• 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.

TentLabs Directly Heated Tubes Supply

Rod Coleman

Member
2004-07-25 6:55 pm
UK
Re: DIY current heating

Rod Coleman said:
Of course there's a DIY option!

http://www.diyaudio.com/forums/showthread.php?s=&threadid=38248&perpage=30&highlight=&pagenumber=2

This circuit is still in constant use - with same 300Bs, and the filament voltage is still 4,9V, nearly 4 years later.

Piotr compared it to other filament circuits in a linestage, and liked it best.

If you try this current source, or maybe use the Tentlabs, you will not go back... AC heating sounds REALLY bad in comparison.


Yup, it is very, very good. Thanks Rod for sharing!
 

Rod Coleman

Member
2004-07-25 6:55 pm
UK
GM70 design notes

GM70 is 20V 3A, so that will need a very BIG power supply!

start with Trafo 24V, with very high VA (power) capability, maybe 500VA. Maybe use a 24V control panel Trafo (get these used from salvage company or industrial seller).

Use 100V 10A schottky diodes ( try the ONSEMI.com site for their sample program, get 10 different diodes, 25pcs each for $15 US)

then add high quality capacitors 10000 or 15000uF/35V, like Panasonic TSUP series. Parallel capacitors a good idea - make sure ripple current handling is MINIMUM 8 Amperes.

This supply should give you about 27V with 2V peak-to-peak ripple. Use powersupply Designer II to check that the parts you buy will give around these figures: 26V to 30V is needed. The dc output you get is very strongly dependent on the trafo resistances, so measure these (pri and sec) and put these, and the open-voltage into PSUDII. If you get too much voltage, use a wirewound resistor to drop it to 28V.

Set R3= 3,3K and R2= 27K or 24K to give about 2 to 5V across the top transistor.

Each power transistor radiating 15W, so a high power transistor is needed, look for TO-220 packaged parts with at least 60W handling. The BDX53C is a darlington with Cob (typically) 100pF at 5V Vce. Use an isolating sheet and a very large heatsink (maybe 1 deg C/W)

test the circuit for 1 hour using 4 parallel 24 ohm (50 Watt) wirewound resistors as a load - don't connect to your high-priced triode until you are sure the transistors/heatsink can handle the heat.

Next, adjust the current sense resistor. Use a 0,22ohm TO-220 packaged part, and add 3 or 4 43 ohm leaded resistors in parallel. when the circuit is hot, cut the 43 ohms out, 1 by 1, until current is exact.

good luck in building!
 

Rod Coleman

Member
2004-07-25 6:55 pm
UK
So the voltage across Q1 isnt that critical(I mean for other filiamentvolatges (2-7volts)can R2 or R3 be made adjustable?

2V across transistor gives 6W dissipation in that transistor.

7V gives 21W disspation in the transistor when the filament current is 3A (for GM70), so this means you need a bigger heatsink. But capacitance Cob is smaller at 7V and this gives better high-frequency noise reduction.

So the value is not so important, but needs to be at least 2V, so that the input voltage is isolated from the filament properly.

At 7V, a very big heatsink is needed. So best way is to build the supply and measure the voltage - then reduce to 2 to 3V using a wirewound resistor.
 

Rod Coleman

Member
2004-07-25 6:55 pm
UK
What do you think of a switched supply? I am thinking about a 24VDC one adjusted for 26VDC

The 230V powered switching units that I have tested all have a leakage current from the 230V input (line) to earth. In a DHT amplifier, we usually connect the filament terminals to earth through the cathode resistor.

Now, the leakage current will flow through the cathode resistor and cause biassing errors and noise - so that would make the design very difficult. If you have some switching units, put a current meter between the output and 230V earth (start on a high range!) and measure. More than 100uA will be too high, I think.

Beware also of high frequency noise. At least, use a 4A common-mode toroidal choke (get them out of old PC power supply units!) and feedthrough capacitors to filter the noise. The switcher needs to be in a metal box, with the feedthrough caps mounted in a hole in the box. Murata BNX series filter units are also very effective. Ordinary capacitors have very little effect above 10MHz, and the switcher emits noise up to 300MHz or more...
 
start with Trafo 24V, with very high VA (power) capability, maybe 500VA. Maybe use a 24V control panel Trafo (get these used from salvage company or industrial seller).

500VA for filaments!? I suppose if surplus is cheap and chokes arent.

