Modified Heising blocking capacitor question

[KE6DF]

This question concerns the plate blocking capacitor between the modulation transformer and the modulation reactor.

I was looking at the schematics of a couple of 250 watt broadcast transmitters -- the RCA 250L and 250m.

The 250L uses a pair of 828s as modulators in AB1. And a pair of 810s as RF finals.

The 250M uses a pair of 813s also AB1. With 813s as RF finals.

In the 250L, the "bottom" of the secondary of the mod transformer is connected to the blocking capacitor, and the other side of the blocking capacitor is connected to ground. The "high" end of the modulation transformer is connected directly to the final plate lead coming out of the modulation reactor. A 12 MFD capacitor is used.

In the 250m the "bottom" of the secondary of the mod transformer is connected to ground. The other end of the mod transformer is connected to the blocking capacitor and through the capacitor to the plate lead of the modulation reactor. This capacitor is specified as 1 MFD.

Unless I'm missing something, it seems like these two circuits are electronically similar as far as the AF is concerned.

So why the big difference in the blocking capacitor -- 12MFD vs 1MFD?

I've always heard numbers in the 1-2 MFD range.

Is this capacitance just not critical?

Clearly one would expect you could get by with a lower voltage rating in the case of the 250L, but why the difference in capacity?

Would something like a 4MFD capacitor work in both cases?

Thanks
Dave

[W1RKW]

2 to 4mfd should be more than enough.   The other thing is you don't need a cap that has a tolerance of thousands of volts.

[N2DTS]

When I tried this setup, the capacitance had a large effect on the audio sound, every amount of capacitance sounded different.
I gave up and just use the RCA mod iron with current through it...


Brett

[k4kyv]

When I tried this setup, the capacitance had a large effect on the audio sound, every amount of capacitance sounded different.
I gave up and just use the RCA mod iron with current through it...

I'm sure that made it sound different, too.  Did you try running a frequency response test with each of the various capacitances vs running current through the transformer?  Most broadcast modulation iron is not designed to carry DC through the secondary winding.  To do so tends to saturate the core and reduce low frequency response. It could make the mod xfmr more prone to crap out, as well.

I have tried anywhere from 2 to 16 mfd in mine, and couldn't tell any difference in frequency response.  Some have said they got a better low end with less capacitance.  But I don't think it is critical.  My HF-300 rig now uses 4 mfd.

With the blocking cap between the bottom end of the mod xfmr and ground, an excessively large capacitance could generate a transient every time the plate voltage is turned on.  The bottom end of the xfmr is effectively grounded by the capacitor until the capacitor is charged, and that charging current generates a voltage spike across the inductance (the secondary winding), which is induced back to the primary winding.

This problem can be eliminated by returning the blocking cap to the +HV of the plate supply, but then, the audio is returned through the output capacitor of of the power supply.  This effectively reduces the coupling capacitance, since the blocking cap and power supply cap are in series.  And any transients, noise or residual ripple that appear at the power supply output will modulate the final.  Returning the cap directly to ground allows the audio circuit path to be completely independent of the HV supply, and allows the mod  reactor to act as an additional element of power supply filter to further reduce ripple and noise output from the supply.

If the capacitor is returned to ground, it should have a DC voltage rating similar to that of the power supply filter capacitor, since the full DC voltage appears across it once it is fully charged.  If it is returned to the +HV, theoretically it doesn't need much of a voltage rating, since the only voltage to appear across it is whatever audio voltage drop might appear across the capacitive reactance plus the DC voltage drop across the resistance of mod reactor, but I would still use one with a substantial voltage rating for a safety margin in case a low AF transient somehow occurs.

[AB2EZ]

The best low end flatness is achieved as follows. [This method also avoids having a high peak in the low end response at very low frequencies (subsonic), that could lead to various types of problems... including subsonic oscillation and damage to the components of the modulator]

Step 1. Calculate the modulation resistance, R, that is presented by the r.f. output stage to the incoming modulating signal. R = average plate voltage / average plate current. For example, if the plate voltage (B+) is 500 volts, and the plate current (at carrier) is 100 mA, then the modulation resistance is 500/ 0.1 = 5000 ohms.

