or Connect
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › YANSSMH (Yet Another Not So Starving Student Millett Hybrid)
New Posts  All Forums:Forum Nav:

YANSSMH (Yet Another Not So Starving Student Millett Hybrid) - Page 2

post #16 of 33
Originally Posted by Goobley View Post

I'm back again, with a metamorphosed amp. Yes some of the cables are longer than they should be however, they are mostly and have all been braided (the ones that aren't braided are not for testing reasons.)

Well I have some good news and some bad news, however I hope someone may have answers for some of it.

As you can see it is entirely point to point over a ground plane (zinc and foil). All cables have been kept as short as possible.
My main issue at the moment is a high pitched noise (around 8kHz in the left channel). This is the side with the two coupling capacitors, I don't know if they could be considered the issue, but no signal flows directly next to them. I still have a fair amount of other background noise (even when in grounded metal box) which may (probably) be due to the smps as it's about the same in both ears. I could live with it if I managed to clear up the uneven and very wearing noise in the left channel.

Do you think that the capacitors are likely to be a the left channel problem? And if yes, how should I go about resolving the issue?
Thanks for all the help I've had so far,

What do you mean when you say the left channel is the "one with the two coupling capacitors" ? As far as I can see the circuit is completely simmetrical.

post #17 of 33
Thread Starter 

The two big 680uF capacitors (lower right hand corner of the photo are on the same side as the the left channel, it's the only difference I can find. Sorry for not being clear.

post #18 of 33

Those two capacitors, C1 and C6, are power decoupling capacitors. They decouple the tubes and mosfets from the power supply. They create a reservoir of power, and work both ways: (1) They give the item downstream a reserve of power to give out in case of sudden need, diminishing the effects to the power rail, like a voltage drop. (2) They keep their voltage when the power rail goes down or up, and will either source or sink power to keep the power rail steady.


C2 and C4 are what we call the coupling capacitors. They are the link between the tube's output and the mosfet's input. These filter out the DC and let the AC signal go trough, allowing the tube and the mosfet to be biased independently.


So yeah, it's easy to get their name confused, and it can confuse people even more when you mix them up. wink.gif


Coming back to the two decoupling capacitors. I cannot see behind them in the picture, but something tells me you have them connected to only one channel. Though when I think about it, it's not really possible, since both channel share the same power rail anyways...


I'm guessing the two red wires are the MOSFET drain? Why aren't both of them connected together?


Which of these resistors is R13? Is it hidden?


Is the big black wire coming from the right and connecting to one of them the PSU V+?


Like that? This would be a correct wiring.




Have you verified the voltages at various points? Since stereo amplifiers are symmetrical, you can compare voltages from one side to the other to see if something's wrong. It's a good place to start.

post #19 of 33

Aside from Kim's questions about the wiring of those decoupling caps, I also see what seems to be a small ceramic disc cap just to the right of the decoupling caps. What is that part?


Over on the left side of your build, I can see that the braid is made up of 2 ground cables (red and blue) and 2 signal ones (green and brown). From what I can make out from the schematic those signal and ground cables are the output from the MOSFETs.  Here I'd like to point out a some things:


1.- The output from the MOSFETs is a low impedance signal and thus almost immune to noise. Braiding and/or shielding this cables is not really necessary.


2.- Braiding or twisting cable is only really effective if it is symetrical, cable of the exact same gauge is needed for that.


3.- On the other hand, the amplifier's input is high impedance and very sensitive to noise. The input signal cables MUST be kept as short as possible and twisted/braided/shielded or kept flush to the ground plane to minimize noise pickup. In your build, I can see a couple long runs of cable arching over your build and connecting to the grid stopper resistors. This is bound to pickup noise. Additionally, I can't see where the ground from your input signal is connecting, if there's no input ground coming in, the signal cables are "floating" and that can only compound the noise problem.


Finally:  cool way to build your own terminal strips, I really like your idea.


EDIT: I also forgot to say that your gate stopper resistors, the ones soldered directly to the MOSFTET gate pin, look perilously close to your ground plane, particularly the one in the upper section of your picture. It would be a much better practice to insulate it at least with electricians tape.



