Hey, if the 80%+ number from the previous article is true, I can see the on demand stacking up favorably to a cast iron beastie.. of course, it's not much cheaper, and it uses a ton more electricity and pumping power. But again, it's a solution without a niche: they don't make on demands for oil, and if you're using gas you're STILL looking at mod/con if efficiency is desired or tank water heater if economy is desired. This "middle of the road" solution is not a "good compromise" between the two, it's just a waste of time and money... perhaps, with the exception of larger load systems which would outstrip cheaper tank water heaters. But in those systems, efficiency makes the most sense, and you would no longer want to compromise on that in all likelihood.

I'm again willing to say that the on demand does better than I expected... but so did the tank, I would previously have pegged them both at about 75% in a heating app and really they seem closer the low 80's. Great!

AFUE is a bizarre test, agreed: but it's bizarre in a way that actually works for this comparison; less than ideal temps for the mod/con. And they still test in at 92% plus.

No one notices bad efficiencies ;) the country is full of double, triple and quadruple sized cast iron boilers and leaky furnace ducts and people just think it's normal. The fact that lots of people are doing X doesn't really compel me much. Especially with the "quality" of the people doing X in this case, in most cases... it doesn't inspire me. I don't know anyone I would consider actually competent who specs on demands in heating applications. Not that it means it's bad; it just means that because it's being done it doesn't mean it's good either.

I'm still not really sure I buy all this "internal mechanical" theory though with low mass mod/cons running reset either. The boiler is capable of reclaiming heat from itself by running without firing (that's one major beauty of decoupling heat plant firing from direct heat demands, and relating it to water temp instead); heat imparted to the rest of the boiler isn't necessarily lost until it's emitted through the jacket, and when that happens, it's highest during steady state ;)

That even helps our cast iron buddies; you can, in effect, leverage the mass of the boiler as a buffer tank of sorts. You cycle the boiler up 20 degrees and draw it down before firing again. That, IMHO, is the primary benefit of reset; not just heating everything up as hot it can go until it satisfies a demand, then ending and letting all that heat bleed out into the surrounding space.

Less important with low mass boilers to be sure; but I also have to think it has an effect on "cycling inefficiency" with such boilers.

All of this is simply to say why I would not consider a mod/cons efficiency massively derated until some 3 minute burn cycle is achieved and why, while I do use buffer tanks and advocate for them in most cases, I am not losing sleep over it when units like the Prestige offer a hedge against cycling.... with reset ;) Not that you don't need a buffer here with it, I don't know your situation specifics for zones and sizes and the like and I will spec a buffer with a prestige from time to time. But in more typical scenarios, at least, I don't need much.

Question for you though since you're doing such intense research: HOW are the on demands getting nearly 10 to 1 turndown ratios, when the boiler industry is incapable of hitting anything more than 5 to 1?

Thanks for the very rigorous conversation by the way. I've learned a lot so far.

It's hard to say where the crossovers are in the AFUE test on mod-cons between cycling losses with length of burn (IIRC 9 minutes on, 20 minutes off, or something close to that) and the sub-optimal return water temps. If your burn cycles are signficantly shorter (even at the same average duty cycle), like say 3 minutes I suspect the efficiency falls off a cliff. If the installed system is set up to guarantee minimum burns of at least that duration, along with dramatically lower return water temps, you'll likely clobber the AFUE ratings in actual performance. It's much harder (but not impossible) to beat the AFUE with cast-iron boilers.

Jacket losses per cycle as a percentage of the total ARE reduced as the duty-cycle increases- but it's a significant 2nd order effect (while the Raypak model is only first-order approximation to keep the math simple.) Cycling losses with mod-cons (AND on-demands) also include flue purges, and the shorter the cycle, the greater percentage is thrown away on both the purge chilling the heat exchanger, and heating up the appliance mass to bel lost during off-cycles. Purging losses probably exceed jacket losses on well-insulated mod-cons, even on relatively long burns.

