hv caps. I decided on a 600vdc 30uf film for the first cap. I'm using a 6DE4 bypassed with 1.3kv uf diodes for rectification/slow start. It's using the 5v taps (heater floating for now), and so voltage comes up very slowly to 524. I'll lower that later, and or elevate kt88 cathodes.
IMG_0781.jpeg

Fit right nicely in between the rectifier and can cap. Next is the stock 1.5H 50ohm choke, and then on to the combined can cap which measures 105uf. There's B+. After that, a pair of 5w 15k resistors =7.5k to drop the B+ down to 408v, as well as a beefy 470uf 450v electrolytic.
IMG_1043.jpeg

Then on to the board where it goes through additional rc stages to feed the input and LTP tubes. Everything is spot on, input 6sn7 4.1ma 145v, driver ltp 4.2ma (each) 208v, kt88s at 50ma 518v with plenty of spare room bias...Except the expected cathode voltage of the ltp was 10.5v and I'm getting 7.7v. But I know CCS can confuse that, and I know it's working at the current I set, so I'll wait to see if it still has enough headroom or not. The blue wire, resistors, and green safety cap (bottom right) is where I took advantage of a free solder terminal to make the shortest b+ reference I could. Resistor to terminal, back to ground with resistor and cap. Tapped at junction. 6.3v heaters elevated to 25v. You can also see I added stoppers on the grids as well as the screens of the kt88, 1k and 82ohm, respectively.
IMG_1044.jpeg

What else...the extra terminal strip spaces will be used up when I install the LED bias meters. The original switch, I've thought about using as a 4/8 switch. They're new, not original, rated for 6a125v. But I'm still weary to trust them for dynamic spikes...extra components soldered to pcb are bypass caps and temperature compensation for the lm334. The negative bias diode and cap I placed off board, and so I just turned that spot on the board into another rc filter. Now it is 470uf>680ohm>100uf>1k>100uf for the bias supply, and works a treat.

I'll get an updated schematic drawn up, and some initial reports soon. I never did get around to testing different dampers for DC drop specifically, but getting refreshed with them, it seems probably the same per current as the ac charts in their sheets.

Time to try it, thanks all

Loren
 
Thank you 🙏🏼 . It's dissipating 1.45W or so, less than 3/4w each, so I shouldn't have any issues there. The one supplied was this tiny 3w "flameproof", that likely would have flamed 😅

Gave initial testing this evening. No smoke. Sounds good, stable, dead silent, that was nice. Too much gain, but really close. Bottom end seemed a little lacking? But again it was close, and the gain mismatch made fair assessment impossible. I've got GE 6550s in there now biased at 50ma, and running UL with around 515v on the plates. This quick edit shows the current values/measurements on the driver circuit board. And here's a couple photos from up top as well.
IMG_1061.jpeg
IMG_1062.jpeg
IMG_1063.jpeg


Loren
 
  • Like
Reactions: huggygood and aroS3
I agree about the NFB. Even on the 16 ohm tap I thought it was too low. I didn't measure after the change so not sure how much I ended with but using the 16 ohm tap and halving the 6.8k by paralleling another 6.8k I thought it sounded better, especially in the bass.
 
  • Like
Reactions: kb2wyl
Check the NFB ratio. R8 ( 6.8k) might be a little high, also NFB should be connected to 16ohm tap to use the full winding.
To low NFB will get boomy bass and high dist, very high sensitivity as a by product.
Hi Peter. I thought the same thing. I know some people like the 8 because "it's the one I'm using" I like the 16 because that's how the transformer was designed originally. So yes, that's where it's placed. The 6.7k (my schematic said 6.8) was just what I had close. I erred towards the lower side because some of the other versions of this circuit (Dynaco 70, VTA ST-120 and M-125, etc) all used lower values, and I figured I'd be fine tuning it regardless. Thanks-

Loren
 
  • Like
Reactions: huggygood
Almost forgot, I wanted to post the final power supply schematic since I had posted another earlier in the thread. Again rectifier tube is a 6DE4. Heater's -5.45v, taken off the 5v tap, floating. The actual rectification devices are these NXP damper diodes (I know, ironic) I have had success with. Look at the specs, still impressive to me for an all plastic package.. Voltages are almost identical to PSUD modeling🙂
IMG_1074.jpeg


I've always shied away from slow rise times in the power circuit. But in this instance, I didn't see an issue. With the capacitors I had, either a 30uf 600v, or the 470uf 450v cap, had to go last in my main CLCRC supply. So I simmed it, and I thought about it. The driver stage here has headroom, and I understand it is primarily pulling from the last cap, which in this case is the 47uf on board, correct? (And the 102uf can cap for output stage) I don't think the slow ramp up to voltage for the PCBs should be an issue. Obviously everything here can handle the extra initial current. The 6DE4 takes 25sec to warm up, so in practice, there really isn't any. Please someone correct me if I'm wrong. It even felt wrong doing it, ha ha, but really I didn't see cause for concern. For what it's worth this is also why both the on board caps are the 47uf, and not changed to something larger. After 470uf and those big Rs, 47uf was more than sufficient,
IMG_1073.jpeg
I figured.

Thanks all,
Loren
 
I should've checked that. Another thing I've had to get used to- It's so much more frequent that parts are replaced by newer models, and then discontinued entirely. Honestly, 4007's in a well designed supply "sound" OK to my ear. So, any time I'm buying HV diodes that look superior to them, I don't worry about much else. I've had these at least 6-7 years now.

The replacement that digikey suggests https://www.digikey.com/en/products/detail/stmicroelectronics/STTH812D/1117517
looks more than acceptable. 1200v, 8A, and at 50+ only $0.77 each (otherwise $1.25 ea). But if it were me I'd just be searching through lots, of hv diodes, at my internet resources. On the hunt for ultrafast and soft recovery, high junction temp capabilities, you likely could find even more economical choices than the $0.75. For 2/channel, that's not bad 🙂

Loren