The AM Forum

Here's the current scenario:

Four 6146s in push-pull are driven at about +8dBm by a 600:5000CT transformer. The transformer is a military item with a reported 400-5000 c.p.s. response, and it sounds it. A significant improvement is realized when the secondary is loaded with a 5k resistor, but the voltage swing drops to an unusable low level.

I'm thinking that another stage of amplification is needed, such as from a dual triode. The task is to pick one out. I have a high quality 600CT:15k line level input transformer that shows -1 to +0dB from 20-30k.

The four 6146 grids to be driven are biased at -47 Volts and each pair of cathodes have 5 Ohm resistors with bypass caps to ground, primarily for current monitoring.

I've attached a schematic of my proposed approach. Thoughts for a tube?

Because the anode load is a little high, 6SN7 at 6 - 8 mA per section That will give you 33 k, 2 - 3 Watt anode resistor and B+ of 400 - 500 V or  22 K at B+ of 250 VDC Plenty swing to drive the grids at low distorsion. Please add the grid leak in your diagram ;)

do not bypass the 5 Ohm cathode resistors of the 6146 tubes. Bypassing with a cap that has less than 5 Ohms Xc is difficult. And the tube has some neg. feedback with a non-bypassed resistor which improves linearity at cost of a little extra drive. With the 6SN7 there is plenty drive available.

Ja, grappig ;) Let's use a 47k resistor from each grid to ground since the transformer lacks a center tap on the secondary.

The Sylvania Tube Manual suggests a plate impedance of 6700 Ohms. If I used that value for plate resistors, there would be a 67 Volt drop @ 10mA, thus requiring a 164 Volt supply (7 more Volts are lost in the cathode resistor for grid bias) which could come from the OA2 (150V) regulated supply for the 6146 screens.

Here is my suggestion using a 6SN7 or other low Mu (20) triode, with a stage gain of 14.

It would be helpful if you supplied more info such as a block diagram of the complete system.


Phil - AC0OB

Here is my suggestion using a 12AU7A triode, with a stage gain of 17, unbypassed.

Bypassing the 168 ohm cathode resistor with a 100 uF@16V electrolytic will get you more gain at the expense of increased distortion.

What you need is a decent amount of voltage drive into the grids of the 6146, so you want the load resistance of the speech amplifier to be about 5 to 10X the source resistance, which is what you see here.

+8dbm = 1.95V rms into 600 ohms.

1.95VX1.414X17 = 47V Peak per plate.

It would be helpful if you supplied more info such as a block diagram of the complete system.

And I trust this is for personal use.  :)


Phil - AC0OB

I'm leaning toward the 6SN7 solution for a variety of reasons including that an octal socket will fit perfectly in the hole used for the previous input transformer (I lack Greenlees).

For B+, my choices are the 150V regulated supply that the 6146 screens use, the 600V used on the plates via a voltage divider, or locating another tap strap for the 50k/100W bleeder.

Interestingly, this circuit can be "breadboarded" for testing using an octal relay socket with screw terminals. That might be a first, lol. Thank goodness for my EICO 1030 power supply. Just love that...

I posted a block diagram although it may not be what was asked for.

Well, well, well.  Some pineboard style testing produced interesting results using Phil's 6SN7 input circuit (attached).

B+ had to be limited to 150V in order to keep the plate voltage from exceeding 90V (per the manual). Only 2.25V was developed across the cathode resistor (suggesting a combined plate current draw of approximately 5mA) which is half of what the manual suggested yet the tubes tested "good" (I have two).

The grid resistors were changed to 25k each which loaded the input transformer properly and resulted in a -1 to +0 dB from 20Hz to 10kHz with -1dB @15kHz!

This arrangement produced a 70V PtoP output across a 200k load with 0dBm input level.

The next step will be to add a -4dB H-pad on the input for isolation and S/N improvement.

Does the input xfmr have a CT on the secondary??  What input xfmr are you using?  no name?? or brand name?

No, unfortunately. It came out of an LPB 25C AM transmitter. The 600Z primary has a CT, which could be used for remote switching by putting DC between the balanced line and ground to operate a relay. I believe it has a 15k secondary based on voltage ratio measurements. Pretty good specs, though!

My circuit was based on the 6SN7 Characteristic Curves for plate voltages of 150V@8 mA and a cathode bias of 3.5 V, and a 600 ohm 1:1 isolation transformer. The voltage Gain of the stage was 14 (23 dBV) with the values given.

The primary of the transformer was already loaded.

The cathode resistor should be 220 ohms for a cathode voltage of 3.5V@16 mA.

Where did this 90V Vp requirement come from? ???


Phil


BTW, have you looked at your total current/power budget? If this is a 4X4 6146 transmitter the estimated powerbudget is 700V@1.25A > 1kW if you include filament power?

