Dentron MLA-2500 Repair.  


Origins


The University of Maine Amateur Radio Club (UMARC) Dentron MLA-2500 HF power amp found its way back to the UMARC radio shack in the late 1990's or early 2000's.  It had spent many years in storage at the department of electrical engineering at U. Maine.  At the time EE personnel claimed that:

 ``... it doesn't work ... it's broken...''

we (AA1ZB & KE1JH) figured that was OK we would fix it.  After all, given the attitude of the EE department toward all things electrical (whether they belong to them or not), they probably would not have given it back if it had worked.  True ownership was, and remains, unclear but it is believed that it does indeed belong to UMARC.  No documentation came with the amplifier.  Any documents were lost over the years it spent out of UMARC control.

Unfortunately UMARC membership has ebbed and flowed... mostly ebbed.  Thus "control" over club hardware and club space has been spotty at best.  Be that as it may there remains a few dedicated people who try to keep things going, despite the University beauracracy's attempts to kill the club.  As of early 2010 the club's website is down, and I haven't heard from anyone connected with the club for months.

Dentron MLA-2500 Checkout circa 1998

The MLA-2500 was checked out at LASST (Laboratory for Surface Science and Technology), and then again at the UMARC shack.  The power cord was given the correct plug to match the outlet at the UMARC shack.  Since the UMARC shack already had a 240V outlet wired, this is a strong indication of the ownership of the amplifier.

With the power feed set, the UMARC high power RF "paint can'' load was filled with "baby oil'', and everything was cabled together.

The unit fired up, tuned, and ran fine, although the output power seemed low at an indicated 600W.  We had no manual at the time so we assumed that the unit wasn't built for more than 600W out.  A watt meter, which measured beyond 300W, was not available.

It was surmised that the folks at EE may have tried to drive the thing at signal generator type levels.  Such a power amplifier requires far more drive than a signal generator can produce, if it is to amplify effectively.  Perhaps they also neglected to toggle the relay port to put the unit in Transmit mode.  Who knows.

In any case the unit worked fine.

Work of Feb 2009

Several years ago the University FM (Facilities Management) crew plowed one of the windows open during winter snow removal.  This deposited a quantity of dust and dirt on much of the bench space, as well as the equipment.  How much water ran into things is not known.  The discovery of the window being open was long after the fact.  Given the high ambient temperatures in the shack, it wouldn't take long for the snow to melt and then evaporate, even with the window open.

UMARC personnel were reluctant to operate the MLA-2500, for any length of time, without a cleaning and a checkout.  It also appeared to some that the unit may have some sort of fault.

The amplifier spent a good deal of time in this undetermined state.

At length, AA1ZB offered, and undertook, the cleaning and checkout of the unit.  Unfortunately this ended up taking many, many months due to other commitments. 


MLA-2500 Version

It turns out there are at least two versions of the MLA-2500:
This was discovered when looking for a manual on line. 

One of the more obvious differences between the two versions, is the way the two 8875 valves have their Cathodes tied together.  On the MLA-2500 the two Cathodes have an anode to cathode diode pair connecting them.  In the MLA-2500b the Cathodes are tied directly together with no diode pair.  The two versions also differ in some of the connections of their respective Grounded Grid implementations.

Since these differences did not involve the broken RFC, it made little difference which version the unit was.  There may be other versions as well.  Once a representative schematic was found, no further search efforts were undertaken.


Repairs 

(Inside The MLA-2500)

The guts of the Dentron MLA-2500 are not all that complicated, and can be viewed in figure ?.  The design is divided into two parts.  As viewed from the front, the right side houses the power supply components.  The left side houses the RF deck.

Top View
Note that in figure ? the plate contacts have already been freed, and the fan baffle has been removed.  In the manual the watt meter calibration potentiometer is the one near the back wall.  The other potentiometer is for the ALC calibration.

HIGH VOLTAGE!

The warning placard mounted on the transformer is no joke.  Careless snooping inside the amplifier will kill you.  Keep in mind it takes quite a while for the voltages to decay after the power has been turned off.  Use an appropriate high voltage probe to determine if the Plate supply voltage has decayed to a safe level, before you touch anything.