Edited by the_equalizer - 10/1/12 at 3:14pm
post #20 of 33
Thread Starter 

Right, I may not be able to play with my amp, but I can answer a few of your questions aimed at helping me :).


With regards to the gate stopper resistors I've been very careful with them, the amp as it is in its current state is merely a 'test bed' situation and has been running more upside down than the right way up.


Somebody's gonna have to explain to me how to handle the signal input though... Because if I have my RCA jacks on the rear face and my pot on the front I'm struggling to think of a way (using only newtonian mecanics) that I can go from the back to the front with onty a short run of cable.


Thanks for the pointers about the capacitor names, I know a bit about this stuff but I normally make it up as I go along, not having had any formal training. (I'm not yet in uni, and the electronics in the french crriculum is limited to Daniell batteries and simple lightbulbs and switches!).


The small ceramic disk capacitor is 100nF that seems to seriously attenuate noise from the PS, it however makes no difference to the noise in the L channel.


The annotations that Kim added are equivalent to what I actually built, but as I had 40 mins free earlier I drew up my PTP schematic, it's a bit hard to follow but I wanted it to be very similar to the photo.



I drew a purple box around the areas that I though would be most likely to contain potential issues, but I have no real science to base this off.


I measured some voltages and they were very close (as far as my dmm can discern, then again, it's not very accurate when you get up to 48V because the scale goes from 20VDC to 200VDC, it's showed itself to be to very close tolerances when I've tested it against better ones but I should really replace it.)


The noise in the left channel is probably around 7-9kHz and remains at quite a constant note, it doesn't wobble or oscillate much. I don't whether that extra info may help with diagnosis?


With regards to the terminal strips, they're made with trademarked, copyrighted and registered non-OFC finest quality and purity 2.5mm solid core house wiring cable attached to 13.5*13.5mm beech batons. I'm thinking of going commercial... but I think next time I'll try and even out the spacing... :D


Thanks again



post #21 of 33

Thanks for the schematic, it will certainly help in helping.


If you're sure your power supply is a big source of noise, I repeat my recommendation from some posts ago: get a Cisco 48V supply. Once you get your amp smooth and quiet with it, then add your own supply and debug it. Right now you don't know if the noise is coming from the amp, the PS or both and if some change you do to the amp actually worked as there's still noise coming in from your PS.


Maybe I did not look closely enough but I think one of my questions remains unanswered by your schematic: where are the input grounds connected? I see the output ground clearly labeled but as to the input I only see one individual wire per channel going to the grid stopper. Floating (unreferenced) inputs are prone to cause problems (noise among them)


About going from the back to the front, use the shortest possible run of shielded cable (or twisted pair); shield connected to ground at only one end (preferably the RCA jack end). Now my build is far from exemplary but you can take a look at the two twisted pairs that run from the RCAs in the back to the volume pot in the front in the picture I posted in the regular SSMH thread.  Yet another option is to do as Mr. Millett himself did in his own build and put the RCA jacks right next to the volume pot :)


About your gate stoppers, good to hear you're taking care of them. It is still a better strategy to just drop a piece of electricians tape there. For the cost of it you avoid:

1.- having to constantly take care that the resistors don't short out to ground every time you move something around

2.- possibly having to replace a blown output mosfet because it did short out


Finally, I'm sad to hear you've trademarked your terminal strip design as I was already planning on flat out plagiarising it for a build I'm planning to do as a Christmas gift to a relative. Would you consider open sourcing your design? To go right along with the design of the amp itself?  :)



post #22 of 33
Thread Starter 

Hi all,


I am henceforth open sourcing my terminal strip design, but would respectfully ask that anyone who tries it makes sure that they use solid core house cable rather than any of that audiophile ofc stuff! :)


With regards to my amp, I'm on my last few ideas here - almost at the giving up stage, which'd be a terrible shame.


