BTW I also asked that same friend of mine if he knew of any good papers on WI, perhaps published in SAE or the like. He said there is not much there since for the most part the OEM Engineers have not been playing with it, and neither has the racing community. Plus in racing if someone finds a performance advantage they are usually motivated NOT to publish it for obvious reasons. That is why a lot of the WI information is from the WW-II era piston engine performance development like the article that Joe linked to. I suspect that the information was closely guarded during the war, but was released once everyone switched over to jet engines and it did not matter much anymore anyway.

Is his name Mr. Flow Bench? :jerkit:

I have a gutted manifold as well but I decided not to put it on because of what I said. I don't want stupid shit to happen to my motor because of stupid shit.

Have you tested this at the typical injector pulsing frequencies of 4000-6000 RM (30~50 Hz)?

WI was used in F1 during the turbo days. Granted the didn't even turn it on until like 30PSI, but they were running like 60+PSI.

I would listen to this post before trusting sound engineering judgement...

No I think that is your name? Seriously, what is the basis for the assumption that the manifold is going to be a problem? I'll wager that my guy knows more about manifold design than most of us here. So his assessment that it is OK carries more weight than the assessment that it is not OK, at least until someone comes up with something better than changing the stock configuration = doom and gloom.

So, in the mean time, let's keep the manifold aspect of this out of the discussion unless it relates to the WI.

Not yet. Have you?

agreed, thanks

The Adaptronic has built-in knock sensing and retard. Just wire in the knock sensor. However the tuning aspect of it is nothing beyond threshold versus RPM. I think Jeff's filter circuit could be helpful at the least. I also had a knocksense with the MS that seemed to work well.

EDIT: OK nevermind, found the thread assuming you mean this one:
https://www.miataturbo.net/forum/t41877/
Yep I already did something like that a while back using a cheap radio shack amplifier.
https://www.miataturbo.net/forum/t22322/
I can either listen in on its speaker, or I send the output to one of my stereo inputs, where I can apply 16 band equalization. Boosting the upper midrange frequencies seems to make it easier to hear knock. In fact, I installed two knock sensors on the motor. One for the Adaptronic, and one for the electronic det can. That way the input impedances do not interfere with each other.

Agreed. I like the idea so I will let you know if I find them anywhere, though I would have looked at the same places you did for starters. Maybe Omega though their pricing tends to be high.

Here is a flow switch from Omega that would work though the price point is too high
Liquid Micro-Flow Switches

Originally Posted by JasonC SBB
Have you tested this at the typical injector pulsing frequencies of 4000-6000 RM (30~50 Hz)?
Not yet. Have you?

Of course not. If I had, I'd have stated my results instead of asking stupid questions.

Re: knock - the miata's main knocks frequency is 13 kHz in my testing. However I recorded knock off-idle. If someone has knock recordings under heavy load I would like to analyze those waveforms.

I posted a 2nd order bandpass filter schematic centered at 13 kHz.

What does Jeff's circuit look like?

I think it is just an audio amplifier. I went through that thread quickly however. I think the filter I was remembering is yours.

Do they make an electronic check valve? Basically a valve that closes or opens a circuit when it's activated? I would think you could tie this in with the PWM so that anytime the PWM is active, it would make sure the check valves are opening. If they aren't opening, then the nozzle is clogged causing a high back pressure. I would think that would be much simpler and cheaper than a flow meter (assuming it exists).

Correct. I left filtering out of the mix due to the fact that I'm trying to keep cost low and the price of small batches of custom made PCBs is a bit prohibitive. As an "entry level" knock detection device, an amp works suitably well. Worlds better than not having anything at all, but not as nice as a phormula or a gizzmo.

I think what you are describing is the same as what gospeed is suggesting.

Agreed. I am happy with my electric det can setup. Reliable, sensitive electronic knock detection is tough, but it is pretty easy to hear it with this setup. It cannot be on all the time of course but for occasional use, like for tuning, it is great.

Yes, exactly the same, yet still different from the $200 valve you posted. That looks to have enough resolution to serve as a low flow meter, it's really just another flow meter. I guess you could use something with a turbine wheel in it, but they will all be just as expensive.

I'm thinking of a contact-based ball type one-way valve that simply says "hey dummy, there's no flow here" or not. We don't care so much about knowing the actual flow rate, as that's all calculated ahead of time, and dependable as long as there are no busted lines or clogged nozzles.