The turndown ratio is a function of the heat exchanger design and how far you're willing to let the efficiency slip at low fire. In most mod-cons, below some minimum firing level the turbulence level starts dropping off dramatically leaving a laminar-flow resulting in a much lower heat exchange efficiency- the flue temps actually start to RISE(!) (I have a PowerPoint file clipped from a web-seminar stashed away somewhere that explains it in detail- can't seem to find it right off hand). The knee at the low-fire end is pretty sharp on most condensing heat exchangers, and dropping the efficiency from 95% to 75% (or less- depends on the design) with a 10/1 turndown would be considered an efficiency-disaster. The designs of copper tube boiler heat exchangers are less efficient overall than mod-cons on the higher-burn rate, but have a slower flatter derating curve at lower (and higher) flow. Dropping from ~85-86% PEAK combustion efficiency to 78% at both highest-fire and lowest fire is much less of a disaster for something rated with an 82-84% average. It's generally true with mod-cons that their lowest efficiency occurs at the highest fire, but the peak efficiency is well above the lowest- somewhere around 30% of max fire is the typical sweet-spot. (But manufacturers will never publish that info- you have to sniff it out from academic engineers who actually test & measure stuff.)

Looking at the curves on Raypak's site it looks like their heat exchangers are designed to deliver best efficiency at max fire, but only have 4/1 turndown, which is different from the feedback I've gotten as to the sweet-spot of on-demands from the sales-engineers (who NEVER put anything in writing, of course! :-) ) They're hinting that like mod-cons best combustion efficiency (operated as a HW heater) is somewhere in the lower end of firing range, but more like 15-25% of full fire. (Who knows what that means when you bump it up 40-60 degrees, or whether they're just making it up on the spot? :-) )

This is exactly the sort of discussion I was looking for when I asked someone to talk me down on it (you're doing a good job- thank YOU!) Thinking outside the box is good for getting perspective sometimes, but there are often good REASONS why it's outside the box too, eh? :-)

Sorry to brng this back to life, but Dana i am curious if you made a final decision on your system.  Are you going tankless for radiant with a heat exchanger for DHW?

I'm in a similar situation so trying to become as informed as possible.  My heat loss and design climate is very similar to yours, just i'm doing a suspended slab with either RHT or ROTH panels, and my sq footage is only 1200 or so..  My water temp wouldn't need to be as high as your staple up so i'm not sure how this affects efficiency of tankless versus mod con.

In plain English.

If you design your radiant panels to satisfied an accurate heat load and design water temperatures below 140F you will be in condensing mode most of the heating season, recovering 10% of the energy that would otherwise go up the flue (using non-condensing technology of any kind.

Match this with the logical and proven 'system' efficiency gained by a modulating gas valve driven by outdoor reset and the whole exercise becomes elementary.

I specify a Mod/Con replacement for a mis-applied tankless water heater every week.

Water heaters are not boilers.

That's only sensible if recovery of 10%, or even 15%, of your heating load would ever, ever, ever pay back the additional cost of a mod/con install.

In small, low load houses, that is often not the case.

You are right of course, but most of my customers go for the Green then comfort and ROI last. I do what they want.
I also heat domestic hot water, offer sealed combustion, small footprint (wall print) and silent operation.
I think most people would be happy with a 10% (even retrofits are closer to 15%) return on investment in the current market.

You know I can think past the incomparable ModCon boiler...hhehehehee

I have no problem with people going green for green's sake. As long as it is an informed decision. And you're mixing metaphors because saving 10% to 15% energy usage is not a 10% to 15% return on investment. To do that, you would have to say the $5 to $10k installed cost for the mod/con perhaps plus indirect install would return $500 to $1500/year, and that is only feasible if you have a real heat load. The whole point is that the return on investment is actually much, much lower on a low load system.

As a card carrying Green, I recognize payback is not everything. But it is something to a lot of people, and so they need the facts to decide.