Step 2. Calculate the frequency at which the reactance of the Heising inductance equals R (calculated in step 1). The Heising inductance is the inductance of the Heising reactor in parallel with the magnetizing inductance of the modulation transformer.

For example, the modulation inductance of the Ranger's modulation transformer is about 7 Henrys. If you use a Heising reactor having a value of 20 Henrys, then the parallel combination has a value of 5.2 Henrys.

The frequency at which the reactance of the Heising inductance equals R is given by f= R / (6.28 x L).

For example, if R = 5000, and L = 5.2 Henrys, then f = 5000 / (6.28 x 5.2) = 153 Hz.

This frequency represents the lower 3dB rolloff frequency of your modulation if you use a large capacitor (see the next step)... but you will get a lower low-frequency cutoff (by a factor of about 1.4) if you use the optimal capacitor (see the next step).

Step 3. Having obtained f (from step 2), calculate the value of capacitance that has impedance R (from step 1) at frequency f.

C = 1 /  [6.28 x f x R]

For example, if R=5000 ohms and f = 153 Hz, then C = 1 / [6.28 x 153 x 5000] = 0.2 uF

If you choose C to have this (small) value, then the frequency response of the modulation will be flat down to 153 Hz, and the roll-off will start at 153 Hz (instead of being 3dB down at that point). The parallel combination of the Heising capacitor and the Heising inductance will have a resonance at that frequency; and the Q of that resonant circuit (including R) will be 1.

If you use a larger value of capacitance, the resonance will occur at a lower frequency, but the Q will be much larger... possibly leading to a tendency toward subsonic oscillation... and associated damage to the components of the modulator.

In my home brew, plate modulated amplifier I have the following values:

Modulation resistance: 1700 volts / 300 mA = 5667 ohms

Heising inductance = 40 Henrys [A 50 Henry Peter Dahl Heising reactor in parallel with the 200 Henry magnetizing inductance of the Antek toroidal ferrite modulation transformer]

Frequency, f, at which 6.28 x f x 40 = 5667:  22.6 Hz

Required capacitance = 1 / [6.28 x 22.6 Hz x 5667 ohms] = 1.25 uF

I used the calculated value of capacitance in my modulated amplifier... and both the amplitude and the phase of the measured modulation frequency response matched the textbook calculations.

Stu

[WU2D]

I use a 2 uF Microwave Oven capacitor in my ARC-5 Modified Heising; stellar sounding AM I am sure, except nobody hears me because I am mobile...

[k4kyv]

If you choose C to have this (small) value, then the frequency response of the modulation will be flat down to 153 Hz, and the roll-off will start at 153 Hz (instead of being 3dB down at that point). The parallel combination of the Heising capacitor and the Heising inductance will have a resonance at that frequency; and the Q of that resonant circuit (including R) will be 1.

If you use a larger value of capacitance, the resonance will occur at a lower frequency, but the Q will be much larger... possibly leading to a tendency toward subsonic oscillation... and associated damage to the components of the modulator.

But how does this affect phase shift at lower frequencies, in the above case in the vicinity of 153~?  Without delving into the mathematics, it would seem that using the larger value would result in less shift at frequencies above resonance, whereas using the calculated value might give a smoother roll-off at subsonic frequencies but result in more phase shift as you approach roll-off.

One thing you want to avoid as much as possible is phase shift at the extremes of the low end, since this can alter the waveform and thus the amplitude of peaks, resulting in overmodulation and a canting effect on waveforms that resemble a square wave.

A certain amount of canting is inevitable unless the modulator is DC coupled to the final; neither a capacitor nor a transformer, nor an inductor shunted across the load resistance, can pass DC.  This problem often causes brute-force speech clipping to result in more overmodulation on voice peaks instead of less.

It would be interesting to run a sweep using a signal generator with various coupling capacitances and observe the frequency response and how low you can go before the sine wave begins to show noticeable distortion.

