Vintage mic power supply - filament regulation

[22busy]

Love this forum and I'm really impressed with, and a bit envious of, the knowledge you guys have.

I'm hoping someone can help me understand a bit about the workings of this vintage tube mic power supply (Telefunken Ela M250), particularly what's happening with the transistor regulation of the filament voltage. I'll try to attach a schematic below.

My transistor knowledge frankly sucks, and anyone willing to shed some light on this would be greatly appreciated.

Couple of questions:

What's going on with that transistor and zener? How is the regulation being accomplished?

Am I reading the schematic correctly that there should be a 9.5vdc drop across the transistor? How critical is this value?

Anybody know of a modern drop in substitution for an OD603 transistor?

thanks,
Bruce

http://www.mymastering.com/schematics/Elam%20251E%20schematic.pdf

 

[dale116dot7]

That voltage drop is 'nominal'. What has to be correct is the current. That needs to be adjusted to 175mA. This circuit is a constant current source. The PNP transistor with the zener diode sets the voltage between the base and the power supply filter capacitor. This also sets the voltage between the base of the transistor and the power supply filter capacitor. The resistance value between these two points is the current. See: http://en.wikipedia.org/wiki/Current_source Your circuit is just upside down when compared to the schematics given here. PNP replaces NPN, and the zener diode flips, and the input voltage is upside down too.

I would probably use a TIP127 darlington to replace whatever transistor is in there if it's bad.

 

[PRR]

> Am I reading the schematic correctly that there should be a 9.5vdc drop across the transistor?

That's what it says. As dale says: No, that is NOT what you want.

> What's going on with that transistor and zener?

Look further. There is a resistor. Actually a trim-pot. It is vital.

The drawing of the pot is funny to my eye. The intent is clear.

Transistor Base-Emitter junction is in series with ~~50R resistor.

Zener (biased through 1K) holds 6V across the transistor-resistor combination.

Transistor Base-Emitter voltage is uncertain, anything from 0.1V to 0.9V. However this is all small compared to 6V off the Zener; and we can trim the ~~50R resistor.

Taking a rough-crack: 5.5V across 50R is 5.5V/50= 0.11 Amperes.

Look at 6072 data-sheet. Heater sucks 0.15 Amps. Close. (Note that a full 6072 draws 6.3V 0.3A -or- 12.6V 0.15A, but this sick puppy only connects one-half of the twin-triode so is 6.3V 0.15A.)

Now go back to the 50R: it is trimmable. ASS-ume it is supposed to give 0.15 Amps. 5.5V/0.15A= 36.7 ohms, which is a very plausable setting for a 50R pot.

ah-HA! The thing makes 0.15 Amps, no matter what voltage is needed to make that happen (within limits).

You can do the exact same thing with a tube. Hold the grid up at +60V, put 10K cathode to common, hold plate high enough to stabilize, you get near-exactly 60V/10K= 6mA at any plate voltage.

> a modern drop in substitution for an OD603 transistor?

It is a non-fussy application. The practical path is to use any darn Silicon PNP jellybean. The Vbe will be ~~0.7V instead of ~~0.3V, but that 0.4V is small-change next to the 6V Zener, and you can trim.

The key criteria is Power. And the voltages are not adding up for me. We start from 24VAC, so expect up to 35VDC, less due to 20R resistor, sy 30V. But counting up from the 6.3V we have nominal 9.5V and another ~~5.5V or 21V.

Guess high. If the transistor dropped the full fall from 35V to 6.3V that is 29V. Load is 0.15A or 150mA, plus 6.3mA in 1K, 160mA or 0.16A. 29V*0.16A is 5 Watts! Taking the nominal "9.5V" we get 1.52 Watts.

So this sucker needs to be on a heat-sink. So any Silicon PNP "power" transistor. (A Darlington would work, but its ~~1.3V Vbe might need futher re-thinking, the one old part worked, a single transistor "should" be fine.)

If the Zener and pot are sick too, there are 3-pin power ICs which can be programmed as constant-current sources. Do the math for 0.1563 Amps, use 5% resistors, it will be close-enough to make the 6072 happy, and no unreliable trimmer.

Or, with modern cheap caps and 24VAC to start, I might just C-R-C-R-C-R-C filter 35VDC down to 6.3V in 0.15A. We "need" low-low-low ripple because the heater power runs next to the mike-output wires in the cable. But 3 of 60R and 4 of 2,200uFd will put the ripple out of sight. They could have done that originally. I wonder if they were told "this is 1960, put a transistor in already!"

 

[22busy]

Wow, now I feel not very smart, to say the least.