There are things like paddle switches but they are too large, at least the ones I have seen. That device is not quite a paddle wheel flow meter in that it does not measure flow rate. It just uses the movement of the wheel to trip a switch. I agree it is still too expensive though. I threw it out there as more of an 'FYI' than a 'here is something'.

I'll see if I can find something else. There are so many practical minded geniuses where I work that somebody may know of something offhand that would do the job. I get a lot of those "Oh yeah I know just what you need. Go here and get this..." eureka moments courtesy of my colleagues. The mechanical requirements are so simple there has to be something cheap out there somewhere.

For the PWM variable flow usage, I would basically have to set it at the lower flow threshold so that when I am running at 20% DC (or whatever the minimum flow rate for good atomization is) it will let me know if I have an obstruction. It would basically be a necessary but not sufficient indicator of proper flow. Still very useful though.

Yes, I did. I fixed the link, and here it is again: http://ntrs.nasa.gov/archive/nasa/ca...1993093245.pdf


In a perfect world, I'd think you'd have one HSV per nozzle, with the nozzle screwed directly into the HSV. Failing that, you'd use semi-rigid metallic tubing.

HOWEVER, I'd love to see ZX do a test on his system, while it's still out of the car, where he drives the HSV at various duty cycles at its intended operating frequency, and makes two observations: First, does the volume of water actually emitted match the predicted flow based on the PWM table, and second, is the atomization quality consistent even at low flow rates? If both of these tests return true, then party on. If not, then I'd make the case for re-testing withrigid metallic tubing and compression fittings.


This argument is inconsistent.

I did this test on the bench with everything level, so there's no vacuum or gravity at work. And it wasn't just a few drips- the nozzle was very clearly spraying after the pump was shut off. You nailed it when you guessed "pressure equalizing to atmospheric", however you need to figure out the reason that there was any pressure to equalize. If the tubing was in fact non-expandable, then as soon as the pump shut off there's be no more pressure to equalize. Water is non-compressible, at least, not at the pressures we're working at. The reason there was pressure was that the tubing was expansing slightly and acting as a pressure reservoir. Someday I'll measure it with a micrometer.

Oh, and the tubing is nylon.


That's very true- this stuff was classified back in the day. There has been some recent work done on the subject (I've got one relatively modern SAE paper on it somewhere) however you're correct in that WI sort of fell out of vogue for some reason.

Like I said, I've got a whole library of this stuff, I've just been too damn lazy to collate it and publish. One of these years...

Ok. Assuming everything you stated happens with all setups, all I see is a positive more so that a negative. Assume the pump is running, and the HSV just went into the closed part of its cycle. Since the tube acts as a pressure reservoir the nozzle continues to output a fine mist under near full pressure for the few ms until it is open again for the next cycle. All it is doing is extending the spray to be continuous instead of truly pulsating. The only way to fully stop any pressure reservoir and leaking would be to have not just a hard line but also a check valve right at the valve as well. Even with a hardline eliminating the effect you speak of, without a check valve water will be siphoned off after the pump cuts off. It is not really a big deal for that though. I will agree that you want to maintain max pressure at all times though and minimize any dips.

A true HSV that is like a fuel injector would be able to cycle around 200Hz. Do you think that the pressure drop over the maximum 5ms time period until the next cycle is that significant? My guess is that it is not.

Yes I most definitely want to do those very tests and report the results for you guys. I have the whole bench setup/procedure in my head and I also want to try PWM frequency as another variable. All I need is a way to bench trigger the solenoid.

So I had another thought on that. Joe maybe you can help. As mentioned previously I found a shareware PC app that creates a PWM square wave with variable frequency and DC. So, is there a quick circuit I can rig up with a transistor that would switch the transistor in response to what is essentially a PC audio out (headphone jack) signal? It would have to handle about 12W (P=V^2/R) at 100% DC on the load side. I would just have the transistor either switch to ground or switch to 12V DC, and vice versa the other end of the solenoid coil. I think 300 Hz would be the highest freq I would even think about trying, more like 200 Hz which is what I believe the valve is rated for. Too fast/too many cycles for a mechanical relay but no problem for solid-state.

This has to be a pretty simple analog circuit I can build with cheap Radio Shack parts right?

Cheap and radio shack parts don't belong in the same sentence ;)

I'm not disagreeing here. Honestly, I'm having a hard time visualizing the effect of this arrangement. It will certainly contribute to turn-on lag, but I don't know what its steady-state characteristics will be. That's why I think a test will be valuable.

