The AM Forum

I picked up a free DX100 yesterday from a dusty 20 year rest in a loft. Removed the knobs and cleaned it, and vacuumed and brushed down the chassis. It is acutally in really good shape physically. Maybe i can actually get a sig on AM (as I never seem to get finished my other two AM rigs, one a Junkston 500 restore and the other a Continental 1kW BC rig).  

Now I have questions for the experts:

Are many DX100s on the air, any idea? I ran an Apache as a novice and new general back in the early 70s, loved it back then. Gads, even used the SB10 for SSB! It is a direct descendent of this rig, easy to see.

I have read Timtrons' list of audio mods, as well as those by Ed AA8TV. Fixing the modulator for wider response and adding neg feedback seem like excellent goals. I noted that the original Heath design drove the modulator into clipping and then filtered it with build-out LPF capacitor across the transformer secondary, sort of a crude modulation maximizer.

Where does one find 1625's anymore? Is there an easy sub? The apache used the 6CA7s i think, which were expensive hi fi tubes.

The book and description show the DX100 (not the DX100B but the original two tone silver pyramid-faced rounded-corner box)  having a very wide range of output loading up to 500 ohms or so, with a set of switched capacitors for coarse loading and a fine loading variable cap on concentric shafts. Mine doesn't seem to have this at all, the front panel is marked for the coarse and fine controls, but only the variable cap exists, I don't even see the location or components for the switched Caps. Anyone have a clue about this mystery?

Finally, I need to replace the plastic grey knob for the drive control pot, and the outer concentric knob for the tuner on the same. They are missing/cracked. Anyone got a dead DX100 that they would part these out from?

I hope to get this big box on the dummy load today, assuming that the electrolytics will reform OK. Then will consider which of the mods to do for audio. I prefer to do outboard processing, with balanced 600 ohm circuit. I assume that the first 12AX7 stage may be unncessary then.

Thanks for any advice your collective wisdom can give!

John, Ed Seib, VA3ES compiled a nice booklet of mods and hints for both the Apache and DX-100. Interestingly enough, I just found it today when looking for some 'wall paper'. Also, if you can, contact Ray, WD2AFJ. He did the mods on my DX-100B and the reports I get are very good. Email me at w3slk@uplink.net and I'll send you Ray's address. I'm sure he can help you out.

Ed's compilation includes/features the Tron mods. I wouldn't mess with most of the other audio mods (w/the exception of the feedback mods) as many are half measures or just plain incorrect.

Tron's "turbo" mods reduce or eliminate the modulator clipping due to an impedance mismatch. The will allow you to run more audio without get W-I-D-E like most stock DX-100s do. I've heard an Apache with this "turbo" mod and some other Tron mods done to it and it was the best sounding Apache I've ever heard - hands down.

I've seen 1625s at fests. Fair Radio used to have them. They are identical to 807s except for the 12v filaments. I would think they would be fairly common as the mil bought tons of them and few amateur radios used them other than the DX-100 (there was 50's/60's vintage linear that used 4 of them).


In my experience, the DX-100 is the most commonly used AM rig. There are tons of them on the air!

Lemme know if you're good on QRZ.com, and you can expect a package.
I have a couple dozen or so moldering in my attic and I'm happy to donate a pair to the cause.

73 John

I'm good in QRZ.com. Thanks a bunch.  I tested the 1625s and both are the 1959 vintage, one has 1/2 the mu of the other, so i though might as well put in a closer matched set.

73
K5PRO

I've not tried this on a DX a hundred but it works well on a viking 2 and a 32 v.......use your outboard audio processing gear to drive the grids of the modulators thru an audio transformer......bypass the stock speech amp completely...no drilling and blasting and good audio.......good luck

John, that knob you need, is it the same as what was used on 50's Heathkit test equipment?  Gray plastic, skirted, one-piece molded?  If so I'll include a couple with the toobs.

73 John

Thanks guys. The knobs I am looking for is the concentric pair, the outer one is 1/4 inch shaft, grey fluted, and the inner one is a smaller diameter (3/16?) shaft which is used for RF drive I believe.

I glued the outer one back together last night, will see if it holds. The inner one is missing altogether so I am using something off an old Scott Hifi which doesn't look OEM but fits. If you have the authentic knobs around, would love them. Thanks again. Tonight I will finished getting the 'lytics replaced or reformed, and find out why there is one broken resistor inside. Looks like the audio stuff is well covered.

BTW, I looked in the 1959 ARRL HB, and the 807 modulator is all there, with the 12AX7 preamp and all. Only they used a 6C4 triode for the driver while Heath chose to use a 12BY7 triode connected in the DX100.

Wonder why Heath switched to the 6CA7/EL34 for the TX1 Apache? It is a more expensive and lower plate dissipation pentode. Maybe they wanted to get away from using plate caps and get all the HV under the deck or inside the RF cage?

Somewhere I remember Tek scopes used little red knobs.  Maybe they would fit or you could adapt.  I think these are hard to find.

Because they could get the 1625's for a pittance. 1625 were manufactured in such quanity and were pretty cheap on the surplus markets, made them a good candidate. Somewhere in my shack, I have a bag with about a dozen new ones. Originally, they outfitted the Apache's with Mullard EL34's whether they were less expensive than their American counter-parts remains a mystery. MY TX-1 has the Mullards installed. I do have some Tesla EL-34 and GE 6CA7's yet to try. My hanging question is why not the venerable 6L6??

The 6L6 had similar plate dissipation rating, but had only a 360 or 500 volt plate rating depending on whether it was 6L6 or 6L6GC. The EL34 had 800 Volt rating. Apache ran the modulator off the same B+ as the pair of 6146s.

Marshall used the EL34s in many guitar amplifiers. They had a peculiar distortion when overdriven, which created their distintive Marshall sound.
The Marshall wall of sound.

Vacuum Tube Valley magazine (issue #2 and #16) ran nice articles on the reasons for EL34/6CA7, and one of them was to get around RCAs beam power tube patents.

http://www.vacuumtube.com/

RE the missing fixed laoding caps:


I think that it was common to replace the switched fixed caps/one section variable with the DX 100B's triple section loading cap.   My second, recently acquired DX 100 is the original non-B model.  It has some mods,  including no fixed loading caps,  and the B's triple-section loading cap.