Gotta take some time to study your responses so I can make some sense of them. They're a bit over my head but I really appreciate all the help.

Thanks a bunch!!!!

 

[CJ]

You could rip all that out and build a new power supply.
Maybe connect the tube for 12.6 volts so the heater is balanced.
This would eat some heat also.
The 3 watt dropping resistor could come down.

Maybe use a single rectifier into some huge caps, no regulation at all.
This would bring down the DC Voltage a bit, hopefully, which means less heat.
However, there is dc on the transformer, which means do a heat check after a while.


Use a pi filter, a big cap, a resistor or choke, and another big cap.

A choke would have to be matched to drop the right amount of voltage.

So use a resistor.
Very dependable.

Silicon regulates good, but the chips do not like spikes for some reason, go figure.

 

[22busy]

Very good suggestions. Thanks again. I'm sure I'll rely on many of them when I build the new mic and PS.

For now though, I need to understand the basics of how to calibrate the existing system. The transistor works but the heater voltage is a bit low (5.9vdc) and I'm not sure if the current is correct.

Here's a photo of the variable 50R resistor. I've seen this general type of resistor before, but never one that had 2 simultaneous adjustments. So, I'm a little in the dark about how to adjust it.

http://www.mymastering.com/schematics/PS_photo.jpg

A couple of practical questions:

How would you go about setting the two variable resistances on the 50R resistor?

I can easily measure the 6.3vdc, but what's the best way to actually measure the 175ma?

 

[PRR]

> the heater voltage is a bit low (5.9vdc)

That's quite fine; may even be Right.

For best power output, you like to stay within 20%, preferably 10%, of nominal 6.3V.

But this is not much of a power amp. The tube could pass over 8mA without strain, and 2.5mA at 50V; it is asked to pass 0.5mA. It does not have to be heated 100%. 5.9V is only 7% down on nominal 6.3V.

And it is often observed that tube noise declines if heater voltage is a shade low. Dyna ran 12AX7 at closer to 11V than 12.6V.

And a special situation: this is nominal 6.3V 0.15A 0.945 WATTs of heat in the same small case as capacitors and a poly diaphragm. That's why you don't want to light both sides: you'd double the heat. (You'd use a single-triode, but in the Mini line most solos were low-price $9.98 table-radio fodder unworthy of a fine microphone.)

The actual spec for all these tubes was "works in a car radio". Won't burn-up speeding with the "6V" dyno churning 7V, will keep playing with the engine off and battery declining to 5V. Of course car-radios don't accumulate a lot of hours, and most of them are in the 6.0V-6.6V zone, and car-radio tubes may die from potholes before any electrical parameter fades away.

5.9V is Not A Problem. It may even be the Factory Setting, determined after the drawing was made, based on a year and a half of Field Experience.

> I'm not sure if the current is correct.

Within tolerance, if voltage is right then current is right, and vice versa. For General Use, you may either apply 6.3V or you may apply 0.15A. In this case, low current/hiss/heat, you might aim for 5.9V -or- 0.14A. And if you are reading 5.9V, I would leave it alone.

> Here's a photo of the variable 50R resistor.

Yow!

> never one that had 2 simultaneous adjustments.

You paid a nickel extra for a clip. If you had 250V and needed 220V, 180V, and 90V, and were not fussy, you'd put 3 clips on one resistor, bump them up and down under load until you got what you want.

The correspondence to the schematic seems clear:

WHY they did it this way is not clear to me. They get a 5R power resistance "free", but if the constant-current function works, adding resistance does nothing. If you slide the two taps together, you get "infinite" current and a burned-up heater. It is NOT for adjusting by idiots, or even well-meaning experienced techs who don't quite grasp what the designer is doing.

Got 5.9V? Leave it alone.

The red/orange fin-thing to the left is a selenium rectifier. Mildly toxic. Wash your hands. If it ever goes up in smoke, don't breathe much, air the room, wear a mask while you brush the debris out. We used to play with these things, we didn't die, but NOW they tell us it is a bad idea, and they may be right. I'd think, after you find the wires, a 100V 1A Silicon Bridge would work just as well and a wee bit better. But this one is plenty big for its load, it may live forever, and I'd rather let sleeping selenium lie than poke it to get it out.

BTW, I suppose you know a 6072 is a high-price 12AY7.

 

[22busy]

For best power output, you like to stay within 20%, preferably 10%, of nominal 6.3V.

OK.

But this is not much of a power amp. The tube could pass over 8mA without strain, and 2.5mA at 50V; it is asked to pass 0.5mA. It does not have to be heated 100%. 5.9V is only 7% down on nominal 6.3V.

