• 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.

Power supply issues with Borbely style 12AU7/MOSFET hybrid

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Thanks for the comments and suggestions, Bear and M.Gregg - very useful.

looks to me that the amp shown has no AC drive on the lower mosfet. is it suposed to be a SE amp?

regardless it sounds like the mosfets are turned on almost fully if the rails are pulled way down.

so you need to measure the gate to source voltage, not to mention the current through the resistors.

the bias on the lower mosfet is developed by the zener that sits on the negative rail. First off you have to be sure you have the mosfets mounted in the proper orientation, the source is going to be on the rail side, not to the output. Is it? Is it the proper type? Or are they reversed (likely). BOTH polarity, N or P channel, AND physical pinout have to be correct!!

I spent quite some time worrying about this, and am confident that the MOSFETs are physically connected as they are shown in the schematic I posted. The schematic is based on the one in the AudioXpress article, and I did find a couple of errors on that schematic (e.g. NPN symbols used for the PNP devices in the current mirror). In that schematic the Source of Q1 is attached to the output, which as you point out would be incorrect (but see below).

The zener sets the bias directly so it has to be the proper voltage for the particular device you chose to use, it may be different than the 5.1v zener.

I've compared the datasheets for the device I'm using (IRFP140 -http://www.datasheetcatalog.org/datasheet/irf/irfp140.pdf) and the one used in the article (IRF540 - http://www.datasheetcatalog.org/datasheet/irf/irf540.pdf and they appear to me to be very close (although I am by no means an expert).

Since the upper mosfet's turn on voltage is set by the tubes via the adjustment in the tail of the LTP, the other half needs to be correct before it will balance for "0vdc" at the output.

You can test the lower mosfet alone, take the tube out, and disconnect the upper mosfet's resistor. Place a resistor to ground at the output... anything near 8 ohms will be ok. Now if the lower mosfet is biased up correctly (how many mA or A does it call for?) you should see the proper current through either resistor. D1 & R15 are a simple voltage divider.
The current through Q1 is supposed to be around 3A ... Q1 is referred to as a "constant current source" in the article by Cozza. Q1 is shown in that article as having the Drain connected to the rail. Perhaps this is correct after all - I am becoming more confused:confused:

A short quote from the article, in case it helps clarify the design: "The output stage is composed of one or more P-channel MOSFETs in a single ended Class A common source configuration similar to the Nelson Pass Zen amp. Its drain current is supplied by constant current source Q4 (Q1 in my schematic) which develops a 3A idle current using the specified value of R14 (R2 in my schematic)"

Btw, this is essentially similar to the "It Thinks It's A Triode" SE Class A DC Coupled amp I put out in the mid 1990s... except I used a current mirror for the lower mosfet and a pair of Jfets for the input. The trick is to wrap a DC servo back to the adjustment transistor in the LTP tail! Of course you have to set up the amp for static dissipation before engaging the DC servo... I biased mine using 3 Hitachi Mosfets per rail in the output at 3.5 amps and only ~30vdc on the rails. This toasted out about 120watts of pure class A heat, and needed a heatsink with efficiency and FINS...

That sounds very interesting (and hot!) - do you have a link to the schematic for it, please?

Along those lines I don't understand how this amp can be biased properly with 45 volts on the rails?? And I don't understand how it will maintain DC balance - it MUST have an output capacitor as shown.

_-_-bear

Yes, I think it needs a hefty capacitor on the output otherwise it would damage the speakers I suppose. Needless to say I'm not yet at the stage where I've actually connected any speakers to the output ;)
 
Disabled Account
Joined 2010
I don't think for one minute that you have only 3 A through Q1

If this were the case the 25A supply would hold up and the fet would get hot!

It looks to me as if you either have "diode" conduction or the FETs are flat out..

What voltage do you have across the lamp?

What voltage do you have on V1 anode ref to Gnd

Do you have a link to the original circuit?

What I am saying is if you have the full voltage minus a few volts then none of the fets are controlling..and the voltage is just appearing across the lamp
IE its only the lamp stopping the supply being "shorted out"

I would expect to see a volt drop across each FET representing the supply voltage with controlled current flow..


Regards
M. Gregg
 
Last edited:
I don't think for one minute that you have only 3 A through Q1

If this were the case the 25A supply would hold up and the fet would get hot!