1625's were dirt cheap on surplus market,  and Heath often used good parts from the surplus market.

Here in the West,  the DX100 is quite common,  there were a ton made,  and they seem to be well designed ... go forever.

73  GL    Vic   KF6RIP[/quote]

Thanks for the tip on the loading cap revision. That makes sense. What threw me was the front panel labeling still showed COARSE and FINE where the old knobs would have been.
John

John - I'll get those 1625s out tomorrow.  Try as I might I can't seem to escape work to get to the post office.

73 John

Just a thought aloud: What was the major difference between the DX-100 and -100A? Now there's a rig you don't hear too often.

If you heard one, it would be in your dreams. They never made a DX-100A.

DX-100 1955-1958
DX-100B 1958-1960

There was also a version made for European export.  I believe the letter designater was "U".  I'm unsure of the circuit difference in the "U" model.
(http://www.shopingathome.com/Heathkit_TX.jpg)

W3JN

Thanks for the 1625's - they arrived yesterday! I am on my way to getting this DX100 on the air. Got the electrolytics replaced/reformed yesterday.

This is a great gang of AMers on this list.
73
John

Any time, John... glad to help.  Better you use them than they molder in my attic for another 10 years.

73 John

UPDATE on DX100 REFURBISHMENT:

Replaced the HV electrolytics with 2 x 210 uF 450 VDC
Reformed the Med Voltage electrolytics (20 uF x 3)
Replaced the Bias electrolytics

Found about 4 unsoldered and same # of poorly soldered connections, some buried deep near the chassis underside. The guy who build this thing, bless his soul, had bad vision or a bad week. Seems like some areas were fine, then others were a mess. (Esp under the 6146s and near the AM/CW mode switch - lots of melted wire insulation)

Replaced the line cord with 3 wire, and added a single fuse to hot side, and replaced the 0.005 uF RF bypasses on the AC line with UL approved bypass caps.

One 1625 was low mu (1800 vs 3000 on the other). When fired up, one tube had a very hot spot on the edge of its plate, glowing brighly orange. . Thanks to a generous contribution by John W3JN, I was able to find another good tube, and match both of them.
 
It tested out with plenty of RF, 120-130 watts easily, also lots of drive to the 6146s. The audio was abismal. Response was flat 500-2000 Hz and below 400 Hz was non existant. Distortion bad (7% THD) when driven above 50% mod. The S/N is about 48 dB when the AF gain is zero, and some background hum is evident at higher levels. Not too objectional though. The S/N became 70 dB when filters switched in on the Tek AA501 audio analyzer, when the AF gain is turned up - the noise is dominant from the 12AX7 mike preamp stage as expected. Carrier shift is 4-5% at full mod.

I spent the day yesterday ripping out the audio section, which is buried behind a little copper divider plate and the front panel mike jack. Recapped the coupling and cathode bypasses, and added a feedback loop from the 500 ohm tap of the mod tranny back to the second 12AX7 cathode. Tried the AA8TV feedback mod (has a voltage divider and then a series electrolytic to break the DC path and pass audio).

Audio response was better, but still not stellar. At least it could modulate at 100 Hz now. Removed the above feedback mod and replaced it with a 1 Meg series R, and a large cathode bypass capacitor (47 uF) per Timtrons posted modification. No series cap or shunt 200 Ohm resistor in this feedback circuit.

Final audio response is very good. It is about 3 dB down at 100 Hz, using 1 KHz as reference 0 dBm. 400Hz to 3 KHz is flat. Out to 6 KHz about a dB drop and it goes out futher. However, there is some ugly distortion on the modulated envelope on scope. Seems like opening up the audio and adding the feedback has worsened this aspect. Carrier shift is much better, maybe 2% at full mod.

I plan to go ahead and fix up the 6146 screen circuit, adding a fixed screen voltage source in addition to the resistor to the modulated plate voltage. On the web I found the 60:40 ratio circuit for an Apache. After this, I may consider replacing the 5R4's with solid state rectifiers as I have a pair of Semtech SCH7500 devices. However, there is something sentimental keeping me from doing it, seeing those gorgeous rectifier tubes (OK - they don't have the blue glow of 816s). I think the carrier shift and the ability to get a bit more out of the DX100 would benefit from SS rectifiers for higher and more stable 800V.

I might neutralize it, a la TX1 Apache scheme. It does seem to have a bit of quirky response to final output tuning, on the grid side. I haven't been up to 10 meters yet, still playing into a load on 80 and 40.

Its been a fun project, as I never anticipated having a DX100, it sort of fell into my lap (ouch!). I hope to get it on air this week once I figure out the source of audio distortion.

Does anyone have the section of the manual on operating the DX100, and the voltage charts for the tube elements? These seem to be missing on the BAMA manual.

What is your email address John?  I have a voltage chart and can scan.  Will also scan the tune up procedure for you.  What the manual says about tune up is 5 ma on the grids, and 250 ma on the plates unless you plan to operate phone.  On phone load to less than 250 ma.  My manual doesn't specify, so I load to about 230 ma.


Are you good on QRZ.com?

Jim
Yessir, I am OK in QRZ.com. I downloaded the TX1 apache manual from BAMA, as it appears very similar. I finally realized that Apache advertized 150 watts AM "input" and DX100 was rated 100 "Output". Marketing trick. THey both have same plate voltage and current. It appears that the TX1 had a different mod transformer, as it had 6CA7 which needs about 11K ohms versus 6K for the 1625s.

73
John

Can anyone tell me about the Electric Radio article in April 1993, on the DX100? Who wrote it, what was the content (modifications or just history)?
THanks.

John, I looked at QRZ and it did not list your email.  Whazzup with it?

