[Goobley]

I wondered if any of you clever people would be able to explain some stuff regarding the mosfets to me.
 
I am aware that the mosfets are operating as source followers and that one biases them to be above their ohmic range. (With the heaters as one of the biasing resitors)
 
If I understand correctly that would mean that the mosfets would be permanantly fully on, with all the power flowing from drain to source and therefore into the heaters. Now most of us are running 12 or 17v circuits for the tubes and it's pretty clear to anybody that 12 =! 48. Does that mean that the heaters (which are more or less lightbulb filaments as far as I'm aware) are purely current dependant? Do they just need their 12V*0.15A=1.8W? or does the mosfet only provide them with the appropriate power/voltage? If so how does this work? (I've been unable to find much information on this subject despite a fair amount of googling, it's always too simple or too hard for me to understand.)
 
Another part of my question is once again related to the source follower nautre of the fets. We apply plate voltage + amplified music ac to the fet gate (high impedance therefore low current) and we get a low impedance, high current output from the source. The question is probably the same as the last, how come we just have a fully open transistor but instead get a buffered version of the original?
 
If somebody could explain this topology or could point me to some decent reference material on the subject I'd be very grateful,
Cheers,
Chris

 

[the_equalizer]

No, the MOSFETs are not fully on, they are 'just enough on', that's how the source follower works, as I mentioned in a previous post.
 
The source resistor, in our case the tube heater, effectively acts like a current sensor, that provides 100% negative feedback to the circuit, self biasing the source in relation to the gate to a point where just the exact amount of current for the particular gate to source voltage to be self-sustained.
 
As I said in my post, if the gate rises, more current goes through, so the voltage drop across the source resistor rises, so the source voltage terminal rises so the gate to source voltage again decreases to just that point where the current required to sustain the gate-to-source voltage is let through.
 
The reverse happens when the gate voltage decreases, the gate-to-source voltage decreases, less current goes through, the voltage drop across the source resistor falls, so the source voltage falls. Again to the point where just that current required to sustain it is let through. It can't let more current through because then the voltage drop across the source resistor would rise, thus diminishing the gate-to-source voltage, which will turn off the MOSFET, which shuts down the current, then the voltage drop falls, gate-to-source rises, the MOSFET turns on, ad infinitum.
 
It's a self-sustained bias, a negative feedback bias.
 
So you can see that the voltage at the source, tracks the voltage at the gate. And since you have a lot of current going through the MOSFET you can sustain that voltage across low impedance loads too. That's why the signal at the source terminal is a copy of the signal at the gate. That's what a source follower does.
 
The negative feedback effect created by the source resistor in relation to the gate also helps explain in part the low output impedance and high-linearity of the source follower.
 
As Kim Laroux mentioned in a previous post, the tube heater is a resistor that, when you put 12.6 volts across it, will have 150ma going through it, or viceversa: push 150ma through it and you'll have 12.6 volts across it. All this once it reaches it's operating temperature.
 
So knowing this, and the fact that the MOSFET gate-to-source turn on voltage is ~4 volts, you set the gate biasing voltage divider so that when idle, the gate sits at about 4 volts more than the tube heater voltage. Then the rest is done by the self-biasing effect I described above and in my previous post.
 
So it's not the MOSFET per se that provides just exactly what the heater requires, it's the combination of the heater and the MOSFET in this negative feedback arrangement, the source follower topology, that makes it all work.
 
cheers!

 

[KimLaroux]

the_equalizer explained all that on the last page.
 
You can't really "bias" a MOSFET to operate outside of it's ohmic region. This is dependent on the drain-to-source voltage and the drain current. In the MSSH, Vds is ~35 V and Id is ~150 mA. This puts the MOSFET far into it's saturation region.
 
To understand how source follower works, you have to understand a couple of things:
 

• MOSFET are controlled using voltage, and no current flows trough the gate.
• By design, they are turned on by applying a positive voltage between the gate and the source, Vgs. The voltage needed to fully turn it on, Vgs(th), is around 4 V in this circuit.*
• In the MSSH, the gate is biased at a fixed voltage using a voltage divider between the power rails. This means the gate voltage always stays fixed, regardless of how much current goes trough the MOSFET.*
• The heaters will use as much voltage and, consequently, current as is available to them. This means they will use 10 V if that's all they've got, but if you give them 48 V , they will burn 48 V. They are rated 0.15 A at 13 V, so they'll use 0.3 A at 26 V.

 
When we switch on the MSSH, there's 0 V across the heater, the source of the MOSFET sits at 0 V, the drain at 48 V* and the gate at, say, 17 V*. Since there's 17 V between the gate and the source of the MOSFET, it turns fully on and conduct as much current as the PSU can supply (#1, 2). The heater starts pulling power, so the voltage drop across it rises (#4). This pushes the source voltage of the MOSFET up. Now if the MOSFET was only a relay, the heater voltage would simply rise until it hit 48 V... or until the heater blows.
 