It looks to me as if you either have "diode" conduction or the FETs are flat out..

What voltage do you have across the lamp?

What voltage do you have on V1 anode ref to Gnd



Regards
M. Gregg

Quite correct - I don't have anywhere near 3A. More like a hundred mA with the lamp in series. With R2 replaced by a 50 Ohm resistor (which is equivalent to the lamp resistance), the power supply stabilizes at around 35-0-35. A 1 Volt 1kHz square wave applied to the input at R20 produces a 25 Volt swing on the output at Q2 drain with no load ... so the amplifier is amplifying :) There is some overshoot on the leading edges of the square wave, and the level of that distortion can be adjusted by changing the plate currents through the tube via R14.

It seems very close, but something is wrong, so no cigar :snail:
 
Disabled Account
Joined 2010
Quite correct - I don't have anywhere near 3A. More like a hundred mA with the lamp in series. With R2 replaced by a 50 Ohm resistor (which is equivalent to the lamp resistance), the power supply stabilizes at around 35-0-35. A 1 Volt 1kHz square wave applied to the input at R20 produces a 25 Volt swing on the output at Q2 drain with no load ... so the amplifier is amplifying :) There is some overshoot on the leading edges of the square wave, and the level of that distortion can be adjusted by changing the plate currents through the tube via R14.

It seems very close, but something is wrong, so no cigar :snail:

What voltage do you get across Q2 at idle in this situation?
 
Do you have a link to the original circuit?

What I am saying is if you have the full voltage minus a few volts then none of the fets are controlling..and the voltage is just appearing across the lamp
IE its only the lamp stopping the supply being "shorted out"



Regards
M. Gregg

There is an almost identical circuit here: http://digilander.libero.it/essentialaudio/hybrid_circuit.htm which I have also been using as reference material.

hybrid_schem.jpg


In fact, I suspect that the author is Cozza at this Web site, but I'm not sure, i.e. the AudioXpress design and this design are by the same person.

In the schematic shown at that link, Q4 is inverted from how it is shown in the original AudioXpress article, and is orientated the same as in my schematic:

6720847903_5724b897db_o.png
 
Q1 must be an N ch Mosfet.
The Source must be on the negative rail.

The other side, the + rail MUST have a P ch Mosfet.
Source to rail.

Then it will work.

Assuming you have the pins on the actual mosfets in the correct positions.

Once you have the lower mosfet correctly installed and it is not fried - THEN you can measure the Vdrop across the series resistor and use Ohms Law to determine the current. The supply, assuming it has sufficient VA (does it?) will drop only a few volts.

The bias voltage across the 0.5ohm resistor.
E = I * R so 3 * 1 = 3 then divide for the 1/2ohm resistor = 1.5.
1.5 volts across the resistor!
Ahhhh... so the difference between the Vdrop on the resistor which is ALSO the voltage on the Source and the bias voltage is 3.5 volts @ 3 amp bias!

Now the question is, will the mosfet bias on? Is that enough bias voltage, or too much... may be. Many of the IRF device want more... have to try it and see...

The transformer.
Let's see: P = I * E P = 3amps x 45 volts = 135 watts!
You need a transformer for stereo that will be > 300va minimum.

Which is why the segment of the article you posted called for "one or more" mosfets.

I suggest rewiring the transformers for half voltage - possible depending on how the primaries and secondaries are configured - if you are only going to use one mosfet, unless that mosfet has a really really high current rating, and you look at the SOA curve on the spec sheet and see how many amps they show it taking continuously at 45volts... or add more mosfets after you get ONE pair to bias up correctly. You will need LARGE heatsinks and very good thermal connection to them.

First, please follow my instructions and get the lower one to operate properly.

_-_-bear
 
Last edited:
Q1 must be an N ch Mosfet.
The Source must be on the negative rail.

The other side, the + rail MUST have a P ch Mosfet.
Source to rail.

Then it will work.

Assuming you have the pins on the actual mosfets in the correct positions.

Confirmed: Q1 is N-channel, Q2 is P-channel. Q1 source is to the -rail. Q2 source is to +rail. And the pins are connected correctly, as per the datasheet.

Once you have the lower mosfet correctly installed and it is not fried - THEN you can measure the Vdrop across the series resistor and use Ohms Law to determine the current. The supply, assuming it has sufficient VA (does it?) will drop only a few volts.