The schematic indicates the output of the HV secondary is 800V.  If this is an RMS value that would put the peak, rectified DC voltage at 1,131 volts.  The plate voltage meter on the front of the MLA-2500 indicates in excess of 2000 volts.  This suggests that the transformer HV secondary may be a center tapped 1600V winding, with each side of the secondary running at 800V.  The peak voltage of 1600V RMS is 2.263KV.

If either of these voltages goes across your chest it will be like having a heart attack.  BE VERY CAREFUL.

Clean Up

Surprisingly there was little evidence of dirt or water damage.  In the end all that was needed was a quick vac. and a little dusting with a cloth.  Other than that things looked none the worse for wear.

Both tubes were removed and dusted with a cloth and Isopropyl alcohol.

When re-installing the plate terminals it became evident that the resistor of the RFC (Radio Frequency Choke) on the outermost valve (vacuum tube) had cracked, see figure??.  When this had happened is unclear.  It could have cracked when the valved was removed for cleaning, it may have already been cracked.  Since some UMARC operators felt the unit was faulty, the resistor may well have been cracked before the cleaning.

Broken Right RFC resistor

The intact RFC, for the plate circuit of the innermost valve, is shown in figure???.

The length and width of the RFCs was measured as shown in figure??? and figure??? respectively.  The length was about 21mm.  The width was about 8mm for an inner diameter, and 10mm for the outer diameter.  There were about 5 turns of 16AWG wire, evenly spaced over its length.

Left RFCUsing the outer diameter, the approximate inductance for such an air core winding is about 558µH.  On 160m the inductive reactance would be about 6.3Ω.  On 10m the inductive reactance would be about 105Ω.  Given these levels of reactance the 100Ω resistor must serve to decrease the Q of the inductor at harmonics of 10m.  Thus helping to dampen parasitic harmonics.

Since the chassis seem so clean, no further disassembly was attempted, just a quick vac. and wipe down.

If it aint broke, don't fix it.

Broken Choke

Which choke (RFC5 or RFC6) was broken is not clear from the schematic.  If there was a marker as to which valve went where, it was overlooked.  Clearly the RFCs themselves were not marked.  In the schematic no details of the values of the the two elements that make up the choke are given.  Nor is it suggested that they are the same.  Unfortunately the one manual found, for the MLA-2500b, listed the two as "Choke Parasitic''.  This was not very useful, except that the parts list did suggest that they were the same.

LengthOfRFCsmall.JPGInspection of the two chokes showed that they were both composed of a 100Ω, 10W carbon composition (CC) resistor with 5 turns of 16AWG around its length.  So both RFC5, and RFC6 were clearly supposed to be similar.  Given the construction of the two chokes it seemed unlikely that mutual capacitance between the turns and the resistor body were a factor in the choke's operation.  This also seemed to be suggested from the schematic, which showed only a resistor and inductor in parallel.

WidthOfRFCsmall.JPGShould Both RFCs Be Replaced?

Since both valves operate with their plates tied together, with an RFC between the plates and the common connection, it seemed likely that symmetry would be important.  On the other hand, at realistic plate voltages and currents the voltage across the RFCs would be swamped by whatever action the valves themselves were generating.  For this reason it was deemed that absolute symmetry between RFC5 & RFC6 would not be required.  There would only be a need to keep the two relatively matched.

The Replacement Parts

The only true replacement was the resistor.  The inductor is the same coil as the original with the resistor removed.  It was slightly reformed to fit the perf-board mount, and the TO-220 style 30W resistor.

The new resistor is a Caddock MP930, 30W\footnote{When used with appropriate heat sink.}, 100Ω, 1$\%$, low inductance part.  As of Feb 2009, traditional 10W axial CC resistors are not stock items in standard supply houses.  A modest selection of metal oxide resistors in this range are available but none are necessarily low inductance.  This Caddock part is known to be low inductance and seems to be well supported by at least two supply houses (Newark, DigiKey...).

The connections to the plate electrode clamp and the inductor tower were made using common 14AWG ``Romex'' type household wire.  This was the closest thing on hand to the original 16AWG wire.