I can't seem to avoid it. In addition to the datasheets I've found on line, these books concur for Class A service:

"Characteristics And Typical Operation": 90V @ 10mA per section
  ~ Sylvania Technical Manual

"Characteristics And Ratings": 90V @ 10mA per section
 ~ Essential Characteristics for Receiving Tubes - General Electric

*However*, the 1976 ARRL Handbook shows the 6SN7 as: "250V @ 9mA" on each plate which seems much closer to what I was seeing but didn't want to push. Of course, how could such a stout tube have to run at 90V on the anode. Weird.

No worries on the power budget. B+ is 600V, screens are regulated at 150V, and idle current per pair is 70mA. The RF deck only produces 120W of carrier, so these four 6146s are living in Cushville.

Hmmm. So there’s gain to spare. I might change the 6SN7 grid resistors to 330 Ohms and use D.C. blocking caps on the input. This will never be driven by anything other than a b’cast processor.

Wow Clark you are just full of surprises   ;D

So is this circuit

1) two 6146s in ClassAB1 Push-Pull driving a single 6146 or

2) two 6146s in ClassAB1 Push-Pull driving a set of paralleled 6146's?   ???

3) four 6146s in ClassAB1 Push-Pull paralleled driving a set of paralleled 6146's?   ???

4) four 6146s in ClassAB1 Push-Pull paralled driving a set of four paralleled 6146's?   ???


I don't know how they came up with those 6SN7 numbers, must be for some low noise, low gain audiophile circuit.

What was their plate load RL and supply voltage???

Both the GE and RCA data for ClassA show running the plates at 250V@9mA, with a -8V bias, but that was in the less linear region than running them 150V@8mA with a Vk of 3.5V.

My goal was to run the 6SN7 with the least plate diss. and in the most linear region.

If you want 90 V on the plate then the cathode needs to be at ground since according to the tube curves, the cathode bias should be = 0V for 90 volts@10mA.

I still do not see or understand your rationale for loading the heck out of the secondary of the 1:1 input stage transformer, if that's what it is.

Most processors I have ever delt with have an output level adjustment to lower output voltage levels.


Phil

Ah, okay. A few clarifications...

This project is to effectively drive four 6146s where V1 and V2 are in parallel, V3 and V4 are in parallel, and those sets are in push pull. That’s the modulator. The RF deck is four 6146s all in parallel driven by a Gates BC1G Oscillator/Buffer circuit.

The secondary of the input transformer needs to loaded to a point in order to flatten out the response otherwise the lows roll off at 100Hz and the highs show an upward curve past 16kHz.

Pretty sure you can put 450 volts on a 6SN7.  I run one in my HB xmtr at 300 volts on the plate.

Another “Ah-Hah” moment.

Yes, 90V @10mA and -7 for bias. My error was trying to derive that bias from the cathode resistor. Can’t be done with an Ep of 90, trust me. So, the next step is to see at what plate voltage I get 20mA combined and a 7 Volt drop across the cathode resistor. In theory, wouldn’t it be 350 Ohms?

Clark:

Fun project. I wonder:

1. If you can eliminate the driver transformer and replace with a tube phase inverter? This will allow more system negative feedback.

2. While a left-turn, consider a derivation of the WA1GFZ MOSFET driver. This will allow lower source impedance to the 6146 grids and thus less grid current distortion:

http://amfone.net/Amforum/index.php?topic=23632.0

73,
Dan

To my opinion do NOT use 90 Volts, too low. No Idea where that comes from, but it is not correct. I should go for a minimum of 250VDC B+ in order to have a loaded output swing with sufficient headroom to stay away from the distortion zone with max drive required for the 6146   Low distortion = headroom in signal capability. That also gives you the possibility to add some negative feedback and linearize the design by adding non-decoupled cathode resistors to the 6146 tubes Attached the datasheet of the 6SN7

I was going to let this go for the evening but curiosity got the best of me and I had to fire up the EICO power supply for some more tests. Having broken through the glass plate voltage ceiling, I was able to get 7.5mA per plate at 225 V at the anode.

I’ve attached a schematic showing the circuit powered from the 600V supply. The Volts shown next to the resistors represent the drop.

Perhaps the last step will be to determine what current draw is needed to fully drive the four 6146s and settle on a cathode resistor that yields 7 bias Volts and appropriate dropping resistance.

Thoughts?

Clark, I am not sure where your numbers are coming from but here is a tabulation derived from the 6SN7 Tube curves for various parameters, keeping in the most linear portions possible, and which is the only reliable way I have found of getting into the ballpark:

Vs = 300V
Vp = 90V
Ipeach = 10 mA
RPeach = 21k
Vk = 0V
Rkt = 0

Vs = 300V
Vp = 150V
Ipeach = 8 mA
RPeach = 18.75k
Vk = 3.5V
Rkt = 219 ohms

Vs = 350V
Vp = 200V
Ipeach = 10.5 mA
RPeach = 14.3k
Vk = 5.5V
Rkt = 262 ohms

Vs = 400V
Vp = 250V    
Ipeach = 12.5 mA   My Pick of the Litter ;D
RPeach = 12k
Vk = 7V
Rkt = 280 ohms

