What the seller forgot to mention #4 – why the Erchang makes you choose: fish icons or the water column

This is part four of the series in which I check how much of what you see on a fish finder screen is a measurement and how much is a drawing. In the previous episode the nameless black ball XJ02B turned out to be hardware that sends a real echo cross section, only far too coarse to judge anything by, while the app draws the bottom from a picture by default anyway.

Today the Erchang XF07C and XF07W: a ball the size of an orange, 90 grams, in a Bluetooth or Wi-Fi version, driven by the Erchang Sonar app. On the box two frequencies, 125 kHz with a 45 degree cone and 330 kHz with a 14 degree cone, and a depth range of 60 metres. This is the top shelf among cheap sonar balls and the first of those described here that promises something the others did not have.

In the previous episode I promised to check whether this family of sondes really has two frequencies. It does. Genuinely. And that is not the only surprise, because the Erchang is the first sonde in this series that draws its echogram from a real measurement and not from a picture. Except that in the factory settings it wipes that very measurement off the screen to make room for the fish icons, and that is exactly the subject of this episode.

The Erchang XF07C sends 140 bytes, and inside there are 120 real echo measurements

The data frame has a fixed length of 140 bytes and arrives about six times per second, exactly as in the ball from the previous episode. It is the same format used by the whole family of cheap Chinese fish finders.

• 120 bytes are consecutive measurements of echo loudness, from the surface down to the selected depth
• 2 bytes are the depth, in tenths of a foot
• 2 bytes are the temperature, in tenths of a degree Fahrenheit
• 3 bytes are the sonde’s finished verdict: target depth and its size on a scale of 1 to 3
• 1 byte says which transmit channel this particular ping used, and that is new compared with the previous sondes
• the rest are start and end markers, battery level, current range and the checksum

The conversation is clocked by the sonde, not by the phone. For every correct frame the app replies with a short, ten-byte settings frame: noise filter, sensitivity, lure light, range and the choice of transmit channel. The phone initiates nothing on its own except the first packet that establishes the connection.

Inside, the sonde counts in American units, and the conversion to metres and degrees Celsius is done only by the app. For the record: metres = reading times 0.0305, and degrees Celsius = (reading minus 320) divided by 18.

Two frequencies are not a marketing trick – the Erchang has two separate transmitters

In cheap hardware a declaration of two frequencies is often just a label: the same transmitter, the same work, a different icon on the screen. Not here.

The Erchang XF07C has two separate transmit circuits, each with its own receive path. The first taps out a pulse lasting 382 processor ticks, the second 144. At a 48 MHz clock that gives 125.65 kHz and 333.33 kHz, exactly what the spec sheet promises. The sonde even keeps a separate queue of detected fish for each channel.

It always transmits on one channel at a time. A byte in the settings frame selects which: 0 is the wide 125 kHz permanently, 1 is the narrow 330 kHz permanently, and 2 is the automatic mode in which the sonde hops by itself, according to its own counter and not to your wishes: after five consecutive fulfilments of the condition it goes narrow and returns after two hundred ticks.

What the narrow 330 kHz channel really gives you

Parameter	125 kHz (45 degrees)	330 kHz (14 degrees)
Dead zone right under the sonde, 5 m of water	13 cm	4.8 cm
Circle of uncertainty at 5 metres	4.1 m	1.2 m
Circle of uncertainty at 10 metres	8.3 m	2.5 m
What it is for	searching, a wide field of view	precise indication, resolution

The narrow channel means three times less horizontal uncertainty and a three times smaller dead zone. Vertically nothing changes, because there are still 120 slices for the entire range.

The price for this is physical and cannot be avoided: a higher frequency fades faster in water and lights up a smaller patch of bottom, so on the narrow channel you have to hit the spot more precisely. The sensible way to work is therefore: search wide, look narrow.

The manufacturer’s numbers here are in fact consistent with physics, unlike balls declaring 90 degrees at 125 kHz, where the transducer would come out smaller than the wave it transmits. At 125 kHz and 45 degrees the element works out at around 15 mm, at 330 kHz and 14 degrees around 18 mm. Both fit inside a 68 mm ball.

One hundred and twenty slices for any depth, or why at nine metres the image loses half its detail

The sonde always cuts the water column into 120 slices, no matter whether there are three metres of water or sixty. There is not a single field in the frame that would declare that number. Always 120.

It switches range by itself, in steps, along a ladder of 3, 6, 9, 18, 27, 37, 46 and 61 metres. The larger the range, the thicker the slice, because the same 120 pieces have to be spread over a longer section.