1) I remeasured the voltages and in some cases they vary by up to half a volt (only of the valve pins though)


Pin:    LeftV   RightV

1        23.7     24.2

2        bugger all (0.02mV for both)

3        1.2       1.3

4        12.4     12.3

5         nil

6         see 1

7         nil (same as 2)

8         see 3



All the other voltages - MOSFETS etc are all the same across both sides and the PS reads as 47.9V under load. Does anyone see any issues with these voltages or want any other readings to help diagnose?


2) My final idea is to move the pot, and rca jacks to the top to be on the ground plane like on the original design, I'll make the cable seriously, crazy short and see what happens.


I'm out of ideas, other than, next week I'll be getting an old oscilloscope, so if there's anything I should look at with it (other than the PS) to diagnose problems then please tell me what. :)


Many thanks again for all your advice and encouragement (if I decide to give up I'll be building the diy version of the Graham Slee Novo, as it's solid state, not too challenging and as everything comes in the kit there are very few unknown elements to account for  -edit maybe a different solid state amp kit actually given some of the reviews I've read...  If I get lucky I may get both working :)).



Edited by Goobley - 10/5/12 at 11:32am
post #23 of 33

Variations from tube to tube and resistor to resistor (unless hand matched, and even then...) will cause variations in the measured DC voltages around the tube by even as much as 10%. You are well within acceptable tolerances. When people screw this part up it usually reads all or nothing where it should read in the middle.


Idea #2 - RCA jacks near the front of the amp are fun. I find front & top mounted RCAs a bit of a pain in actual use - front and side is about equally good as far as quality is concerned, but much easier to actually use & build. 

post #24 of 33
Thread Starter 

And it was going so well...


I have moved everything including the pot and RCAs to the top; keeping the signal lines right on the ground plane cleared up a lot of the noise. With my new toy (oscilloscope), I was able to determine that the power supply should be good enough (40mVpp ripple) even if not as good as a linear. I was also able to determine that the imbalanced noise was probably linked to the poor job I did soldering the links across the sockets as some of solder joints were very close to each other.  I had however forgotten that I'd almost buggered one of the mosfet legs the first time I soldered it - just moving the wire attached to it after unsoldering snapped it clean off. This has happened to me with voltage regs before and I've been able to more or less stick them back together unfortuneatly this fet isn't having any of it as the leg snapped off quite deep and and the plastic casing keeps oozing a liquid that means my solder won't stick.


It's fair to say I'm rather fuming now, as there is nothing I can do (no electronics shops within a hundred miles, and I'd end up paying 10€ for just a mosfet and postage in france).




I'm going to have to make some decisions about whether to continue.

post #25 of 33
Thread Starter 

Well it's me again, I've been playing around with my scope on this amp again today and I now KNOW that the noise is coming from the power supply (this is where you can say 'I told you so', the_equalizer :D). As the noise at the power supply input is almost identical to the noise at the output. One thing that surprised me is how good the power supply ripple rejection is on the valves. For the duration of this test I connected the RCAs to my main source (Pure i20), but with nothing playing, just to make sure I had a clean audio input. With regards to the valves I ran an Electro-Harmonix 12AX7 in one channel and an RFT 12AU7 in the other. These pictures speak for themselves (in all cases horizontal is 1ms/div and vertical is 5mV):



12AX7 Output (taken from valves side of 2k resistor between the plate coupling (?) capacitor C2 and the mosfet gate)



12AU7 output (taken from same place as the AX7)





Mosfet output, from its source (pretty much identical for both channels):





The power input to this amp looks identical to this last picture (same amplitude and frequency appears the same).


As the noise is the same at the power input and the output I'm going to conclude that the noise is purely from the power supply.


N.B. I know that the frequencies that can be seen in the oscillograms are not audible, however this signal also seems to be the ripple on top of a different ac signal of much lower frequency. Working on a 1ms scale however made it easier to evaluate the amplitude.




I'm looking at building a power supply along these lines (as I said in the main thread):




(I'll be increasing the main capacitance though as I'm drawing more than the original design).


I'll be doing a few tina simulations over the next few days before I buy anything though, unless anybody has a different supply recommendation.


Well at least we're fairly sure about the origins of the noise now,

cheers all,


post #26 of 33
Originally Posted by KimLaroux View Post

There's at least 4mm of aluminum between the NFB's transformer and the wiring inside the SSMH.


