On Building A 1937 Modulator

[w8khk]

Hello Mark,

I would like to offer some assistance toward reducing the hum in your speech amplifier project.  As a reference to some of the suggestions I will make, you might want to have a look at this thread, which illustrates many of the discussion points in another speech amplifier:
http://amfone.net/Amforum/index.php?topic=46696.0

Although there are many individual steps you may take to address every possible cause of hum, first it might be beneficial to identify the stage where the hum is originating.  Likely it is the first stage, a very high-gain pentode microphone preamplifier, but to be sure, a few quick tests are in order.  Starting at the last stages, you may short the grid pin to ground with either a test probe, screwdriver, or short wire, and monitor the output hum level.  If it disappears, then you know the hum is being generated in the grid circuit of this stage, or a previous stage.  Working back from the pair of 6C5 driver triodes, short the arm of the volume control feeding the other 6C5.  If the hum is gone when shorted, it is being passed from the plate of the 6J7.  Finally, short the input jack terminals, and monitor hum level.  It may not disappear completely, due to some wiring issues.

You mentioned the hum level changes as you move the filament transformer about near the amplifier chassis.  While it is prudent to ground one side of the filament circuit to the chassis in RF amplifier circuits, this practice can lead to hum problems in high-gain audio amplifiers.  

Another root cause of hum in audio is the use of the chassis as the ground return for each successive stage.  This hum is caused by different 60 cycle AC potentials at different points of the chassis, usually caused by alternating fields from nearby power transformers.  The chassis is acting as a transformer secondary winding, with different potentials all about the chassis.  

A clean, hum-free power supply is imperative.  Measure the hum level on your DC supply with a 10-1 scope probe, with the scope set for AC, not DC coupling.  Electrolytic capacitors have become larger in value, smaller in physical size, and lower in cost over the years.  The values used for power supply filters in the 30s and 40s may be considered marginal by today's standards.  Feel free to increase the filter capacitors by a factor of up to five, for much-improved hum control and dynamic stability.

It appears you have used a twisted-pair of wires to distribute filament power; this is the proper technique for audio circuits.  It also appears you did not ground either side of the filament source, which is also correct technique.  Lead dress is also an issue when working with low-level stages.  The proper method is illustrated on your wiring of the first 6C5.  The pair of wires should wrap around the outside of the socket, avoiding wires that capacitively couple hum to the signal pins of the tube.  Also correct is your method of extending the filament source from the 6C5 to the 6J7, as you have brought the wires from the 6C5 in parallel, around the same side of the tube, avoiding a current loop that could inject hum into that circuit.  The incorrect method would be to run the wires to the second tube from around the opposite side of the 6C5, which you did not do.  The twisted pair cancels the electromagnetic radiation of the filament wires, and it is best to run this twisted pair far away from other components, such as resistors and capacitors.  Preferred method is to run that pair in the corner of the chassis where the flat surface meets the chassis sides.  Incorrect wiring is illustrated as the wires are run over the top of the 6J7 socket, instead of around the socket.  It is difficult to achieve proper lead dress and interconnections with some of the older, heavier wire, so it is suggested that you try wiring all the filaments with smaller, more flexible wire, similar to the green filament transformer leads.

While it is always preferable to power the filaments of low-level stages with DC, it is not always practical.  Acceptable performance may be achieved with AC, and several other techniques may be used to minimize the hum level.  One is to connect the filament supply to the chassis through a balancing potentiometer, where each end of the variable resistor is connected to one of the transformer secondary leads, and the wiper is connected to the chassis.  This hum balancing pot is then adjusted for minimum hum.  Another potential cause of hum, especially in the lowest level stages, is an AC current between the filament and the cathode.  This effect is caused when the cathode is positive with respect to the filament, such that diode current flows, as if the filament is the cathode, and the actual cathode behaves as the plate of a diode.  One way to avoid this is to elevate the entire filament circuit above ground, with a positive potential of 25 or 30 volts.  This may be applied either to one side of the filament transformer secondary, or to the wiper of the previously mentioned hum balancing pot.  In may applications, this positive potential may be obtained from the cathode bias resistor of the output stage (not in your case, as the 6L6 tubes use fixed bias and the cathodes are grounded).  Another method is to add a low-current voltage divider from the speech amp B+ supply, and filter the tap with an appropriate capacitor.  If the filament is kept more positive than the cathode, there will never be diode current flow between the filament and the cathode.  All suggestions in this paragraph are finer points, to be applied after the root cause of the hum is corrected.
Wiring length and grounding is likely the culprit in your amplifier.  

