COULPLING CAP FOR HEISING

[W2PFY]

Maybe JJ already asked this. What is the correct value for the cap. I use 6 mf. Is there a such a thing as too much? I use my cap at the top of the reactor with the B+ at the bottom of the choke.

[K1JJ]

Hi Terry,

We should get Tron in on this one.

I was told that 1 or 2 ufd is ideal, with 2uf the max to use, but can't remember the reason. If I were to guess, I'd say that using a larger value cap might parallel resonant with the mod transformer secondary at a particular audio frequency in the usable range and create ringing. Maybe 2uf and broadcash mod transformer inductance generally resonate way above 20kc, so is safer etc. dunno. But if it did, a severe case might create large enuff voltages to cause hi frequency arc gap arcing at normal power levels. I understand the BC industry uses 1-2 uf usually.  Maybe someone else here knows the reason.

Though, like you, I ran a high value [8 uf] for years before I found out. I replaced it with 2uf but cud not see any difference doing tone tests...

Anyway to comment on your other question, I think that using the cap at the BOTTOM of the mod transformer secondary winding to the power supply neg return will halve the HV requirement and do the same job.   Refer to my link for the schematic on the last post  requesting cap ratings.  The jury's still out whether or not the AUDIO HV appears across this Heising cap, along with the DC, in this config..

Either method works OK -  I'm just trying to reduce the size of the HV rating, cuz I found a great 2 uf cap that is not rated to handle the HV X 2.. only 1.5 X.

I'll axe Tron on the air if he doesn't check in here - and let ya know the answers to both questions, OM.

T

[k4kyv]

I ran 16 mfd for years, but later changed to 4 mfd.  Could not tell any difference. My cap goes from the bottom side of the mod xfmr secondary to ground, with the top side connected directly to modulated HV line. 

The larger cap would take longer to charge up, thus making a bigger transient spike when the HV is first applied and the bottom of the mod xfmr winding is effectively shorted to ground.  Otherwise I could see no reason why it would matter, except for the possibility of resonant effects with the mod iron or reactor,  but with normal values of inductance, it should make no difference once you exceed 2 mfd or so.

Most broadcast transmitters run 1 to 2  mfd, but I have seen some schematics showing higher values in the vicinity of 8 mfd.

[Bacon, WA3WDR]

Heising has a series cap and a resistive load with a shunt inductor.  This makes it a high pass filter section.

The performance of this high-pass section depends on the inductance of the modulation transformer, the inductance of the Heising reactor, the load impedance, the plate-to-plate load impedance presented to the modulator tubes, the plate resistance of the modulator tubes, and the value of the coupling capacitor.

When I ran Heising, I had triode modulators (810s), a VM-5 mod transformer (lots of inductance), 1500V plate, 300 mA (5K load), 50 Hy of Heising inductance, and 30 uF coupling capacitance.  It worked fine, but now I think the 30 uF was higher than it needed to be.

I'm playing with an old version of a simple simulation program (Electronics Workbench).  It looks like too much capacitance will extend the bottom end deeper, but there will be a dip, then a small peak way down low.  The larger the capacitor, the lower that peak is, but is always below the midrange response, and it gets lower and lower as the peak frequency gets lower.

It looks like insufficient capacitance makes the rolloff higher in frequency, and slower.  With a triode, it is a simple rolloff; with a tetrode you get a peak before the rolloff.

The right amount of capacitance will make the bottom end very flat down to a reasonably low frequency, and then you see an 18dB per octave rolloff below that.  With 20 Hy transformer inductance and 20 Hy Heising inductance, this value was... 0.4 uF!  (I was surprised how small).  Response was very flat to below 100 Hz, then -3dB at 50 Hz, with 1K plate resistance.  With 10K plate resistance, this produced a 2.5dB peak at 50 Hz.  I tried different capacitor values, but I couldn't get rid of the peak without getting a dip.  A little more capacitance extends the bottom but there starts to be a dip and then a peak a little lower before cutoff.

Using 50 Henries for the transformer and 50 Henries for the Heising reactor, the optimum capacitance was 1uF, and the -3dB frequency was 17 Hz with 1K plate resistance, and I got a 2.5dB peak at 20 Hz with 10K plate resistance.  Again, I couldn't get rid of the peak without getting a dip.

I'm not sure about the accuracy of these simulations, but they agree with other tests that suggest that the optimum capacitance value is probably in the vicinity of 1uF.

