Millett "Starving Student" hybrid amp - Page 421

I never said the MOSFET were current controlled. I said the source current was dependent on the gate voltage when in saturation mode, which I believe is the case here.

I did not say I put 48v across the heaters. I said the MOSFET went fully ON, so I guessed it was like hooking the heaters directly to the power rails. I forgot about the drop needed for Vgs, and I don't know how much it was. But you're right the heater surely did not see as much voltage.

Quote:

Originally Posted by KimLaroux 

I never said the MOSFET were current controlled. I said the source current was dependent on the gate voltage when in saturation mode, which I believe is the case here.

 

Quote:

Originally Posted by KimLaroux 

The tube is not "holding" the source current at 150mA: The MOSFET is biased to allow only 150mA trough the tube.

This is CCS operation. In fact the whole paragraph I cut this out of makes no mention of the fact that its a source follower and works based on voltage controlled voltages. This paragraph only references voltage controlled currents as in a CCS. 

Sometimes looking at the problem from a different angle helps. sometimes it hurts. Where are you going looking at a voltage problem from a current perspective? 

Quote:

Originally Posted by KimLaroux 

I did not say I put 48v across the heaters. I said the MOSFET went fully ON, so I guessed it was like hooking the heaters directly to the power rails. I forgot about the drop needed for Vgs, and I don't know how much it was. But you're right the heater surely did not see as much voltage.

Quote:

Originally Posted by KimLaroux 

the voltage across the heater was close to 48V before it blew.

I actually agree that the heater probably saw fairly close to 48V before it blew. I disagree with your values of Vgs. 

Quote:

Originally Posted by the_equalizer 

 

 

Shamazo, your measurements are definitely strange since they're not lower but much higher Might it be that you reversed R2 and R4?

 

 

 

Shamazo: If you're seeing 19 volts at the heaters, what's the voltage at the MOSFET gates in your build?

 

 

 

cheers!

I'll measurethis soon and edit this post, knowing myself I probably did reverse them.

Quote:

Originally Posted by kchapdaily 

yesterday i plugged a set of gr06 into mine, and with the volume knob all the way down, i actually picked up what sounded like an fm frequency. some random song i had never heard before was playing quietly, and when i turned the pot up, it would go away. tried it with my he-4 and it wouldnt pick it up. guess it was the high sensitivity of the iems. has anyone ever heard something like this happen?

I had this when I forgot to wire in the two 2K R3/R9 resistors. They need to be soldered as close as possible to the mosfets.

@KimLaroux and nikongod: Gentlemen, thanks for your replies. After reading them several times carefuly the error of my ways dawned on me. And it was so simple and obvious! I had it in front of my eyes all this time but I refused to see it...  Of course the quiescient source voltage rises when you raise the gate bias. It is supposed to, the MOSFET is working as a source follower. Oh my! it's so simple... increased current through the load tied to the MOSFET source causes an increased voltage drop across it ... how could I be so blind!

@KimLaroux: As nikongod pointed out, decreasing the current set by the CCS plate loads will not really reduce the gain, it'll only change the operating point of the tube. A quick and simple way to sensibly reduce the gain of the tube stage is to remove the cathode bypass electrolytic capacitors. I removed them from my build a long time ago to make it more manageable.

About R13: I did not remove it. It is still there, but basically doing nothing but providing a place to hook the bypass cap and the gate bias voltage divider.

Great observation on the CCS transistors: I concentrated on Vceo and completely disregarded noise values.The 2N5087 is a great choice, it's the same part used in the Millet MiniMAX (which I've built) and it works perfectly

About measuring errors, it is indeed a matter of meter input impedance. Decent meters (e.g. Fluke) have an imput impedance in the order of 10 Megohms for voltage measurements, just like my old Kyoritsu FET VOM. You can read a bit more about the meter input impedance problem here and here.

cheers!