The grid lead to the 6J7 cap is properly shielded, and properly grounded at the input jack end.  But it should be as short and direct as possible. so cutting and reconnecting the coax is recommended.  All wiring in the lowest level stages should be as short and direct as possible.  Coupling and bypass capacitors that have a "banded" end, the band indicating the outer foil layer, should have the band end either connected to ground, or, if a coupling capacitor, the banded end should connect to the lower impedance driving source, and NOT the higher impedance grid end of the circuit.

I have never achieved success controlling hum levels in a high-gain amplifier when using the chassis for a signal ground return.  Either a ground bus wire, or a single, insulated ground wire, may be used to provide the ground return for all the audio stages.  This technique avoids any potential differences adding to the signal content impressed on the grid of each stage.  The ground should connect to the chassis AT ONE AND ONLY ONE point.  This is typically the point of the lowest signal level, which in your case is the microphone input jack.  As illustrated in the attached photo, it is suggested that you consider running a piece of #12 or #14 solid copper wire, grounded at the mic jack, and passing over each tube socket, supported along the way as necessary by terminal strips.  Now the grid return resistor, cathode resistor, and any bypass or decoupling capacitors, may be attached above each tube socket to this noiseless ground bus.  The bus proceeds to follow the signal path to the last stage, and it is at this point that the power supply ground is connected.  This connection prevents the varying current from the output stages from impressing a feedback signal on the ground bus, preventing the corruption of the clean ground at the input stages.  This technique alone can resolve many of the hum problems when using AC on the filaments.  Have a look at the linked thread above for details and resulting hum measurements when using this technique.  

I love the octal tubes and the old-school techniques.  I hope some of these suggestions will help you resolve your hum issues!  Good luck and have fun.  If it worked perfectly the first time, you would not learn anything, but identifying and resolving these problems yields great satisfaction in this hobby!

73, Rick

[wa2fxm]

Wow. Thank you Rick for such a detailed explanation of AC hum problems and your suggestions for resolving them. Re-doing the filament wiring was in the back of my mind. And I figured the 6J7 grid lead would be shortened up in the final build. I have to decide if I'm ever going to find a proper chassis for the rebuild or go with two smaller chassis' and start the rewiring now on this one. I've never taken a close look at the Heathkit IP-17 output so I will put that on the to-do list. The filament transformer is center-tapped to ground. I tried using a balancing potentiometer across it by lifting the ground. It actually made a very slight difference. But as you said that's one of the finer points. I'll start with re-wiring the filaments. Then see about adding a buss bar ground. Thanks again for your suggestions.

Mark - WA2FXM

[w8khk]

Hi Mark,

You are very welcome, I am happy to help.  Starting with the filament wiring is a good plan, but do not be discouraged if that one step does not resolve the issue.  Those octal tubes are easy and fun to work with, and they sure are pretty when it all comes together.

One thing I totally forgot to mention.  It is possible that a bad vacuum tube is causing the problem.  Although it is very uncommon, it is possible that there may be a cathode-to-filament short in one of the tubes.  A tube tester with short-test capability will isolate this issue.  A short there will cause the grid-cathode signal to be modulated with the 60-ccycle hum.  I ran into that problem on an HRO receiver, a new tube fixed it; I wondered why I didn't think to try that first. 

I wouldn't worry about a chassis until you resolve the issues and the amplifier functions as you desire.  When I was a little kid, I remember my dad built a high-quality Williamson audio amplifier on a scrap chassis, and did not purchase a new chassis with an enclosure until he got it running perfectly.  That way there were no extraneous holes, scratches, etc. in the new chassis when he finished the project.  By the way, the amplifier I built in the thread referenced in my first response was built on a used chassis with lots of holes, but by carefully planning the layout I was able to re-use almost every hole, and all the punched socket holes.  Plenty of time to make it pretty after the experimenting is finished.   You will often find very good deals on new or almost new chassis at ham fests.  New chassis from distributors are obscenely expensive these days!

Rick

[wa2fxm]

Working on eliminating the AC ripple in the modulator I redid the filament wiring and even added a grounding bus but I was getting conflicting results when switching around between various power supply connections. I finally found a good 50K dropping resistor for the UAT-2 power supply so I installed it and decided to concentrate on getting it to work. I'm getting decent voltages but I'm not sure about the ripple and noise that I'm seeing.