Yeah I know their stuff is really inflated. :)
But if it is only $20 worth of crap I'm not worried enough about price to go to digikey or the like.

Agreed. One good test is worth a thousand expert opinions. Give me a good circuit design and I should have it up and running pretty quick. I cannot work on this this weekend, but next weekend should be no problem.

Something like this would work: DPRG: A Simple PWM Circuit Based on the 555 Timer

Agreed. But assuming that the PC PWM output approach works then all I really need is the right-hand part of the circuit, right? That is, the transistor and whatever resistors etc. to support it. I know squat about transistor switch circuits. I could google and read up on it but I am hoping someone knowledgeable can save me that trouble and sketch something up off the top of their head.

The audio method you speak of would trigger the irf fet. You would need that and a dc power supply that can handle 12W. Also might need a heat sink on that fet. For power supply, if you have a pc power supply sitting around that would work perfectly.

this is the datasheet for that FET: http://www.audiolabga.com/pdf/IRFZ46N.pdf

You will at least need a diode (unless you care to fry your headphone jack) and probably a 1k resistor as a load to make it switch to the headphone jack. I'm not sure what kind of signal you'll get out of it. You might want to use a scope to see what the app will produce and compare it to the FET datasheet.

Wait...

You're planning to drive this with your Adaptronic, right? Can't you just do that?

Well I considered that but that is inconvenient for a few reasons. One I do not have a spare harness for it, and two, I do not have a means to stimulate it, like a jimstim or whatever the MS uses. I am not sure that I can get it to fire the outputs with the laptop tuning software with no trigger inputs. I am going to look into that though.

Oh. I assumed that the Adaptronic could generate its own free-running sync pulses for injector testing like the MS2 can.

Even still, assuming it's currently wired up and running (the Adaptronic to the engine, I mean) you could just run the one extra wire for the HSV and simply extend the wire over to the bench. Have somebody hold the engine at a constant speed, set the HSV drive to whatever setting you want to test.

Oh Snap! Was that a shot across the bow from the MS camp? :giggle:

Well, one, the engine is not even in the car right now, and two, I want to be able to run it independently of the vehicle. I am working on something that would require testing in a lab with no car.

ZX-Tex you're better off using the parallel printer port rather than the headphone output.

The headphone output will have a hipass filter output so the pulse train will be very distorted. Itll be a pain to build a circuit that can correctly read it to drive the MOSFET.

Good idea but the shareware software I found generates an audio signal.

I had a friend build me a very simple transistor circuit once that plugged into the headphone jack (audio signal) on a radar detector. No PCB board, just wires, solder, resistor, transistor can, maybe a diode, and shrink tubing. Built it in less than an hour with parts lying around the lab. It triggered a bright LED cluster whenever the radar beeped. I used it so I could see the radar detector go off when riding my bike (too loud to hear) and it worked great. I basically need a higher wattage version of the same thing. I'm tempted to use that setup but am thinking I might overload it.

Unfortunately that guy, an uber electronics tech, is not around here anymore so I can't tap his skillz for this. I was hoping one of you guys was uber enough to duplicate his deed :)

Ok, here's what you do.

You've got the Adaptronic itself, right? Since it's not capable of doing something that a $200 ECU based on a 25 year old 8 bit microcontroller with 512 bytes of RAM can (Ok, that's the last one, promise), we'll help it along.

Hook the HSV up to the ECU. Configure it for a single 36-1 VR crankwheel. Then, download the software and build the circuit on this page: Crankshaft Timing Signal Wheel Simulator

For the transformer, you don't need to order from Digikey. RadioShack P/N 273-1380 will suffice. You can ignore the transistor and resistor- all you need is the 1/8" plug and the transformer.

With this, you can drive the Adaptronic to any RPM. I assume you know how to simulate various MAP conditions on it.

or you could just build the 555 timer circuit...

+1

or is that, +555 = 1110.

SMACK! :owned:
Good thing it has some other awesomeness to compensate for this relative weakness.

OK I am liking that. Cheap, easy, freeware driven. I do not think I will even need the MAP sim to get this to work since the PWM output can be proportioned to whatever IIRC so I can just scale it from RPM.

I am late to this discussion and am only here to offer a little info on PWM water delivery in the ag sector. Case IH uses it on their patriot sprayers. It is called AIM Command. It might give you some ideas. You set the pressure you want to deliver the liquid at and it adjusts the pw to match the desired spray volume vs speed.

I am late to this discussion and am only here to offer a little info on PWM water delivery in the ag sector. Case IH uses it on their patriot sprayers. I'll look up the name. I'm on the iPhone and it's a bit clunky. It might give you some ideas.