With the values shown in my schematic, the supply voltage drops to 10-0-10, and I haven't measured the current through R2 under those conditions. With R2 increased to around 50 Ohms, the current is around 100mA, and the supply settles to around 35-0-35.

It seems to me that, if I use the specified value of R2 of 0.5 Ohms, the current is too high, and the voltage supply droops.


The bias voltage across the 0.5ohm resistor.
E = I * R so 3 * 1 = 3 then divide for the 1/2ohm resistor = 1.5.
1.5 volts across the resistor!
Ahhhh... so the difference between the Vdrop on the resistor which is ALSO the voltage on the Source and the bias voltage is 3.5 volts @ 3 amp bias!

Now the question is, will the mosfet bias on? Is that enough bias voltage, or too much... may be. Many of the IRF device want more... have to try it and see...

If the current through R2 is high, that implies Q1 is biassed on, correct? Or am I missing something here?


The transformer.
Let's see: P = I * E P = 3amps x 45 volts = 135 watts!
You need a transformer for stereo that will be > 300va minimum.

This is the major uncertainty in the components I'm using: I don't know what the specifications of the transformer are! It is a large, heavy unit that came out of an old piece of tube-based HP test gear. It has a primary and a centre tapped secondary - no other windings. It's made by PACO for HP, and although I have the model number, I can't find the spec. online anywhere. Perhaps it is not big enough for the job after all, which would be a blow.

Which is why the segment of the article you posted called for "one or more" mosfets.

My intention was to get the circuit running as shown, then parallel another pair of MOSFETs to get the power up.

I suggest rewiring the transformers for half voltage - possible depending on how the primaries and secondaries are configured - if you are only going to use one mosfet, unless that mosfet has a really really high current rating, and you look at the SOA curve on the spec sheet and see how many amps they show it taking continuously at 45volts... or add more mosfets after you get ONE pair to bias up correctly. You will need LARGE heatsinks and very good thermal connection to them.

The P-channel MOSFET I'm using is rated at 18 Amps and a max power dissipation of 125 Watts. Both MOSFETs are clamped to the metal chassis (but electrically insulated from it).

The SOA for the P-channel FET:

An externally hosted image should be here but it was not working when we last tested it.


and for the N-channel:

An externally hosted image should be here but it was not working when we last tested it.




First, please follow my instructions and get the lower one to operate properly.

_-_-bear

Thanks so much for helping with this - I am learning as I go along, as you can perhaps tell :eek:

I am now worried that the transformer may not be up to the task - but if this were the case, wouldn't it simply get hot (it doesn't) and eventually burn out?
 
I made some measurements of when Q1 is turning on. I disconnected Q1's gate resistor, and I put a 50k pot between the -ve rail and the ground rail, with the pot's wiper connected to Q1's gate.

With the wiper at the potential of the -ve rail, Q1 was off, and current through Q1 source to drain was zero. Power supply was at 45-0-45 in this condition. Then I increased the gate voltage: here are the results:

Q1 Gate Volts Q1 Ids Power Supply
0 to 3.40 0 45-0-45
3.5 30mA 35-0-35
4.0 200mA 7-0-7
>4 ~200mA 5-0-5



This looks like Q1 is turning on at a bias of around 3.5V, so the 5.1V of the Zener ensures that it is always on.

The problem appears to be that the power supply voltage drops drastically as soon as there is any appreciable current through Q1.

I'd welcome any suggestions as to what may be wrong with the power supply.
 
Last edited:
Ok this is simple.

You need the low value for the series resistor in order to be able to drive a speaker.

Let's test the transformer and power supply.

What you need to do is to remove the power supply rails from the amplifier.
Now you will put a dummy load resistor between either rail and ground.

You can start with something like 100 ohms or 50 ohms. At minimum 5 watts (but at that wattage you only have a few seconds to take the measurement!!)
Measure the voltage across the resistor.
That tells you the current.

Next reduce that resistance down, incrementally and you SHOULD be able to put an 8 ohm 100watt
resistor across the rail to ground and only see a few volts of drop!!

A 50ohm high wattage resistor is ideal if you have a pair.
First in series = 100 ohms
then one alone = 50 ohms
then two in parallel = 25 ohms

Any two resistors of similar value can do the same thing,the actual value is not important, just the scale ranging down.