Water depth	Range it will jump to	One slice	Dead zone 125 kHz	Dead zone 330 kHz	Circle of view 125 kHz	Circle of view 330 kHz
0.5 m	3 m	2.5 cm	12 cm	4.5 cm	0.4 m	0.12 m
5 m	6 m	5 cm	13 cm	4.8 cm	4.1 m	1.2 m
10 m	18 m	15 cm	15 cm	5.7 cm	8.3 m	2.5 m
20 m	27 m	23 cm	20 cm	7.5 cm	16.6 m	4.9 m
40 m	46 m	38 cm	30 cm	11 cm	33 m	9.8 m

Two things follow from this table that spoil the mood.

At nine metres the image suddenly loses half its detail. The water goes past 9 m, the sonde has to jump to the 18 m range and spreads the same 120 slices over a section twice as long. The slice grows from 7.6 to 15 centimetres, and half the screen is water that is not there.

On the wide channel at 10 metres the circle of view is over 8 metres. A fish shown on the screen may be standing four metres to the side. A snag three metres away looks exactly like the one under the boat. Only the narrow channel shrinks that circle to two and a half metres.

The first 0.8 metres under the Erchang is its own ringing, not the bottom

The transducer behaves like a bell. While it shouts, and for a moment afterwards, it is deaf, because its own echo drowns out everything.

And now the key thing: the sonde does not know where its own ringing ends. It discovers that anew every time. After transmitting a pulse it waits for the signal to fall below a threshold, then for it to stop falling, and only the first echo after that wait is declared the bottom.

In water shallower than the length of that ringing the sonde therefore measures the end of its own shout, not the bottom. In this unit that comes out as a stubborn 0.8 metres, regardless of whether there are 20 or 50 centimetres of water. Measured on the water and consistent with how the sonde looks for the bottom.

When the Erchang says “fish” and why in shallow water it will never say it

Fish detection happens in the sonde, not in the phone. The app takes the finished verdict and draws an icon. The complete list of conditions under which the sonde declares a fish:

• the corrected echo is louder than the threshold set by sensitivity
• the target is at least 0.61 metres above the bottom
• the target is deeper than 0.67 metres

That is all. No shape analysis, no comparison with the neighbouring measurement, no tracking of a target between pings. The decision is made on a single sample from a single ping, exactly as in the ball from the previous episode.

From the last two conditions follows something no listing mentions: in water shallower than about 1.3 metres detecting a fish is mathematically impossible, whatever the settings. At 1.5 metres the detection window is 22 centimetres.

On top of that comes gain correction with depth. Over the first few dozen centimetres the sonde subtracts about 200 from the echo on a scale reaching 255, in order to suppress its own ringing. Where it is blind anyway, it becomes doubly blind.

There is also a trap for those who crank sensitivity to maximum. The table of detection thresholds has entries up to 90 percent. Above 90 it reads zeros, so the threshold disappears and everything gets through, rubbish included. That is not the highest sensitivity, it is running off the end of the table. A sensible maximum is 85 to 90 percent.

Fish size is computed from echo loudness in three bands, not drawn at random as in the Lucky FF918. Except that the loudness is corrected beforehand according to depth and the sensitivity setting, so the same fish at two different depths will get two different sizes. The number is repeatable only at a constant depth and constant settings.

The conclusion is the same as in episodes two and three. A floating leaf, a stick, a mat of algae, an air bubble, a thermal layer, a suspended line and a weed tip more than 0.6 metres above the bottom: each of those gets a fish icon. And weed lower than 0.6 metres above the bottom and a log lying on the bottom have no representation on this screen at all.

Fish icons or the water column – in this app you cannot have both at once

The Erchang Sonar app really does draw its echogram from real measurements and not from a ready picture as the sondes from the previous episodes. And that is where the compliments end, because it is not an achievement, it is the minimum: with a 45 degree cone and 120 slices for the entire range, even that real image says very little.

The catch is that the fish icon button does two things at once while claiming to do one. Turning the icons on, which is the factory state after installation, adds a layer with little fish and at the same time orders everything above the bottom to be cleared from the echogram: the app finds the position of the bottom, walks upwards while the signal is strong, gently fading the bottom edge, and zeroes all measurements above it.

Hence that characteristic picture: a white background, the bottom contour and little fish hanging exactly where the echogram had an arch. There is no third option in which you see both.

Turn the fish icons off and you get the full echogram with arches, because the water column wiping stops working. And here is the second thing worth knowing, which only shows up on the water: an arch is made by anything that deformed the pulse even slightly. A leaf, a bubble, a branch, a weed tip, suspended matter after rain. There are then plenty of arches and that is an honest picture of what the sonde hears, only it should not be read as a list of fish.

In other words the fish icon adds nothing. It hides the evidence: instead of an arch you could judge for yourself, you get the sonde’s verdict and a cleared background on which there is no way to check that verdict.