Probably would've taken me over a week staring at that schematic to catch that. But I do think I know why. The tube is used primarily as an impedance converter to bring the capsule impedance of several hundred megaohms into a range that's manageable for the output transformer to do the final conversion to about 200 ohms. So, the amp probably doesn't need to be capable of much power handling.

And it is often observed that tube noise declines if heater voltage is a shade low. Dyna ran 12AX7 at closer to 11V than 12.6V.

Hmmm. I know Neumann in their U47 mic puts 36 volts on the filament (a VF14 which has a 60 v heater). Maybe that's the same approach AKG is doing with this ELAM mic. Noise is definitely a potential problem with these mics and underheating could be a good method to minimize it.

My main concern then is determining what effect, if any, underheating the 6072 filament has on the tube's linearity. I'm pretty sure Neumann designed the U47 circuit intending the filament to be run at 60%. I just don't know if AKG optimized the ELAM circuit for underheating.

5.9V is Not A Problem. It may even be the Factory Setting, determined after the drawing was made, based on a year and a half of Field Experience.

You're probably right.

Within tolerance, if voltage is right then current is right, and vice versa. For General Use, you may either apply 6.3V or you may apply 0.15A. In this case, low current/hiss/heat, you might aim for 5.9V -or- 0.14A. And if you are reading 5.9V, I would leave it alone.

Probably the best advice. I'm just really tempted to measure the current just to see how close to spec this 45+ year old mic is.

If you slide the two taps together, you get "infinite" current and a burned-up heater. It is NOT for adjusting by idiots, or even well-meaning experienced techs who don't quite grasp what the designer is doing.
Got 5.9V? Leave it alone.

Can't argue with good advice. BTW, I'm sure it's obvious I'm not a tech (idiot? no doubt, at least intermittently :roll: ) . . . just way too curious to leave well enough alone. But, I bet I'm not the only one around here with this affliction.

Again, really appreciate the sage advice!

 

[Gus]

Some tubes do not underheat well. Some 6.3 fil tubes underheat to about 5.1 some have problems under 5.7VDC from my limited testing. This is due to the cathode metal alloy and coating used as best I can tell from reading.

More people should look look at some of the older schematics and think about why it was done that way.

Measure the current? easy way use a meter.

First setting before you power the tube fil I would use load resistors for the heater and plate supply. I would guess for the heater 6VDC and 150ma, using ohms law resistance need as a load = 6VDC/.15 amps

resistor load power rating needed > I^2 R or E X I

Connect the load and adjust for 6VDC at the load resistor(s) you should be close for the final setting of the tube. I would let it run for 10 min then adjust for about 6VDC at the fil and check it a few times over say an hour to make sure it is stable.

As Dale posted a TIP127 would be good to use TIP transistor numbers are common and often easy to find. Read up on heatsinks you will need one for the transistor used.

 

[dale116dot7]

[quote author="CJ"]Use a pi filter, a big cap, a resistor or choke, and another big cap. .... Silicon regulates good, but the chips do not like spikes for some reason, go figure.[/quote]

If you're not using active filtering, you gotta get the filament supply very clean and hum-free. With active filtering (using an IC or transistor), that part of the circuit does it for you. Even ground paths and capacitive coupling across the transformer of the heater supply can introduce hum and noise into the tube of a mic. I found that with the Amveco transformers. Something with a shield works much better.

When you crank up the gain, do you hear hiss or hum? If there's hum, you gotta fix it, cuz it'll show up in the mix, especially when you compress the snot out of something you recorded with it. Since recording gear has such a low noise and hum floor now, you hear it if the mic has a buzz. My Altec Coke Bottle needed a total PSU revamp because the original, even with new caps, had a quiet buzz. I took CJ's approach of passive filtering, but I ended up with something like a three-stage C-R-C-R-C filter with around 35,000 uF on each cap. And I had to use a transformer with a shield and ditch the toroids. I don't like running 60 Hz transformers in half-wave if I can help it, although most power supplies I design do - those are single-ended switchers.

In automotives we use transient voltage suppressors all of the time. I don't quite understand why we don't put one across the input filter cap on these designs. I run a SM8A27 TVS on my engine controllers, it starts to clamp at 27V and will conduct 120A (briefly) while clamping at 45V. That should address most of CJ's very justified concern about the fragility of semiconductors.

 

[Gus]

To add to Dale's post. I find the heater supply design part to be more important than the plate section.
FWIW when I built the sony c800g type circuit and found a low noise 6au6 I ended up using super caps at the final filtering of the heater supply to get rid of heater supply noise and low level ticks(maybe the edges from diode conduction) as I got the noise down by selecting tubes.

