LDG AT-1000 autotuner acting strangely on some bands?

LDG autotuners are famous products worldwide, and for good reason: they are very well designed and built, providing a convenient solution when needed. There are very few complaints about them, as a quick Internet survey showed me. Why did I search, you say?

The other day a friend sent me an LDG AT-1000 that was acting up on some bands, failing to provide a match or even maximising SWR instead of minimising it, but working fairly well on other bands. He had been told it's the microcontroller's fault, and LDG even sent him a replacement chip (thumbs up for that kind of customer service), but the problem remained exactly the same with the new chip.

A quick check showed he was quite right: on 80, 40 and 30 meters the tuner was acting as a mismatcher, rather than a matcher. On 20 meters and higher bands, it worked rather well, although it seemed a bit "sluggish" (how's that for a scientific term?).

After examining the circuit in and out, I was sure that all of the relays, inductors and capacitors of the matching network were OK. I connected a dummy load at the output of the unit and saw that the autotuner indicated significant reflected power on 40 m, although there should have been none (I had it in the "bypass" position). "The SWR bridge circuit must be unbalanced", I thought - but why was it unbalanced on a specific frequency range? A Bruene circuit, if misadjusted,  behaves progressively worse as the frequency is raised - here, it was just the opposite! Nevertheless, I proceeded to balance the SWR bridge. For this, I cut the wire connecting the SWR circuit to the input of the tuner PCB, and connected a precision dummy load right after the bridge PCB with a very short piece of coax. I set an RF generator on 30 MHz and nulled the reflected power reading using the trimmer capacitor. Curiously, it was off by just a little. Hmmmmm...

Fig. 1. The bridge PCB in its original state.

After reconnecting everything back, I checked again on 40 meters with the dummy load at the output of the tuner, which was in the "bypass" position. There was a significant reflected power reading on the tuner's meter again - but NOT actual reflected power, as another SWR meter between the generator and the tuner showed me! Furthermore, on 20 meters there wasn't any reflected power indication on the tuner's meter with the same conditions! So something was clearly amiss with the bridge - but all of its components had checked out OK! Curiouser and curiouser!

Fig. 2. Look ma, no screws!

Fig. 3. The hi-tech insulators.

 After a considerable period of head-scratching, the light went on. The fact that the bridge was balanced on 30 MHz but not balanced on 7 MHz showed that at the lower frequency the RF currents must have been taking a detour. The only way I could visualise that happening, was through the grounding posts and grounding wire connected to the SWR meter PCB (Fig. 1 - click on the photos to enlarge). So I removed the two screws affixing the PCB to the posts, also disconnecting the wire this way. I checked again on 80 and 40 meters - and bingo! No reflected power indication any more! The bridge nulling remained excellent across 1.8 - 30 MHz. When I touched the PCB to the posts, the reflected power reading jumped up again. 

So, I modified the bridge structure by insulating the PCB from the grounding posts, as shown in Figs 2 and 3. I used a little square piece of thick paper at each post, hot-glued to the PCB and posts to accomplish my goal. I completely removed the grounding wire. After re-nulling the bridge (just to be sure), the tuner worked perfectly on all bands, with no abnormal readings at all. It even produced a 1:1 match feeding my 20 m quarter-wave ground plane at 80 m!

So, if your AT-1000 is horsing around on some bands, this trick may well work for you too. LDG might want to have a look into this matter and modify the design. For me it was yet another good reminder that RF currents, given half a chance, rarely pass from exactly where we would like them to!

Addendum: I just purchased a very nice LDG Z-817H tuner for my bicycle HF hamming during the summer! It has a different structure at our point of interest (the Bruene directional coupler is located on the main board) and a series of tests with various types of load conditions showed no tendencies for whimsical behavior. It successfully matched whatever I threw at it and had a nice time doing it. The directional coupler stays balanced throughout the operating range. Well done, LDG!

Good news: YAESU has added the DC blocking capacitors at the IF filters of the FT-857D!

Fig. 1. The new FT-857D PCB.

Sometimes I think it's mothing short of a miracle that we can buy new technology products (like our transceivers) at VERY reasonable prices (not exactly so when I got into the hobby, 30 years ago - a handheld transceiver was a very expensive item in my country, it cost more than the monthly wages of a public employee).

In sharp (and delightful) contrast, recently I bought an Agilent E4406A Vector Signal Analyser,  an HP3586A Selective Level Meter and a Hagenuk Digiflex LAN Time Domain Reflectometer from the surplus market, at what must be a tiny fraction of their original cost! Mind you, the E4406A was about 45000 euros less than a decade ago! Long live companies like NOKIA (that's where my wonderful E4406A came from) and the ever-faster changing industry standards!

Fig. 2. The Serial Number.

But I digress... It seems that the nasty filter problem has caught the attention of the manufacturers! Tomi from Romania has sent me a photo of the PCB of the FT-857D he bought recently (Fig. 1 - click on the photos to enlarge), in which we can see that the DC blocking capacitors have been added! Thumbs up for Mama Yaesu! 

Maybe they do that for all their products now. I am sure other manufacturers have also taken steps to cure the nasty filter plague, but I haven't seen any concrete proof yet.

Tomi also sent me the serial number of his rig (Fig.2) - to save those that would like to add the capacitors in their recently-built rigs from the trouble! Thanks, Tomi!