But if you remember #2 and #3, it's impossible for the source to rise all the way up to the rail. The gate is biased at 17 V, and stays there. The MOSFET will turn off if Vgs goes bellow Vgs(th). As the heater voltage rises, Vgs falls. When Vgs approaches Vgs(th), the MOSFET starts to turn off, and chokes the heater. An equilibrium state is reached, where Vgs = Vgs(th). If your gate sits at 17 V, the source will never rise over 13 V or so.
 
Since we use a tube heater as a source follower, we're limited in the source voltage and current.
 
* For the sake of simplicity, I disregarded audio signal, voltage rise at power on and that Vgs(th) depends on Id...
 
Edit: Aahh.. You posted while I was typing this. 

 

[the_equalizer]

Quote:

<snip>  
Edit: Aahh.. You posted while I was typing this. 

 

 
So Kim, are you planning to put the CCS plate loads into your amp?

 

[KimLaroux]

Yes, but I don't have the transistors in stock. I don't feel like placing an order at Digi-Key and pay 8$ of shipping for 2$ worth of parts. I'll just wait a couple of months.
 
Last week end, I had another round of modifications:
 

1. Changed the pot for an 50K RK27.
2. Removed R16/R17.
3. Changed the RCA for better ones, as the cheap ones I had were an insult to the word "RCA". One had the outside too large to plug a connector in, and the other had the inside too loose to make a proper connection.
4. Moved the signal ground of the RCA to the star ground in the middle of the amp.
5. Rebuilt the whole PSU. Moved the output cap away from AC. Rewired in balanced fashion, i.e. no star ground in the PSU.
6. Twisted the wires from the AC in to the transformer.
7. Removed cathode bypass caps.
8. Added 47K resistor to R2/R8. I originally used 330K as this was all I had in stock. With the latest information, I realized it was useless to have the MOSFET biased so high, so I lowered it. Which makes me think, I should add 390K resistors to my to-buy list...

 
 

 
 
I tried it using the 12AX7 as I didn't dare turning it on for the first time with my new German NOS tubes. And was it a good idea! There was a cold joint in one pull-down resistor, so the corresponding MOSFET went fully ON. Did you know current production Sovtek tubes can take 40 volts or so on their heaters without flinching? hahaha! The tube glowed purple, it was beautiful. It didn't seem to care much, and it's still working perfectly.

 
I haven't tried it yet with the new tubes, so I can't comment if the modifications improved the SNR. I tried it with the 12AX7, and noticed that if my NFB-12 has it's volume over 50%, the noise out of the DAC drowns the noise from the MSSH itself. But these were 12AX7... I'll test again with the 12AU7.
 
I also got some real tube shields, so I'll be trying these too. I'll report back if they are better or worse than my Faraday cages built out of speaker grill mesh. 

 

[the_equalizer]

Wow... impressive list of work you did there! I'll keep my fingers crossed that all that work solves your noise woes.
 
cheers!

 

[Makiah S]

love how invovled you guys are... might build this one day... still intersted in buying Hand Built from Head-fi'r ;3

 

[the_equalizer]

Mshenay, please post your "want to buy" messages in the "sale/trade" forum.

 

[Goobley]

Thanks very much to you both, I'm aware that the equalizer had already explained this and I've read his post several times, however lacking certain pieces of knowledge, I couldn't quite put the whole puzzle together. However thanks to these newer explanations I actually really understand what you're on about and have been able to read those previous posts with new eyes. The comment on tomb's site by Dsatavisk about only needing to change one resistor to change to compatible tubes also makes sense now. These last few posts have been extremely helpful and will help me think whilst I go around diagnosing once again this weekend. I have to say that I'm certainly looking forward to a university setting where I'm going to be able to learn things like this from the people around me
 
I should be getting hold of a friends linear regulated bench supply this weekend that is as clean as a whistle and should help my find where to look for noise.
 
Thanks again for this great help, it's certainly hard to get one's head around when I've only used a transistor as a switch before,
Cheers,
Chris

 

[Goobley]

Edit: Deleted as I was talking rubbish and worked it out for myself

 

[Makiah S]

Ahh, have a friend who is looking to help me build it, so I'm STAYING. I'm looking at building one and my questions are, for around $30 what upgrads would you suggest, and as far as changeing the sound, I'd like to have deep detailed bass [+12db would b nice] and transparent highs, and ofc the Tubey Lush mids! 

 

[Goobley]

Quote:

Ahh, have a friend who is looking to help me build it, so I'm STAYING. I'm looking at building one and my questions are, for around $30 what upgrads would you suggest, and as far as changeing the sound, I'd like to have deep detailed bass [+12db would b nice] and transparent highs, and ofc the Tubey Lush mids! 

Around $30 over what price? This amp has evolved considerably since Mr Millett created the original design, I doubt the tubes you're going to be using will cost you $2 either... Of course I think it's a good project to do, and the standardish build would be this:

However depending on the tubes you use, you may have the change the voltage divider values, this divider will work for the standard 12v heater tubes (12A_7). With regards to any upgrade beyond this, I'd suggest you get a decent potentiometer (ALPS etc.), and make sure it looks pretty (if that's your sorta thing), this amp works very well as is. As has been discussed a little recently the caps C7/C8 can sometimes be a bit troublesome even if they help give better bass (I removed them from my build today and believe it improved a bit). I'd get these caps and then try with and without them to see how it works in your case.
 
