Dear Tony,

The short answer is 

"Without" (IE turn Momentum OFF, CV3 = 0 and CV4 = 0)

The lag and "dither" than momentum adds to the motor-control circuit masks fundamental voltage/speed behaviour,
thus making it rather difficult to accurately, empirically confirm "speed step X = inches/sec speed Y" accross various locos.

As an clear procedural example, specifically for original ST Tsunami1 decoders,
I reccomend reviewing the very first steps in the "Bruce Petrucca Method" of motor-control matching from
https://sites.google.com/site/markgurries/home/decoders/decoder-motor-tu...

(scroll about 1/2 way down the page to the section titled

Bruce Petrarca (AKA MrDCC / Litchfield Station) Procedure:  )

Happy Modelling,
Aim to Improve,
Prof Klyzlr

I have over 30 consists of 3+ locomotives on my railroad, and have set up many more.  I always tune the locomotive performance using BEMF, start voltage, etc, to get the locomotive to run as well as it can before I set up any consists....that way, setting up the consist is much easier.  Note I said "setting up the consist" and not speed matching.  Having the locomotives perform smoothly and well is critical is you're using braking, too, because binds and momentary power losses stop the loco (but not the consist...) cold and destroy the braking effect.

Even on the prototype individual locomotives respond to the throttle differently, and a little back-and-forth buffetting is very prototypical (at least on the 1970s era locomotives I model).  So, the main goal in setting up a consist is to get the speeds as close as possible but not to sweat the small variations between locomotives.  And when pulling a load, those small differences are evened out by the dynamics of the train.

To set up a consist, I don't usually change the momentum or brake settings (I use the same settings for accel, decel, and brake rate on all road locomotives).  Instead, I put the entire consist on the mainline next to my computer (you are using DecoderPro, aren't you?) and, using program on the main, adjust the start and max speeds of each locomotive.  I set the consist up in the command station, then open the throttle very gradually using 128 speed steps, and adjust the start step on the speed table of the slower locomotives until they all start about the same speed step. 

Then I open up the throttle and observe how quickly they accelerate, and which run faster than the others and adjust the top end of the speed table of the faster ones down until they run well together.  In some cases I need to adjust the top end up, though.  Because I generally set up my locomotives with a top end speed of around 200 to 240 (valid speed table values in DecoderPro are 0 to 255), I generally have room to adjust the top end, too. 

In very rare cases I do need to adjust the momentum to match the other locomotives, but that should be a last resort.  I try to keep the fiddling necessary to set up a consist down because I like to rearrange my consists periodically and don't want too many locomotives with CVs set up in unique configurations.

The key to remember is that a slight variation in the performance of locomotives in a consist will not significantly affect the performance of the consist, and won't damage the locos or decoders. Obviously, you don't want to have any locos slipping their wheels, but a small amount of mismatch will not hurt anything.

"I have over 30 consists of 3+ locomotives on my railroad..."

Jeff: So yours are somewhat 'permanent' consists where the same locomotives are essentially teamed with the others, and not a free-roaming approach where any locomotive might get consisted with any other at any time?  Please confirm.
 

I don't rework the consists very often, but re-doing or revising one -- or making up a new one -- goes quickly with the method I outlined.

Keep in mind that a perfect speed match is very difficult to achieve (and maintain) because motor running characteristics change during an operating session as the motor warms up, and over time as the motor, bearings, etc, wear.  

Just to clarify - I model the state of Victoria in N scale so I have no sound decoders. Consists are made and unmade often (as with the prototype), and there are (unavoidably) several brands of loco and decoder so no two locos have the same CVs. In addition, many of the locos have been butchered to fit their American chassis into scratchbuilt Victorian outline shells (Aussie loading gauges are smaller than American ones) and sometimes something in the drive train seems to suffer. So running characteristics can vary widely. I certainly don't expect to be able to get loco speeds exactly the same - I'd still be on the first one if I wanted to do that. In addition, loco speed settings seem to drift a bit over time which doesn't help.

Anyway, I'll do some speed matching with CVs 3 & 4 set to zero and see how it goes.

Tony in Gisborne, Australia

In my modest opinion, speed matching should be done with both, forward and reverse, momentum off.

Momentum is to be added after having locomotives speed matched, specially when long trains are being pulled.

Harry, Montevideo, Uruguay (South America)

Momentum shouldn’t affect speed matching, IF you wait long enough for the locos to settle to a constant speed before doing the matching.  Turning off momentum beforehand and turning it back on afterwards shouldn’t have any effect on the end results, it should just make it quicker and easier to do the matching.

Momentum off is obviously the consensus.

Tony in Gisborne, Australia

Here's the process I use to speed match, spelled out as a step-by-step procedure, with everything fully explained:

locomotive-speed-matching-made-easy

Momentum off is obviously the consensus.

For the initial speed match, yes. It’s much easier to get things basically matched up if you don’t have momentum, especially if the locos and the decoders are not the same brand.

I like to use a location on the layout with a couple of long parallel tracks such as a long passing siding or yard to compare the two locos I’m looking to match.

Make sure both locos are warmed up first by running them for five minutes. Next I consist the two locos together but put them on separate tracks. I then make sure both locos barely start to move after just cracking the throttle. I like to use a throttle such as the NCE hammerhead that lets me select speed steps, so I move the throttle to step 1 ... once set properly the locos should just barely move and remain more or less parallel.

I have both loco decoder pages open in JMRI and I’m playing with the starting voltage setting here. I tweak each one's CV2 until they match as closely as possible on the starting voltage. Then I move on to top speed, CV5.

I tweak each one's CV until they race down the track more or less side by side when I move the throttle to the max speed step (speed step 28).

Next, I set mid-speed CV6 to a value halfway between what’s in CV2 and CV5 on each loco. I set the throttle to speed step 14 and watch that both locos more or less run side by side. I adjust as needed to get them both to run mid-speed side by side.

Once things are matched without momentum, you’re not quite done, though. Set your desired momentum levels in CV3 and CV4 and then test again. Unless your loco and decoder brands are identical, you will need to tweak things. Avoid changing speed settings, try to stick to adjusting momentum settings only.

Being slightly out of sync is okay, but being way out of sync is not. If the locos are way out of sync it’s likely because the decoder brands are different. Either don’t run these locos in a consist or try again with the same exact kind of decoder in both locos.

So you're programming on the layout.  Does that ever scare the beejeebus out of you?  Because that scares the beejeebus out of me.  I think I have forty locomotives give or take on the layout. It would be easier to speed match with them running side by side and tweaking them as they run, but somehow I think I'd wind up with forty locomotives in that consist. 

Randy