Here's where to find the complete ER Index. Have patience, it takes some time to load:
http://www.qsl.net/n9oo/ersearch.html

The DX100 has turned into a fine 'phone transmitter. Following mods were done in addition to my earlier posting/list. Removed the 60:40 screen voltage resistors as those were not so effective for this particular rig. I was seeing a lot more negative modulation than positive (5-10% near full mod), and thought it might play a part. However, I also tried conventional screen choke modulation with a supplemental resistance, with less effective results. Using a dropping resistor from fixed B+ for self modulation of the screen (6146) was not effective at all.

I aborted the negative feedback around the mod transformer and speech amplifier, as it had only minor improvement, needed a lot more tweaking.

Final best setup was empircally determined:
Capacitor across the mod transformer of 0.0022 uF 6 kV ceramic.

Two 210 uf Caps for the HV, giving 105 uF total capacitance.

32K made of three 10 W resistors, 10, 10, and 12K, for the screen dropping resistor from the modulated B+. 1200 pf mica across this, and another 1200 across only 20 K of it.

Neutralization added to finals, with a wand poking through the shield from the driver tuning capacitor. (very helpful on tuning the grid and plate circuits)

33.3 volts of zener diode bias on the CT of the driver transformer secondary. 15 k resistor loading also. 750 uF 'lytic to ground.

6146 Grid metering fixed, moved clamp tube control pick off to other side of low ohm resistor.

Solid stated the HV rectifier, using Semtech SCH7500 diodes, an RC snubber across choke and MOV across primary HV transformer, per Timtron modification note. Inrush limiter on HV transformer. I left the 5R4s in the sockets for sentimental reasons, but disconnected their filaments.

The incremental results were good, and the final step was to replace the old 6146's with a NOS pair of GE 6146B's. This made a significant improvement in the peak modulation capability, and also in the carrier power. Before it struggled to produce 100 watts output on AM, and at 10 meters was less. Now it easily produces 120 watts AM with fidelity, and can go as high as 150 on CW. I can finally get 100+ positive without reaching negative limit first.

The modulators are running a lot hotter than before, like 60 mA per tube idling, and plate voltage is up around 800 and constant now. Thats a lot of power on those small tubes, but the efficiency is worst with no modulation of course. I may add a tiny Pentium CPU cooler fan under the 1625s to help out.

Proof of performance from 400 Hz to 7 KHz showed THD between 1.4 and 5% depending on modulation level. The low end suffers from inadequate iron, so at 100 and 200 Hz, the waveform is lousey. The response is very good from 200 Hz to 6 KHz, within 0.5 dB of flat from 400 Hz to 3 KHz, and is 4 dB down at 100 Hz and 2.2 dB down at 7 KHz.

The CW side of the DX 100 is still rough, having key clicks and some chirp, esp on 10 meters. But I will keep the switch on AM I think.

I plan to put this rig on the air, with little audio processing, and a jury rigged TR switching. I have nothing on AM to date, so finally....

What ever became of Ed Sieb, VE3ES? Brian WB2JIX, has a few Apache's/DX-100's that he would like to do mods on. I told him of Ed's fine compilation and gave him his email address. He told me his email got kicked back. I think this is a fine collection of modification/theory and should be shared. Anyone know where he is?

Hi all, we're back after a long hiatus. (Don't ask!, Well you can ask, but I'll tell you in a private email, or in a PM.)    I don't know why Brian's email got kicked back;  while I do have a new email addy, all my old ones are still working.  Hmmmm...

My compilation book has been "hung up" for a few years, because I was waiting for new input from Tim, on his "new, improved, Apache/DX-100 mods" which he indicated he was going to publish, back in  January of '02.   I haven't seen those mods anywhere; still waiting on advice from Tim.

My email address has changed, but you can still contact me at
va3es<at>rac.ca

I'd love to get another edition of the "Apache/DX-100 mods book" out. Maybe I'll just reprint what I already have, and go with that.   And yes, it still includes a 90 minute tape of Tim  waxing eloquent on his theories of modifying those two rigs.

"It's been a l-o-o-o-o-o-ng time, Sonny!"

I made many of Tim's suggested mods to my newly acquired and refurbished DX100. The one nagging thing about the rig is the way that Heath derived the screen voltage for the 1625 mod tubes from the center tap of the HV supply divider/bleeder. It is easy to get 475 volts across one of the 450 VDC rated 'lytics (brand new ones in mine) when modulating at high levels. I'm lucky, i suppose, in that the surge rating is 525 VDC, but i am not certain that the regular rises in voltage are considered momentary surges. As Tim says, the screen current through the bleeder/divider unbalances it significantly, both in TX1 and DX100.

The problem is worsened when solid-state rectifiers are used in place of the 5R4s. By they way, this mod is highly effective in 'stiffening' the HV supply, getting more output, and less carrier shift with mod. I tried modifying the bleeder, with a fixed 56K 10 watt resistor across the upper leg, to reduce its value slightly and unbalance the divider across the two 'lytics. This seems to help, as now the divider voltage goes from <50% of HV to above 50% as the plate switch is on and mod is applied. However, when changing the divider, the screen voltage on the 1625 is excessive, so my idling current is nearly 150 mA! This causes a dark orange on the plates and will shorten life if I leave this. I added a small Pentium PC fan at half speed (running off FW rectified 5VAC from the old 5R4 filament winding) to help move air around the modulators. I will need to play around with the resistors and lower the screen voltage. Maybe Tim's suggestion of running the screens off the med HV supply though a PTT relay are not so far out?

I am adding an audio line input connection in mine, to the second 12AX7 section, to eliminate the noisy preamp, and allow feeding balanced 600 ohm +10 dBm levels from a Volumax 4300 limiter. Does anyone have a schematic or manual for a Vmax 4300? I need to recap the thing, as it is all original from 1970. It sure has lots of 4.7 uF coupling capacitors in it.

John, are the "Tim-tron mods" you describe, the one from the AM Window site,  http://www.amwindow.org/tech/htm/dx100.htm  ?

If so, those are the ones I transcribed off the tape I have.  Tim also goes into depth about mods for the Apache, but I never got around to transcribing those.  Maybe this Spring.

Hi Jay!