The plate supply across the dropping resistor is 550VDC but showing a HF 20mV ripple riding on a varying 150 -> 400mV ac signal. The screen supply off the resistor tap at 460VDC shows the same but slightly smaller signal. The bias supply had 8uF paper filter caps which I replaced with modern electrolytics. It was running at -25VDC with a solid 5Vpp 60 Hz ripple on it. When I upped the electrolytics to 40 and 47uF the ripple dropped to 0.24Vpp and the bias went up to -30VDC.

1. What am I seeing on the two HV lines? Is that "normal" noisy (rural) AC house voltage? Or are these small "normal" variations that won't effect the audio quality?

2. It's still not clear to me how the bias voltage is made. Without totally tearing the power supply and transformer apart I can't be sure about the windings and center taps being used for the bias voltage. Do my dotted lines on the schematic represent how it should be connected? I'm assuming the original design calling for 8uF filter caps was good enough to market a decent sounding modulator. Both modern 8 uF electrolytics are good. Is 5Vpp of ripple on the bias supply really "good enough"? Or should I be looking for something amiss in the bias supply wiring?

Mark - WA2FXM

[K8DI]

The HV ripple is fine. Convert it to a percentage, 400mV is .4V/550V is 0.07%. It could be ten times bigger and still be fine.
But the bias…. First, something like one of your red lines has to exist, or it wouldn’t work. The filament(cathode) has to connect to something like ground. You can take an ohmmeter from filament to chassis with power off…it should be a few ohms at most.  Id guess the windings look more like taps near the ct of the HV winding.  Also, a 40uF capacitor input filter on a type 80 is way too high. You’re killing the tube.  Peak current will exceed the ratings.  There’s far less issue with adding capacitance on the HV due to the swinging choke, that makes it a choke input filter.

Ed

[wa2fxm]

Measuring at the 80 socket pins across the filament winding I get 0.2 ohms, and 10 ohms to ground from each pin. Of the 15 leads coming out of both sides of the power transformer 2 go directly to chassis ground. Both of them have solid connections.

I've used 4 different 80 tubes with no difference between them.

I disconnected the choke and it measures only 3.5H. The Utah catalog says a 4661 should be 10H. With the 8uF capacitors back in I clipped in a spare 16H choke and the ripple is down to 1.5Vpp. I put two of them in series and it went down to 0.9Vpp. 

I have in my notes (which means I copied it from somewhere) that "normal" ripple in a Class B modulator is under 1%. With the 16H choke I'm down to 6.3% and 32H is 3.6%. But again I have to assume the Utah engineers chose their component values to produce a modulator that worked decently.

 What am I missing, or misunderstanding?

Mark - WA2FXM

[w8khk]

Hello Mark,

I just read your recent posts, and I had some questions regarding the origin and accuracy of the bias supply section of the power supply.  Your latest post may have cleared that up.

I also read Ed's post, and I agree with most of his points.  Your questions about the dotted lines to the bias supply filament transformer are valid.  Grounding the center tap would provide no output at all from the bias supply, because both the cathode (filament) of the 80 rectifier, and the positive return of the filter caps and bleeder also go to ground, therefore there would be no source of AC, other than filament voltage.  It would be very inappropriate to connect the filament winding to either end of the high voltage winding, for two reasons.  First, with a high voltage supply in excess of 500 volts, the potential from the center tap to one end of the HV secondary winding would far exceed the safe value for the 80 rectifier and the bleeder of 1000 ohms, and the 450 volt rating of C4 and C5.   IF THE CIRCUIT IS CORRECT AS DRAWN there would be no output from the bias supply, and the only way to resolve that without an additional transformer would be to couple one end of the HV winding to the 80 filament winding via either a large-value resistor, or use a capacitor such that the reactance would drop the voltage with minimum dissipation and heat generation.  BUT since you do have output from the bias supply, and your resistance measurements reveal 10 ohms resistance from either 80 filament pin 1 or 4 to ground, I suspect the schematic diagram is incomplete, and there is likely a winding within the transformer, from one end of the filament winding to one of the two grounded leads, thus providing the requisite AC source for the bias supply.  To verify this assumption, you might wish to remove both the 5Z3 and the 80 rectifier tubes, then power up the transformer, and VERY CAREFULLY measure the AC voltage from the chassis ground to the 80 filament on either pin 1 or 4 of the 80 socket.  TO DO THIS SAFELY, I would use clip leads to pre-connect the meter, and do not touch anything while power is on; just view the meter and I would suspect you would see between 100 and 200 volts AC between those two points.   This would confirm the missing winding in the schematic, and negate the need to make any other connections to the filament winding that powers the 80 rectifier.