Vs = 400V
Vp = 250V    
Ipeach = 11 mA
RPeach = 13.6k
Vk = 8V
Rkt = 364 ohms

Vs = 450V
Vp = 300V
Ipeach = 11 mA
RPeach = 13.6k
Vk = 10V
Rkt = 455 ohms

I hope this helps

Phil

I am sorry Phil don't completely agree that this is the best way to arrive where you want to be. You have to calculate back from the requirements of the tube drive. The 6146 need 100Vpp grid to grid at 600 V anode to be driven. That means 50Vpp per grid. In order to stay away from the distortion zone, the voltage drop over each anode resistor of the 6SN7 should be MINIMUM. twice that = 100 Volts. Due to the load of the grid resistors which is recommended to be less than 200 kOhm, say 100 kOhm for 1 tube, 47 kOhm for two tubes in parallel, take 50% more (for anode resistors in the order of 22 kOhm), so the voltage drop should be 150 V over each anode resistor. It is best to stay away from bottoming the tube, so better operate in the higher anode volt ranges. Say 250 V at the anode = 250 + 150 = 400 V VD
The recommended anode dissipation is 2,5 Watts, 3,5 Watt MAX, so you can run the tube at 10 mA per section at 15 kOhm anode resistors or e.g. 7 mA  with 22 kOhm resistors if you want to run a little cooler. It is a driver, and if you take sufficient margin at the the 6SN7 output voltage like in this example, you don't have todrive the 6SN7 more than 50% of the max output and you will stay nicely linear. No need to run the tube up to 12,5 mA

What I gave him was a set of possible circuit values from the curves.

Each to his own philosophy of approaching circuit design. :)

I gave three possibilities with Vsources (Vs) of 400V to 450V.

Each one had at least a 150V margin to the top of the PS, at least a 50 volt margin for signal, and low output impedance if he decides to go with ClassAB2.

I have given a number of possible schematics so may I suggest you post a schematic
of how you think his driver should appear.  Of course the final decision as to how he will implement it will be up to the OP.

Phil - AC0OB


 

That’s what clip leads are for! The last components to be soldered will be the plate dropping and cathode resistors.

I had a way to do this before posting here, but it’s always beneficial to broaden the options and build a discussion that could inspire others.

Ok Phil, misunderstood, I was misled by the first  3 examples. In my opinion, the anode voltage of the first 3 is too lo to have ample signal margin. My pick is the 400 V 11 mA, though I should run a little cooler, a little less current, but that is entirely a personal opinion. Interesting discussions.
I made a simple distortion meter and learned a lot from using that in all kind of amps. From those experiences I ended up with the above philosophy. That was for hi-fi amps driving 2 x KT77 in UL

There's no question that this has been a fascinating project...thank you all!

I've been testing using Phil's original schematic (below) and have achieved the target 100V across a 100k load with a 300V supply @ 13mA, 5V of cathode resistor bias, and approximately 5.4VAC between the 6SN7 grids.

Is this fantastic or is there something I'm not seeing that's not right?

Seems fantastic to me ;D

Well, let's see. 5V Vk/470 RK ohms = 10.5 mA total circuit current for the two triodes which infers 5.3 mA plate current for each triode, which according to the tube curves is in the ballpark.

That would mean about 205 volts should be appearing on the plates and in a linear portion of the APC curves.

Now, 13mA total circuit current through 470 ohms = 6.11 volts Vk. 6.5 mA current through each RP yields a voltage drop of 117 volts, so each Vp should see (300 - 117) = ~ 183 Volts. This is also in the tube curve ballpark, and in a linear portion of the APC curves.

What are the actual plate voltages and component values now being used?



Phil

Addendum: May I suggest one more tweak to move the operation into an even more linear region:

 

Back in the saddle after a few busy days...

A lot was discovered by putting the circuit in the transmitter and running some real-world audio through it. First, and the most surprising, a *lot* less gain is needed. I was thinking 100V PtoP, but placing a VOM across the 6146 grids revealed that the needle wandered around the center of a 50V scale at full modulation. And that's using -6dBM of drive!  Padding, please!  Those four 6146s only need to produce 60-80 Watts of audio.

Other than that, things sounded pretty good. What's "real-world audio"? Andy Williams' "House Of Bamboo" for example. You all are too young to appreciate that ;-)

Seem all correct to me, First of all, you need 50Vtt at each grid to get full max output, for 4 tubes 100W+ So if you need 60 - 80 Watts, approx 80Vp-p between the grids is sufficient which is approx 50V RMS. In addition, the driver design has about 100% signal margin to stay linear, so plenty of gain to drive the 6146 quad. I like the end result, you run the 6SN7 cool, long life and more reliable.
I advice to limit the drive in an audio stage directly after the mic., compressor limiter or so, Don't attenuate after the 6SN7 or lower its gain, the signal margin gives you superior audio quality.
Congrats with the result!!