The DX-100U was offered well into the early 60s, possibly to utilize leftover  components after the DX-100B was discontinued in the US.  Notice also that the "U" model has 6 bands instead of 7 (missing 11m maybe?), and updated meter and knobs.  I'm guessing that the "U" designation could've stood for "universal" and might've had multi-voltage transformers but this is just my speculation (maybe they were only 230V and the "U" stood for UK).  Personally, I kinda like the newer paint scheme on the "U" instead of the 50s "science fiction theatre" look of the original.

I'm visualizing the DX-100U being used alongside the (rare in the US but plentiful in UK) RCA AR-88D receiver in a typical early 60s AM station in UK.

How come this forum (and thread) hasn't awakened any former Heathkit employees to jump in and add 2 cents?

Eric

Hi Eric,

 Good Question, I'd be interested to hear from them as well..

I think people should try fixed or regulated screen voltage, and a moderate value resistor between the screen supply and each screen grid.  One resistor to one modulator tube, another resistor in series with the other modulator tube.  Probably about 4.7K to 6.8K in series with each screen would do it with 1625s.  If possible, to make up for voltage drop on peaks, raise the screen regulator about 50V higher, and readjust the control grid bias.

AM Press/Exchange article (http://www.bruhns.us/T-P_Amp_Lin/T-P_Amp_Lin.html)

Ed, yes those are the same Tim Tron mods for the DX100 that I am talking about.

I made substantial gain in the problem of screen voltage on the 1625s after looking at the Apache schematic off BAMA this weekend. Seems, Heathkit engineers figured out that using the HV divider to get a cheap screen supply for the mods was a bad idea. They used a relay and ran it off the medium HV power supply, which has a different secondary voltage from the one in the DX100.

I tried it, and rearranged the power supply to lower the voltage from 380-400 on mine to about 290, with choke input instead of cap. (Another suggestion from Timtron). The modulators now have a fairly steady screen supply, and I reset the bias to 27 volts with some zeners to get an idling current of 70 mA. Tubes are much happier!  I used a Magnacraft W88 AG8 relay i had, to switch the HV to the screens (at the AM/CW switch). Mounted it in the HV side, where the former 5R4 sockets are. This gave me one more added feature, a set of NO/NC contacts for running other things beside the antenna T/R relay (like RX mute) which i wired to the rear apron accessory socket, after removing the 500 ohm audio output connections there.

Regulating 280 volts or 250 gets more tricky in the limited space i have, would need a few OA2s or a stack of zeners. I have one 200 volt motorola 1N3550B 50 watt stud mount zener which I tried on the VFO earlier, could stick someone on the end of this to get it up to 250 or so.

I am interested in hearing how much success people have gotten with adding feedback around the modulator, either from the 500 ohm tap back to the front end, or just around the 12BY7 triode-connected driver. I need to try it, as part of the quest for East Coast sound, although I am in the Southwest.

If you are going to mess with a triode connected 12BY7 consider using a 12B4 instead. You won't find a miniature tube triode with a lower plate impedance, a must for good dynamic regulation and the resultant low distortion drive.

Steve - last night I did just that. Strange you should mention it, I noted the 12B4 was the minuature version of the old 6B4G and 2A3 in my RCA tube databook. I made a special socket adaptor so I could A/B test the two tubes as drivers in the DX100. It is a 9 pin tube plug and socket mounted vertically, so that I could make the wiring changes in the adapter. There are only a few wires needing to be moved to fit a 12B4 into 12BY7 socket. Both tubes look incredibly similar, and have same filament ratings.

What i discovered is that the 12B4 does provide better drive to the 1625s when grid current flows in AB2. The 12BY7s crapped out about 50 V p-p on each grid. The 12B4 went out to 100 V p-p before noticable distortion cropped up. It does have lower gain however, and the 12AX7 is required to run at higher levels to make up for it. Oddly enough, the overall THD (without feedback) is best with the triode connected 12BY7, not the 12B4! It was consistently better. I think the 1625s and mod transformer crap out before the difference in drivers is noticable.

I added Neg feedback from the plate of the 12BY7 to the plate of 12AX7 #2, through 200 K and a 0.33 mylar cap. The DX100 is very sweet now, with THD running about 0.25% across the band up to 80% mod. It didn't required global feedback around the mod tranny.

Also I spent the time to try adding the series R to the screens of the 1625's as suggested here. Very slight improvement, but the loss in power  
made it impratical since I am running only 300 V for the LV now, and using it for the screen supply. I tried as high as 4700 ohms on each tube, in many steps from 910, 1200, 1500, 2200, etc. I ended up pulling it out and adding a simple 10 ohm to each screen as a place filler in my chassis.

 The biggest improvement by far was to lower the output Z of the 12BY7 and leave it in, with the NFB i mentioned. I think I'm done with the DX100 mods, unless I can find a better turns ratio mod transformer that fits. I am hooking up an old UTC VM4 tonight, to empirically find the best match, and see what it comes up with.

Very interesting. Just goes to show you, it's the complete system that counts, not one component or section of the schematic. Sounds like you have the DX-100 sounding about as good as it can without major surgery like a different mod tranny and driver tranny (or elimination of the driver tranny and some direct connect approach). Probably not worth the effort, unless as a learning exercise.

Have fun with the Dixie 100.

Correction to my last posting: The THD figures are 10x what I said, not 0.25% but around 2.5% and higher. Oops, cockpit error. Still I swapped the 12B4 in and they got worse. Stick with the 12BY7.

Hi John,

Havva couple of DX-100's,  like 'em,  and have thought that for their weight class they do a very good job and seem quite bulletproof.   The Apache TX-1 is prob a bit more sophisticated,  but lacks 160 Meters.

Thanks for all the info on your progression of mods,  and for starting this thread.

C U on the Bands,   DE   Vic   KF6RIP

I'm addicted to playing with this thing. The XYL thinks so. She keeps hearing tones and bleeps in the late night hours. The dawgs are restless too. For me its a learning experience on RC coupled audio amps and pentode/beam power tube modulators. Triodes in class B are so much simpler and sweet sounding...  My first AM rig was an Apache though, when I was WN4PRO, over 30 years back.