The second layer: sensitivity decides how many weak echoes survive at all

Regardless of the icons, each of the 120 measurements in a column is reduced by 100 minus the sensitivity setting, and finally the contrast curve zeroes everything below 24. These two things add up and decide what still makes it onto the screen and what disappears.

Sensitivity in the app	Subtracted from the measurement	Contrast curve threshold	How loud an echo must be to draw anything
51 %	49	24	73 of 255
70 %	30	24	54 of 255
85 %	15	24	39 of 255
100 %	0	24	24 of 255

The sensible setting is fish icons off and sensitivity somewhere around 85 to 90 percent. Not 100, because above 90 percent the table of detection thresholds inside the sonde runs out and rubbish gets through as well.

The app carries a ready recorded echogram, but it sits behind the Simulator key

Inside the app file sit 43,200 numbers: exactly 360 columns of 120 measurements, loaded into memory at every start, before you connect anything at all. You can see a surface band, the bottom contour, fish arches and a second return. It looks like a recording from a real transducer and it certainly is not a measurement from your water.

Material with a file like that inside is asking for an accusation, so let me answer the question that has to be asked straight away: can that recording reach the live screen instead of a measurement?

It cannot. It is played back only in Simulator mode, which the user switches on with a separate button, and in that same mode incoming frames are not parsed at all. There is no path in which the demo data mixes with the measurement or replaces it when the signal is weak. That much honesty must be granted to this manufacturer.

A different objection remains, though. The simulator button sits on the device search screen right next to the rescan button, so it is easy to enter before anything connects, and on the sonar screen nothing tells you it is a demo. The image from the recording is also far denser than what you will see on the water with default settings, because the recording was made without water column wiping. That same file is ready-made advertising material and I suspect this is where the impressive app screenshots in the sales listings come from.

Password 12345678 shared by every unit and zero pairing

The Wi-Fi version sets up its own network with a name starting with ErchangFish. The password is hardcoded in the app and it is 12345678. One for every unit in the world.

The sonde listens on a fixed address, and the entire authentication consists of the phone sending one short greeting packet. Whoever sends it first gets the data stream. There is no pairing, no code, nothing. Being within range of someone else’s boat means being on its network.

For an angler this is mainly a curiosity, but it is worth knowing if several identical sets are working next to each other at a match.

Details you cannot see on the screen

• The checksum does not protect the image. It is verified only after the reply has been sent and it guards only the depth, the fish, the battery and the temperature. The hundred and twenty echo measurements go through even from a corrupted frame.
• The narrow channel gets a fixed amplitude boost of 20. Physically justified, because 330 kHz collects less energy, but it is a fixed offset instead of a curve.
• The fish icon stays on the screen long after the sonde has stopped seeing anything.
• Saving the connection setting is decoration. The app records which protocol the sonde used, reads it at startup and immediately throws the result away. On every launch it guesses again, hopping between variants roughly every 2.8 seconds until something arrives. Hence the few seconds of waiting for the first reading.
• The contrast curve cuts off the bottom 9 percent of the scale. An honest noise threshold, but you need to know that you will never see the weakest echoes.

Check it yourself, on your own water

1. Turn the fish icons off and raise the sensitivity to 85 percent. With the icons on, the water column is wiped from the image; without them, arches appear. The same water, the same second, a completely different picture. While you are at it, count those arches and compare with the number of icons you saw a moment ago.
2. Force 125 kHz, take a screenshot, then force 330 kHz in the same spot and take a second one. On the narrow channel there should be fewer arches but more reliable ones, because the circle of uncertainty shrinks threefold.
3. Go onto a shallow below a metre. You will see a stubborn 0.8 m no matter how much water is really there.
4. Switch on the Simulator and compare it with what you see on the water. It is that image that circulates in the sales listings.

Verdict: two stars out of five, and no update can fix it

What I rate	Rating
Sonde hardware: two channels, 125 and 330 kHz, short pulses, a proper transmitter	★★★★☆	4 / 5
Depth: solid above a metre, below a metre there is nothing to look for	★★☆☆☆	2 / 5
App: an echogram from a measurement, but either fish icons or the water column	★★☆☆☆	2 / 5
Image resolution: 120 measurements for the entire range	★☆☆☆☆	1 / 5
Security: one password for every unit, no pairing	★☆☆☆☆	1 / 5
Is it worth buying as a fish finder	★★☆☆☆	2 / 5

The reason for this verdict is one and it outweighs everything written above: 120 measurements for the entire range is simply too few. In 10 metres of water one slice of the image is 15 centimetres, in 20 metres already 23. A fish, weed and a snag all have to fit inside a layer that thick to mean anything on the screen.