Point taken.

Two well informed customers, with less-than 1200' homes, going green this week.

The geo-thermal guys don't seem to have your scruples or their customers are even greener than the ModCon crowd.

I think it is like radiant floor, the value is a subjective decision best left to the informed buyer.

Green is important to me, but at the current time efficiency, up front costs and payback is as well.  When i can afford to do a solar compliment, probably 5 years down the road, i'll be able to feel better about myself from a green perspective.
And seeing i'm confined to oil, my mod con selections are limited, and all of them are expensive, $4k plus, and then additional if i want to add DHW.
I can do a oil fed tankless for DHW and radiant, like the Toyotomi 180 with Taco X Block for $2500.  Efficiency is around 86%.  Given i need low temp, heat loss is only 30k, and heated area is only 1300 sq ft, this seems like a win win.  Am i missing something?  If you think i'm making a mistake please let me know.

As much as i would like to have a Mod Con it just doesn't seem to fit the bill.

Oil is an entirely different animal. I frankly would use a conventional water heater and I like your application of the X block. Annual burner maintenance is essential of course.

(Sorry for the radio-silence- sometimes I actually hafta take time to design stuff for a LIVING, rather than mega-slackin' on web forums... ;-) )

You're fond of dead horses, eh? :-)

I'm going with a reverse-indirect heat exchanger due to it's buffering capability in a micro-zoned already pretty-small heat load.   The indirect tank will act as the point of hydraulic separation, with all zones running as independent (separately pumped) secondaries off the tank, not the boiler.  That archicture kills several birds with a single stone,

Control of the boiler is agnostic of the state of the zones, slaved to the tank's aquastat only.  This frees up the definition of the heat source (or multiple sources) considerably.  Since I was practically given a Takagi KD20 (basically a sealed-combustion version of the T-K2), that's what I'll run with.  If it craps out in a couple of years it'll still have been worth the entertainment value of the experiment, if not, well, great!  (My wife might not find it as entertaining if it craps out on some night when it's -10F out and I can't get replacement parts for a week, but that's MY problem... ;-) )

Since I'd be suffering the standby loss of an indirect no matter what, and the standby losses of a standard tank HW heater are atrocious, it wasn't too tough to come to this.  I checked out Rob's recommendation pretty thoroughly, and although tank HW heaters have only slightly lower combustion efficiency than a tankless, they have MUCH higher standby losses, and even though nearly perfectly matched to my design-day heat load, it's economic to go with the more expensive option.  Modeled, the performance difference between a the tank & tankless over a year is larger than the difference between a tankless and a mod-con. The tank HW heater is a lot cheaper option though, but since I'm out only ~$1375 in hardware between the indirect + the KD20, with better than $200/year better performance (at current NG pricing) if it lasts 5 years it's more than paid for itself (and it's more fun to design around.) Combustion efficiency isn't the same as operating cycling-efficiency- 2-3x the standby loss KILLs at anything below a 30% duty cycle, where it would live for half the heating season or more.

Since it's output temp is fully programmable, I can pretty much right-size the burner no matter what the flow rate- just set the temp high enough to get the BTUs/hr I need. It doesn' t have to go with a heavy-hitter of a pump on the primary- I just need enough to guarnatee ignition- the highest head it'll be driving is the tankless- the heat exchanger/buffer's head isn't very much a at all (and nothing like 300' of PEX) It'll modulate some, despite designed-in turbulence in the buffer tank (for better heat-exchanger efficiency on the DHW coils) that doesn't allow huge stratification of return water/boiler water. The biggest efficiency gains will be from guaranteed minimum burn cycles (from the buffering) combined with the low standby losses of the tank.  Modulation is a secondary, but still significant factor in a system like this, but it's really all about the low-mass burner combined with low standby losses.  By being able to PROGRAM the burn rate, you can fine tune it (even adjust it seasonally, if you care to.)