I tried out a UTC VM4 modulation transformer, massive 300 watt thing, in place of the stock Heathkit tranny. Walked it around all of the ratios from 1.5  - 3:1. Then discovered that the VM4 was breaking down, could smell ozone and also hear audio arcs inside. Switched to my newer CVM4, and as expected, the bottom end is quite extended on the DX100 with it. It buzzes my speaker when I tone modulate it. However, also found that big iron was quite distorted at around 5 KHz, at greater than 50% mod. Almost seemed like the modulator was breaking up. Possibly a parasitic? I had 12 inch leads going out of the chassis and from the plate caps out to the remote ungrounded transformer. I was careful not to touch anything while testing but the power switch. The test confirmed that the 8000 ohm primary impedance of the Heathkit transformer is limiting peak (pos) output of the 1625s. (2.86 ratio). At a more optimal 6900 ohms, there is slight improvement. At lower Z, the modulator crapped out with higher distortion. Other night I finished solid stating the power supplies, replacing the 6AL5 with a pair of 1N4005 diodes for bias. Now I have space all over the place for 'improvements'. This DX100 isn't ever going to be converted back to stock. Since the CW mode is limited by the marginal VFO stability and clicks and chirp, I removed the phone jack for the key, installed a stereo (tip, ring and sleeve) phone jack in its place, and ran shielded audio cable over to my UTC A21 line input transformer which has been mounted beside the audio driver transformer. Now i have a convenient way to stick in processed audio at 0 dBm levels without using that first 12AX7 noise maker.

Another small but significant improvement came about after realization that there is no such thing as matched JAN-1625s, due to the vintage of the tubes and that most of them being used are either NOS or used in Korea. I stuck a 50 ohm cermet pot on a bracket, and fed the wiper to the modulator cathode current shunt, and each end to the cathodes of the tubes. With this in place mounted next to the shielded audio compartment, I am getting about +3 volts of self bias for 120 mA peak current in each tube, added to the -26 volt zener regulated bias on the grids. Injecting a 1 KHz tone at 80%, there is a definite null in distortion that can be found.
------------
With all this said and done, one appreciates why the Apache used EL34/6CA7s instead of 1625s. They are rated for 800 volts, have substantially more emission and cathode current capability. If I tap the secondary of the DX100 mod transformer at the 500 ohm point, ratio changes to 3.48, which would present about 10K to the modulator plates for a nominal 130 watt RF output. 6CA7s need 11K for AB2 at 750 volts to deliver roughly 100 watts. The Apache has the correct iron for that. Johnson, of course, went the 6146 modulator approach, and that is certainly an option but it  needs a 6K transformer, or lower even.

Wow John, talk about re-engineering.  You gonna furnish a kit?  How much improvement in THD from stock is there?

I noticed significant high-frequency distortion from my VM-5.  Looking at the waveform, it appeared to me that the upper half of the waveform was slightly delayed with respect to the lower half.  With a high frequency tone, I saw a big crossover distortion on the rising side, but none on the falling side of he waveform.  I don't have the VM-5 any more, but I suspect that the problem was unequal leakage inductance in the two halves of the primary winding.

I saw a similar, but lesser high frequency effect in my A-3894 transformer.  I reduced this distortion substantially by introducing delay in the less delayed half-cycle at the phase splitter.  I just put a small capacitor from one output of the phase splitter to ground. Of course, this isn't possible when a driver transformer is used.

It might help top-end distortion to try a small inductance, probably in the range of 5 to 20 millihenries, in series with one plate lead to the VM-4. If it makes high-end distortion worse, try it in the other plate lead.  Has anybody ever tried this?

Watch the voltage ratings, because the inductor will be at modulator plate potential, and also there will be some inductive kick from the pulsing class B current.  Also the little inductor needs to handle half of the push-pull plate DC.

I tried two new pairs of modulator tubes in my DX100 'test bed'. Built a dummy load out of parallel 25K and 10K ohm 100 watt resistors to simulate various RF stage loads on the secondary of mod tranny. Goal for each was to reach 120 watts into an optimal load, with reasonable distortion numbers. I was able to coax 110 and 115 watts from each pair. The 6CA7/EL34, 6146A and 6146B were tested in AB1 class, no grid current, eliminating the stringent requirement of having low output Z for the driver (12BY7 or 12B4 depending on which I was using).

Calculations of the preferred plate to plate load Z for EL34s and the Amperex and Telefunken datasheets showed needing about 10-11K for full output (100 watts) with 800 volts of Ep, 400 of Eg2. Measurements proved that wrong, as i saw 25 watts of undistorted power with this impedance presented to the plates. When i went back to using the full secondary winding of the DX100 transformer (2.86 Z ratio) and played with many combinations of resistors, I found that 8500 ohms is about optimal for these pentodes. I wonder why there is a discrepancy on the datasheet,all manufacturers copied it.  Thats exactly what the DX100 transformer presents to the plates if you load it for about 130 watts output. After tweaks to get about 38 volts of neg bias on the grids, and a solid 400 VDC on screen, the EL34s really worked well. This is far from the operating point that they see in a TX1 Apache however, with the 3K:11K transformer it has. THD is 1-2% for 50% mod, and rises to 4-5% at high mod. No feedback yet. Running up to class AB2, the distortion rises slowly, and even clipping is relatively 'soft'. Screen current is about 25 mA for both tubes, higher at 100% mod. Best idling current (Ip) is about 60 mA.

With 6146A, the optimal Z is lower, similar to the 6K or so on the datasheets perhaps. I tried 8500 again, and it played fine. The bias voltage is higher (42) for idling. Eg2 is only 200 instead of 400. Used an external power supply to fine tune screen voltage. THD is slightly better in mid band, but the low end of modulation capability is more limited, as the transformer effect (distortion) was worsened when using the beam power tetrode style tube. I tried a pair of 6146Bs but the bias voltage needed to be near 50, difficult to reach in the DX100 without more changes to the bais resistor string affecting other stage bias too. Audio was similar with the B's to the A's. 6146s had a harder clipping point, probably where the plate voltage swings below screen, and the beam forming plates are not doing the same function as suppressor grid in the pentode, to minimize secondary emission at the screen. Also, the effect of AB2 operation is more noticable with the 6146 on the scope.