A good test for phase distortion is to reduce the plate voltage on the modulator well below its normal value (in the case of plug-in rectifiers in a full wave circuit, pull out one of the rectifier tubes or diodes).  Now drive the hell out of the modulator until flat-topping is clearly visible.  It doesn't matter if the bias is wrong for the lower plate voltage; all you are looking at is the crests of the envelope waveform.  Observe the scope as you sweep down in frequency, making sure the sweep rate on the scope decreases to maintain a full audio cycle on the screen.  Phase distortion begins to occur at the point where the flattened peak begins to tilt from the horizontal.  But do this with the modulator plate voltage well below normal; otherwise this is a good way to blow mod iron.

An alternative to the above test if the audio section up to the modulator grids has excellent frequency response well below the expected roll-off of the modulator, is to simply feed a square wave signal into the audio chain.

I have always taken the intuitive assumption that the larger the inductance in the reactor and the larger the blocking capacitor, the less the phase shift near the extreme low end.

[AB2EZ]

Don

I simulated the Heising circuit using LTspice IV, with a sine wave input. See attachment 1.

With C1 set at 10 uF, I lowered the frequency until the amplitude of the voltage across the load dropped from 50 volts to 35 volts (3dB down). This occurred at 65 Hz.

Attachment 2 shows the input signal (100 volts) and the output signal (35 volts).

The phase shift (the output leads the input) is about 2.5 milliseconds (59 degrees) with the 10uF capacitor

Then I changed the value of C1 to 0.5 uF (which produced the largest output for values I tried in the range of 0.3 to 1 uF).

Attachment 3 shows the input signal (100 volts) and the output signal (45 volts).

The phase shift (the output leads the input) is about 4 milliseconds (94 degrees) with the 0.5uF capacitor

Best regards
Stu

[AB2EZ]

Here are the LTspice IV simulations using square waves

For each attachment: top = 10uF capacitor  bottom = 0.5uf capacitor

Attachment 1: 300 Hz square wave
Attachment 2: 160 Hz square wave
Attachment 3: 65 Hz square wave

[Opcom]

Does the inductance of the modulator iron, loaded as it is with the resistance of the modulator tubes and the resistance of the PA (albeit coupled through a capacitor) have ann effect on the calculation and the ultimate frequency response, or is this effect negligible because the mod xfmr looks like a resistor to AC by virtue of its coupling to the modulator and PA tubes?

[AB2EZ]

The effect of the mod iron, at low frequencies, is represented in the calculation by the magnetizing inductance of the modulation transformer (shown as 7H for the case of a Ranger).

[K5UJ]

I'm looking at the schematic of the Sparta 701 I have to see how they did it.  This is a 1 kw rig with pair 4-500 in RF and modulator.  They operate it with the bottom of the mod. trans. secondary grounded through a 8 uF 4 KV cap.  The B+to the finals goes through a 65 H 500 mA reactor.

Rob K5UJ

[k4kyv]

Stu's charts and graphs seem to confirm what I said about phase shift.

@ 300~, the peak-to-peak amplitude of the square wave, as shown by the lengths of the vertical sides of the waveform, is 49 volts.  With 10 mfd of capacitance, the canting effect, which appears as the slanted portion of the top and bottom of the waveform, offsets the initial rise and fall of the square wave by about 16 volts.  With 0.5 mfd, the canting increases to about 25 volts.

The effect is more dramatic at 65~.  With 10 mfd, the canting effect is about equal to the p-p rise and fall of the square wave, so that the sides of the square wave are completely offset by the canting effect.  With 0.5 mfd, the canting effect (66v) exceeds the rise-fall of the square wave (38v) by about 74%, and the resulting waveform has little resemblance to the original square wave.

Using a large coupling capacitor allows the low frequency response to roll off below a certain point due to the capacitive reactance in series with the load, and the combined inductive reactances in parallel with the load.  Reducing the coupling capacitance to the point where the circuit becomes broadly resonant at what originally was the -3dB point produces a resonant peak that more or less compensates for the natural roll-off that can be expected due to the practical limitations to the amount of inductance of the modulation transformer and of the modulation reactor, but as can be expected with any kind of resonant circuit, phase shift occurs near the resonant frequency, even though the frequency response curve of the modulator may appear to be flat down to a lower frequency with the smaller, but resonant, capacitor.

