As I understand it, in residential systems, tankless water heater are commonly paired with a buffer tank to deal with the "cold sandwich" and finite capacity of tankless heater1. This seems reasonable, but I'm curious as to why the only config I see is a buffer tank at the output of the tankless. That is, why is a buffer at input of tankless heater seemingly not an option?

It seems to me that keeping a buffer of hot water at the output negates one primary benefit of a tankless system: energy-efficiency. As such, the system is obligated to keep the reserve hot 24/7, thereby generating energy waste.

On the other hand, keeping a reserve at the input, the system can selectively heat the reserve when the demand is expected to be the highest (a couple of hrs a day?) and neglect it at other times. And in the worst case, if demand is unpredictable, the system could simply be programmed to maintain a hot input buffer 24/7, just as with buffer at the output. But in either case, an additional opportunity for efficiency with an input buffer is by maintaining a lower reserve temperature.

Under limited circumstances, I think an input buffer tank may not be an adequate solution to the "cold sandwich", but assuming the system could be programmed to minimize the problem during certain time of day, in a typical residential system, is there a really compelling reason to keep buffer at tankless output? Is there historical or empirical data to back the choice to put buffer on output?


A buffer tank must be on the output side of a tankless water heater. If you put the tank on the input side a bunch of unintended consequences happen:

  • The buffer tank will do the bulk of the water heating, until it is overwhelmed.
  • As long as hot water use continues, the buffer tank will continue running as it tries to keep up with raising the incoming water to its set temperature.
  • You might still get a "cold water sandwich" as the buffer's hot water supply runs out. That's because your tankless will be sitting around doing nothing at all while the hot water runs, because the water's already heated! When the buffer runs out the temperature drops and the tankless kicks on (but might need more time to warm up fully).

On the other hand, when you put a buffer tank on the output side of the tankless, sustained use of hot water will shift most of the work to the tankless. It will be heating the water before it reaches the buffer, and the buffer will do nothing during those extended uses of hot water where the tankless is most efficient.

If you want to add a timer on the buffer, the way to do that would be to both shut off the buffer tank's heating and route the water around the buffer (rather than through it), using some electronic 3-way valves.

Please note, however, that there are better solutions to the cold water delay problem. For example:

  • Use a demand-activated recirculation system. You have to train your family to press a button a minute or two before they want to start a shower, but it's pretty nice having the water hot immediately after.
  • Configure your tankless water heater to use a recirculating mode that keeps its heat exchanger warm. This will consume more energy, but is designed to ensure it's pre-heated during periods of likely demand. My Navien CH-240, for example, can be configured to run in recirculating mode during specified 30-minute intervals via its remote control panel.
  • The 3 way around the buffer when off or below temp makes sense to me I had not seen this in the past but could use the idea. + , I usually just add a second smaller point of use tankless where it is needed but this provides another option. – Ed Beal Nov 28 '17 at 17:34
  • @Rhura Seems to me that 2 of 3 "unidentend consequences" listed are more problematic with buffer on output, not input. The 3rd is unlikely given rate of change in storage water temp is likely slow. By putting hot water store at input, means buffer can be made, with timer+record, to heavy lifting only when needed. Other times, it is effectively a big fat pipe. Hot water store on output must maintain temp ALL the time, even when not needed...or you add complexity like your "3-way valve" – codechimp Jan 9 '18 at 12:56
  • Why? If a tank is attached to the tankless output, then the water coming into it will be pre-heated by the tankless heater and the tank won't need to do much. – Shimon Rura Jan 9 '18 at 21:14
  • Note that a much better approach for a buffer tank -- rather than adding a second heater -- is to use an insulated, non-heating tank along with a thermostatically controlled recirculation pump. – Shimon Rura Jan 9 '18 at 21:16

I have never seen a buffer on the output of a tankless system this would be a waste of energy in my opinion , adding a tank on the input side with a timmer would make sense to me , I have added a second smaller tankless at the point of use when the whole house tankless was not enough. If the buffer was on the output side during the time the timer was off it would take a lot of water going through the tank before it got warm.

  • See Shimon Rura's snswer for why an input side buffer tank doesn't work. Personally I can't wait till he day that we do away with centralized water heating and simply make it a function of the faucet. Sure that changes the way a house is wired, but over time there is massive savings, especially in climates where waste heat doesn't help much of the year. – Tyson Nov 28 '17 at 16:11
  • Agreed that is a good option if using a tank. – Ed Beal Nov 28 '17 at 17:36

TLDR: that link is crai. Just ignore it.

No. Just no.

The rest of what you're saying is totally right. The opinion you linked falls in the "everybody's got an opinion" category. I could not agree with it less. Just goes to show any opinion can get published these days.

His entire thesis rests on a single use-case which I don't believe actually exists, and would be easy to mitigate if it did exist.

His use-case is that hot water use discontinues for a short time, and then is used again. In that time, the water in the water pipe he insulated does not cool down, and (for some reason??) the water inside the heater's own heat exchanger does cool down. (Why now?) There is no reasonable explanation why the heat exchanger would cool faster than the pipe.

But even if there was, there'd be a simple cure: don't insulate the pipe. Duh. If the heater is cooling off so fast, ambient temperatures must be low. If ambient temperatures are low, let the pipe cool off too. Now you have the normal situation of cool water in the pipe until the hot water reaches the point of use.

And one more thing. Why is ambient temperature so low? Doesn't water piping have to be inside to avoid freezing??? I thought he also mentioned he liked to install in crawlspaces. Note the insulation above his head in the crawlspace. This all adds up to him installing in the sunbelt, San Diego or Miami, where he thinks stuff won't freeze. That's a weird use-case all by itself.

His aversion to electric heaters is another warning sign. Sure, they require a lot of service, but it's no big deal if you have it. In fact it works better because you can put the heater at the point of use (since they're so cheap, you can have two) and dispense with the slug of cold water entirely. No slug of cold water in electrical wire.

Note also his interest in radiant heating and Taco this and that. Either this guy's a builder (which I suspect from "our") knocking together hundreds of homes to a stock formula... or he's very rigid in his thinking and is finding inventive ways to resist good ideas. This isn't that hard, Europe has been doing it for decades.

Bottom line: feel free to install a tankless. Don't bother installing a tanked also. If you're that worried about it, omit insulation between heater and point of use.

  • What is this responding to? Maybe something got deleted, but I don't see a link in the question or any related comments. – Shimon Rura Nov 28 '17 at 3:48
  • @ShimonRura End of first sentence in OP, 1 character wide, check it out, you will laugh... – Harper Nov 28 '17 at 14:15

Buffer tank on the output side works wonders.

I have a Rinnai tankless water heater and a recirculation system, and I was annoyed by the cold water sandwich effects. They happen when cold water from the mains hits the tankless water heater and it doesn't have time to react to change in the temperature.

A small 10 liter tank on the output side solved this issue entirely. I can still detect a small temperature dip when the cold water passes through the heater, but it's entirely ignorable.

  • I'm sure an output buffer tank works to reduce cold water sandwich. The question,however, is whether a buffet tank on input is MORE efficient and/or have more drawbacks a compared to tank on output. – codechimp Jan 9 '18 at 12:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.