In addition, getting 6146s to fit into the ceramic octal sockets that were used in replacement for the 1625 sockets, was tricky. I reused the sockets from the now gone 5R4 HV rectifiers. The holes for 6146s would have to be cut much bigger diameter to allow these fat tube bases. So I used some Vector tube extender adaptors, which raise them about 2 inches. Changes to the wiring are small, as both tubes are nearly the same, except for screen and plate connections, as long as the wiring to the cathode/suppressor is chosen to be the common pins for both types.
Filament current (1.25A) of the 6146 style favored it over the EL34 (1.5 Amps) when reusing the DX100 filament tranny winding, which is only good for 4 ampsl.  

I stuck with the EL34s for now, and have optimized the situation to get best audio performance with them. They don't have the vintage plate caps, but they also don't need a separate zener regulated 200 volt supply for G2, and can use the low voltage B+ (~400 if using the stock filter arrangement with capacitor input, and solid state rectifiers) through a relay linked to the HV switch.

Next week, I plan to test a multitap SNC modulation transformer that I just got off Epay, to fine tune things. The amazing thing is that the Heathkit DX100 transformer has held up to quite a bit of tone testing at 90% modulation. Adding a Modified Heising choke and capactor would improve things immensely all around, esp the low end. I read on Peter Dahls transformer listing that his Apache modulation iron is heavier and larger, but I am not certain if the Heathkit original is also.

It appears that using 6CA7/EL34s in the Apache is less optimal for high peak and low distortion modulation. The lower impedance ratio of the DX100 mod transformer works fine.

I hooked up the SNC 5P532 modulation transformer I got off epay (it was unused new old stock in the box, probably from the 1950s - a good vintage to stick into a DX100). First tested the 2.84:1 Z ratio which is nearly same as the stock DX100 iron. This proved to be uneventful, except that the SNC makes more noise (talks back) and is bigger and has a bold 'bottom end'. I was getting about 120 watts with 8500 ohms transformed back to the pair of EL34s. THD and response was darn good, like 2-3% at mid band, and worse at the ends - esp the bottom. Response was nearly flat (+0/-0.5 dB) from 100 Hz to 10 KHz!

Switched to 2.12:1 Z ratio which would be ~ 6.5K for the pentodes. Power came up, THD slightly worse. Could get about 1850 V p-p audio, which would give it a lot of positive modulation capability. The audio was providing 130 watts of sinewave with good performance into a load.

Then I tried a ratio < 2 and the THD was worse, the modulator was screaming at me, got 2 kV p-p audio across an equivalent test load. This was pushing it far, the positive capability is nearly 150%.

For a while I tried pulsing audio bursts to see what sort of tilt and overshoot the rig had, but the torture test opened my 25 ohm cermet pot in the cathodes of the pentodes (my distortion null/balance pot).

Hooked up the secondary to the mod transformer to the 6146B DC feed, and processed my voice from a microphone with a CBS Volumax to limit the negative peaks to under 100% to protect the transformer. It confirmed the peak capability for 2.12:1 ratio (125% positive reachable) and at 2.84:1 (115%).

I plan to test the rig with the stock transformer once again, now that i have the audio processed from a mic as well as tones. I can live with 115% positive and occasionally higher peaks. It would be a lot easier to use the little black Heathkit tranny than to kludge in the SNC, it does't fit.

I did note that the EL34s are quite picky about their screen voltage, and 400 is much better than 380-390 for lowest THD/IMD. These tubes like it hot. But I could not let them soar up to 430 volts when the HV is switched on and there is no audio. Cutoff bias was then inadequate and a warm orange glow appeared on one plate. As the heated the idling current rose, and they tended to want to run away slowly. To stabilize things, I came up with a very simple screen voltage series regulator which should be of benefit in pentode audio amplifiers where there is no chance for reverse current out of the G2 element. It consists of an 800 volt IRF820 mosfet, connected drain to the power supply (varying between 405 and 445 volts)  and source to the pentode screen grids. A 20 K 1% resistor runs to the gate, which is grounded via a 400 volt leaded zener diode. This regulates the screen close to 402-403 volts over power supply variations and with 0 - 50 mA screen current variations.

Thats about all, the DX100 project appears to be nearly done. As soon as I finished the audio chain I hope to try 75 and 40 in the late afternoons or mornings to get some on air experience with it. I am hearing a lot of Texas, Arkansas, Oklahoma, Colorado, Arizona, and CA from my locale near the Rio Grande.

Very cool, John.  Now you need to write the whole thing up in an article that Gary can post on the AMFONE part of this site, or HUZ man can put on the AM Window.

73 John

Putting this thing to bed now, after all the open loop measurements, time to close negative feedback loop and see what that does. Running the DX100 mod transformer is not quite as nice as the SNC transformer, with respect to THD and response, but its a compromise. I can still get 115% positive modulation using the 6CA7/EL34s, they have a lot of cathode emission!

Last night though, fireworks in the shack. Boom. Pow. Oh crap, I thought I had destroyed my Heathkit, at least the HV supply. Plate current was pegged, and it would blow the line fuse (5 amp fast blow) quickly, not to mention loading down the 800V to about 600. I unhooked the yellow wire from the power supply to the CW/PHONE switch, and it worked fine unloaded. Relief! Then I high potted the HV lead to the finals but my 450 volt power supply didn't uncover any shorts to chassis. I hooked it back up, must be a defective 6146B? I had been running this thing on its side for a month, at high power with high modulation, so it was a possiblity that I had a plate to suppressor or screen short. Tried another tube, same explosion. Went back to an older 6146, same thing. This time it left a smoke trail though. Turns out I had not calculated the voltage rating for a 1500 pf mica cap I put across the screen dropping resistor for the finals. I just stuck a 500 volt cap in, thinking the screen voltage was only concern. That capacitor is used for high audio frequency phase compensation for the screen bypass. 1500 seemed like the best value for THD for both 1 and 5 KHz audio. That capacitor has the brunt of the modulated B+ coming in one end, which is 800 + 950 (for asymetric peaks) and about 185 on the other end at the 6146B screens. Lets see, thats over 1.5 kV across a 500 V cap! Duh.