There has to be a lot of audio voltage across the coupling capacitor around cutoff.  Definitely use a high-voltage cap if you use less than 10 uF or so.

[K1JJ]

Thanks for taking the time to do that analysis, Bacon.   

I don't think anyone has published info like that about the Hesing cap before... at least that I've come across in my limited search.  Interesting about the 1ufd being a rough optimum and the various peaks, etc.


You mentioned there has to be a lot of audio voltage across the cap during cut off... Could you elaborate on that?  ie, what max voltage appears across the Heising cap when it is in the bottom of the mod xfmr secondary to ground with 3KVDC and 100% modulation?

I have a 5KV at 2uf available for the job and was hoping there wud be only the 3KV dc there and no significant audio HV....  I COULD use a pair in series giving 1uf at 10KV, but am running out of room.... .

T

[Bacon, WA3WDR]

Hi Tom,

I may be able to measure the voltage across the Heising cap in the simulation, but essentially the filter works by the phasing of the voltages at various frequencies.  Around resonance, the circulating currents in the L and C result in relatively high audio voltages across them, and the phase shift causes these voltages to add to the applied voltage in odd ways.  And the small coupling cap is deliberately introducing a resonance right at the beginning of the low frequency rolloff, to bump the response curve up at that point, extending the response down lower.  It is very possible that the peak audio voltage across that cap can exceed the supply voltage at some frequencies.  Come to think of it, current in the capacitor may be an issue as well.  Also, if audio comes through at frequencies below cutoff, the modulator will be unloaded and the filter output will be very low, so a lot of audio can be across the coupling cap. And turn-on transients may cause inductive kick effects.

Geez, don't let this scare you away from Heising; people use it all the time with good results.

My 30uF coupling cap consisted of three 10uF 600V oil caps in parallel, and I left the mod transformer secondary connected to B+, so there wasn't much DC on the caps at any time.  But the resonance of that hookup was waaaaaaay low, so even my 30 Hz furnace rumble didn't bother it.

There's another issue too.  If you apply negative feedback from modulated B+ back to a lower level stage, then a high-value Heising capacitor is probably necessary to prevent phase shift problems.  If the coupling cap is huge, then you mostly have R-L effects with transformer and Heising inductance and load and plate resistance, and the resonant peak is very low in frequency, and very low in amplitude relative to midband response as well.  But if negative feedback is applied from the mod transformer output, it would look like lower plate resistance to the filter, and it would control modulator output voltage below the filter cutoff.  Then, a smaller Heising capacitor would be OK, but of course there would be significant audio across it.

Now, if I can learn to use Spice or something, and cross-check these results.  Also I should measure some transformer inductances and throw actual measured values into the simulation.  Close examination of the plate resistance issue is in order, and exactly how to simulate it, because it definitely affects the results.  An actual set of measurements from a real radio would be good, too!!!

[K1JJ]

OK Bacon -

I wud say the consensus is that at certain times there will appear SOME audio voltage across the Heising cap, in addition to the DC. So, looks like my 5KV cap for a 3KV B+ is barely marginal.

I'll try it with both rigs (4X1 final and 813 final) and see what happens.

Yes, Heising is all I've ever used in the past - the transformers usually demand it. Just that I used more readily available caps that happened to have a high value in both voltage and capacitance. Now I'm stuck with a lower HV value at 2uf for now.

If you actually do a simulation of the HV across the cap, as you mentioned, please let me know the results.

Thanks again and 73,

T

[W2PFY]

Also, if audio comes through at frequencies below cutoff, the modulator will be unloaded and the filter output will be very low, so a lot of audio can be across the coupling cap. And turn-on transients may cause inductive kick effects.

I heard Bob W2ZM talk about zeners or diodes across his mod transformer. I don't think it's the KEEP ALIVE circuit that Steve QIX invented. Dirk and others use it and think it's great. I guess an unloaded mod transformer could result in a blown thansformer.

 

[k4kyv]

Phase shift at the bottom end can have the undesired effect of changing the shape of the audio waveform.  Examples include a near-square wave (produced, for example by clipping effects in an earlier stage, intentional or unintentional), or the asymmetrical waveform of a  human voice.

The square wave will appear canted, with an increase in amplitude at the leading edge, and a dropping off of the amplitude at the trailing edge.  Thus a waveform that would not exceed 100% negative modulation will be shifted enough to overmodulate on negative peaks.  That is why, with speech clipping, the phase shift following the clipper must be reduced to a minimum.  Even without speech clipping, similarly waveforms may occur in the audio fed to the modulator.  The result of this overshoot would be lower average modulation density if overmodulation is to be avoided. 