This is not a software flaw and no update will touch it. The number of bins is baked into the sonde, and the frame format does not even have a field to declare it. Two transmit channels narrow the circle of uncertainty and that genuinely helps, but they do not change the thickness of a slice. For comparison: sondes that draw a sensible image of structure work with several hundred measurements per column and at higher frequencies.

Summary – what the Erchang does not say out loud

• Switched-on fish icons wipe the entire water column off the image. What remains is a white background, the bottom contour and little fish where the arch used to be. There is no third option in which you see both.
• With the icons off, an arch is made by anything that deformed the pulse: a leaf, a bubble, a branch, the tip of a weed stem. The image is honest, but it is not a list of fish.
• Any reflection in the water column can become a fish. Weed near the bottom and snags have no representation at all.
• Below 1.3 metres of water a fish cannot appear, and below 0.8 metres the sonde measures its own ringing instead of the bottom.
• Target size is computed but not comparable between depths and settings.
• At nine metres the image loses half its detail, because the sonde jumps to a range twice as large.
• At 10 metres on the wide channel the target is somewhere in a circle 8 metres across. The narrow channel shrinks that circle to 2.5 metres.
• The Wi-Fi password is the same on every unit and there is no pairing at all.

And the other side of the coin, because credit is due: the Erchang XF07C has the best hardware of the three units I have described so far. Two genuine transmit channels matching the spec sheet, shorter pulses than in the nameless ball, meaning a smaller dead zone, interrupts disabled during transmission so that nothing smears the pulse, depth measured properly above a metre and target size computed rather than drawn at random.

And one thing that has to be highlighted separately, because in this series it is an exception: the bottom on the screen is a real echo, together with the second return and the structure below the bottom. The cheaper sondes from the previous episodes lay a ready texture there so that it looks nicer. The Erchang does not do that and deserves a point for it.

It is at the same time the most instructive episode so far. In the previous ones the hardware was weak and the bottom image was entirely a drawing. Here the hardware is the best of the three and no good comes of it. Out of the box the image is wiped to a white background with icons, and once the icons are off you do get a real echo, only blurred into slices of 15 centimetres and into a circle eight metres across. Better hardware, the same effect on the screen.

As a depth sounder on water deeper than a metre the Erchang works and that is the whole of its value. As a fish finder, that is for looking at the bottom, at weed and at fish, it is not worth spending money on: two stars out of five and that is not spite, it is the result of the numbers above. If you are buying it with an image of bottom structure in mind, you are buying a screenshot from demo mode.

What I did not test and therefore do not write about: whether the ball uses one transducer at two wavelengths or two separate ones, the real sensitivity to small targets, and a comparison with branded gear on the same water. I describe only what I could establish: what this sonde sends and what the app does with it.

Sources and material worth checking

• NOAA – What is sonar? – how a fish finder works: a transmitted pulse, the echo return time, the conversion into distance
• NOAA – how sound travels in water – about 1480-1500 m/s, the constant every conversion from echo time to depth relies on
• FAO Fisheries Technical Paper 240: Fisheries Acoustics – beam angle, target strength and the reason why the loudness of an echo says nothing about the length of a fish
• Episode 3: the black ball XJ02B – the same frame format, the same family of hardware, a different app
• Episode 2: Lucky FF918 – for comparison, what a sonde that sends no echo at all looks like

In the next episode: a set with its own screen, no phone and no app. It looks the most serious of the whole line-up, and it has the bottom image baked into memory: PUPOPAN – the fish finder where the bottom is a picture from memory.

What equipment I use and on what basis I test this

Every device described in this series was bought with my own money and is my property. I examine them in order to know what really reaches the angler’s hands and what I can demand from the hardware I design myself. The analysis covers one specific unit and the firmware version I found in it; other batches, versions and variants of the model may behave differently.

Observing, studying and testing the functioning of a program in order to determine its principles is permitted for a person who lawfully acquired the device: article 75(2)(2) of the Polish Copyright Act, which implements article 5(3) of Directive 2009/24/EC. Reading the transmission concerned exclusively my own transmitter and my own receiver. I do not publish source code, memory dumps, keys or tools that would allow anyone to copy someone else’s software – only conclusions about what the hardware measures and what it does not.

Names and trademarks belong to their owners and are used solely to identify the hardware under test, within the scope of critical analysis and the right to information. Photographs from sales listings are used for the same purpose. If a manufacturer or a seller believes any conclusion is wrong, I will gladly repeat the measurements on a unit they provide and publish a correction.

That is how it works here: before anything goes into our boats, it passes across this bench. And if something does not pass, we say so out loud, even when nobody wants to hear it.