I don't necessarily recommend this to the less-adventurous- parts & support might be hard to find.  But being an engineer who still has some plumbing skills I'll chance it. YMMD.  There is a local Takagi distributor, but I have no idea how deep their spare parts stock is. It can work if you can find (or are willing to undertake our own) local tech support, but your average heating & plumbing contractor is apt to think you nuts whenever there's a problem (and tell you all about how you should have gone with whatever boiler manufacturer took him golfing & sailing last year. :-) )

In the current subsidy-climate, I'll get more fuel-use reduction for the delta-$ between the kludgey tankless/buffer system and a mod-con + indirect by putting the money into planned upgrades to the building envelope.  If at a later date I install a mod-con in it's stead, if the buffer isn't stratifing sufficiently I can better isolate the return water with a small buffer (5 gallons would do) in series with the buffer. If I set the buffer tank aquastat to 130F and can get a 15-20F delta-T or better out of the bigger zones, combustion efficiencies will still be in the low to mid-90s.  I might get into the high-90s with a lower temp system, but at under 700 therms/year, who cares? The difference between 93% and 96% average combustion is something like 20 therms, and dissappears behind other system losses (or gains from alternative thermal inputs to the house. Thermal air panel, anyone? :-) )  To appreciate how subtle  difference that is in my circumstance, consider this:  A standing pilot on a tank HW heater burns 75-85 therms/year.

I'm hoping to eventually drive the system solely with a micro-cogenerator (as yet unavailable), but there's currently more incentive (subsidy, and easily attackable state-of-current insulation & infiltration reality) to continue to invest in boosting the efficiency of the building envelope, more so than getting the last 10% of efficiency out of the heating system.  Setting it up as a generic thermal buffer, future changes to the burner/heat source are easy. 

I must be the slowest guy in the world, because I'm *still* not getting how the *measured* efficiency of a tankless vs a tank.. in the very study you posted to this forum... shows that the difference is very small. How then, did they fail to measure standby loss in that study?

I thought we were trusted people who measured stuff dana ;)

Combustion efficiency is a measurement of efficiency at steady-state burn- it specifically does NOT include cycling loss or standby loss.  The as-measured combustion efficiency (not seasonal efficiency) in the eKoComfort experiment was well over 80% for the tankless, under 80% for the tank.

Standby losses factor very heavily in how rapidly efficiency degrades with duty cycle.  (The AFUE type test vs. a steady state combustion efficiency test.)  The standby loss of a tank is much higher than the standby loss of a tankless, and the burner size of the tank is fixed, whereas with a tankless in a heating app it can be adjusted to match peak load, increasing it's duty cycle, deriving even more AFUE-type test efficiency. 

The fact that the total seasonal NG consumption between the two similar apartments in the eKoComfort test were similar obscures rather than highlights any real performance differences.  The apartments had differing/changing numbers of occupants, and no attempt was made to monitor total hot water consumption or actual indoor air temp, etc.  Even with those large measurement-error factors, the tankless beat the tank by about 10% IIRC, but that aspect was an apples & pears comparison, not a controlled experiment, hardly definitive.  The true performance delta was left unresolved.

The ekoComfort article noted (but did not quantify) that the efficiency of the tank system degraded significantly more rapidly in the spring/fall shoulder seasons than did the tankless, which is consistent with how this would model based on differences on cycling losses.  But it was a different heating system configuration than mine too, eh?  Buffering the tankless in a micro-zoned radiation should make it run  at LEAST 5-10% more efficiently than in the single-zoned airhandler system they had set up.  In my heating system it would have the high-mass advantage of the tank, with far lower standby losses.  The standby loss of the tank heater itself is comparable to medium-mass cast iron boiler.  (Actual standby losses of differing HW heater types can be inferred from  test data reports from CA utilities in HW heating tests using different HW use profiles, which is how I came up with a standby loss for a generic tank HW heater. I modeled the standby of a generic as better than a cast-iron boiler, but may have been too generous). The standby losses of a tankless is by it's very nature extremely low, comparable to a low-mass mod-con (but I use RayPak's numbers for standby on a somewhat higher-mass copper tube boiler).