You should be able to get approx. 240 Watts out to the antenna with a modulating power requirement of  ~ 130 watts for 100% modulation.

Neat project.


Phil

Hartelijk bedankt!


That's where the power budget comes in as I'm now limited by the power supply. Are there any 6SN7 curves that permit a B+ of 150V? I'm running 250V now but I'd prefer to tap off the screen supply for simplicity. Remaining at 250V means adding a power divider from the 600 HVB+ or locating a second clamp tap for the 100W bleeder.

Ho Hakiwisilaasamamo Waswasimamo

Here is another potential Power Supply using a Toroidal core transformer. I have used a number of these in various projects.

Two of the HV secondaries are paralleled to provide 1.24 A.

Phil - AC0OB

Should be able to run the speech amp off the 600V supply through a simple dropping resistor and a cap.  Another 15ma load on the supply shouldn't be much of a issue.

But what do you do on KeyUp when the supply goes to 725V? Do you let the 300V rail go to 425V and have the tubes heat up?

We can't neglect the effect of the higher voltage on KeyUp.

One can regulate it with 0A2's or zeners.



Phil - AC0OB

He runs the tubes at 5 mA each, 425 V Vd shouldn't be a problem.

I was going to suggest a couple of 150 volt gas regulators in series.  In my HB xmtr the 300 volt supply is keyed along with the 600 volt supply.  My xmtr is designed in a way that only allows the voltages to rise less than 10% on key up.

By my accounting he's running each tube at 6.5 mA. I later suggested a move to 7.5 mA for each tube, for more linearity, by decreasing the cathode resistor to 330 ohms.

But that was not the point. The point is you have to consider no load voltages when working with power supplies and branch voltage rails.


Phil

I agree that 7,5 mA will be a little more linear, that's about the current I suggested in the very first post. When the 600 Volts jumps to 725 with key-up, the Vd for the 6SN7 will jump from 300  to approx 380V, when you use a series dropping resistor, not to 425V, Seems no problem to me.
But off course, the solution with a stabilizer will be the cleanest and prevents plops due to charging/de-charging coupling cap's during the jump which I doubt are noticeable because the 6SN7 drives a balanced amp. I should just leave it as is, but that is my personal opinion, KISS

I have found this true in experimental breadboarding. At one point I thought a transformer didn't care about being loaded, i.e. driving no grid current, why not just consider the ratio? No load is easier on it than a full or double load, right?

But as it turned out, the frequency performance was seriously affected if the load was not reasonable to the transformer's design. Alas, one of the imperfections of a 'perfect' transformer, and I mean good UTC parts not a $2 unit.

Transformer experts please comment as to why loading is important.

May have more to do to the load on the secondary reflecting a load back to the primary and the source.

http://www.kandkaudio.com/wp-content/uploads/PSW_WhitePaper_Download_Chapter_6.pdf

See Figures 1,2 and 3.

Effectively you are broad-banding it by loading it.

But having to put a very low resistance load across an audio transformer secondary indicates to me that the transformer quality is questionable and that it has excessive internal capacitance.

The reason I use these transformers:

https://www.edcorusa.com/wsm_series

in the 600 ohm to 600 ohm (1:1) series is that they have very low internal capacitances, center taps, and good frequency response without having to load them excessively.

Another approach is to do something like this for a transformer without a centertap;



Phil

It is indeed an old problem of spread inductance and capacitance. In the years 20 Philips did wind tube coupling transformers with resistance wire to flatten the resonance!!
I always load the sec until the high resonance is hardly noticeable, not more. If you want more low, lower the DRIVE impedance to the prim. The lowsof a correct loaded transformer roll off due o the inductance of the prim which you can overcome by driving at a lower impedance.
I suppose you are not looking for highs up to 20 kHz. otherwise you need more expensive transformers.

When Murphy strikes, he's calculated and effective...

I had the new 6SN7 driver all wired up in the modulator chassis and was giving the transmitter a dynamic test when I noticed how thin and tinny the audio was. Hyellified, I started testing audio at each stage after the processor. It didn't take long to find out that my *prized* 600:15k input transformer now had an open secondary.

So, that's that until I can locate replacement iron. Fortunately, it's small iron...

Ok now I understand why you were getting such a high gain and had to pad the secondary. I recall you mentioning two different input transformers but I thought you were using the 600:500 ohm transformer.

You had a 5:1 step-up situation. +8dbm in would have resulted in 10 volts at the grids of the 6SN7.

That 15k secondary would have a lot of distributed capacitance and inductance.


Phil

Some ideas
Many Philips radios from the years 50 / 60 have small AF output transformers that have a negative feedback winding for approx 500 Ohms There you have 500 - 7 kOhms. I used these, worked very nice
Microphone transformers, sometimes in the base of CB table mic's
I may have one for you, but my stuff is still in my sea container until my house is ready. So that is the least probable at short notice

Nico, bedankt! I found and installed another with reasonably good specs when loaded carefully. I also discovered that the power transformer secondary has an unused center tap, so with the full-wave bridge, that would make a useful 300V source.