The other critical issue of connecting the 80 filament winding to one end of the HV secondary is the fact that the HV is switched OFF via a switch between the center tap and ground.  This works fine when the winding only provides positive high voltage.  But if that winding also has a path to complete the negative supply section, lifting the center tap from ground WILL NOT disable the HV output, rather, it will still provide output but with the center tap free, the negative bias supply output will be increased substantially, further aggravating the lower voltage filter cap and the small 1K bleeder.  So, it retrospect, it is very fortunate that the "hidden winding" not on the schematic is providing the requisite bias AC source to the 80 rectifier.

I concur with Ed's assessment of power supply ripple, and your hum problem likely lies elsewhere in the circuit.  As I suggested previously, it might be helpful to determine from which stage in the chain the majority of the hum is originating, by selectively grounding the grid of the 6J7 and the intermediate 6C5 driving T1, then you could even ground the primary of T1 at the right side o fC6 without causing any harm.  If either the 6J7 or the 6C5 stages are causing the hum, look at both the ground return for R1 and R10 path for ground loops, as well as measure the ripple at the top of decoupling caps C3 and C6, and screen cap C2, all of which could allow hum to be amplified.  Also, do not forget to check the tubes for internal cathode to filament shorts.  This will cause hum, and no attempt to cure it will be successful, short of tube replacement.  Sometimes the easiest cure is not attempted until the end!

In response to your question about the red dashed line at the top of the power supply schematic, there are three possibilities, which I alluded to in a previous post.  First, if there is no ground connection to either side of the filament winding, common practice is to ground the center tap of the main amplifier filament winding, thus balancing the AC voltage on each filament wire to ground.  A refinement of that would be a 100 to 500 ohm pot, with the ends tied to the filament winding, and the wiper grounded, which can be adjusted to balance for minimum hum.  A third improvement along the same lines, would be to connect the pot wiper to a slightly positive voltage, thus raising the filaments positively above the cathode, thus eliminating any AC current between the filament and the cathode (diode effect) when the filament is in a negative transition with respect to the cathode voltage.  In most cases, when using cathode bias, the cathode is positive with respect to the chassis, and that makes for a conducting diode between filament and cathode when the filament is more negative than the cathode.   This slight DC bias for the filament hum balance pot may be easily obtained by adding a 2K or 3k ohm resistor between the bottom of bleeder 50K R1 and ground, then pick up the positive voltage for the pot wiper from the junction of the two resistors.  

I hope some of this is helpful to you!
Rick

[K8DI]

Some schematic clarification.  Attached ar two photos of quick hand drawings. The first is the two ways to make a simple positive supply, and the second is for negative (bias supplies).  Note that one of the negative versions requires a tube with actual cathodes rather than the directly heated cathode/filament as cathode more typical of diodes, like a 6X5.

In all cases, there are two windings. One is filament and one is the ac voltage you’re rectifying.  It could be that that winding is part of or taps on a winding with other purposes, as well.  

Rick is likely correct that your schematic is incomplete or wrong on how the bias supply is wired.

As far as ripple, with the amount of filtering you are doing, there should be much less. I will add more ideas in another post, I just got called into Sunday breakfast.

(Editing to add/get back to after breakfast...)

also attached is a pdf of the output from the Duncan Amps PSU simulator. This simulator is generally respected, although somewhat limited for output options.  I ran the stock component values, then the 40uF with the stock choke, then the 40uF with the 32H.  I assumed a HV winding resistance of 20 ohms, based on your resistance measure. This could be off, but would not make a huge difference.  The most filtered version shows 4 millivolts ripple on 30 volts DC -- 0.013%, or, none for any practical purpose. That is not at all what you are reporting as measured, which to me says either some component (including the transformer) is bad, or something is not connected. So... as I said, there should be much less. 

Besides figuring out the actual schematic, some investigation as to the condition of the transformer is warranted -- for reference, as I have no idea, is 30v bias in the ballpark for this rig? should it be much higher/could it have shorted turns in the transformer?
Ed

[wa2fxm]

Ed and Rick,

From chassis ground I measure 61VAC from one 80 filament pin and 54VAC from the other with both rectifier tubes out. With no 5U4 tube and the 80 in there's a 40VAC sawtooth on the input filter cap and 5VAC ripple on the output at -25VDC. So, there must be a separate 60 volt winding on the transformer being used for the bias supply? And as to whether the proper bias voltage is even being produced, RCA manual says  -22.5V grid bias at 360 plate volts for 6L6 AB2 push pull service. And the 1940 ARRL Handbook says -25 volts at 400 plate volts. That sounds like ballpark to me.