The standby loss of the buffer tank can be made almost arbitrarily small, but the standby loss of tank-heaters from flue convection and the insulation-gap at the burner will always remain.  This is why gas-fired tank heaters fall off an efficiency cliff in low HW-use profiles. By contrast, in HW heating tests tankless heaters only become less efficient when low volume short-draws (handwashing,etc) become a significant fraction of the total volume use (where they still beat the tanks overall, independent of use-profile.)  In the buffer-tank-heat-exchanger configuration, efficiency robbing short burns just never happen- ever!  (Not so in the eKoComfort air-handler combi system.)

So yeah, I DO trust peops who measure stuff, but I don't cherry-pick datapoints or infer too much from single measurement, don't assume that steady-state combustion efficiency will be anything like the average seasonal efficiency when used to serve a particular load (particularly when the design-day load is smaller than the burner output!) , nor to I take too much from the uncontrolled aspects of the experiment.  I tend to trust the math, when I think the model is close enough to count. I'll take the data every time, if the collection methodology is adequate.  I trust that the combustion efficiency measurements in the eKoComfort article, if not their absolute numbers- as long as the CE tests, were conducted comparably, using the same equipment & technicians, their relative (not absolute) CE is known.  The rest of the relevant information they didn't really quantify for us, (but utility engineers in California, and boiler engineers at Patterson Kelly & RayPak did, or at least close-enough. :-) )  It's the standby losses that make the REAL difference in the anticipated heating-system performance, not the raw combustion efficiency, even though the tankless has the advantage there too.

But the tank is a lot cheaper up front fer sher!

that's all fair, but I think you're overly discounting the fact that flue losses must be pretty much static in a tank, or it would reflect directly in the combustion efficiency numbers... as your usage increases, as in any heating load (decreasing in the shoulder seasons, as you note) situation, the static value (relatively) of the standby loss does not increase by a similar rate.

Your article placed the tank at 80% plus in the final total: the tankless, only about 3% higher, and that was attached to a heat emitter that could actually dump the tankless's modulated output in real time, only short cycling off the thermostat, not on water temp like it would in most radiant situations. I think you might be overrating the amount of short cycling on that system as well.

In short, I think you're "double dipping" on shortchanging the tank. The article was not measuring steady state efficiency, and variability in the domestic demand.. compared to heat load usage... should be fairly negligible in the yearly energy usage calculation. Nothing measured should be taken to the "one percent resolution" level of credibility anyway.

I'm very comfortable that both operate in similar ranges on a yearly energy usage standpoint when asked to do heating. Unless, of course, you have a short cycle issue on the tankless, then it would be worse ;).

From that, also don't forget any differences in pumping energy in your comparison as well. That could easily total $15/month difference in favor of the tank water heater... big pump (a la 26-99), vs very small (15-18) variable speed pump, at least if piped as typical for these units. Not huge, but noticeable.

Mayhaps you're mis-remembering the article? 

The numbers on the tank were not the final total efficiency, and was the measured steady-state combustion-efficiency!

http://dsp-psd.pwgsc.gc.ca/collection_2007/cmhc-schl/nh18-22/NH18-22-106-108E.pdf

"Combustion efficiency is an indicator of the performance of the system
while the burner is operating. It does not account for any system losses
during standby or any losses in the distribution systems (for example,
heat lost from a hot water storage tank, pipes or ductwork). Site
measurements to determine combustion efficiency were taken with a
portable combustion analyzer and are summarized in Tables 1 and 2."

(emphasis mine)

In the tables (difficult to place here), the Rinnai hits 84-86% during DHW loads, 83% during heating, compared to 79-80% for the tank. Not much of a difference, you say?

