Inspired by the discussion in the recent topics "NEW SYSTEMS - A MEGA FLOP !!!" and "Marklin 60172 Booster mfx Delivery", I have made some experiments.

I have been able to add mfx feedback to a 6017 booster. It works and I can register loks on the boostered section. It is just a lab experiment so far, and I don't know the reliability of my current circuitry, but I am quite confident that it can be developed to a fully functional and reliable solution.



The mfx feedback data is sent by modulating the current draw. I pick up the current draw variations in the boostered section with a signal transformer. Then I amplify the signal and let it cause a corresponding current draw in the section directly fed by the Mobile Station.

Great work! What's the difference between a lab experiment and real operation if you can register loks on the boostered section? None I would assume.

What do you mean by the voltage lever shifter? I am only familiar with using the Delta Controls 6604(5) as booster. Do they have such a VLS inside?
Regards, Henk.

Yes, I see.

What is the propagation delay - relative to the packet gaps.

i.e. any chance of the amplified signal (from a mFX decoder) still driving when the MS exits its "listening" mode?

Perz,
Could you tell more about the characteristics of the mfx signal sent back by the loco decoder: amplitude, frequency...
Do your transfo+amplifier act also as a filter?
Sorry if this has already been answered in an other post, but up to now, I was not interested in mfx and did not follow this topic.
But your post makes me curious.
Thanks,
Fred

I have made some more measurements. I was able to extract the mfx feedback signal from the sending lok, and measure it. I have tested two loks, the 37555 and the 39980.

The 37555: 140 mA current variation (peak to peak), 52 kHz frequency
The 39980: 120 mA current variation (peak to peak), 55 kHz frequency

Conclusions from this:

a) The feedback signal consumes significant power (2.4 to 2.8 W peak, 1.2 to 1.4 W average while transmitting). If something goes wrong so that the transmitter is on permanently, it is likely to burn the decoder. I don't think the circuitry is designed to handle this high power dissipation permanently. This is also consistent with Rainer Müllers findings that a continuously turned on feedback signal could burn the decoder.

b) The feedback frequency does not seem to be so exact. The RDS system uses 57 kHz but it seems sufficient to be somewhere in the range. I can note though that the 39980 which is transmitting with a frequency closer to 57 kHz will continue to register when I have made the conditions so bad that the 37555 just stops to register, even though the 39980 transmits with somewhat lower amplitude.

I have tested the sensitivity of my current "lab" circuitry. It detects current variations down to 10 mA. So there is a 12 to 1 margin. Good enough.

I have tested the sensitivity of the receiver of the MS. It detects current variations down to 13 mA. With 11 mA the loks will not register. So my detection circuit is somewhat more sensitive than the MS, but in the same range.

In my "lab" setup, I have used a 1 kOhm resistor to feed the return signal to the MS. This gives 20 mA. To match the signal strength of the loks, the resistor should be 150 Ohm. I will then need a resistor with higher power rating and also a stronger transistor to drive the signal.



Yes, it does, but it is a filter with a rather wide frequency range (low "Q value"). I haven't calculated it exactly, but it picks up all signals in the frequency range used for forward and reverse digital communication.

During my test I have got the "mfx lok won't register" problem twice. Maybe not so strange since I am varying the conditions until it stops working, and then you get all kinds of erroneuos behaviour.

However, I used another method to cure the problem than what Märklin proposes. Instead of using a second MS, I register another mfx lok in the MS. Then I delete that other lok from the MS again. Then the lok that had the problem will not have the problem any longer.

I think I can force bad enough conditions to be able to reproduce the non-registration problem rather easily. If I get a little bit more time I could then also try to see what kind of signalling that goes on/does not go on in this error state. That would hopefully shed some light on the root of the problem.

I have now put a description of the design on my homepage: http://www.persmodelrailroad.net/mfx_boost.html.

Edited by user 01 September 2012 20:51:37(UTC)  | Reason: Not specified

No, I haven't. I actually never really built it. I just wired it up on an experiment board of a kind where the components are held by springs and not soldered. I do not intend to use it myself. I just wanted to test the principle to verify my findings about how mfx works. The schematic is however designed according to "worst case" design rules, including margins for component variation, voltage variation etc. and I made some tests with "tilted" component values to verify that the margins are not just theoretical.

Ok, I thought you had. I have gone through the diagram, and have a few questions, if you dont' mind.

1. Are V1-V2 the same as V3-V4?
2. S1-S3 what would you use for straps?
3. X1-X4 do you just solder the wire straight to the the circuit board?

This what I found from the diagram, anything wrong?

CODE NAME QUANTITY MODEL other
C ELECTROLYTIC CAPACITOR 2 4.7uF 50V
C CERAMIC CAPACITOR 2 0.1uF 50V .01uF 50V
D 3 TERMINAL POSITIVE REGULATOR 1 78L05 5V 0.14A 3-Pin TO-92
D DUAL VOLTAGE COMPARATOR 1 LM393 ±18 Volt / 36 Volt 8-Pin MDIP Rail 
Q NPN TRANSISTOR 2 BC547B
R RESISTOR (BBYG) 10 1K .5W 5% 1K .25W
R CARBON FILM RESISTOR 2 2.7K .5W
R RESISTOR 3 4.7K .5W
R RESISTOR 1 22K .5W
R RESISTOR 1 82K .5W
S STRAPS 3
T (IFT) TRANSFORMER 1 01:40
V GP RECTIFIER DIODE 4 1N4007
V HIGH SPEED DIODE 4 1N4148
V ZENER DIODE 1 1N4733A 5.1V 1W 1N751A 5.1V 500M



I used http://www.jameco.com as a reference source for most of the material.

Thanks.

Per.