I have an old Altec tube type 4-channel mic preamp with built-in bass and treble controls.  I found that if the bass is boosted with the control, a phase shift occurs at the low end, so an asymmetrical voice waveform ended up with the higher amplitude peaks in one direction at the low end, and in the opposite direction at the mid frequencies and higher end.  The only way I could maintain consistent asymmetry was to keep the tone controls set for flat response.

I suspect a low value of Heising coupling capacitor could produce the same effect.  I would go with the "brute force" method of using enough coupling capacitance that the low-pass filter cutoff frequency is at least an octave below the lowest audio frequency likely to be contained in the audio modulating the transmitter.

My homebrew transmitter uses 50 hy's of inductance and 4 mfd of capacitance.  I run 2000 volts on the plate of the final, and the coupling cap is rated at 3KV.  The top end of the mod xfmr secondary is connected directly to the modulated B+ line, and the bottom end of the cap goes directly to ground, so the final amplifier HV appears across the  capacitor, just as it would in a HV power supply filter cap.  After more than 30 years I have never blown that cap, which is an ancient wax paper cap, not oil filled. The mod xfmr is from a 250-watt RCA broadcast transmitter.  I never measured the secondary inductance.  The original RCA BC transmitter used 50 or 60 Hy's of modulation reactance.

When I first built the transmitter, I used 16 mfd of coupling capacitance, but later changed it to 4 mfd because I was concerned that the larger capacitance would hold the modulation xfmr secondary at near zero HV potential for too long when the HV is first applied, as the cap charged up, and I was worried about the transient effect that might cause.  In practical operation of the transmitter, I could not tell any difference in normal operation, or when I swept the frequency response using a signal generator.  The low frequency rolloff begins at about 40~ as evidenced by a noticeable clipping and tilting of the peaks of the sinewave when the modulating frequency is brought below that frequency.

I suspect if you use a low enough value of coupling capacitance to have that peak and valley in the response curve at the bottom end of the expected frequency range of the modulating signal, you will have a lot of phase shift at the low end, and quite a bit of audio voltage across the coupling cap.

The UTC Linear Standard audio transformer series catalogue recommends designing the amplifier, and using appropriate audio transfromers, to assure a flat response at least one octave above and below the expected audio frequency range of the audio signal.  They speciically mention this in regards to undesirable phase shift.

For example, if you plan to use "communications quality" audio at 300-3000~, you should design the amplifier to be flat 150- 6000~.  If you are looking for hi-fi voice at 80-5000~, you would need a flat response of 40-10K~.  For hi-fi music reproduction, 20-20K~, the amplifier would need to be flat 10-40k~.  That explains the ultra wide frequency response, far beyond the range of human hearing, claimed for some of those LS series transformers, and similar ones manufactured by other companies.

A  rule of thumb for the modulation reactor is 8 Henries for every 1000 ohms of modulating impedance.  For example, running the final  2KV at 500 MA would present a 4000 ohm load to the modulation transformer.  The reactor should be at least 32 Henries.  I would assume the inductance of the mod xfmr secondary would be at least equal to that of the reactor, although I have never been too concerned with it if the modulation transformer was designed for BC xmtr use.  2 to 4 mfd should be sufficent for coupling capacitance.  I seem to recall reading somewhere that the coupling capacitance should be such that the capacitive reactance in ohms is equal to the load resistance in ohms at the lowest end of the frequency range. 

Following the above rule of thumb, and using at least 2 mfd of coupling capacitance should get the peaks and dips at the bottom end of the response curve well below any audio frequency you want to transmit.

I remember Tim telling me about a 5 kw Gates that used a rather small inductance in the mod reactor and he noticed some strange irregularities in the response curve down below 40~.  As I recall,  he did some shuffling with coupling caps and mod xfmrs, and was able to smooth out the response.  He could undoubtedly elaborate on the subject if asked.

[K1JJ]

My homebrew transmitter uses 50 hy's of inductance and 4 mfd of capacitance.  I run 2000 volts on the plate of the final, and the coupling cap is rated at 3KV.  The top end of the mod xfmr secondary is connected directly to the modulated B+ line, and the bottom end of the cap goes directly to ground, so the final amplifier HV appears across the  capacitor, just as it would in a HV power supply filter cap.  After more than 30 years I have never blown that cap, which is an ancient wax paper cap, not oil filled.