Mark - WA2FXM

[W7TFO]

Once you get it running, an easy way to see distortion is to use a dual-trace scope.

Input to the modulator goes to trace one.
 
Output (with suitable attenuators) goes to trace 2.

Run a sine wave thru the properly loaded modulator, and superimpose both traces on the scope.

Distortion will manifest as fuzzying or splitting of the pattern. 

A useful realtime tool for diagnoses of problems.

73DG

[K8DI]

From the various descriptions and measurements and such, my best guess is the attached schematic, for the stock setup. Simulating this has a bunch of ripple, 4%.

After running sims and messing with values, I’d head here…

First cap, a new 10uF 50v part. Choke, you said you had some 16H around. Use one as long as its dc resistance is between 100 and 500 ohms. For the second cap, 100uF at 50v. Adjust the bleeder to make the output 25 volts. To a degree, the bleeder and the choke act like a voltage divider. As far as ripple goes, the sim, with a 1500 ohm bleeder, is 0.3%. If you build this and it’s still got a ton of ripple, then the only conclusion I’m left with is that the transformer is bad.

To Dennis’ comment on the scope, comparing the in and out…it’s kind of important to do as he said, properly load the output…. I have a handful of 1k, 100w and one 500 ohm 50 watt resistors that can be arranged to make most modulator loads. The nice part of using resistors for the load is there’s no HV on the output to contend with in the measurements, versus measuring in place while modulating a final. You can also overdrive the piss out of it without causing RF mayhem or unloading the output, as happens when 100% negative mod occurs.


Ed

[wa2fxm]

First cap, a new 10uF 50v part. Choke, you said you had some 16H around. Use one as long as its dc resistance is between 100 and 500 ohms. For the second cap, 100uF at 50v. Adjust the bleeder to make the output 25 volts. To a degree, the bleeder and the choke act like a voltage divider. As far as ripple goes, the sim, with a 1500 ohm bleeder, is 0.3%.

Jeesum Crow Ed, you nailed it.

I pulled the bad choke, and outboarded the (too big) Stancor 16H. I replaced the filter caps with 10 and 100 uF, and left the 1K ohm bleeder in. At 117VAC on the Variac: 100mV ripple with -24.9V at Pin 2 of the output plug. By my math that's 0.4% ripple. Where can one find this PSU simulator software?

Mark - WA2FXM

[K8DI]

Where can one find this PSU simulator software?

https://www.duncanamps.com/psud2/index.html

It’s older software. It runs fine on Windows 7. I’ve not had a reason to try it on 10 or 11.

Ed

[w9jsw]

Runs fine on 10/11...

[wa2fxm]

Do I need to rebuild/replace this grid cap on the 6J7? I'm getting an audible and measurable hum at the 6J7 plate. If I touch the braid the hum almost disappears. If I touch the cap the hum goes way up. The grid cap is from the junkbox and doesn't seem to have any metal shielding. I soldered in a piece of shielded audio wire to the interior metal clip that connects to the tube grid contact. I added the braid on the outside of the wire to bring the wire shield closer to the cap. The shielding is connected to a copper bus bar that I added that connects at one end to the mic input ground lug. Shouldn't the cap also have some metal shielding? Or should I be looking somewhere else to reduce the hum that's making it's way to the grid?

Mark - WA2FXM

[W7TFO]

There are vintage caps that snap on the top of a metal tube like a 6J5, to offer shielding for the grid cap.  I've seen silver and black ones.

73DG

[w8khk]

Hello Mark,

I looked at your photo, and the bus bar looks like a good improvement to the grounding in your amplifier.  I assume the bus bar connects to the chassis only at one point, preferably to the lowest level input point, which would be the microphone connector on the front.  

In order to minimize the hum generated in a high-gain amplifier stage, such as the 6J7 pentode, it is also important to ground all the components that affect the cathode to grid signal to this stage, to a common point. That would of course be the ground bus, with the following components grounded to the bus as closely together as possible.  An inch or two either side of the tube should not matter.  Grid return R1, cathode return R2, cathode bypass C1, screen bypass C2, and plate supply decoupling/bypass C3, are these critical components.  Since the grid resistor is 5 megohms, it may be difficult to eliminate residual hum when no source is connected, but with the microphone connected with a shielded cable, the residual hum should be reduced.