RONG! That's raw combustion efficiency. The standby losses of tanks are more than 4x that of tankless burners (easily derived from other sources.)  The only time the two are at all comparable is while actively burning at high duty cycles, which is still less than 20% of the year even were the burner is perfectly matched to the design-day load.

They went on to say:

"Overall thermal efficiency was calculated for various time intervals during
the project by summing the space and water heating energy outputs from
each system and dividing the total by the energy input. The performance
of the two systems was similar during periods with high loads, with the
overall performance of both systems falling with reduced system loading.
However, the efficiency of the tank-based system fell faster than the
efficiency of the tankless system as the load was reduced. This may be
attributed to the constant standby losses from the storage tank, which
has a greater impact on system efficiency when the loads are low."

(again, emphasis mine)

The only time they were similar is at high duty cycle during the peak of the heating season.  So how much of a difference does this add up to on an annual basis?

In my case it turns out to be plenty.

A clearer, simpler measurement of annual standby losses for a forced-draft tank heater lives here, and is consistent with other sources:

http://www.builditsolar.com/References/Measurements/MotorOnTime.htm

200 therms/year standby losses, compared to under 50 for the tankless- it's a large fraction of the annual fuel usage when you're burning well under 1000 therms/year.

With a tank you start out 3% less efficient during active burner cycles, then blow another 10%+ on standby losses.  It ends up being a greater annual difference in fuel use than going with an 83% combustion-efficient low-mass burner like a tankess vs. going with a high-efficiency low-mass burner like a mod-con.

The pump load argument is also overstated here. To add up to $15/month difference in my market (18cents/kwh) means the primary pump would have to run 24/7 at 116watts.  A: it won't run continuously. B: it won't take over 116 watts of pump in the first place.  It only needs to have sufficient flow to reliably induce ignition on the tankless- 2gpm would be plenty (Takagi sez 0.75gpm), then the delta-T (or more precisely, the output temp of the tankless) can be programmed accordingly.  At 2gpm the output of the tankless serves my design-day load handily, set to 150F,  since the return-water temp will still be 120F or less.  If the tankless is higher head than that, crank it up to 180F- the distribution losses from the piping between the tankless & the buffer will be low, and the hit in combustion efficiency minor.  It would only need a monster pump on the primary if the water temp out of the tankless needed to be the same as in the secondaries, but here the buffer tank/heat exchanger is the hydraulic separator, with significant mass & mixing at the top of the tank.  I'm sure it takes a much heftier pump to make it work with micro-mass hydraulic sepatation, with a primary that HAS to run whenever the secondaries are operating though.  (Remember, by design the primary on this architecture only responds to the tank temp, and is ever agnostic of the state of the zone pumps.)

Your water heater pump will run quite a lot in most cases; you're buffering, and it's got a fairly low min mod. That translates into longer burn cycles, which was your goal to begin with, was it not? if your particular unit doesn't need higher flow rates then so be it, typically though just to run this class of heat sources you're talking about a 26-99 to get much of anything out of them. They are generally expecting 60psi for just 3 or 4 GPM, after all. But if you have those numbers where you need it... rock on ;)

On the article, you got me cold, I was misremembering that article. Thanks for re-posting it.

But that tank draft loss situation in the second link was for a year-round, apparently continuous domestic recirc system: 200 of the 260 gallons of "wasted" energy they recorded was because of the recirc system, they said. That's year, round, so you would cut that in half if you called he heating system a "recirc" system, and in their case, recirc itself is wasting heat (the hot water line is unfinned baseboard) that yours is not because you are actually using the heat that is recirculated.

In fact, the post purge of the tankless seems to (by your numbers you just posted with the 50 therm "standby loss") seems to be pretty similar to the 60 gallon reminder of "waste energy" on the tank that was not directly attributed to the increase in firing caused by the recirc.