Your comment above is what I was looking for, Don!  In that case, think I'll try the 5KV Heising cap with the 3KV B+ service.

Interesting comments on the 1000 ohms = 8 H for the reactor.


BTW, Tom/KLR sent me an email about pi net impedances and finals, but I changed computers and lost it.

What are the formulas for plate modulated plate impedance [ when figuring plate impedance for a mod transformer]... and also for a pi net?  I know E/I, but which one uses a factor like .55? to calculate? 

T

[nu2b]

Bacon and Don made some good points on the Heising situation.

If you consider the problem just from a filter viewpoint, the easiest approach is to come up with a highpass filter as Bacon described. A good approach is to design a Tchebychev filter with very low ripple that matches the ballpark mod and heising reactor values. A 0.005db ripple filter with
9000 ohm load (3KV at 0.333A) and a 60Hz cutoff  results in the following values:
 
  Lp1= mod secondary=43 H
  Cs2=Coupling Cap=0.33uf=8.06kohm at 60Hz (close to Don's 9K estimate)
  Lp2=Heising reactor=43 H

These can be scaled to about 120 Hz with 21H and 0.66uf (close to Bacon's results)

Remember that this is a Tchebychev response, so the response is not down 3db at the cutoff frequency but only down 0.005db at 60 Hz. The above filter is down -1db at 35 Hz -3db at 30Hz and drops like a bandit 18 db per octave below that.

At the cutoff frequency, the reactances are such that the filter starts to act like a low Q (Q approx=1) resonant circuit and all reactances are drawing about the same current as the 9K load. This results in the capacitor voltage being 3000 +/- 3000= 6Kv rating required at 60Hz. If the input freq is dropped to 30Hz the mod tranny does start to unload but the series C has twice the reactance and could see up to +9Kv to -3Kv during the swing. Up at higher frequencies the cap has low impedance and the resulting cap swing is neglible.

Another problem to look into is the transient response when plate voltage is applied. With the correct 0.33uf coupling cap the transient response is good with only a few hundred volts overshoot. With a 4uf (too large) cap the overshoot on the cap can reach +4.6Kv and the 9kload will go to +3.7kv. This is with a perfect ramp up to 3Kv in 40ms.

If we now couple this with a real full-wave bridge, choke input power supply using say a 20H and 10uf filter plus 100k bleeder,and look at the combined transients of everything the PS transient(1/2pisqrt(LC)=11Hz)) adds about a 1.4/0.9*3Kv=4.7Kv(plate tranny Vpk) initial kick to the whole affair. Spice shows a +4.9Kv max.

If during this transient an additional transient mod peak at 100% (a quick on the PTT low freq belch at 30 Hz) occurs the peak overall .33uf cap voltage can get to 4.7+4.7=9.4Kv.

I would go with no more than 0.33u to 1uf coupling cap at no less than 10Kv.

Stay away from bigger caps not only for the reasons above, but also because the low freg ringing will probably cause an imd muddy effect when mixed with the higher audio freqs during transient voice peaks.

The power supply transient can be minimized with a double clicker step start of about 100ms delay (for 10u). If a larger cap is used, increase the delay proportionally.

Regards,
BobbyT 

[nu2b]

Heres the spice waveform with fw-bridge transient and 100%mod startup transient.
It shows the cap voltage is equal to the output V with a 90 deg shift

BobbyT

[K1JJ]

If during this transient an additional transient mod peak at 100% (a quick on the PTT low freq belch at 30 Hz) occurs the peak overall .33uf cap voltage can get to 4.7+4.7=9.4Kv.
I would go with no more than 0.33u to 1uf coupling cap at no less than 10Kv.
Regards,
BobbyT 

WOW!

I think many of us will have to take another look at our Heisng cap HV ratings! 

Thanks for the analysis, Bobby. I will be putting a pair of caps in series now, giving 1 uf at 10KV.

As always, a good piece of work there, OM!

73,
T

[WA1GFZ]

baat Tom Vu u mot runnin 9kohms

[K1JJ]

baat Tom Vu u mot runnin 9kohms

The 813 X 813's rig is running 12K.  The 4X1 / 833A's maybe 6K.  [That's using a simple E/I -  or is there a 0.55 factor to multiply for plate impedance?]