I am not sure why you are seeing less hum when touching the shield of the grid cable.  Perhaps something is not securely grounded?  A great deal of additional hum would be expected when you touch the grid cap.  From what I can see in the photo, you are doing everything correctly.  Shielding the grid lead as closely as possible to the grid cap terminal is the proper procedure.  While grid cover shields were available, they are not absolutely necessary to remove all hum.  

I offer a graphic example of a preamp my father built in 1953, which originally had three mic inputs and one phono cartridge input.  In the 60s, we later modified it to provide four mic inputs, as well as mono and stereo output mixes.  It was used many times each year for recording concerts and musicals, to Ampex tape recorders, and no hum or other issues were ever a problem.  I attached a front view, as well as a top inside view.  You will observe the input stages are all 6J7 tubes, the two on the left have the press-on shields, while the other two do not.  The shield is grounded via the metal tube shell, which should be grounded by pin 1 at the tube socket.

The top cap connectors use shielded coaxial cable, with the shield connected to the bus wire, and a separate bus is used for each channel, each grounded only at the microphone input connector.  The grid wire connects to the toggle switch you see behind the tube, and this allows the input to be connected to either the mic input matching transformer for balanced low-impedance mics, or to either a TRS or Amphenol single-ended input connector on the back of the chassis.

There were no power transformers in this pre-amplifier, as all power was derived from the power supply in a 20 watt Williamson.  The pre-amplifier was fully enclosed in a steel cabinet,  No trace of hum was ever experienced with this configuration.

These 6J7s were wired as triodes, by tying the screen grid, suppressor grid, and plate all in parallel.  The next stages were also 6SJ7s, wired as triodes as well.  Even though much less gain is achieved as a triode, cascading three in series provided lower noise, and lower distortion, for high fidelity applications.

If you wish to try an alternative configuration, avoiding the grid top-cap on the 6J7, you might consider using a 6SJ7 tube as a substitute.  Electrically, the 6SJ7 is an equivalent replacement for the 6J7, with the "S" in the nomenclature indicating the tube is "single ended", meaning all connections are made with base pins, no top cap required.  Unfortunately, changing to the single ended tube, in this case, means that all pin assignments are different except pins 1, 2, and 7.

I don't know if you are testing with a mic connected or the input open.  You might try plugging in a shorting plug, or a plug with a resistor of 47K or less.  With this plugged into the input, if the hum is gone, then the hum is either coming form the open circuit input, or the mic and/or cable.  If the hum remains with the input shorted by zero to 47K, then there is still a problem with the input stage.  It may be an electromagnetic field from a nearby power transformer, as the chassis is not shielded for hum.  This hum would be caused by an AC potential between the 6J7 cathode and grid pins, incompletely filtered screen or plate voltage, or the return filters for screen or plate voltage not at the same ground potential as the cathode or grid return resistor.  It would not hurt to try paralleling additional capacity to C2 or C3, while monitoring the hum, with the grid input terminated.  It looks like you have made significant progress and are getting close to completion.

73, Rick

EDIT:  Looking again at your last photo, it appears the shield braid is only a couple inches long, and does not extend the full distance of the grid lead.   In order to minimize hum, the shield braid MUST cover the entire length of the grid lead, from the top cap insulation, all the way to within a fraction of an inch of the input connector terminal.  The shield braid then should be securely connected to the ground bus or input connector ground terminal by the shortest possible path.  Please try this and let us know how it works for you.

[wa2fxm]

I don't know if you are testing with a mic connected or the input open.  You might try plugging in a shorting plug, or a plug with a resistor of 47K or less. With this plugged into the input, if the hum is gone, then the hum is either coming form the open circuit input, or the mic and/or cable.  

With a 47k ohm shorting plug in the mic connector the ripple on the 6J7 plate does indeed disappear. Down to 100mV.  I have not been using a mic with a shielded cable for testing. I've been using a 400Hz 20mV input signal from a Siglent signal generator with an oscilloscope probe hooked onto an unshielded 1/4" plug in the mic connector. With this setup the hum was audible under the decent 400Hz tone at the 6L6 grid socket contacts. At the 6L6 grids I measured:

Filaments On:  100mV clean 60 Hz ripple
HV On - No Signal - Gain Min: ~200mV ragged hash ripple
HV On - No Signal - Gain Max: 10Vpp    "     "     "
HV On - 400Hz 20mV Signal: Hum clearly heard under good 400Hz tone.

I switched the input cable to an old piece of 75 ohm video coax with a cluster of adapters at the mic input end. While the no signal hum was still definitely there, it was inaudible with the 400Hz tone.