The transmitter survived the ordeal, which says good things for the solid state rectifiers as well as the screen grids of my 6146Bs. They took quite an overload for a few seconds each time I tested. while at it I thought about the peak voltages and replaced the main plate bypass cap, one of those red mica things which is marked 0.001@2kV. I used a ceramic doorknob, type 858, 1000 pF@5kV. It measured capacitive on the vector impedance meter up to 29 MHz, so is perfect for this application - and has a lot more headroom for peak voltage.

The DX100 got negative feedback installed, overall loop from the 500 ohm tap, through a divider made of series 450K HV resistor plus a 39K 1/2W to gnd. This reduces the DC component to 68 volts, allowing a large electrolytic in series to block it from biasing the cathode of 12AX7A stage #2. If I used a ceramic capacitor for HV rating (i.e., 0.01 - 0.05 uF 1.6 kV),its application in the feedback would give a bass boost. With the wimpy DX100 mod tranny, this isn't a great idea although it can be 'xperimented' with using audio shelving EQ later.

With these components i have about 6 dB of moderate feedback. This lowered the THD to 1.5% typical at 50% mod, and about 3% at 80-90%. The freq response is nearly board flat to 10 KHz, with slight rolloff at 100 Hz and below.

I am toying with the idea of installing a small filament transformer (6.3@0.3A) that I found at Ed's Black Hole last week. It would power only the 6AU6 VFO regardless of whether power switch is on, similar to what Apache did. I am wondering if long term stability would improve with this? Already have the VFO tube running off 150/110 volts DC regulated now instead of 400/150 as original DX100. There is room next to the LV choke under the deck near the LV power supply cap.

Hallicrafters did the same with the SX-101 and SX-115.  They also  bolted a power resistor to the chassis continuously powered off 110 to keep the surrounding components warm and dry (damp basements apparently being an enemy of stability in the 50's and 60s).

73 John

John, Have you rechecked optimum load impedance with the feedback loop in use? Great nuts and bolts writeup.

This has been an excellent series, and  a real boon to us DX-100 fans.  I've copied all these posts, and will certainly use the info when I next upgrade my DX-100's.

Thanks John.

Dave
Thanks for the question, no I haven't checked the optimal Modulator load Z after adding NFB. I hope it isn't different by much. Its too hard to get back in there and hook the modulator back to my load bank, since it was all hacked together and hanging in air above the TX which was lying on its side.

Last night I cleaned up all the air circuitry under the deck, and shortened the modulator plate leads from the tranny to the sockets of the EL34.

I have removed that little copper plate that shields the audio from the rest about 20 times now. Gotten real good doing it. Getting it back on is a chore trying to get a nutdriver back under there. While I have it open right now, I am trying to get all the filament circuitry wired so that the 12.6VCT winding is balanced with full load on each side. With the EL34s I maxed out the filament power. I noted that in my DX100, the audio filaments (both 12AX7 and 12BY7) are wired only to the "X" filament string, not X and Y as shown in the schematic. I wonder if this was due to the builders discretion or some change Heath made in the design? I would think that running off both sides X and Y would have given lower hum in the mic preamp. When I turn up the gain, there is noticible background hum. I measure about 50 dB s/n with my line input circuit- which comes in on the cathode of the second 12AX7 section, along with the NFB connection.

Another thing, since I am already slightly over 4 amps with the filament current overall, I started looking at ways to reduce it. That LV transformer runs warm, but I don't know what the filament winding is doing inside. I found that someone had used a #47 pilot lamp on the VFO and a #44 inside the meter. The 47 draws 150 ma and the 44 draws 1/4 amp. I got a 14 volt bulb to try, which will be quite dim inside the meter, but drawing only about 50 mA instead. If i add the small Stancor 6.3@0.3 transformer that I plan to do, that will relieve the overload on the transformer. It will power the 6AU6 all the time, like the Apache.

I plan to take some pictures of the various mods in place, and put all of the reports into one article, either online or in ER. Since a lot of stuff for DX100 has been discussed before, its really regurgitation. But I did do a lot of measurements to back up what I tried, and some of the improvements didn't do much towards making measurable differences. For instance, tying the screen voltage of the 6146B to both modulated B+ through a dropping R and to the fixed B+, to eliminate that last bit of nonlinearity. Or using a 15 Hy reactor for self modulation of screen. The benefits of these were lost in the overall performance of the modulator when coupled to the 6146.

I use the following equipment for the work:
Harris AM90 mod monitor
Homebrew voltage divider off the coax
Bird 600 watt load
Tek 7603 scope with dual plug in
Tek AA501/SG505 THD/IMD test set
Tek 7L5 Spectrum analyzer
Werlatone directional coupler
Mini circuits splitter
CBS Volumax 4300A for asymmetric audio tests

John, that's what's sorely lacking in many articles either in ER or elsewhere on the net.  Lotsa info on what resistor to clip, etc., but not backed up by solid data using quality test gear as to the results of said mods.

Hats off to you for documenting this.  I suspect this will become very popular.

73 John

Yeah, what he said. Big problem with mods is the users like to pick and choose and leave certain things out. Your before and after measurements really show why and not just how. Great job.
The thing about feedback is that if your mod lowers the optimum load impedance then it would be useful to someone who has tetrode or pentode modulators and only has (or desires the use of, for positive peak capability)  modulation iron with a low numeric ratio.

Dave, I hear what you are saying. My experience with the DX100 using an outboard multitap SNC transformer was, that the optimal Z to the pentodes is not black and white, like the data books for the EL34 state (11K for AB1 at 100 watts or so with 750 volts of Ep). As I got lower and lower, towards 6K, the peak positive cability got higher and higher, and the THD rose slowly. I suppose that 11K would have had the lowest THD, but also lower power and it was clipping (plate swing limited) at 100%, no headroom. For an audio amplifier, this would be perfect, as you would know you have a 100 watt amplifier, no more than that for clean sound. But for needing headroom for peak modulation (as long as the negative side can be limited with audio processor or a negative limiter at high level) there is a compromise to get more audio voltage, while living with a little more THD.