Love those Philips radios, and Grundigs and Lowe Optas... heel goed!

My next task is to determine why adding negative feedback causes the screens to go into super-conduction regardless of polarity. Supersonics?

Troubleshooting homebrew equipment designed and built 50 years ago can be quite a challenge!

Drawings, schematics? Modulation section? Speech Amplifier?


Phil - AC0OB

My next task is to determine why adding negative feedback causes the screens to go into super-conduction regardless of polarity. Supersonics?

May be RF oscillation
That's the reason I don't like neg. feedback, especially when there are inductors in the circuit. Just get rid of decoupling caps across cathode resistors, that gives you a stable and nice negative feedback per stage. I never used neg. feedback in my HI-FI tube amps other than that.
Fine to hear that you found an other iron.  :D

when you use the center tap of a full bridge loaded transformer to have a lower DC voltage, DON'T forget to put a diode in series with the center tap!!

I've built many of this type of power supply,  never needed a diode in series with the CT.  The FWB diodes are already in series with the CT.

Fred

When you don't put a diode in series, you lift the AC winding to a DC voltage! In this case, your AC winding will be 300VDC

Doesn't matter,  The winding is already 300 volts above ground with FWB rectifiers.  What is at the CT is a 300 volt DC pulse at 120 Hz.  You don;t need any diodes in series with the CT.  The FWB diodes are already in series with the CT.  Current is only flowing in one direction from the CT.

The xfmr winding does not have any DC voltage on it.  It's all AC voltage.  If it's a 600vac winding then there is always 300vac at the CT.

OTH you must use a choke input filter off the CT.  If you use a cap filter or even a CLC filter then the high voltage output also becomes a cap input filter.  So then you could use a diode in series with the CT.

I never use cap input filters for xmtr service.

Fred

I am sorry Fred, you are wrong in this case.
The center tap in your opinion should be connected to a filter cap for the 300VDC. So there is NO ac at the centertap, the 300V filter.cap short-circuits the AC to ground. When the center tap is directly connected to a cap charged to 300 VDC, the upper half of the winding + THE 300VDC will charge the HV caps to 600V or deliver the current to the 600VDC user in case of an choke input filter for the HV. This charging current comes from the top value of the half sec + the 300VDC in the cap, and NOT from the whole winding (until the 300V cap is partly discharged) so the cap gets a VERY HIGH surge discharge current every time the HVcap  is charged or the HV user draws current, when the HV delivers current to a load.. That results in a high ripple at the 300V when you draw current from the HV When you don't draw HV current or the current by the 600V user is MUCH lower than the 300 V user, no problem, but that is NOT the case.
You have to put a diode in series with the center tap in this application OR the 300VDC should be a choke input filter, but than you have less voltage than 300VDC for the low voltage users..

No that's not what I said.  I said that the CT should have a choke input filter.  If you use a cap input filter on the CT then that will cause the HV filter to also become a cap input filter, even though there may be a choke input filter on the HV.  You're saying about the same thing, only saying it differently.

If you use choke input filters you don't need a diode off the CT.  If you use a cap input filter on the CT, then yes, you should put one or two diodes in series with the CT to prevent the HV from seeing the cap on the CT.

A cap directly on the CT will charge higher than RMS voltage.  This higher DC voltage will bias the CT higher than the RMS value.  This will cause the HV to also rise to a higher voltage.

That's why I never use cap input filters for xmtr service.

Fred

Ok You are right. Now I read a gain, we indeed tell the same at a different way.
For the CT I normally use a diode and a cap input to have more voltage, smaller and simpler and because so far I used it as a low current supply. Saves a choke
This tread seems also good to improve my English.. ;)

Your English is not that bad,  I've seen worse.  But, I'll re-write your first line.

OK, you are right.  Now I read it again, we indeed say the same thing in a different way.

The rest of your post is OK FB

Fred

Thanks for the lesson Fred, Highly appreciated

I'm simulcasting this with another thread but wanted to share the solution that's working well. The 250V point supplies the 6SN7 (15mA) while the 150V point supplies the four 6146 screens (35mA) and OA2 regulator (15mA). The power transformer doesn't seem to notice the 65mA drawn from the center tap.

By the way, Nico's English is quite good. If this thread were all in Nederlandse, then I'd be in deep trouble as dat is niet gemakkelijk ;-)


Ugh. The original schematic was drawn in pencil and the paper is wearing thin from all the erase-and-change that's been going on. Need to have things settle before sitting down to the drafting table! And by the way, I'm hesitant to show that the modulator uses a (G-A-S-P!) capacitor input supply (here come the torches, pitchforks, lectures, and pontification).

No problems from my side, the current draw is low. But I should put a diode instead of a resistor in series with the CT. What is your opinion in this case Fred?

Haha  ;D  And I too think Nic's english is pretty good.

You must have one Honkin' transformer.  :)

As long as your ripple voltage at the 250V point does not result in hum on the 6SN7 plates you should be good to go.