Any progress ZX? Wish you were closer so I could help you out. I'm very interested to see the results for my build.

Sorry nothing to report yet. Reason being I got back to work on the LS1 Miata and made some good progress on that, but not this. I'm also really interested in the results so don't worry I'll have something as soon as I can.

Anyone knows a formula to calculate flow out of multiple nozzles?

E.g, four 70cc(1gph) nozzles, driven by a 150 psi pump, there is a checkvalve, too.

I am in the process of converting an AEM progressive kit into a port injection setup. Best I can find are the 70cc nozzles.
Instead of fabbing an aluminum manifold, I'll do what ZX-Tex has done, and go with the three Y connectors. Much simpler, less volume.
I also have a pretty sensitive knock warning system. (knockgauge.eu) Works incredibly well. I am planning on relying on that for system problem detection till I can come up with an alternative.

Try this web site for info. I don't have enough info to figure it for you. I would need to know what pressure your nozzles are rated for and how much pressure you will be running. I have a spreadsheet for this, but it is on my computer that my power supply died on this saturday.

TeeJet - Spray Pressure Information

Just multiply the flow amount by the number of nozzles. This is one thing I actually deal with as a farmer... ;) The flow pressure will be the line pressure minus the pressure inside the manifold. I am assuming you are not using a fuel pressure regulator for the WI that references the manifold pressure.

If that is too confusing you can find charts on that web site that will give you flow at various pressures for certain nozzles, but different type nozzles are rated at different pressures. Here is an example.

http://www.teejet.com/media/cfefeb5c...LoRes_p020.pdf

The TP80015 is rated at .15 GPH (yours is .166) at 40 PSI. You can use that chart to get a rough idea if your nozzle is also rated at 40 PSI. They also have metric charts on this site and there are european manufacturers.

Hope that helps some.

Thanks.

Using the formula you have provided, once the boost pressure is factored in, I end up with 327.6 cc/min. total amount for all 4 nozzles.

Works well for me.

327.6 cc/min = 5.2 gph


Should I subtract the check valve cracking value, too?
In that case, the 5.2 gph flow rate shall decrease further, of course.

What's an AEM check valve rated at?

The pressure for flow calculations needs to be taken at the nozzle with the system running. The ag sector has all kinds of controller options for these applications. Most are pressure based controllers and are tied to speed, but there is no reason they could not be tied to an RPM signal or MAP reference. However finding a flow meter for such a small amount might be difficult. Raven makes a chemical injection system that might be able to do it. It is called the "Sidekick" I think.

UPDATE:

I have the Aquamist HSV and the PWM control working. It turns out that triggering the HSV using the Adaptronic without the engine running was very easy :facepalm:

As mentioned before the Adaptronic has outputs that are easily reconfigurable in the interface software. All I did was set the HSV output to be PWM (0% to 100%) in response to the throttle position sensor (TPS). It worked very well. Moving the throttle varied the duty cycle on the HSV. I set everything up in the engine bay, including the laptop, so all of it was right in front of me. I watched the on-screen gauges in the Adaptronic interface to keep track of what the DC was. So if I wanted to do it on the benchtop I could do this though I would have to pull the ECU out of the car and wire in a spare TPS. Not too bad I suppose.

As far as how the HSV WI setup works, after playing around with it for a few minutes here is what I have seen so far:
- I am having trouble with one of the four WI nozzles. It looks like there is debris in the nozzle and I cannot get it cleaned out.
- I really think it could use an accumulator between the pump and the HSV. The pressure fluctuation from the pump is problematic for maintaining atomization at lower duty cycles. I think an accumulator would help this out.
- The accumulator effect in the hoses downstream of the HSV is negligible.
- There is definitely a minimum DC at which the mist is well atomized. I am wondering if it would be worthwhile to go to smaller nozzles and up the pressure at the pump.
- Distribution between the nozzles looks good visually.

UPDATE:

The nozzles were not clogged. I had a blockage problem with a fitting which is now fixed.

Rather than mess with the accumulator I took a different tact. I sold the standard 150 psi pump (that DO included with their kits a while back) and bought the 250 psi pump. Not so much for the increased pressure, but for better flow. The 250 psi pump has an internal bypass pressure regulator instead of a switch regulator like the 150 psi pump. So the 250 psi pump once powered up runs continuously instead of pulsing on and off like the 150 psi pump. I tested it tonight with the HSV and the flow is a whole lot smoother at lower duty cycles. Much better.