I find this amp to have a nice rather classic valve sound with a bit of mosfet/SS sheen/polish/sparkle, I like it a lot, I find tubes have fairly emphasised bass anyway (may just be the circuits I'm used to), but if you want a full 12dB you're either gonna have to build some filters or use an eq on your source. (IMHO it shouldn't be the amp that makes such a large difference to sound, subtleties yes, bass control etc. (not emphasis), but over +/- 3dB I'd expect to have to eq).
 
Hope this helps, I'm sure others will have different suggestions for you too.
 
 
 
On a different note, does anyone see any reason why a power supply like the one below wouldn't work?

 
(Ignore the microphone stuff obviously)
I'd be planning on significantly increasing the capacitance and using a 317HV instead but it looks pretty solid to me (nothing revolutionary obviously). It looks like a fairly decent circuit particularly as there's a place not too far from me that'll custom wind me a 2*20v transformer, which would be almost spot on dropout for the 317HV.
 
Cheers
Chris

 

[scootsit]

Quote:

Quote:

Ahh, have a friend who is looking to help me build it, so I'm STAYING. I'm looking at building one and my questions are, for around $30 what upgrads would you suggest, and as far as changeing the sound, I'd like to have deep detailed bass [+12db would b nice] and transparent highs, and ofc the Tubey Lush mids! 

Around $30 over what price? This amp has evolved considerably since Mr Millett created the original design, I doubt the tubes you're going to be using will cost you $2 either... Of course I think it's a good project to do, and the standardish build would be this:

However depending on the tubes you use, you may have the change the voltage divider values, this divider will work for the standard 12v heater tubes (12A_7). With regards to any upgrade beyond this, I'd suggest you get a decent potentiometer (ALPS etc.), and make sure it looks pretty (if that's your sorta thing), this amp works very well as is. As has been discussed a little recently the caps C7/C8 can sometimes be a bit troublesome even if they help give better bass (I removed them from my build today and believe it improved a bit). I'd get these caps and then try with and without them to see how it works in your case.
 
I find this amp to have a nice rather classic valve sound with a bit of mosfet/SS sheen/polish/sparkle, I like it a lot, I find tubes have fairly emphasised bass anyway (may just be the circuits I'm used to), but if you want a full 12dB you're either gonna have to build some filters or use an eq on your source. (IMHO it shouldn't be the amp that makes such a large difference to sound, subtleties yes, bass control etc. (not emphasis), but over +/- 3dB I'd expect to have to eq).
 
Hope this helps, I'm sure others will have different suggestions for you too.
 
 
 
On a different note, does anyone see any reason why a power supply like the one below wouldn't work?

 
(Ignore the microphone stuff obviously)
I'd be planning on significantly increasing the capacitance and using a 317HV instead but it looks pretty solid to me (nothing revolutionary obviously). It looks like a fairly decent circuit particularly as there's a place not too far from me that'll custom wind me a 2*20v transformer, which would be almost spot on dropout for the 317HV.
 
Cheers
Chris

A millet hybrid can be built pretty inexpensively, too. I opted to go that route as the zillions of variations on the starving student gave me some pause.

 

[Makiah S]

Quote:

Around $30 over what price? This amp has evolved considerably since Mr Millett created the original design, I doubt the tubes you're going to be using will cost you $2 either... Of course I think it's a good project to do, and the standardish build would be this:

Ok so let's start from point A, the Basic Parts needed are what... were talking about building the Simplest version of this Amp following the very old and outdated compents. Now once we have established what BARE minumum is for this project [because I'm sure we can all agree on SOME varition of lowest quality and cheapest build on this amp] so again once we established the CHEAPEST way to build this, let's add +$30 to that budget to make some improvements!
 
I was told a few times that a basic Millet Hybrid cost around $100, so my question is... do I build a simple Millet hybrid for $100 or maybe $130 OR do I buy a Bravo v2 for $70 and Mod that with around $30 worth of parts and improve on it... because I've heard both options and people are telling me that at the end of the day... wheather you build a Millet amp for around $100 maybe more, OR yoy buy and mod a V2 for around $100 [or more] they both will have about the same Price to Performance Ratio... as far as the skillz in building and modding I have that taken of... just trying to decide the BEST option for a realitive Cheap Intro Hybrid Tube amp
 
Build a Basic Millet or Mod a Bravo V2

 

[scootsit]

I don't know what a Bravo is. My girlfriend uses a Millett daily and loves it. With the buf634 mounted directly to the chassis, it has gobs of current, and sounds pretty close to my mosfet max. I also customized it to her aesthetic. It sounds like you are asking for a decision that only you can make.

Here's the thread I made with some pictures: http://www.head-fi.org/t/598987/millett-hybrid