EEVblog 1493 - The MacGyver Project - Part 2

[EEVblog]

Powering up the Banshee ultrasonic leak detector LED display to see how the display is multiplexed turned out to be very interesting!
With special guest debugger, Sagan!

[JustMeHere]

Drive this with a SPI running at 2v?  It looks like a MOSI, a CLK, and a CS. 

[firewalker]

I believe everything has to do with the case. You could build a spark igniter inside the case and they wouldn't mind.

Alexander.

[NivagSwerdna]

On the PCB it is marked HDSP-U113... now all you 7-segment fan boys will recognise that as a Broadcom Device... https://docs.broadcom.com/doc/AV02-3584EN

Now on those babies Vf is 2.1V and 15mA/seg at full brightness...  A supply of 2V doesn't cut it.  And those segments all appear evenly lit... and 88888 looks as bright as .40.db. and v.v.

There is still some mystery to be solved here Scooby!

[NivagSwerdna]

Could it be you have your knickers in a twist?

@20:49 you notice that during the shift register refresh the CH3 signal goes high... that probably represents being OFF for the duration of the data refresh to avoid artifacts... the actual driving is when this is LOW?

(The free range bits also suggest a shared data line?  "Free Range" is a great term... might have to adopt that)

[NivagSwerdna]

According to... https://www.instrumart.com/assets/Net-Safety-Banshee343-manual.pdf

Display is dimmable... 12 settings... would be interesting to see how that is achieved.

[Peabody]

I think this method of driving the 7-segments makes perfect sense, and is not at all "how ya doin".  The current out of an HC164 gate is a function of the voltage at Vcc versus the Vf of the segment diode.  For a particular Vf, you set Vcc to whatever level draws 50mA through Vcc when all eight segments are lit, and as long as that's 1.9V or greater, it will work fine, and won't violate any absolute maximum.  The displays are then driven constantly at 6.25mA per segment, with no flicker, and there's no need to refresh unless the displayed values change.

The downside is you then have to generate the 2V power supply, and the clock and data lines also have to be 2V.  Unless you need the 2V for something else, it would probably be a lot easier to just add the resistors and drive everything at 5V or 3.3V.

[Peabody]

According to... https://www.instrumart.com/assets/Net-Safety-Banshee343-manual.pdf

Display is dimmable... 12 settings... would be interesting to see how that is achieved.

You could vary the voltage of the 2V power supply line for the HC164s.  With fixed Vf in the segments, changing Vcc would change the current through the segments.

I guess in theory you could also limit the current flow through the NPN switching the ganged CC lines, but there would be heating issues doing that.

Anyway, my guess is they just modify something in whatever circuit they use to generate the 2V supply.

[Kleinstein]

Modern ARM based µCs and FPGAs have no big problem running with 2 V.  If really needed, there are level translators, e.g. from 2.7 V to 2 V.

For battery operation a low voltage helps, as it also reduces the power. So just need a regulator from some 2.5-4 V to 2 V.

The free range data may be SPI bus data for other devices. The display part has no extra CS and thus somehow has to turn off the clock instead of a CS signal.

The shared data make me think this is more likely from the µC than the FPGA part. The FPGA could use a seprate data line just for the display, while the µC may only have 1 SPI interface available.

[tigrou]

Can somebody explain me shortly what's the trick being used there ?
AFAIK in Part 1 Dave found out that only one segment should be lit at a time (eg: using multiplexing), because lighting all segments at same time would require too much amps flowing trough ground pin.

EDIT : by "segment" I meant one digit.

[Kleinstein]

They use a very much reduced supply voltage and this way reduce the current to a safe level. The tricky part with this is getting a constant current for all segments, as the current depends on the individual outputs R_on. So it relies on the matching between the 74HC164. It is a bit tricky to estimate how well they match. They should definitely use chips from the same manufacturer / series.
The actual needed voltage may depend on the exact series / reel.  There may be a way to fine adjust the supply voltage (in SW or with the switches).

Using 1 segment at a time helps to get an even more even intensity, but it also makes the pin current limitation even more of a challange, as the current per output is limited to some 25 mA, so half the VCC current limit. The pin current limit is more like a reason why the sequencial mode is not a good idea.
The system used with low voltage uses all the supply pins at the same time.

[EEVblog]

Can somebody explain me shortly what's the trick being used there ?
AFAIK in Part 1 Dave found out that only one segment should be lit at a time (eg: using multiplexing), because lighting all segments at same time would require too much amps flowing trough ground pin.

The trick is the low supply voltage.
Everyone assumed they must be multiplexing the digits or even individual segment, but there is no multiplexing happening at all. It's just a straight normal direct drive of each segment.

[EEVblog]

Could it be you have your knickers in a twist?

@20:49 you notice that during the shift register refresh the CH3 signal goes high... that probably represents being OFF for the duration of the data refresh to avoid artifacts... the actual driving is when this is LOW?

Yes, that's what I said. When it's low all the LED's are being switch on. They swithc it off durign the data update to prevent flicker.

[Peabody]

According to info posted by NivagSwerdna in Reply #5, the display can be set to one of 12 brightness levels.  If they were multiplexing, they could vary the On time.  But they aren't.  So maybe they are varying the supply voltage.  If that's done digitally, I'd like to see how.