Filaments On: 80mV clean 60 Hz ripple
HV On - No Signal - Gain Min: ~100mV ragged hash ripple
HV On - No Signal - Gain Max: 6Vpp   "     "     "
HV On - 400Hz 20mV Signal: No Hum heard under good 400Hz tone

The no signal hum was definitely variable if I touched the chassis or moved my hand around the 6J7 grid cap. So I added a tin foil hat for the 6J7. This reduced the No Signal - Gain Max ripple down to 2Vpp, clearly reducing the audible no signal hum.

Measuring at the 6J7 plate
Filaments On: 3mV clean 60 Hz ripple
HV On - No Signal - Gain Min: 1V ragged hash ripple
HV On - No Signal - Gain Max: getting audio howling in the speaker
HV On - 400Hz 20mV Signal: 1.8Vpp clean 400Hz signal with absolutely NO ripple.

Also, to clarify some other points, the copper bus bar does indeed connect to the chassis only at the ground lug of the input connector, and ALL components with chassis connections are tied only to the bus bar. And the 6J7 grid lead shield DOES extend the entire length of the wire. The braid at the cap end is soldered to the coax shield which extends all the way to the input connector and is soldered directly to the input connector ground lug. The braid was added to cover the unshielded stub of wire that came with the cap.

So now I have 100mV of ripple, amplified to 6V with the gain advanced, which I can't hear when the audio is switched on. Is this "residual hum" that you mention, or should I be doing better than this?

Mark - WA2FXM

[w8khk]

Mark, I believe your modulator is working fine and there may not be any hum problem.  You have not yet tested with a microphone, or determined at what level you need to set the volume control to achieve the desired modulation percentage.  With the extremely high input impedance, and the high gain of the amplifier with the level control advanced, some hum is to be expected.  The fact that hum is gone when you bridge the input with a 47K resistor indicates that the hum is NOT generated in your amplifier, but it is coming from outside sources when running the grid at five megohms input impedance.

The fact that you hear more hum when placing your hand near the grid cap or chassis indicates that you are providing a capacitive connection to a high impedance input, and the hum in that situation is to be expected.  I apparently did not need the grid cap shields on my father's amplifier for two reasons.  First, the input impedance was around 50K, the output of the microphone matching transformers.  Second, the preamplifier was fully enclosed in a metal shield, avoiding most capacitive and inductive ingress of hum at the input grids.  Had I use the amplifer without the shield cabinet, near the power supply, I expect I would experience the same hum you are.  

The Siglent is a fine instrument, and the 60 cycle hum of course is not originating in the test instrument.  But it is possible to entertain a ground loop, and hum, when attenuating the signal to the minimum level at the generator, then running the cable to the mic input, in parallel with the scope probe.  In order to avoid that problem, a common practice is to assemble a resistive attenuator, or pad, that reduces the generator signal level around 100 times, and place that attenuator right at the mic input.  A couple resistors at the mic plug, at the end of the coax cable from the generator, would let you run the generator at 100 times the necessary output level, and attach the scope probe right at the generator output, not at the mic input.  I expect this would reveal a clean 400 cycle signal, hum free, indicating the amplifier is working properly.  

When I test a modulator, separately from the RF final amplifier, I usually configure a resistor, equal to the modulating impedance of the RF final amplifier, the resistance being the final plate voltage divided by the final plate current.  Attach this resistance, of adequate power handling capability, to the secondary of the modulation transformer.  You may first test with the Siglent generator attached to the mic input with the attenuator plug.  Assuming this test is sucdessful, attach your microphone and determine whether you have hum at the output, using your scope at the output of the modulation transformer. If you still have any residual hum, try positioning the power supply as far from the modulator as possible, and, if necessary, try enclosing the modulator chassis in a metal chassis or box.  These tests will reveal whether you have sufficiently reduced any residual hum.

It has been common practice when using low level, high impedance microphones, to mount the first pentode preamp stage in a small metal enclosure, at the operating position, away from the modulator and power transformers.  The handbooks have many examples of this method.  You will likely not need to resort to this method, if the output level of your microphone is high enough to run the level control low enough to avoid any noticeable hum.  If you are using a D-104 that is in good condition, you will likely have plenty of mic output voltage such that hum may be a non-issue, even with no shielding of your modulator chassis.  I think it is time to move your testing to the full modulator, with mic input, either with the resistive load or the final amplifier attached.  If you choose to monitor the final amplifier, even with a dummy load, on your receiver, you may experience hum that is NOT from your modulator or transmitter, due to receiver input overload. Your transmitter and modulator may be just fine, but the monitoring method could exhibit hum that is not on your transmitted signal.