The two pumps can be seen here. Notice how the 150 psi (and 220 psi) pumps have the wires that run to the head of the pump unit. This is for the pressure regulating interrupt switch. Also notice that the 250 psi pump does not have those wires.

BTW the pump difference is probably not a big deal for the on-off setups (no HSV, full flow) because the flow is steady even with the 150 psi pump. I only switched because of the HSV setup. And even the 150 psi pump could be made to work with the HSV if an accumulator was used between the pump outlet and the HSV inlet.

I have been thinking about how to make some kind of video of the injectors running but it might be a tricky thing to capture on a standard camcorder. I was wondering if some benign fluorescent dye in the water and a blacklight would help :)

Very cool. :D

A strobe light, synced to the NTSC framerate. Maybe even a timing gun, controlled by the ECU, with the RPM set to a fixed value.

Interesting. I had not considered that.

OK Joe (or another EE) please check me here. This is a partial cross-post (italic font) from the Adaptronic board; I know not everyone goes over there. See the part towards the end (in bold) about using a SSR to tap off of the fuel injector circuit.
http://adaptronic.com.au/forum/index...sg7444#msg7444

FlyingGrape is right. The Aquamist HSV is fast. In fact I ran it at 200 Hz and it worked fine. Even at 8000 RPM (for sequential mode on a 4-cylinder) the fuel injectors are only firing at ~66 Hz if my math is right.

Slides go for it but I am not going to implement a water/lubricant emulsion system for the WI so that I can use a fuel injector. No thanks.

Cobber that is an interesting idea. Hmmm... The Aquamist HSV has the same resistance as a high impedance injector so it can be driven by an injector circuit. I am not sure if the extra load would impact the other injector in the circuit if I tapped straight off. I suppose I could isolate it with an IGBT or something like that.

I wonder if there is a setting in the Adaptronic that allows an aux output to mimic whatever the injector drives are doing. I do not remember seeing it in the manual or in WARI. That would take care of it right there with a relay to set the threshold like you suggested. There is a two-stage injection setup but I think that still requires the second map correct? I RTFMed but I am not getting it yet on exactly how to set up two-stage injection.

EDIT: OK more on Cobber's suggestion. I could use an IGBT aka solid state relay (SSR) like this one.
http://www.grainger.com/Grainger/items/1DTT2?Pid=search
- The input side of the SSR would be driven by tapping off of one of the injector drive outputs. The output side of the SSR would drive the HSV. That way the SSR/HSV would put a negligible load on the injector drive circuit.
- On the output side of the SSR, instead of going straight to ground, it could be connected to one of the Aux outputs. The aux output could be set for on-off (not PWM) based on say MAP. That way, the SSR does not pulse the HSV unless the MAP threshold has been reached. In fact this could be done on the input side of the SSR for that matter.

IT COULD WORK!
Anyone see any pitfalls here? Before anyone asks, the SSR is plenty fast enough. It changes state in 0.1 ms.

Pros:
- The HSV would automatically follow whatever the injectors are doing, and pulse at the same frequency. Plus a minimum threshold can be set.
- No need to tie up the second fueling map
- Changes in the fuel curve are automatically followed by the injector.
Cons:
- Unless I am running the injectors at 100% DC (bad idea) the HSV will never reach 100% DC. That kind of sucks actually, since the upper part of the HSV DC is where it is the most useful. So I lose range at the bottom from minimum atomization pressure, and range at the top due to max injector DC.
- The WI and FI are not independently adjustable.

Use a 60V 20A or so MOSFET, not an IGBT.

I might be using IGBT incorrectly. According to the manufacturer of that SSR I linked to above it is a FET?
http://www.crydom.com/en/Products/Ca...ard_output.pdf
I'm open to suggestion. I know little about SSRs.

EDIT: Maybe this one? According to the specs it is a MOSFET. More expensive, but can handle 12V DC inputs, and has better packaging.
http://www.grainger.com/Grainger/items/1EGK7?Pid=search

Yeah, at 8,000 RPM you are seeing ~133 revs per second, or ~66 cycles per second. So the injector frequency would be ~66 Hz for sully sequential, or ~133 Hz for a banked system with two squirts per cycle.

The SSR won't work. If you look at the datasheet for it, the rated turn-on and turn-off times are way too slow.

However, you can easily construct a circuit to mirror the main injectors. All you really need is a large transistor and a few supporting elements, and then a means of gating the control system on and off.