Edit:  Actually, it could be much simpler.  If constant drive is the high brightness setting, they would just vary the duty cycle by switching the CC pin.  The lower the duty cycle, the dimmer the display.

[thm_w]

On the PCB it is marked HDSP-U113... now all you 7-segment fan boys will recognise that as a Broadcom Device... https://docs.broadcom.com/doc/AV02-3584EN

Now on those babies Vf is 2.1V and 15mA/seg at full brightness...  A supply of 2V doesn't cut it.  And those segments all appear evenly lit... and 88888 looks as bright as .40.db. and v.v.

There is still some mystery to be solved here Scooby!

Typical forward voltage is 2.0-2.2V at 25mA, max of 2.5V.
Since its running less than 25mA, max might be 2.2V or so? Yeah, its possible they received LEDs in spec and the LEDs are super dim. But these are very highly priced LEDs so maybe that is not going to happen.

Hell half of these display boards could fail production and get tossed out, who knows.

[Peabody]

Below is the datasheet graph showing this part's behavior.  At 2V you would get 10mA.  But that would be too high because with all 8 segments lit up, the total current would be 80mA, which in time would blow out the HC164.  But at a slightly lower voltage you would get 6.25mA, which puts the total HC164 current at 50mA, which meets the absolute maximum.  Then the only question is whether constant drive at 6.25mA would be bright enough.  In the video, it looks acceptable in the lab, but I don't know about outdoors.

[thm_w]

Below is the datasheet graph showing this part's behavior.  At 2V you would get 10mA.  But that would be too high because with all 8 segments lit up, the total current would be 80mA, which in time would blow out the HC164.  But at a slightly lower voltage you would get 6.25mA, which puts the total HC164 current at 50mA, which meets the absolute maximum.  Then the only question is whether constant drive at 6.25mA would be bright enough.  In the video, it looks acceptable in the lab, but I don't know about outdoors.

Yes but that is just the typical response, or averaged response from a few samples, it does not show the minimum or maximum values mentioned above.

[Peabody]

But these displays are tested by the manufacturer and binned according to the luminance they produce with a given current.  In the video it shows these are marked as H group parts, which is a middling category.  So if they always order H parts, the performance should be pretty consistent.

[Kleinstein]

The On resistance of the 74HC164 is going up quite a bit when one gets lower in supply voltage. This is especially the case, when getiing close to the 2 V lower limit.
Chances are the adjusted the 2 V supply voltage and they may very well have some possiblitiy to adjust / trim the voltage if needed.

I could not find detailed data for the more normal logic chips and there is probably quite some variability with series /  manufacturer, but in increasing resistance is the normal expected behavior.
The datasheets usually have data for the speed and much of the delays scales with the on resistance. For the 2 V operation the speed is usually about 1/4 the speed at 5 V and thus expect an ouput resistance of about 4 times the value at 5 V (usually in the 50 Ohms range). Some 200 Ohms from the R_on is a significant part in limiting the current. So chances are the LED current would be lower than 5 mA.

[NivagSwerdna]

Well I have to admit I am very surprised by the design especially as I have been toying with similar for one of my own projects... After extensive googling I see 74HC164 being used in this configuration so sorry I doubted the approach!
You learn something every day!

[Peabody]

I certainly had my doubts too in the beginning, but the more Dave reveals about how this circuit works, the more impressed I am with the engineering.  After all, you eliminate all 40 segment resistors, but still have no excess current issues or segment brightness issues, and you don't have to bother with multiplexing.  And you can adjust the brightness by PWMming the CC transistor.

On the downside, you do have to produce the 2V supply, and the clock and data lines also have to be 2V.  We don't yet know how that was done.  And of course you have five HC164 chips.  If you multiplex by digit, you only need one, but then the brightness would be much lower.

I think today you can find efficient 7-segment displays that are bright enough when multiplexed by digit.  And even at 5V you can eliminate most of the resistors if you multiplex by segment.  But controlling the current by reducing the power supply voltage is still an interesting idea - one I hadn't thought of.

[NivagSwerdna]

...and the clock and data lines also have to be 2V.

According to the Nexperia version of the datasheet you can have the input lines at above VCC as long as current limiting resistors are used.

[Kleinstein]

For the data signal a series resistor is no porblem. For the clock signal one has to be a bit careful, as many shift registers want a minimum slew rate to work reliable. A large resistor may slow down things quite a bit, and the clock signal goes to all chips and thus sees quite some capacitance. One may have to plan with a capacitor in parallel to the resistor for speed up.
Many modern µCs / FPGAs have no problem with a low supply and with some types even 2 V is OK.

[Peabody]

I believe Dave's video showed both the clock and data lines were actually 2V.  Maybe they didn't have to be, but they were.

[thm_w]

But these displays are tested by the manufacturer and binned according to the luminance they produce with a given current.  In the video it shows these are marked as H group parts, which is a middling category.  So if they always order H parts, the performance should be pretty consistent.

Given current and color bins, not given voltage.
But yes, it should be quite consistent.

Chance of biting you in the ass if this LED display went out of production and you needed to put in an alternate part.