...metal can...


...The next thing to do will be to place sheet metal all around the AC wires, and ground it.


drugs r bad, m'kay? you shouldn't do drugs, m'kay? cuz if u do, u r bad, m'kay? 


all metals r equal, but some r more equal than others, m'kay?


Recall that there is a fundamental difference between metals. All are conductors, but some are magnetic (ferromagnetic), like iron, steel, cobalt and nickel, others, like aluminium and copper are not.


Consequently you may find a copper electrostatic screen in a transformer to prevent capacitive coupling between windings, but the magnetic fields in the transformer pass through it virtually unaffected. If you want to have an impact on magnetically induced hum then virtually the only economic screen is a steel case. Otherwise you have to adopt other methods to counter hum, or other magnetically induced (nearfield) noise.


A conductive screen will impact propagating (EM) radiation, but this is different from nearfield.



post #27 of 33
Originally Posted by wakibaki View Post


drugs r bad, m'kay? you shouldn't do drugs, m'kay? cuz if u do, u r bad, m'kay? 


all metals r equal, but some r more equal than others, m'kay? </snip>



I'm not sure it's helpful in this instance to use that tone just because he may not have recognized the difference betwen aluminum and steel when it comes to transformers.  Steel doesn't solve everything and aluminum is not exactly inert when it comes transformer interference.

Edited by tomb - 12/7/12 at 8:54pm
post #28 of 33

Yes I know there's a difference between metals... I'm studying in natural sciences. By sheet metal I meant a ferrous metal.


I may be stupid, but I don't get the joke with the drugs and the silly tone. Lost in translation, perhaps. 


Thanks for the measurements Goobley. I've been wanting to measure mine for months now but still hasn't got the chance, so your measurements come in handy. What confuses me is that the signal out of the tube has less noise than the signal out of the MOSFET. I thought the tubes were most sensitive to noise, as they are the voltage gain stage and so would amplify the noise present on the power rail. The only explication I can see is that R13, C6 and R1/7 together present a much higher PSRR than the MOSFET connected directly to the power rail. Higher than the gain of the tube, even. Anyone can confirm this? Or am I speaking nonsense again?


If it's the case, then it'll open new doors to modify my circuit. So far I reworked just about everything except the MOSFET wiring.

post #29 of 33
Thread Starter 

My answer to the noise out of the tube being lower is as follows, I of course have no science to back this up with. Your CRC filter could well be the correct answer, I have a feeling that it may be a mix of both.


The tubes are voltage gain devices as you say, during this experiment they were amplifying no signal to speak of, however the tubes themselves aren't amplifying PSU noise, they're amplifying what comes in from the grid, which is almost nil. Admittedly this would then pick up the power supply noise again on the plate (from memory the plate load was quite noisy). It seems to me that the C2 may actually be removing a lot of the noise (I'll poke in this area with my scope). So what goes to the gate is therefore fairly clean, not great by any stretch of the imagination, but better than the output. When the gate opens the mosfet the full quantity of dirty power goes through and then contaminates the signal path. This certainly seems to be what I was observing, with very little differences between the two tubes of different mu.


I'll do some poking around and I'm fairly confident that we should be able to get somewhere now.

post #30 of 33

Told you so. Maybe I just thought it real loud...


Try to get some Mu-metal foil to isolate the PS from the rest of the circuit. McMaster Carr sells 4" wide strips by the foot (or is it 6" wide?) - Its expensive, but nothing is better for shielding.


A linear regulator between the PS and the audio stage may help. This is kind of a misnomer - as you learn more you will see that in simple amps with low PSRR the PS is as much a part of the audio stage as what most people call the audio stage! But the MHSS is inexpensive so we do what we can. 


Depending how the noise is "moving" around the amp (if its purely on the power lines, or if it is also being radiated through the air) you could absolutely have more noise at the output of the mosfet than at the output of the tube before it. There is also the fact that big power Mosfets are not generally as electronically quiet as BJTs or smaller fets and have some random noise contribution of their own. 

New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › YANSSMH (Yet Another Not So Starving Student Millett Hybrid)