BTW, I am not trying to hijack your thread or trying to tell you what to do, just making suggestions.    :D


Phil

The original transformer failed years ago so I put in my prized RCA color television transformer. I'm not using the 5 or 6V windings, just the HV and now with CT. Yep, it's honkin'.

No worries and no hijacking here. With all the mysteries of this transmitter, this thread could easily morph into a completely different topic. To fill in some of the blanks for the sliderule jockeys, R1 which comes off the CT is 1500 @ 5W, and R2 which drops to the regulator is 6.8k @5W.

The resistor is part of the filter.  Whether a diode in series with the CT is needed really depends on the resistor value.  A few hundred ohms and it would not be much of a RC filter. Almost more like a cap only filter. So I would add the series diode off the CT.

If the resistor is a few thousand ohms, probably you would not need the diode.  But,  adding the series diode in all cases really doesn't hurt anything.  Diodes are cheap enough.  It is just something that I never do because I always use LC filters.

Fred

I consider an LC filter is absolutely the best Fred, also less stress for the rectifier and the transformer, but sometimes space and availability of chokes drove me to other solutions. I produced many 27 MHz RF generators for plastic welding, 300W, 500W, 1 kW and 2 kW with MOSFET's. The voltage for the driver and the control electronics was derived this way from the CT via a diode. For the other electronics this voltage was fed to a switching regulator to get + /- 12V and 5 Volts. To find chokes for 35V, 2 - 3 amps was not easy and a reliable source for many years was problematic, especially for series production and the size was a problem as well. In the USA these things are more easy.

Clark,

Those RCA TV xfmrs can handle a lot of over load.  Especially if you're not using the filament windings.  I had about 20 NOS RCA TV xfmrs from the 50s.  I'm using two in one of my HB xmtrs.

I once used two to make a 1400v power supply at 300ma load as an experiment.  Worked great.  Both xfmr were about 765vac CT.  I used FWB rectifiers on both xfmrs.  I connected the diode output of one xfmr to the negative terminal of the FWB diodes on the second xfmr.  This added the two voltages together.  I used a choke input filter off the diode output of the second xfmr.  Loaded it down to 300ma.  Ran it for hours.  1400 VDC at 300ma.

The second xfmr was modified.  I removed all the filament windings and increased the insulation around the core.

Fred

Something like this?   :D

If the PDF schematic below is Clarks PS then I can see why there is a weird voltage drop across Rdrop2 since some of the FWB current through ground is going up and through the gas regulator circuit.

https://www.electronics-tutorials.ws/diode/diode_6.html


Phil - AC0OB

This circuit is working quite well, and the sufficiently high input resistance (1.5k) seems to eliminate the need for a choke. If there's any weird voltage at the regulated output it's because I drew the OA2 upside down. Additionally, I cited the screen current incorrectly as the 35mA was for the RF finals, *not* the modulator screens, which is MUCH lower.

The challenge du jour is fabricating a new negative feedback ladder to accommodate the new 6SN7 input circuit. Significantly less FB is needed thanks to the higher gain/impedance of the input circuit.

Before you make that ladder try first to add 2 cathode resistors to the 6SN7 cathodes, eg 150 Ohm or so and see the result in distorsion improvement, I think no other neg feedback is required.

I was too fast with sending, sorry.
The best is to add these resistors and drive the cathodes with a real current source, e.g. a FET instead of a resistor to ground. That will guarantee an exact balance to the 6146 tubes.
So select a JFET that has the approx. the required drain current at zero gate volts.(in your case approx. 15 mA) The drain impedance of a JFET is very high, so the sum current of the two triodes is always the same.  Than add the 2 resistors from the drain to the cathodes. ( this is a stable and simple negative feedback methode improving the linearity of the tube.  That will give you a good linearity and an excellent balance. But all this may be a little overdone, that's more for HI-FI amps, (but still fun ;D)

The feedback I was thinking of is from the modulation transformer back to the audio input. About 3dB is what I had in mind, but because the 6SN7 offers a much higher impedance point for the feedback than the previously used 5k input transformer secondary, the FB ladder I have needs a bit more loss added.

As it turns out, a discussion of feedback is going to bring us right back to the input circuit design. It's now looking as though the 6SN7 should have separate cathode resistors in order to provide a return point for the FB circuit.

There's a single 300 Ohm resistor now. How about 560 on each cathode, 18k on each plate, and 260V source?

I agree with Nic, kill any distortion up front before complicating matters with a feedback system.

The proposed circuit was biased into the more linear portion of the operating curve so distortion should be minimal.

How's about a 300V source and separate, un-bypassed 680 ohm cathode resistors like this:


Phil - AC0OB

Phil,

Check your schematic you just posted.  You have the two cathode resistors jumped together,  right where you have the 5V notation.