From your description of the transformer, it may not be any where near large enough.

Post an image, it will be obvious I would think.

_-_- bear

PS. no the voltage will sag, and yes it will get hot, but in this case it is not even sufficient to run the amp at all, or so it seems...
 
Last edited:
OK Bear, I think you nailed it ... I put a 100 Ohm 10 Watt resistor across both rails and the voltage dropped from 90VDC to 18VDC immediately :mad:

For reference here is photo of the amp with transformer in question mounted on the top at the back.

An externally hosted image should be here but it was not working when we last tested it.


I have quite a collection of old power transformers, so maybe I can find something suitable as a replacement. On the other hand I have a feeling that these older items, being from tube test gear, have secondaries that have HV windings only capable of running a few hundred milliAmps, whereas what I need is Amps.
 
there are a number of suppliers of toroidal transformers. Online. Ebay included.

You will want to buy one that is around 400VA minimum.

You will also need to buy some honking BIG heatsinks!
If you got those mosfets biased up, it would turn the paint brown and the mosfets would overheat quickly and blow up.

Look at some of the F5 builds, and the heatsinks used there. You need at LEAST that much with this design...

So, to get a handle on it, I suggest dropping the rail voltage DOWN to more like <24vdc per rail. This will drop your total dissipation in the outputs to a level where you MIGHT be able to handle it. The reason is that this is a Single-Ended Class A design, which means you get ~25% efficiency. That means that for 25watts of usable output to a speaker you use 75-100watts of HEAT. That needs a BIG heatsink!!

Something like a 9"x7"x1.75" size... PER CHANNEL!! the 1.75" is the fin height.

Ok, so time to rethink the project!

It can be reconstructed with the proper parts as a decent Push-Pull class AB design. Still using the LTP as the input... of course that is a different amp, and still needs bigger xfmr and heatsinks...

This design with 45vdc on the rails I do not believe is terribly practical...

_-_-bear
 
Disabled Account
Joined 2010
Agreed,

Its a shame.. then again could you mount a toroidal Tx under the chassis or double up one under one on top..and a V shaped heat sink maybe..
The blue did look good..:)

You could expand down using a wooden base etc..mount heat sink on the back...

Regards
M. Gregg
 
Last edited:
Agreed,

Its a shame.. then again could you mount a toroidal Tx under the chassis or double up one under one on top..and a V shaped heat sink maybe..
The blue did look good..:)

You could expand down using a wooden base etc..mount heat sink on the back...

Regards
M. Gregg

Thanks for the encouragement - I admit to having been despondent about the project, but am now determined to pursue it, even if it means adding monster heatsinks!

Looking in my parts bin the only transformer I have that is remotely suitable is a humungous 750VA unit with a 120V primary and 450V secondary, without a centre tap.

(You may wonder why I don't just go and buy a suitable transformer - but the idea of this project was to use what I had to hand, rather than e.g. simply buying a kit.)

I'm wondering if I can use this transformer reversed, with an artificial centre tap. That would give me 120x120/450 = 32 Volts AC on the secondary or around 45 VDC rectified. Thus about 22 VDC on each rail. I suppose I could also use a high current diode based voltage doubler to double that if I wanted to, but Bear suggested around 20 VDC per rail would be more reasonable.

Comments?
 
Disabled Account
Joined 2010
You could try it…as long as it’s an isolating Tx not an auto Tx..
In your mind scale it down and use it as a test piece that might become what you are looking for with extra work or you will fail..
Run up the Tx not connected and measure your voltages..
Please think low power to start with..forget higher voltages.. and make sure you have the voltage rating on the supply caps for your Tx.
Don’t be complacent with this project you have possibly of high amps it is dangerous.
Heat will be an issue also so my thoughts are you can run it up slowly using the control on the FETs that you now have..and watch the heat ( I think it will be a surprise how hot it is) that is generated bring it up slowly..
The pot is not a permanent fitment…once you have it going correctly use fixed parts to ensure safety..
Don’t forget bleeder resistors across the supply caps for discharge.. The resistors should give a balanced voltage across the caps and CP. I would make sure You have fusing in the rails..
The dissipation in the FETS will be high..

Regards
M. Gregg
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.