 

[22busy]

Measure the current? easy way use a meter.

I was hoping it would be possible to measure this in place without having to desolder any connections in order to put the meter in series with the heater circuit, but I can't think of a way with a standard multimeter.

Edit: Oops, just realized by using the dummy resistor I wouldn't even have to attach the mic, just insert the vom between supply and resister. Duhhh . . . must get more sleep!

 

[22busy]

Very interesting tips and comments. Thanks everyone!

This is a bit off topic but it makes me think of an old Vox tube amp I've had for years but never used much for recording because, while it sounds great, it hums like a banchee. I checked all the caps in the Pi network with an ESR meter and they're fine, but I can't get the hum out. This makes me wonder whether the power transformer is squirrelly. Maybe I should try running the heater circuit from a separate transformer to see if that's the problem. Interesting!

 

[Gus]

Look at the schematic you posted a link to.

See the 20ohm 3 watt?

Use ohms law
the voltage drop across the 20 ohm divided by the 20ohm true value is what the reg circuit and tube fil total current is

Then measure the voltage across the 1K at the base to ground and subtract that from the value you calculate from the 20 ohm resistor

And measure the voltage across the 1K shunt if 5.9 volts then thats 5.9ma more you subtract

 

[PRR]

> See the 20ohm 3 watt?

Not really. Look where it is: raw rectified un-filtered DC. The waveform there is all spikes. Most meters will not read the DC correctly.

It also includes current in two 1K resistors, though we could allow for that. And cap-leakage, but after 10 minutes this better be negligible.

You have that 2-tap 50R resistor. We know it isn't 50 ohms as-tapped. Use a DVM (200mV max voltage on Ohms). With power very dead, measure the ohms between the two sliders. Now re-rig for Volts, power-up. V/R will tell you the I. This includes transistor Base current which escapes the load, but from experience and other clues we may bet Hfe is over 50 and maybe like 100. So the error is 2%-1%. Measure 158mA to emitter, it's probably spot-on 156mA to collector, 6mA in the 1K, 150mA in the heater.

In Principle, you can load a constant-current source into a dead short. So shove your Amp-Meter needles into the jack. I would NOT do that in this case. For-sure the transistor dissipation and Vce rises relative to a 6V load, and we don't know how much leeway we got. Also simple "constant" current limiters are not flat-constant, the output changes "a small amount" as voltage changes. Since the alleged error is small, a test with additional error is dubious gain for some risk.

You could just load in 6.3V/0.15A= 42 ohm resistor. You could measure current sure; but if you trust your 42R you can just measure voltage and do math.

The "real" test, if this is a Constant Current Source or not, is to try 42R, 56R, 82R, etc. If the voltage is constant, it is a constant-voltage source. If the voltage rises with the dummy-load (within limits), it is constant current. But the tube does not care (it is rated for either source). The reason they did CCS is probably because, for HIGH ripple rejection, it can be 1-part cheaper (or more reliable) than a voltage regulator or passive filter.

> Read up on heatsinks you will need one for the transistor used.

The original was a TO-3 transistor bolted to the chassis. (And it does not look like 1963 to my eye, but maybe non-USA parts were cut different.) He "could" find a power PNP in TO-3, though the very few recent runs were mostly giant 30-Amp parts, which may be soft at 0.15A.

The modern replacement is the TO-220 power package. It looks flat, but if you bend the leads just right two of then drop in TO-3 pinholes and the tab fits a TO-3 case-hole. Cut the center lead (the metal tab is the same connection). All three terminals are then connected just like the TO-3. You can even keep the original mica insulator, though the far end will flap. You can get TO-220 insulators. You can instead get insulated TO-220 packages: now you have to strap the center leg to where the TO-3 case-screw was wired.

The TO-220 has less heat transfer area and less clamping than the TO-3. But the TO-3 never sits quite flat, and has a longer heat-path. If this were a 100-watt application, I'd think twice. But in this situation it is more like 4 Watts, a "60W" TO-220 will not be limited in case-sink transfer. With just a chassis the real bottleneck is thin metal. But a pro-grade chassis can easily throw 4W this way. Also the original Ge transistor went soft at 100 deg C, and a Silicon replacement will hold-in to 150 deg C. Since the actual temp-rise is more like 25 deg C, neither is in any danger, but the Si part is even further from trouble.

> U47 mic puts 36 volts on the filament (a VF14 which has a 60 v heater).

We must hope and assume they knew what they were doing. I'd guess a fairly beefy tube, a need for very low current, and a desire for very low heat in the mike.

> what effect, if any, underheating the 6072 filament has on the tube's linearity.