Fred

Thanks Fred, I corrected it.  :)

Sometimes I need a proofreader to check my schematics.  ;D


Phil - AC0OB

I am sorry guys, I don't agree. Separate cathode resistors will give you less perfect balance IF YOU GROUND THEM.
A cathode resistor in each tube will linearize the tube characteristics, it is the best negative feedback system without any complications. BUT where these cathode resistors meet, DO NOT GROUND. That will give you un-balanced signals, second harmonics. The cathodes should be fed with a current source, an impedance very high compared to the cathode resistors plus 1/S, the cathode impedance. Than the sum of the two triode currents is ALWAYS exactly constant and you will have a perfect balanced system. less even harmonics.
You can reach that in two ways, feed the sum-point of the cathode resistors with a J-FET, that is an almost perfect current source, (take care that the FET has sufficient source - drain voltage to operate. A transistor biased at 15 mA will do as well) or connect with a LARGE resistor to a negative voltage, e.g. the bias supply. When you have e.g. a -100 V bias supply, that resistor should be 100/15 mA=approx 6.8 kOhm, 2 W
In HI-FI amps I use a J-FET with its source at approx -12VDC or more selected for the current I want with zero gate volts.
In your case, I should simply use a 6.8 kOhm to the neg. bias supply (if it is approx -100VDC)

I might not have described by intentions adequately, so here's another shot...

The feedback loop goes from the UTC S-21 modulation transformer in the plate circuit of the four 6146 tubes back to the input of the modulator, which now includes the new 6SN7 driver. My proposal is to use separate 560 Ohm cathode resistors so that the FB loop can be connected to the cathodes.

Don't make me draw this out. You wouldn't like by handwriting...

Its like Phil said

kill any distortion up front before complicating matters with a feedback system

Why complicate with that feedback system, not required and high possibility for oscillations Feeding back over two stages with two transformers in the circuit is really asking for trouble.
Simply each cathode approx 470 Ohm (not critical, more resistance is more negative feedback = less gain) Tie these together and at this point 8K2, 3 Watt to your bias, -120VDC
Very good balance guaranteed, less second harmonics and with the two 470 Ohm resistors you can select the neg feedback, the gain.

Is this the concept?

JMO, but since this an ARS transmitter transmitting under conditions not found in a home listening room with a Dynaco Amplifer, it would seem that one would not need a 0.05% distortion figure for this audio system and modulator. :o


Phil

Thanks, Phil. That's pretty much the circuit I had envisioned although I wish I took NFB into consideration before my pristine soldering job on the 6SN7 octal socket ;-)

Since there's no FB winding on the mod iron, I can either:

1) Take a feed from two unused primary side terminals on the S21 mod transformer and cross them back to the two cathode resistors for balanced NFB.

2) Take modulated B+ from the secondary and loop that back to one of the two cathode resistors for unbalanced NFB.

There's going to be considerable resistive attenuation with D.C. blocking caps in either case.

So, 680 for cathode resistors? The B+ that's appearing at the top of the 18k plate resistors is around 260V which yields a plate potential (to ground) of 125V @ 7.5mA.

Consider this:

Most of your even harmonic distortion will be very low due to Push-Pull

Most distortion products (if any) will appear on the secondary of the Mod. transformer

So I would go back to the single cathode resistor and send feedback to it as suggested below with a DC decoupling cap so as not to upset the 5V bias on the 6SN7.


Phil - AC0OB

I think this xmtr has four 6146s in the final.  With that much plate current the two secondary windings on the S-21 mod xfmr should be run in parallel.

He will need a high voltage cap to isolate the modulated B+ off the feedback circuit.  He could connect to the tap on the mod xfmr to lower the modulated B+ on the FB circuit.  Even doing that he will still need a cap with at least a few KV rating.

He could also run the feedback to the grid resistor.  Dividing the 470K grid into two parts.  This will have a much higher impedance on the feedback circuit allowing a much lower value FB isolation cap.  Still the cap would have be a few KV rating,  but much easier to find a low capacitance cap with the necessary high voltage rating.

Although phase may be an issue.

DZT: I think this xmtr has four 6146s in the final.  With that much plate current the two secondary windings on the S-21 mod xfmr should be run in parallel.

I would be more concerned about the primary's current carrying capability. The S-21 is rated for 115 Watts max, here we have the capability of producing up to 192 Watts:

http://www.bunkerofdoom.com/xfm/UTC_1949/UTC49.pdf    Page 15.

DZT: He will need a high voltage cap to isolate the modulated B+ off the feedback circuit.  He could connect to the tap on the mod xfmr to lower the modulated B+ on the FB circuit.  Even doing that he will still need a cap with at least a few KV rating.may be an issue.

HV cap not needed if he uses a voltage divider attenuator.


DZT: Although phase may be an issue.

For sure the proper side of the secondary will need to be selected.


Phil

Phil.

Right, I overlooked using a voltage divider to ground.

I have a S-21 here,  not quite a big enough mod xfmr both on the primary or secondary side for 4x4 6146s.  The UTC S-22 would be a better mod xfmr for this xmtr, but he probably doesn't have one.