So you are being unnecessarily unkind to the tank, I still conclude (though my initial reasoning was wrong). and 5% ballpark difference on the efficiency, on the kind of loads we are talking about is really fairly inconsequential... if it wasn't, you'd be better off putting in a mod/con.

get a tank, man!!! ;)

Oh yeah, I forgot to mention:  The fact that the KD20 is sealed-combustion adds another handful of percent of as-operated efficiency over a generic tank...

A primary reason I'm opting for this Rube-Goldberg contraption over a cheap tank isn't the raw efficiency, but the flexibility of the system architecture.  With 400lbs of buffer the beast isn't very fussy about the heat source or temperature, about anything can be retrofitted onto it without changing it's function much (even multiple heat sources.)  As anticipated envelope improvements fall into place, standby losses become an never larger portion of the total fuel use.  A well-insulated buffer tank keeps it under control at the outset.

OOPS! I referred to the rong page (and even made quick & dirty calc base on it in my haste. MEA CULPA!! )

The real page was:

http://www.builditsolar.com/Projects/Conservation/Recirc/RecircEnergy.htm

"

With the recirculation loop off: burner on time was 5.1 minutes every 5 hours.

This is 1.0 minutes/hour, or (1.0/60 hr/hr)(24 hr/day)(365 day/yr) = 146 hr/year

 

The water heater burner input rate is 40,000 BTU/hr."

So that's 146 x 0.4= 58 therms/year standby (I thought the other number was high, but my simulation model & related doc now lives on my home machine so I couldn't refer to it quickly.)

...and the standby of tankless + buffer (from other sources not in evidence, just my fallible memory :-) ) is less than 58/4=14.5 therms/year

...for a delta of over 40therms, or about 5-6% of the total fuel use on standby factor alone.

Then there's the addtional ~4-5% difference in fuel use based soley on the 83% vs. 79-80% raw combustion efficiency...

...and another ~5% difference from going sealed combustion...

....all adding up to about a 15% total reduction in fuel use going with the tankless beastie + buffer.

Is going with the Rube Goldberg contraption worth it purely on fuel savings on the order of 120 therms, or $200/year at current NG pricing?

Maybe not.  (It is still greater than the sub-100 therm delta of going from this to a 95% combustion-efficiency mod-con though.) 

But it's more flexible/tunable as the heating load falls with ongoing envelope improvements, more amenable to retrofitting other inputs etc. I like the system architecture of a central buffer & microzones more than the burner itself, but the programmability/adaptability of the burner to the shrinking heat load makes a difference. (I'm not all that far from being able to run the whole shebang solely on a Honda cogenerator, at which point I could toss the Takagi and get REAL payback!  The Marathon cogen is alread too big for both my heat & power loads- it would be a waste.)

Pump power:  Takagi recommends keeping it between 1-2gpm for recirculation configurations:

http://www.takagi.com/?p=installation_requirements.php&page_id=66 (scroll down a bit)

They talk a bit out of both sides of the mouth, since they say 1/12 hp minimum (the Japanese-English is less then clear & complete) then talk about Taco-003 to Taco 009 (or Grundfos 15-42 to 26-96) depending on the head of the intervening plumbing.  The head of the buffer is quite low, as is the 5' of fat copper running between it and the KD20- I think we're safe going with a sub-100 Watter here, but a multi-speed probably isn't a bad idea since they don't specify the head of the heaters themselves.

The lowest modulated input is 19KBTU/h, so yes, it'll run with high duty cycle during the coldest days and weeks, but probably no more than 50% on average December through March, half the power consumption you were talking or less.  (I s'pose I have to compare the draft blower & control power of the KD20 vs. an alternate forced-draft tank too to be fair, but it won't be a $15/month delta.)  The modulation needs to be set higher than the minimum to meet my current design-day load (and to have some reasonable buffer tank recovery speed so's my wife doesn't scream at me from the shower... :-) ).  But it's tweakable.