Linearity is nearly unchanged; we are too far below the tube's emission limit. As Gus says, you -can- get in trouble with underheating, but he cites 5.7V and 5.1V. Depends on tube and use.

Check 6072 data sheets. 5.9V is probably allowed. If not, it is only because test-limits on other parameters are so tight that they did not want to promise low-heat operation also. But this chore is not demanding of any 6072 spec. Darn near any not-dead triode will work.

And you may be using 6072, the high-price model. If G.E. had a batch of thin cathodes, they'd put them in 12AY7s and sell them for gitar-amp uses; the 6072 was supposed to be Industrial and such customers will complain if service logs show a batch of trouble. I'm sure 99% of 12AY7 will also do the job at 5.9V; the real difference is Marketing.

Like scanners. I can buy a Letter size scanner for my PC for $49. A Legal size scanner tends to start at $300. The actual mechanism is the same... an 8-inch scan-bar runs on a string in a case. Another penny of string, another buck of plastic, you have Legal-length. But they know they have a huge market for Letter, and most of us are cheap. The market for Legal is smaller, but most of those folks simply charge-off their expenses to their clients. So why cut the price while you have the customer cornered? (Also a $49 home scanner may do 100 scans a year, and an office scanner will sometimes do 100 scans an hour all day for a week; you have to use a better grade of string.)

Transconductance may drop 2%-3% at 5.9V on a 6.3V heater. This means gain is 97% of full-heat. You might get 0.26dB more output with 6.3V. Do you care? Usually a condenser gives you plenty of signal, a 0.26dB increase just makes you back-off 0.26dB on a fader (if you can even care 0.26dB when your musicians vary +/-5dB take to take).

Direct-heat "filament" triodes "can be shown" to get "more linear" with underheating. There is a toe at low current, and a knee at HIGH current where emission runs out. By bringing the knee down you quasi-cancel two curves. But this should not happen on oxide-cathode receiving tubes where the emission knee is extremely high (10 times normal currents).

 

[PRR]

OH!

The other reason to under-heat is: grid current.

Most tubes have a Maximum Grid Resistor. Grid current is controlled and tested so that a Max resistor will not upset tube bias with a worst-leak tube.

Small triodes usually are rated 1Meg Max.

That's again a worst-case spec. It is well known, and maker-suggested, that in very low-current amplifiers you may run 10Meg. It "upsets the bias" but in a good direction; you can skip a cathode resistor and bypass cap. Wonderful for very low cost.

Your microphone has a 30 Meg grid resistor!

The 6072/12AY7 has no maximum grid resistor spec. This makes some sense: it is intended to be low noise, grid current contributes to noise, they would try for tighter limits, but did not feel compelled to publish the exact values. Many tube-amp plans don't need high Rg; a transformer input may be 5K DC resistance. Usually 1Meg is plenty high. Most general purpose tubes tolerate 5Meg or even 10Meg if you fudge the other resistors a bit. The condenser mike is the main place we need gridleaks over 10Megs.

The high-price German shops used rather exotic or custom tubes for this reason.

But I'm sure 98% of 6072 will bias-up fine with 30Megs. And under-heating reduces most grid current, so at 5.9V you might get 99% working perfectly well.

 

[CJ]

That unlit heater sucks heat from the lit side, not sure how that enters into the equation, if at all.

Put a 0.1 ohm precision resistor in series with the heater.

Measure the voltage drop with a good meter.

150 ma will read 15 millivolts
200 ma will read 20 millivolts

 

[22busy]

This is great!!!!!!!!

Got a few things to ask/discuss but really need to spend some time with the girlfriend before she wises up and leaves me :sad:. Will read, consider and post more soon. Thanks!

 

[Gus]

Good point about the wave shape across the 20ohm

One could maybe use a large cap across the 20ohm (> C= 1 /(2 pi x R x F)to filter the waveform more to DC BUT that would cause a surge current at start up or connecting it without doing crazy stuff like a soft start.

DC current clamps make this stuff easy if you can get the probe around the wire inside the case

I noted some time ago the use of 8 and 30meg grid to ground in 250 251 circuits with 6072a/12ay7s. I am not sure the reason is it for the highpass or is it because of grid current or both. the transformer is small in the 250/1.

That is why I have posted in microphone thread postings when someone is parts changing in a tube microphone are you SURE you want to change that 200meg etc to a 1gig or 3 gig? with the tube being used.

 

[zebra50]

[quote author="CJ"]That unlit heater sucks heat from the lit side, not sure how that enters into the equation, if at all.[/quote]

I've been wondering about the effect of this for a long time now.