I'm not a big fan of UTC S series xfmrs or chokes.  I have a number of them here and never consider using them.  The S series chokes are about the worse of all the common chokes from other makers.  The earlier UTC PA series xfmrs and chokes are very good.  I have more than a 1000 xfmrs and chokes here and did a lot of testing on them.

Fred

Here is one potential Feedback arrangement:



Phil - AC0OB

This is all good info, but, the specs on this transmitter are unique:

RF Supply: 600V @ 300mA (120 Watts carrier)

AF Supply: 650V @ 120mA quiescent

I've attached a simplified schematic of the feedback strategy. The resistors from the mod transformer to the 6SN7 cathodes are comprised of high voltage resistors in series. The total value is chosen to provide 3dB of negative feedback and a very slight increase in bias voltage.

I don't feel that this would present a challenge for four 6146 tubes tasked with providing 60 Watts of modulation, plus headroom. With no feedback, the modulator with the UTC S21 produces this response:

Hz:  20   30  50  100  250  500  1000  2500  5000  10000   15000   20000

dB:  -8   -4  -2    0     0      0    0 (ref)  0       0       0         0         -1     

RF Power Input for a parallel 4 tube 6146B Quiescent = 600V X 300 mA = 180 Watts; Practical RF power output@Quiescent = 180 Watts X 0.74 = 133.2 Watts in Class C.

RF Power Input for a parallel 4 tube 6146B = 600V X 560 mA = 336 Watts;
Practical RF power output@100% modulation = 336 Watts X 0.74 = 248.64 Watts in Class C.

Modulation Power Needed for 100% Modulation >= ~ 124.32 Watts;


For a 4 tube Modulation system in Push-Pull parallel Class AB1, Quiescent input power is 650V X (2 X 48 mA) X = 62.4 Watts;

For a 4 tube Modulation system in Push-Pull parallel Class AB1, input power@100% Modulation is 650V X 500 mA Max. = 325 Watts (assuming Mod. B+ stays at 650 Volts);

Practical Modulation Power Output@100% = 325 Watts X 0.64 = 208 Watts.

Mod. Power levels@100% Modulation are still above S-21 capabilities.


Phil - AC0OB

Probably better to remove at least one 6146 from the PA and two 6146s from the modulator.  This will bring the current load on the TV xfmr down.  It is rated at 400ma with a FW ct power supply.  Using it with a FWB power supply and loading it to 400ma at 600v doubles the VA load on the xfmr.  But these RCA TV xfmrs can handle the overload easily, especially when not using the filament windings.  Although I'm not sure it the filament windings are being used.

This will bring the audio power down,  more in line with the S-21 mod xfmr.

Fred

133.2 Watts / 2 = 66.6 Watts of audio needed for 100% mod, but lets say 80 Watts to be wild.

That should be a non-issue for four 6146s, and since the S21 is rated for 115 Watts of audio, all should be well.


The transformer is probably well within its CCS ratings for playing Honeymooners episodes for hours at a time. The filament windings aren't used, and the modulator screens are regulated at 150V.

Typically the TV xfmrs ran with a considerable temperature rise.

But that was for Quiescent Power, at idle, no modulation.

The itch to finally apply audio may be shortly forthcoming. ;D

BTW, from my experience, the screen voltage for those RF finals is about 25 volts too low, unless you intend to operate this transmitter way below it's capabilities.


Phil

well why not do like the big boys of broadcast did.

Also, having four 6146 in pp-par does not mean having to drive them to full output.

True, but were not talking Broadcast CCS, were talking Intermittent Service.

I once had a friend back in Mo. (now SK) that just loved 6146's and I drew up the schematics for him for a 4X6.

We had the Final tubes arranged in a 6" diameter circle and the Modulator tubes arranged in 4"X4" rectangle.

A 65 cfm muffin fan sucked hot air from the modulator side while a 110 cfm sucked hot air from the final.

However, we ran the Modulator in Class AB1 at 750 Volts.

Did I mention He loved those little bottles?   :o


Phil - AC0OB

For the purpose of this discussion, CCS, please. I built a CW only transmitter to ICAS specs, but other than that, I only do CCS.

The schematics were to show proven methods of taking negative feedback voltage from the modulator plates and applying it to the cold ends of the first audio input transformer windings. Other factors such as power level are not considered in my providing those old BC modulator schematics as one known way to improve fidelity.

Agree totally Pat, as the "ladder" method of feedback was used in most Tube Broadcast Transmitters.

But some people seem to be rather cryptic.


Phil - AC0OB

The feedback ladder is exactly what I had in mind, thus the changeover to the separate cathode resistors given design limitations due to the single secondary of the input transformer. There was no intention of being cryptic, particularly given the generous and helpful suggestions that have been made here.

At this point, I'm experimenting with resistor values and the decision to use D.C. blocking caps or not, depending on how much the resistors increase the bias voltage on the 6SN7 @ ~ 3dB of NFB.