I certainly saw that, and after modest feedback was applied (not 10-20 dB as some have suggested) it has reasonably low amounts of distortion (mostly odd order from the tetrode/pentodes with push pull). I remember doing a proof of performance on a Gates BC1G in the 1970s. It was not much different from the DX100 final results, although it had a LOT of feedback inside, and could not run 7500 Hz sinewaves at 90% mod without tripping off, into a shunt fed BC antenna.

Anyway, about your idea that NFB might lower the optimal modulator load:
The point where I picked up the feedback is the familiar 500 ohm tap on the secondary of the mod tranny. This is after the two EL34s, which are plate to plate connected to the primary. My intuition (from feedback theory) would suggest that it lowers the reflected impedance that the RF side sees, when looking back into the secondary. But that Z isn't important, as the RF side actually sets the Z, i.e., 3050 ohms for 800V/0.27A, the ratio I am using in my DX100. This then is transformed back to the EL34s.

However, there is the effect of damping, which is used to advantage in audio hifi amplifiers by adding NFB. This helps the amplifier to handle speaker loads with 'tighter' compliance from the amplifier. I would think there is a similar effect from modulators towards the RF side of the transformer, but I don't know what it is called or how it is manifested.

I wish I had more time to play with all this. But spring is here, and I still don't have an AM BA TX on my antenna. Static is rising already. The XYL has me working on the garden some, but my pale crouched body prefers the smell of resin and burnt Heathkit wire insulation.

My next project is a Continental 314R1, PDM 1 kW, needing to move up to 3880 from 1230. It is already hifi, sitting next to the DX100 wondering why I have abandoned it for months. That might  be used next fall when the Heathkit cannot cut the QRN/QRM.

I got off on a tangent, which consumed weeks of evenings, with that radio. I rewired all of the filament power to equally load both sides of the existing LV power transfomer (4 Amps on each side, 12.6 VCT). But in doing so, I looked at the 6AU6 VFO, with plans to install a small transformer to keep it powered all the time, like the Apache and Marauder did. I needed to trace those black filament X wires that went through grommets into the VFO and popped out on top for the dial lamp - as I didn't want it powered all the time when the AC switch was off. Curiousity got the best of me first, and I finally opened up that copper box to see what was hiding inside. Wiring was messy, with a rats nest due to the HV, screen and filament wires being about a foot long stuffed in there. I noted that a lot of the RF conductors from the tube to the LC networks were made with tiny bus wire with very little solder in places. This prompted me to try and do better.

Instead of powering the 6AU6 filaments, I removed the tube, socket, and associated components, except for the LC networks. I buit a small vector board and made a JFET VFO similar to what has been in the handbooks. It was still a Clapp (series Colpitts) configuration, with a source follower second stage. After a disasterous attempt to use a current feedback opamp - AD829 - as a gain stage (4x), I built a small board with the priceless (really!) 2N5109 buffer stage. It had a 200 ohm to 50 ohm wideband output toroid, and I ran RG174 coax up to the VHF/XTAL switch and over through the spring shield thing to the 12BY7 buffer. There I installed a terminal strip and a 1:9 ferrite wideband transformer to drive a 450 ohm load and get some voltage there. Through a coupling cap it paralleled the 100K grid resistor for the 12BY7. With about 15 volts Peak to peak, I closed it all up and fired up the DX100, using a pair of 9 volt batteries for the VFO and buffer temporarily (voltage regulators dropping the voltage to 15 and 8 inside the VFO box).

Classic problems cropped up, including too much pulling when tuning the finals, some shift in frequency when the HV was switched on, and a bit of chirp. On the bench the VFO looked and sounded great, but next to the driver/finals in the DX100, it got too much RF feedback. Part of the problem was that the level was too low at the buffer grid. On 10 meters I couldn't get enough drive for >2 mA of grid current. On 80 meters the transmitter was regenerative and ran about 150 watts output, could not reduce the output loading enough!  Given the amount of time spent, and the abysmal performance, I punted the idea and went back to the glow bottle.

Last night I aborted the solid state VFO design and rebuilt the original 6AU6 VFO. I tested it carefully with both 6AU6 and 6AH6, and will post the final verdict here in the future. I am trying to do the right things for stability for thermal drift and line voltage as well as fix that pesky problem of having too much drive on 160 and 80 and barely enough on 15 and 10. If I run up there, and leave the drive pot where it is, then switch back to the low bands,as i retune the driver tuning the meter goes slapping against the peg at >10 mA, which is hard on 6146 grids.

Heath used a 26 uH plate choke for the VFO, which is probably tuned with the stray capacitance to enhance the output. However, all it does is creates a very ugly waveform for the 1.8-2 MHz output, loaded with third harmonics. If I raised it to 510 uH, the low band was sinusoidal, with about 120 volts peak-peak unloaded output and about 40 on the high band. I will look into what can be done to resonate the new layout on the  7 MHz output, and run it nonresonant for 2 MHz.

The moral of this story, if it works, don't mess it up. I continue to write off my time on this project as a learning exercise.

And we're learning along with you John. Great stuff. Thanks for posting.

And we're learning along with you John. Great stuff. Thanks for posting.

And we're learning along with you John. Great stuff. Thanks for posting.

Ah dx100's are my favorite.  i have 2 stations in maine running them,
2 in vermont and one here in az!    one in maine is opened up but
stock. the other one runs 4 1625's push pull parallel, driven by
a bridged couple of 14 dip audio amps. with a multitap mod xfmr
i set the ratio to give me 130-40% peaks

      One in VT is opened up but stock, the other has 2x813's in the final (they fit!) and 2 811's as modulators.  mod xfmr and audio drive is external. it'll put out 300watts+!   i switched the hv supply to capacitor input, replaced the choke with another dx100 plate xfmr to run in parallel
to give it the grunt!

       The one here is a work in progress. I've replaced the 1625's with
a pair of 811's. i'll be making a separate recfier supply off the plate
supply to give me around 1100 volts for the 811's.  i'll drive them with
a power audio chip. this should sound and talk up pretty nice, i hope!

deano!