For a pair of GM70s I would use a 160VA+ 28VAC transformer and a choke input. Hammond 159ZL 2.5mH 10A or similar. An RC section or transistor regulator dropping remaining voltage to 20VDC. The main caveat here is dealing with magnetic leakage. Less dissipation, quieter.

Going regulated be sure there is enough headroom for dropout, 30VAC may be needed with weak/unsteady mains and/or lower performance PTs.
 
design philosophy

For a pair of GM70s I would use a 160VA+ 28VAC transformer

We are trying to build a supply for 20V 3A, with low enough ripple to use passive regulation (gyrator). Using a 24V trafo and 15000uF (as recommended) causes an rms current of around 8,6A - 206VA!!

Using a 2,5mH choke and 28V rms reduces the current to 6A rms or 168VA, which is an improvement if the chokes, as you say, are cheaper, but 160VA would not do. Needs to become 250VA, at least. Still, if you can get a choke like this (it will need at least 7A handling) it is a good idea, for sure.

The reason for the 500VA recommendation is the extremely low source impedance to the rectifier stage, so that the reverse recovery currents are suppressed, the trafos runs cool, and reliability is good.

High Power 24V trafos are common in industrial plant and are easier to find as used parts, compared to other voltages.

An RC section or transistor regulator dropping remaining voltage to 20VDC

The big reason for using a Tentlabs or this DIY alternative is to is the use of a current source to regulate the filament current. The gyrator is not a regulator! The combined effect of the current source on one terminal, and the gyrator on the other, is to steeply increase the impedance looking OUT of the filament!
This is the magic ingredient that makes the sound improve. There are some references to this curious property of DHTs on the Asylum, and other places.

Regulator ICs and bandgap or surface-zener references are much too noisy for this application, and regulator feedback loops reduce dynamic impedance - very undesirable.

If you build this circuit and compare the sound of it to voltage regulation, or RC/LC filtering, you will find the difference is really amazing. Don't take my word for it though, please build!
 
You only have to derate a transformer 10% when using choke input. 200va is plenty. If going strictly passive regulation is better with choke input than capacitor. Peak diode current is vastly lower with choke input than capacitor. That said I was thinking of toroidal PT when I made that suggestion, more might be wanted with an EI.

Some PTs are speced at a much lower temperature rise for a given VA than others.

Nothing keeps you from setting up the regulator as a current regulator if desired.

Unless Im grossly misunderstanding, isnt peak current of a 3A choke input supply, well, 3A?

A passive method for high dynamic impedance is an LCL filter.

159ZL can be had for $22 in the US.
 

Rod Coleman

Member
2004-07-25 6:55 pm
UK
Unless Im grossly misunderstanding, isnt peak current of a 3A choke input supply, well, 3A?

The peak current is equal to the dc current only when the inductance is very high - of the order of 5 Henries for 50/60Hz supply frequency (and practical reservoir values). Below that value, the filter is acting somewhere between a choke-input and a capacitor input filter - and the peak and rms current values increase significantly.

Rather than go through the analysis, try putting the values in Duncan Amplification's Power Supply Designer II (PSUDII) - you can see how the peak and rms current varies with choke and capacitor values. For instance, with 2,5mH you need 15000uF to get 1V ripple, and this involves a peak current of 10,5A!

When you put these values in PSUDII another problem arises - the 2,5mH combined with an output capacitor of 3300 to 10000uF will form a resonant tank circuit near to the supply frequency (50/60Hz). The PSUDII output shows this as an instability of the output waveform. The 159ZL choke is not for use as a choke input filter, but rather as a part of an LC filter applied AFTER the reservoir capacitor. In this position it will work very well for reducing ripple.

A passive method for high dynamic impedance is an LCL filter.

Sure, you can use LCL for a passive filter - but the L values would need to be very large to reach the same impedance level as a transistor current source, and the cost very high. With Hammond, for instance the 3A rated choke in their catalogue is only rated at 28mH. Even if your low frequency cutoff is 20Hz, this is only 3 Ohms impedance, compared to more than 100K for a current source.
 
FWIW: I use a C-L-C-L-C, that is 22Volt AC in, schottky bridge, 10Ku - 60mH common mode - 10Ku - 60mH common mode - 470u. It is as quiet as can be on my 100dB speakers.
The CM-chokes are available from AE-Euope, wound at my specs. They have EI78 laminated core and have two sections of each 15mH. Not that big but effective.
I also use Tentlabs low noise version heater modules in my 10 SE Preamp: very quiet and a real space-saver: the lattter being the real benefit of this design...

Good luck,
Rob