Quote:

Originally Posted by shamazo 

Well I got rid of all of the hum, I forgot to ground the pot.
Now onto the next issue, I only have sound in one channel, I'll do some more trouble shooting later but ATM, I have roughly equal voltages on both channels but at pin 4 the dead channel has 9 volts vs 13 of the working channel.
I tried swapping tubes and that made no difference, so any voltages or resistor values I should be checking?

Whoa! What happened to the 19 volts you had there before ? :)   Good to hear you zapped that hum.

I can think of some things that might be going wrong with that channel:

1.- The MOSFET gate bias voltage divider for that channel has something wrong and it's setting the gate at around 13 volts which causes the source to sit at 9V. Such voltage at the tube heater is maybey too low for the heater to raise the cathode's temperature into thermionic emission and that's why that channel is silent.

EDIT: I remember one builder plugged in a set of 17EW8 tubes (17 volt heater) into his build biased for 12A_7 (12 volt heater) and he reported it worked perfectly So I don't think that channel's silent because of the low heater voltage. Either there is a problem in the gate bias voltage divider and some other miswiring or the MOSFET is bad.

2.- That channel's MOSFET is (somehow) "blown" causing the source to sit at 9V regardless of correct gate bias voltage.

3.- The heater's connection to ground (pin 5 in your build) is 'faulty', offering a large resistance and causing a voltage drop across it, "robbing" 2 volts from the heater.

Good luck with your debugging!

Quote:

Originally Posted by nikongod 

 

The mosfets are configured as simple source followers. The mosfet sets the source voltage and output current is conditional on the load.  

They are not current-controlled devices (ccs's) and should not be thought of as such. Ever. 

 

Still thinking about the MOSFET source follower configuration I now see what nikongod's point was with the above statement. It is indeed the load resistance that limits the current in that series circuit because the voltage drop across it will raise the source voltage to a certain point of equilibrium with respect to the gate voltage. An interesting situation of dynamic equilibrium of Vgs set by the IR drop across the load.

So let's imagine how this works: At amp startup the heaters are cold, their resistance is low, so the voltage drop across them is very low. For the purpose of this thought experiment let's say it's 2 volts, hence the MOSFET source, tied to the heaters, sees this same voltage. Meanwhile the voltage drop across the voltage divider gate bias network is fixed (at 17 volts for the 12A_7 build) so all the time the MOSFETs gate voltage to ground sits at that level so Vgs is 17 V - 2 V= 15 V which is quite larger than the MOSFET turn on voltage and a large amount of current can pass through the MOSFET drain-source circuit. At amp startup this current is so big that it overloads the Cisco PSU causing it to shut down and restart one or more times.

The heaters start warming up slowly, their resistance increasing with temperature, so the current going through them causes an increased voltage drop which causes the MOSFET source voltage to rise accordingly thus reducing Vgs more and more, as the heater temperature and resistance grow, until it reaches the critical turn on value for the MOSFET.  At this point you can imagine that the IR drop across the load makes the MOSFET source voltage to still rise, Vgs goes under the turn-on value, the MOSFET shuts down, no more current flows, the IR drop through the load drops towards ground, hence Vgs goes over the threshold and lets current through, causing the IR drop across the load to rise, hence reducing Vgs to the critical turn-on value, and so on and so forth...  A very nice dynamic equilibrium at play here keeps the MOSFET source voltage *just* right at the turn on value under the MOSFET gate voltage at all times.  Beautiful!

From that it follows that a load with a larger resistance value would need less current to make the MOSFET source voltage reach the Vgs threshold voltage simply because the IR drop across it is larger. Obviously the reverse also works, a load with a lower resistance value would need more current to make the MOSFET source voltage reach the Vgs threshold voltage due to the IR drop across it being smaller.

I finally ran some simulations in TINA-TI, the first one with a load resistor of a value similar to the 12AU7 heaters (R = V/I ,  R = 12.6 V / 0.150A , R = 84 ohms), then one with half that value and a last one with twice that value. The results were just as expected:

So, we can conclude a couple of things from these thought and computer experiments

1. You can't bias the MOSFET hotter by modifying the gate bias voltage divider, you need to change the MOSFET source load for that.
2. It is the tube heater that's holding the current at 150ma, though not quite in the way I imagined   :)

cheers!