T

[W2VW]

Bobby's work here should be archived. Great info. Thanks.

E/I for modulating impedance

(.55) E/I for load impedance class C Tawm.

[K1JJ]

Bobby's work here should be archived. Great info. Thanks.

E/I for modulating impedance

(.55) E/I for load impedance class C Tawm.

Daay ayve, [practicing my southern accent]

So it's (.55) when figgering out the pi net components  and 1.0 for figgering the mod trans secondary load?

The 1.0 factor is used for the Heising cap calculation, right?

T

[Bacon, WA3WDR]

For the load resistance of an RF PA as seen by a plate modulator, it's simple E/I.  So if your PA is running 2000 V at 500 mA, the modulator sees 4.0K.

[W2VW]

Bobby's work here should be archived. Great info. Thanks.

E/I for modulating impedance

(.55) E/I for load impedance class C Tawm.

The 1.0 factor is used for the Heising cap calculation, right?

T

Yes it cares about the modulating impedance.

[Ian VK3KRI]

I dragged up this old thread as I came across some interesting stuff about the mod reactor, coupling cap, and mod transformer as a PI network, allowing lower inductance of the reactor and transformer. When I read this I remembered this thread, I just didn't think it was last year!

Anyway the deal is that by viewing the transformer ,  coupling cap, reactor as a PI hpf the reactance of the reactor & transformer can be 1.41 x the plate load rather than greater than  4x for your normal transformer model.
 
Section Chapter 6 "Amplifier Circuits", section 80"Modulation Trsnformers" of this scan of "Electronic Transformers and Circuits"(24 Mb)  goes into considerably more detail

http://www.pmillett.com/Books/Lee_1955_Electronic_Transformers_and_Circuits.pdf

This is part of quite an interesting collection at :
http://www.pmillett.com/technical_books_online.htm

Including gems such as the Radiotron Designers H/book vol IV
                                                                                                       Ian VK3KRI

[Steve - WB3HUZ]

Smaller values of coupling cap in a modified-Heising arrangement are preferable because they produce a more desirable impulse response. I did some simulation work on this about 10-12 years ago because, I too, wondered why most BC rigs used a 1 or 2 uF coupling cap (seemingly small).

The answer was obvious after running the simulations. The lower valued caps produced a nice clean impulse response with no overshoot. The larger caps did lower the corner of the frequency response curve, but the impulse response got ugly quickly. The larger the cap, the greater the amount of overshoot.

So what does overshoot on an impulse response mean in the real world? Higher frequency components of the transmitted audio will also suffer from overshoot. This means these components will tend to over modulate the transmitter well before they actually should. So, a competent and considerate operator will reduce the overall level of audio (or maybe the high frequency controls on his EQ) to compensate. Thus, a lower level of modulation is obtained, or less high frequency content. Neither one may be desirable, depending on other conditions. At best, the transmitter is not producing the level of modulation it could if properly designed.

My recommendation for modified-Heising is to use the largest amount of inductance you can and the smallest value of capacitance required for the desired low frequency response (phase and frequency). Using the typical 50H and 1 or 2 uF pairing will work FB for most final amp impedances.

[K1JJ]

My recommendation for modified-Heising is to use the largest amount of inductance you can and the smallest value of capacitance required for the desired low frequency response (phase and frequency). Using the typical 50H and 1 or 2 uF pairing will work FB for most final amp impedances.

Dear Steve,

I looked in my Browning Eagle Mark 3 radidio and was shocked to see a 20 uf electrolytic for a heising cap. Will I get more monkey swing by adding a big 2uf oil filled?

T

[Steve - WB3HUZ]

Only if the oil contains PCBs. It's a well known fact that the older BC rigs with PCB caps sounded better than they newer ones with non-PCB caps.

But in reality, on Channel 22, it's better to have the overshoot and attendant over modulation. This way, those on channels 18, 19, 20, 21, 23, 24, 25 and 26 will know when you have the MAUL DOWN and will stay back quiet.

[K1JJ]

But in reality, on Channel 22, it's better to have the overshoot and attendant over modulation. This way, those on channels 18, 19, 20, 21, 23, 24, 25 and 26 will know when you have the MAUL DOWN and will stay back quiet.

You make a good point.  I'd better jack that heising cap up as high as I can to let those mud ducks know I mean business. Can I use an old car battery? Will that be enough?

T

[W3SLK]

Also remember to use Teflon wire since this will increase the flow of electrons to the antenna.