The high-Q resonant point when using the larger capacitor should be well below the bottom end of the rated frequency response of the modulator, and this should be of no consequence unless some stage in the audio chain goes into sub-sonic self oscillation, or it exacerbates an existing condition associated with transients that might occur when the power supply is switched on, either manually or with the T/R relay. I believe the preferred approach, which better preserves the audio waveform particularly when low frequency components are present, is to use the "brute-force" method with a larger capacitor, typically about 2 to 4 mfd, and take measures to correct any self-oscillation or transient problems that may be observed.  This has never caused me any problems, whether using a broadcast quality modulation transformer or not.  But checking high voltages at various points in the modulator and final amp circuits using an oscilloscope may reveal surprises.  Transient peaks and dips are usually too fast to register on moving-coil voltmeters and ammeters, and may account for some of the frequent complaints heard about mysterious arc-over problems with homebrew plate modulated transmitters.

Back in the early 70's I blew about a half dozen kilowatt size modulation transformers, including a UTC VM-5 and a 2 KVA n.o.s. unit designed to go into a ship-to-shore radiotelephone transmitter, while experimenting with high level speech clipping and filtering.  I came to the conclusion that it was voltage peaks developed in the low-pass audio filter inserted between the mod xfmr and final amplifier, whenever high amplitude audio tones occurred near the cut-off frequency of the filter, that was destroying the transformers. I removed the filter and avoided deliberately driving the modulator to the point of self-clipping (a.k.a. flat-topping), and have never blown a mod xfmr since. 

I  suspect the  low-pass "splatter" filter is the culprit in the commonly occurring crap-out of Collins KW-1 modulation transformers.  The Johnson KW uses the same tube line-up and the exact same off-the-shelf Chicago modulation transformer without a splatter filter, and those rigs seldom blow transformers.

[AB2EZ]

Don

In these attachments, I show the effect of turning on the B+. Attachment 1 shows the schematic

In attachment 2: at t=0, I apply the B+; and at t= 5 milliseconds, I input a single pulse, 1 millisecond wide, via the modulator. Top curve: C=10uF. Bottom curve: C=0.5uF

In attachment 3: at t=0, I apply the B+ and then I remove the B+ at t=100 milliseconds; and at t= 5 milliseconds, I input a single pulse, 1 millisecond wide, via the modulator.

Stu

[AB2EZ]

Now here is something interesting (I think):

I increased the time between application of B+ (which corresponds to 50 volts), and removal of B+ (50 volts => 0 volts) to see the effect of turning off the power supply.

Notice that with the 10uF capacitor, the voltage on the load (i.e. on the modulated B+ line) swings negative for a fairly long time (the negative swing is about 60% of the B+).

With the 0.5 uF capacitor, the negative swing is much smaller and shorter in duration.

If the modulated B+ tries to swing negative, it will cut off the output r.f. tube (I should add a diode to simulate this)... leading to a very high load on the modulation transformer.

This behavior would correspond to a design in which the B+ is turned off when one switches from transmit to standby, by opening the primary of the plate transformer.

For example, if the (choke fed) B+ supply has a 10 uF capacitor, and if the modulation resistance of the r.f. output stage is 5000 ohms, then the B+ supply will drop from full B+ toward zero with a time constant of 50 milliseconds when the primary of the plate transformer is opened up.

Stu

[AB2EZ]

I added a diode to capture the fact that the output tube will be cut off if the modulated B+ goes negative. See attachment 1. {Note that I still have the 1 millisecond input pulse starting at 5 milliseconds, which you can ignore on the response curves}

The B+ turns on at t=0, and turns off at t= 250 milliseconds

In attachment 2 (C=10uf) and attachment 3 (C= 0.5uF), the red curve is the voltage on the right side of the diode, and the blue curve is the voltage on the left side of the diode.