Wow thanks for the clear and informative post. Really clears things up. Source follower topology really confused me, as I'm used to use MOSFET as a switch to power electric motors and such.

That makes a lot of sens now. The gate voltage sets the limit, so that the source will never rise over Vg-Vgs(th). How clever!

The idea in this amp is to adjust the bias so that the source never goes higher than the heater voltage, or not too much higher. Raising the gate voltage will raise the source voltage, which in turn will make the heaters pull more current. So there's really no way to adjust Vgs, since it's a property of the MOSFET itself, and the heater will always "push" the source voltage until Vgs(th) starves it. If I got that right, then there's simply no reason to raise the gate voltage, as it does not give the MOSFET any more headroom. Sure it will allow more current for the output, but I think 150mA is enough for most headphones, right?

When I finally get the time to continue working on my amp, I think I'll just revert the bias to what it was so that the source sits around 15v. Damn, to think that I blew a nice Sylvania tube for no reason. 

Again, thanks the_equalizer, really appreciated. 

Yeah this is really great, it's helping me learn so much more than I expected from a simple amp project, I'm sure this info will be useful one day (having had no formal training).

Cheers all,

Quote:

Originally Posted by the_equalizer 

Whoa! What happened to the 19 volts you had there before ? :)   Good to hear you zapped that hum.

 

 

I can think of some things that might be going wrong with that channel:

 

1.- The MOSFET gate bias voltage divider for that channel has something wrong and it's setting the gate at around 13 volts which causes the source to sit at 9V. Such voltage at the tube heater is maybey too low for the heater to raise the cathode's temperature into thermionic emission and that's why that channel is silent.

 

EDIT: I remember one builder plugged in a set of 17EW8 tubes (17 volt heater) into his build biased for 12A_7 (12 volt heater) and he reported it worked perfectly So I don't think that channel's silent because of the low heater voltage. Either there is a problem in the gate bias voltage divider and some other miswiring or the MOSFET is bad.

 

2.- That channel's MOSFET is (somehow) "blown" causing the source to sit at 9V regardless of correct gate bias voltage.

 

3.- The heater's connection to ground (pin 5 in your build) is 'faulty', offering a large resistance and causing a voltage drop across it, "robbing" 2 volts from the heater.

 

Good luck with your debugging!

how can I check if my MOSFET is blown?

I checked the resistance and it was only about 6 ohms.

for the volatage divider would that be something wrong with r1/2 and c2?

EDIT I measured from the positive terminal of the power to pins 1 and 6, on the wokring channel its 37kohms and on the non working its 3 kohms, 

Quote:

Originally Posted by the_equalizer 


I don't know of a way of checking the MOSFET; but just like you seem to be doing, I'd first check the rest of the circuit thoroughly.
3 Kohms is waaay too low. If the value of R13 is correct ( 2 Kohms ) , that would leave only leave only 1 Kohms for the plate load resistor R1 / R7 which, as your other channel mesurements show, should be around 33 Kohms.

R13 is correct, so I am fairly sure I have a short on r7. I'll try and have a look at it and clean it up.

EDIT even if I swap rca inputs it is still only the right channel working, with leads me to believe i have shorted the channels together somewhere.

now I have 14 kohm resistance from positive in to pin 1/6 

both channels have similer volts at the source of the irf510, 

Quote:

Originally Posted by KimLaroux 

Wow thanks for the clear and informative post. Really clears things up. Source follower topology really confused me, as I'm used to use MOSFET as a switch to power electric motors and such.

 

That makes a lot of sens now. The gate voltage sets the limit, so that the source will never rise over Vg-Vgs(th). How clever!

 

The idea in this amp is to adjust the bias so that the source never goes higher than the heater voltage, or not too much higher. Raising the gate voltage will raise the source voltage, which in turn will make the heaters pull more current. So there's really no way to adjust Vgs, since it's a property of the MOSFET itself, and the heater will always "push" the source voltage until Vgs(th) starves it. If I got that right, then there's simply no reason to raise the gate voltage, as it does not give the MOSFET any more headroom. Sure it will allow more current for the output, but I think 150mA is enough for most headphones, right?