Here's a quick-n-dirty sketch of what I've got in mind. It's not complete, but it conveys the basic idea.
http://img38.imagefra.me/img/img38/6...6m_aca6c1b.gif
You put a positive voltage on the "ENABLE" line to turn the circuit on. That allows +12 to flow through Q1 into the second.

When the injector line is "off", the injector wire will be floating up to +12, and that'll hold Q2 off. When the ECU wants to fire an injector, it grounds the injector wire. That will allow Q2 to turn on and conduct through the HSV to ground.

Q3 and its related hardware is a flyback damping circuit which I've shamelessly stolen from B&G's copyrighted schematics for the Megasquirt.

I should probably move Q1 out of the main current path and into Q2's control path, but you get the basic idea.



EDIT looks like I missed some conversation while I was drawing. What, no love for BJTs, Jason? :D

Yeah sorry about that edit :)

I cannot find out if that other MOSFET I pointed to is fast enough. For some reason that spec is not listed by Grainger or anyone else, and I cannot find Dayton's web site (if there is one).

MOSFETs are waaaaaaaaaaaaaaaay faster than IGBTs.
We're talking MHz vs kHz.

Joe your circuit needs diodes added to prevent reverse biasing of the BE junctions... depending if those ECU outputs are pull up and down or one way only.
MOSFETs are easier to drive than BJTs and usually result in lower parts count circuits.
AND when driving solenoids some of the automotive types have Drain avalanche ratings that makes their flyback voltage clamping built in.
I am agnostic and I will choose a BJT or MOSFET as needed.

UPDATE:

OK I have it all in the car and working, even tried it out on a dyno. Some things to report.

- For now I am just using the second fuel map on the Adaptronic to control the HSV. Works great. I'll mess with the other scheme later.

- It seems to only give good atomization at ~60% duty cycle. Below that and the mist becomes more like a spray, then eventually a weak stream.

- I think four of the 2 GPH nozzles is too much. I saw a huge power drop on the dyno when all else was equal and all I did was turn on the water injection. There was about a 15-20% drop at peak horsepower*. This is with 0 deg washer fluid (no detergents) which I believe is about 20% meth. I think it is just too much water/meth, more than twice what the DO calculator calls for. Since I do not have as much usable range as I had hoped for, I cannot run the nozzles at low duty cycles to compensate for their size, which was the plan. They do not have nearly as broad a range as I had hoped.

- With 4 2GPH nozzles, the washer reservoir empties FAST.

- Because of all of the above, I am going to try four 1 gpm nozzles next. Because the nozzles are smaller I should get a broader range of flow control while maintaining good atomization. Also the peak flow (100% DC) will be less, not so much overkill.

- Also, to test out the effect of the hoses post-HSV, I ordered some very thick wall nylon tubing that I am going to try in place of the DO tubing to see if it broadens the usable range. It has a 0.050" thick wall, much thicker than the DO tubing. This should answer the question about whether wall stretch is a significant accumulator.

*Keep in mind on the above power loss, is that the WI is there on my install to provide additional knock margin. I tuned the car without the WI, and added the WI on top of that, without advancing timing or leaning the mixture. That way, if the WI fails, there is less risk of engine damage. If I had advanced timing and reduced AFR with the WI running, I am sure I could have gained back the power and then some.

OK good stuff to report...

- Switching to the 1 gpm valves really worked out well. The usable duty cycle range is now from 30%-100% which is wider than it was before (60%-100%).
- At 100 Hz modulation frequency the minimum duty cycle at which I would consider the atomization adequate was about 30%. When I dropped the DC to 50 Hz, the minimum duty cycle dropped to 17%. Interesting. Although I am still using 100 Hz since it pulses faster than the intake cycling rate at 7500 RPM.
- Switching from thin wall hoses to thick wall hoses made no difference in terms of the minimum duty cycle at which atomization was adequate. So, this indicates that the accumulator effect from hose wall stretch is negligible, at least at 100 Hz PWM frequency.
- The threshold between good atomization and bad atomization was very narrow, only about 2% DC. At the threshold, a very small change in DC brings a large change in atomization quality.

Here are some pictures of what the atomization looked like with the four 1 GPM nozzles. The first shows the atomization at 27% DC, 100Hz. The second shows the atomization at 30%, 100Hz. The images are large but they show more detail. At 27% the droplets are larger and are scattering radially all around the nozzles. At 30% they are much smaller and are being carried off by the wind (away from the camera).

Looks great!