InductorOfNewRFCinPlaceSmall.JPGThis image shows the overall construction of the new RFC. The inductor is below the TO-220 style resistor.  The assembly is mounted on some perf-board, with 14AWG solid copper wire making the connections.

Figure~\ref{fig:NewRFCassemblyInPlace} shows the new RFC in place.  The use of the perf-board frees the resistor and the inductor from any bending loads when the plate contact is pulled off the tube.  The soldered contacts of the assembly are also wired.  This was done in order to prevent the contacts from breaking should the solder joints overheat and re-flow.  This is not likely given the modest currents that flow through the valves.

This image shows the TO-220 style resistor of the RFC in place. The gap between the resistor contacts is wider than it appears in this photo.

ResOfNewRFCinPlaceSmall.JPG

In figure???? the T0-220 style resistor of the new RFC is shown.  The gap between its contacts is wider than this photo suggests.  In any case the gap should be wide enough for any normal operational voltages.  Unless something catastrophic happens the contacts should not arc across the resistor.

Loading


The bulk of the bias Plate current burden of the RFC will pass through the inductor.  Therefore the size of the resistor is much larger than necessary at the bias point.  Typically 1mA - 2mA of bias flows in the Plate, see the 8875 data sheet for 2000V Plate voltage, at 8.1V of Grid voltage.

RF Load

This is a bit more tricky to quantify, without proper test equipment\footnote{RF HV voltmeter... etc} but some educated guesses can be made.

First off the amp is assumed to be operating in class AB.  The valves will be active for more than half a cycle but well less than a full cycle.  Distortion created by this mode will be blocked by the output filters, within reason.

A quick look at the schematic shows that the drive signal at the input (J2) is brought directly to the two cathodes with two parallel isolation capacitors (0.001µF) blocking the DC, and a single 100Ω, 25W resistor making up the balance of the input impedance.  The implication being that the composition of the reactive components in the input circuit will tend to 100Ω as well.  Thus bringing the combined input impedance to 50Ω, throughout the operational range.

Since the valves are driven more or less directly by modulating their respective Cathodes, with the Grids grounded, Any excursion of the Cathodes more negative than their bias points, will tend to shut off both valves.  A full 36W input would generate positive and negative peaks of 60 volts, across 50Ω.  The tubes are biased with 8.1 volts between the Grid and Cathode, via a Zener diode.

With the 8.1 volt bias the tubes get shut off to about 68 volts, when driven with 36W, which is a fairly hard off voltage.  On the other side the tubes are on to about 52 volts.  However, little in the way of dynamic Grid vs. Plate current curves seem to be available on line.  The 8875 data sheet has ``Constant Current Characteristic'' graphs, which are presumably used for establishing bias points only.  So we are left guessing what the valves are doing during operation.

ValvesWithBaffleSmall.JPGIt seems pretty clear from the schematic that the plate voltage of the tubes varies greatly during operation.  The resistor R2, capacitor C5 and RFC2 all serve to isolate the Plate bias voltage from any RF current.  If RFC2 is the ``tower'' inductor (see figure~\ref{fig:ValvesWithBaffleSmall}), just in front of the two 8875's, which seems likely, then it is also likely that the RF voltage appears across this inductor.  Little voltage would appear across RFC5 and RFC6, so the Plate voltage of the 8875's must change considerably, in order to pass RF current through C6.  From C6 the RF current then passes out the matching/filter network on its way to J1.

This means the graphs of the 8875 data sheet are not likely to yield any useful information, under dynamic conditions.  The Plate voltage is not constant nor is the Grid voltage.  Only under bias conditions are both these voltages held constant.  The safe operating area graph is also not very helpful, since the maximum ``pulse'' length\footnote{160m being the lowest operating frequency} is on the order of 0.017µs, and the left most edge of the graph starts at 10µs.  In this context the tube is unconditionally safe, since it is unlikely that the plate current will go anywhere near 6 amps.

The usual caveat to all this is that the amp is tuned properly, and is driving a 50Ω load directly, or via an appropriate antenna tuner.