 

When I finally get the time to continue working on my amp, I think I'll just revert the bias to what it was so that the source sits around 15v. Damn, to think that I blew a nice Sylvania tube for no reason. 

 

Again, thanks the_equalizer, really appreciated. 

Quote:

Originally Posted by Goobley 

Yeah this is really great, it's helping me learn so much more than I expected from a simple amp project, I'm sure this info will be useful one day (having had no formal training).

 

Cheers all,

Thank you. I'm glad you found the post clear and informative. This piece of understanding eluded me for so long that I actually have a feeling of relief now. And to think it was so obvious right from the start. For me too, aside from the musical pleasure it has brought me, building it and tweaking the amp has taught me a lot.

I've kept thinking about the source-follower configuration and doing analysis and simulation trying to understand some other aspects of the output stage. For instance, KimLaroux mentions headroom. At first I also thought that the output's stage headroom depended directly and completely on Vgs. I thought that since the drain is sitting at 48V and the source at 13V then the gate voltage should be right in the middle of that range, that is  13 + (48-13)/2 = 30.5V, then that would give a headroom of 17.5V peak-to-peak and set the quiescent point of the MOSFET right in the middle of class A... Oh, how mistaken I was! I was trying to understand a source-follower (or common drain) as if it were a common source amplifier...

However, armed with what we now understand of the behavior of Vgs in the source follower configuration we can more clearly see the limits of voltage swing available to the output MOSFET.

On the negative swing side we have the situation where no voltage is dropped across the source load, 0 volts. All the supply voltage is dropped across the MOSFET. Since the quiescent (no signal) voltage drop across the load is ~13 volts it follows that the input signal at the gate can go to ~ -13 volts. Probably a little more since the source-follower's voltage gain is slightly less than 1.

On the positive side the theoretical limit of positive swing would be when Vg reaches the supply voltage; when it 'hits the positive rail'. Since the quiescent gate voltage is ~17 volts, it follows the maximum positive swing is ~31 volts. Probably somewhat less since as the source voltage rises following the gate voltage, the drain-source voltage decreases dramatically and it'll reach a region of non-linearity or saturation (given the Vds, the MOSFET cannot let more current through to support an increase in Vs, or support it linearly) before the gate gets to 48 V.

As you can see the headroom is assymetrical and if you read the literature on source-followers (or tube cathode followers for that matter) other biasing schemes are required to allow the MOSFET to swing symetrically; such as a negative supply connected to the source load, or a 'bootstrapping' resistor.

Please note that the above discussion assumes the ONLY load is the MOSFET source load, in the SSMH that is the tube heater. The moment you hook up a capacitor to take the AC and apply it to the headphones load, the available headroom alters quite significantly. I've run some simulations with 32 ohm resistive loads and the negative and positive swings are considerably diminished. I don't have the exact numbers handy but it was something like +15 to -5 V !   This is logical since, from the alternating current point of view, the load looks like 23 ohms (84 ohm and 32 ohm in parallel). As expected, higher impedance headphone loads have less noticeable effects.

Being able to swing 'only -5 V' with a 32 ohm headphone load is probably a bit of a moot point given that even Mr. Millett himself mentions that listening to a 3V RMS signal in a pair of Grados will cause permanent hearing damage. But sill, understanding the inner working and limitations of the SSMH output stage will definitely help in understanding how the amp will interact with difficult loads (like planar headphones) and/or to modify it to improve it's performance.

cheers!

Just to throw this out there again you guys, I'm currently looking at Intro Tube amps and I would LOVE to buy one of these from you guys OR be a tester [for a non explody one lol]  :D as I so do enjoy trying new gear but ^^ I'm happy to see the pics you guys have of them :3. Still though shoot me a pm if any of you have one you'd like to sell :3