[wa2fxm]

Well, it would appear that the hum is residual. Maybe. I finally plugged in the 6L6's and strapped them to the modulation transformer and a spare meter. The transformer is a Thordarson 8470 as called for in the ARRL modulator schematic (BTW this modulator is actually a George Grammar W1DF design from the June 1936 QST). It's a "special job" as they used to say, designed for the 6L6 input. 3800 ohms plate to plate and then your choice of 2.5k, 5k or 7.5k output. The primary seems to have two separate windings so I bridged the two terminals labeled "B" and put the two "P" terminals to the 6L6 plates. When I was running the UAT-1 on CW the average, nominal plate input to the two parallel 6L6 finals was 400 volts at 125 mA = 3200 ohms. I put a variable wirewound resistor tapped down to 3200 ohms across the 2.5k output winding. With the HV on the modulator plate current idles at 60 mA. With the 5k-8 ohm speaker transformer across the secondary, and a resistor pad across that for headphones the 400Hz tone from the signal generator was clean with no hum but with a rushing white noise sound in the background which is also audible with no signal input. I then hooked up the D104 and there was a "slight" hum in the background along with the same white noise. Residual or not you can hear the hum along with the background room noises of window fans and the cat coming up the creaky stairs to the shack.

All of this is with the components thrown together on the bench. I'm hoping that once they are installed in the metal cabinet perhaps the noise and hum will dissipate? I also have not really been paying attention to voltage and signal parameters. The modulator sounds "loud" in the lashed up headphones. With the scope I'm easily seeing +400 V P-P on the output with the volume control not even half way up. I will turn the thing upside down, take some voltage readings and report back.

[wa2fxm]

So. I have a "functioning" modulator based on the June 1936 QST design of George Grammar. All components follow the audio deck schematic. The only differences are my placement of components on a smaller chassis and the use of a different power supply, being an original UAT-2 supply. There has been slight audible hum on the signal to which was added random noise once the modulation transformer was hooked up. So far all testing has been done with a lashup on the bench outside of the chassis. Thank you to Rick, Ed, Dennis and everyone else who has helped me along so far. I am now getting neck deep in this project and appreciate the vast store of knowledge that everyone has been willing to share.

Here is the result of measurements I've taken using the Siglent audio input, all measured from chassis ground. Grammar says "0.005 volt being sufficient to drive the final tubes to full output", so I chose that as an input level. These measurements don't really matchup with his listing of proper AB2 operating conditions. Am I still on the right track?

Mark - WA2FXM

[wa2fxm]

Well apparently I am on the right track. Made my first contact this afternoon on 40 meters down to southern New Hampshire about 120 miles away. The signal was S7 and doing "quite well". But the takeaway quote was "no distortion in the audio". I am more than elated. So now to tie up the loose ends. There's a Tone Control that needs to be added. I'm guessing it was a simple pot and capacitor "bass cut" affair like the ones used in 1930's PA systems like Operadio. I'm going to skip the Mixer No. 2 pot and the Record and 200 Ohm jacks for now. I'll dummy them up until I find an original UAT-2 audio deck. Or maybe just an original schematic to work from. Some day. I need a 500k audio gain pot with a proper shaft to replace the junkbox screwdriver shaft one I've got now. The audio gain right now is pretty touchy. Very easy to crank the gain too high. Is this most likely because it is not an audio/linear pot? Or would a change in value be in order to spread out the gain control? 

Mark - WA2FXM

[Opcom]

Could be a tapered pot. Some pots have a reverse taper as well.

An old trick for working it out is to have two pots, one in front of the other. If the audio's too touchy but its in the right area, then use the other pot to have an additional degree of granularity. You can then measure and get an idea what value pot you need so the wiper's in an easily-adjustable area. At some point hey just try another pot. A 10 turn is also helpful when debugging.

Real nice looking setup by the way!

[wa2fxm]

Time to wrap this thread up. The modulator is officially done (for now). It had it's maiden voyage this weekend on the AWA PM/AM Net on 75 meters. Only 18 watts output but it was powering all the way down through southern New England before sunset. Once the sun was down it was even making it down to SE PA. All reports said the audio was good. I want to thank everyone who chimed in here to help me along with this project. The amount of stored information in your collective heads is priceless. I hope having it all out here in this forum will help future wannabe modulator builders. So, moving on I will be starting a new thread shortly. I acquired this UAT-4 RF Amplifier at NEARFest in April. Once again I will need all the help I can get to bring it back to life.

tnx agn es 73 de Mark WA2FXM