[w1vtp]

When I tried this setup, the capacitance had a large effect on the audio sound, every amount of capacitance sounded different.
I gave up and just use the RCA mod iron with current through it...


Brett

Interesting comment Brett.  I think I have the same mod iron that you are using

Al

[k4kyv]

That photo  looks just like the one in my HF-300 rig.  RCA made two versions of this transformer, one with a gap in the core to carry DC though the winding, and the other cross-laminated to require a heising reactor.  I saw one with the gap in the core one year at Dayton (not to be confused with those crappy 5500Ω, 1:1 ratio MCW jobs released by the jillions after WW2). The turns ratio on mine is 1.55:1, but the only other spec is 2500 volts DC maximum.  Mine came from a 250-watt RCA BC transmitter and is designed for a reactor.  I have used it to modulate a KW DC input to more than 150% without mishap, so the rating must be very conservative.  I have had it in the transmitter since the early 70's.  Just in case, I do have a spare on hand.

[w1vtp]

That photo  looks just like the one in my HF-300 rig.  RCA made two versions of this transformer, one with a gap in the core to carry DC though the winding, and the other cross-laminated to require a heising reactor.  <snip>

Here 's the gap....

[k4kyv]

I'm curious what specs are listed on the nameplate.  I couldn't read it from the photo.

[KE6DF]

I'm curious what specs are listed on the nameplate.  I couldn't read it from the photo.

I must have a better screen on my computer, because it's easy to read here.

The spec plate says:

Type:  900763
Class: Modulation
Pri: 10,400 Ohms
Sec: 4,350 Ohms

That works out to a 1.55 to 1 turns ratio like you said.

There isn't a voltage spec on the plate.

[k4kyv]

I can read it on the photo.  I forgot about clicking the little + sign to enlarge the image.

Interestingly, 10,400Ω is the rated p-p load impedance for a pair of 833A's or 849's.  I'm sure that transformer is not rated for the power level of a pair of those tubes. The one I saw at Dayton had a manual from RCA that included all the characteristic data on the transformer, even down to a diagram of how it was wound, the size wire and thickness of the insulation.  I didn't need the transformer so I passed it up.  Hope someone who could use it bought it.  It would have been a shame for it to go to the dump.

[Opcom]

The BTA-250L transformer PN is 900097-501 drawing no. 15516 and is gapped. I do not recall seeing the Z on the tag of my BTA250 transmitters as in those pictures. The 250L ran at 1650VDC so 2.5KV is reasonable, but I can't make out if this BTA-250 xfmr says 2.5 or 3.5 KV. The BTA-250L used 828's, which would have been happier with the 'book' 16K plate to plate load at 1750V which is reasonably close to 1650. That is for 300W out, much more than needed to modulate a 250W carrier. I did not find any 828 spec for the 10K range but it is vaguely possible. Based on this, I don't think it is from the BTA-250K or L but who knows. (both pics are 250L despite the file names)

The BTA-250L coupling cap is from the low end of the transformer to GND and is 12uF/2KV. The reactor is 45H, 120 Ohms, 2.5KV.


This last one has no tag, but is larger than the one in the BTA-250. I can't say if it is larger than the one pictured, but it looks like it is judging from the placement of the spark gaps and the outer frame bolts.
It has a gap. The only measurement I have right now is the cig pack there beside it which is 2 3/16" wide. It might be from an RCA 1-K, but came without a name plate. I hope someone can identify it.

[Steve - WB3HUZ]

That looks identical (or as nearly as I can tell) to the mod iron in the GE BT-20A. It also used 828 modulators.

[Gito]

HI

To me the circuit that Stu Attach ,looks like a High pass filter , constant - K pi section?
If its true ,than the L and C is depended on Frequency  and R.

So if its a high pass filter we must have the right L and the right C,there are correlation  between the value of L and C.

for instance the lowest frequency that's permitted to pass is 100 Hz than we must have a certain value for C and L,
so not any value  L can be used with any value of C.

Am I right

Gito