5......no.......2.......no.........say what?
Disclaimer: I am writing this with the best possible intentions. I've re-read the answer that I have written and it comes across as very patronising and very rude.....but I promise you, that is only because it's impossible to put "tone" into a post and because it is impossible to be succinct without sounding like a rude and patronising b@stard. Please believe me when I say that I wrote the below in an attempt to help you.
You're a new member and before you get jumped on for asking a pretty blunt question (we're on page 171 of this thread alone .....so clearly you haven't taken the time to read any of the reams of information that is contained on this thread)......I'll give you some variables:
1) What are you measuring the 400W HPS on? PAR? umol/m2/s? Lumens? Unicorn rainbow farts? Serious question though
2) How far will you keep the LEDs from your canopy? (there is a very well known degradation of light/distance equation......oh just look it up)
3) What sort of reflectors/spread are you looking for?
4) How hard will you drive the LEDs?
5) What sort of efficiencies are you looking to gain?
6) What sort of bulb are you currently using?
7) Try asking this question instead:
"I currently have a 400W HPS from "put name of bulb here". It runs at "put % here" efficiency, meaning I get "put actual radiant W value here". As I have to place the bulb "put height above canopy here" above my canopy, the actual PAR strength of the light I am getting at canopy level is "put umol/m2/s PAR value here". Currently the light spread is "put spread size and some indication of difference between edge and centre" and I would like to achieve "put the same/better here" in a space that is "put size of space here".
I would like to replicate the spectral wavelength distribution of my HPS bulb "put the specific brand of bulb specification sheet figures here" and so I am looking at "put the specific temperature of CXA3070 you are looking at here and then also consider the different Bins that are available". In order to get a similar efficiency level, with a similar PAR strength, over a similar size space, with a similar spectral wavelength distribution.....how many CXA3070s would I need and what would the ratio be between the various temperatures?"
Now assuming you have actually bothered to read all of this......my question that I prompted you to ask above is what you just asked the people who have spent literally YEARS testing, evaluating and publishing their results......only all of the information anybody would need to answer your question is missing (hence my "put this information here" prompts).
But allow me to make this blindingly easy.
Unless you know precisely what sort of performance you're getting from your SPECIFIC HPS bulb/reflector set-up, then ANY kind of attempt to answer your vague question will be a lie.