EEVblog #824 - GW Instek GDS-1000B Oscilloscope Teardown

[Stupid Beard]

I know you mentioned it in passing and likely nobody cares, but neon is the arm implementation of SIMD. It is used a lot for vector math amongst other things. Whether they are actually using it for anything I couldn't say.

[Someone]

Similar to the last teardown of the signal generator, the mains supply is using a PCB connection for the earthing except this time we get a good look at it. I'm not sure that meets worldwide requirements for protective bonding, and would at least fail the Australian requirements for PCB earthing traces to be doubly redundant. The trace shown in the video also looks like its marginal to pass the earth bonding test, would you pass 30A through it for one minute for our entertainment?

The Zynq used is the baby of the family, though for the same price you might be surprised how much performance softcores can provide.

[k2teknik]

There is some app for download on http://www.gwinstek.com/en-global/products/Oscilloscopes/Digital_Storage_Oscilloscopes/GDS-1000B
DVM,  digital filter and datalog App is the options right now, maybe they will release some info on how to make you own app's?

[mux]

Just going by back-of-the-envelope product pricing:

The scope is $350, i.e. about $140-175 BOM through typical retailers
The ZYNQ is $60, the hittite ADC is $50 (both octopart 1ku pricing)

How do they get the rest of the BOM, including mechanicals, down to $50? That's insane!

[Tothwolf]

Would have been even nicer if they had conformed to the IEC spec on the earth wire connector, blue is not the right colour for a PE conductor, it should be a green/yellow striped one.

The blue wire is line/mains and not an earth wire.

Neutral here in NZ.

With the orientation they used for the C14 inlet, the pins are X, G, and W. X is hot/service, G is ground, and W is neutral. The blue wire is connected to X and looks like it might be jumpering X to the fuse.

[coppice]

Just going by back-of-the-envelope product pricing:

The scope is $350, i.e. about $140-175 BOM through typical retailers
The ZYNQ is $60, the hittite ADC is $50 (both octopart 1ku pricing)

How do they get the rest of the BOM, including mechanicals, down to $50? That's insane!

Its insane to think that serious buyers of the ZYNQ and Hittite ADC pay those prices when they negotiate for the supply of thousands per annum.

[Someone]

Would have been even nicer if they had conformed to the IEC spec on the earth wire connector, blue is not the right colour for a PE conductor, it should be a green/yellow striped one.

The blue wire is line/mains and not an earth wire.

Neutral here in NZ.

With the orientation they used for the C14 inlet, the pins are X, G, and W. X is hot/service, G is ground, and W is neutral. The blue wire is connected to X and looks like it might be jumpering X to the fuse.

If that is carrying mains directly then the product fails the Australian standards for certain just on that. Looks like there might be a product investigation in the near future for the importers of these.

[mux]

Just going by back-of-the-envelope product pricing:

The scope is $350, i.e. about $140-175 BOM through typical retailers
The ZYNQ is $60, the hittite ADC is $50 (both octopart 1ku pricing)

How do they get the rest of the BOM, including mechanicals, down to $50? That's insane!

Its insane to think that serious buyers of the ZYNQ and Hittite ADC pay those prices when they negotiate for the supply of thousands per annum.

Sure, they'll go down a _bit_, but they're not paying $20 a pop or anything. These chips are expensive even in 10k or 100k quantities. It's a common fallacy to think that component prices, even on individual contracts, go down by a lot. In reality, you're never going to get more than low double-digit discounts.

Source: good relationship with the sourcing guy at a former company I consulted for.

[coppice]

Just going by back-of-the-envelope product pricing:

The scope is $350, i.e. about $140-175 BOM through typical retailers
The ZYNQ is $60, the hittite ADC is $50 (both octopart 1ku pricing)

How do they get the rest of the BOM, including mechanicals, down to $50? That's insane!

Its insane to think that serious buyers of the ZYNQ and Hittite ADC pay those prices when they negotiate for the supply of thousands per annum.

Sure, they'll go down a _bit_, but they're not paying $20 a pop or anything. These chips are expensive even in 10k or 100k quantities. It's a common fallacy to think that component prices, even on individual contracts, go down by a lot. In reality, you're never going to get more than low double-digit discounts.

Source: good relationship with the sourcing guy at a former company I consulted for.

Low value chips don't drop much in price for 10k or 100k. You have to buy millions to see good prices. Chips that never sell in millions, because they are more niche in nature, generally fall in price at more modest volumes.

[mux]

Alright, say that they halve. There's still:
- A reasonable quality 30W power supply
- An LCD
- A bunch of reasonable quality rotary encoders
- Bunch of metalwork
- Bunch of assembly (that's definitely not badly done)

I'm just amazed how they can squeeze that out of a $350 product. I've seen plenty of $350 street price products that have a lot less guts and smarts in them. Somebody did an awesome sourcing job on this.

[Barny]

Will there be an detailed video abaut the scope?
Pherhapes a little compare to the Rigol scope?

I will replace my two channel 150MHz 5000 serie Rigol.
The old big brick with its small b/w screen start to annoy me.

[djQUAN]

I have a possible explanation for the separate processor and front end boards.

The processor board is a high tolerance PCB with better specifications (drill accuracy, clearances, trace widths etc) and is generally more expensive. This is necessary for the BGA package. It can be 4 layer or even 6 layer.

Making the entire huge PCB in this spec is very expensive so they made it into a small sub board and used a cheaper lower spec board with less copper layers for the huge interface boards.

[JackM]

I'd still choose the Rigol DS1054Z over this. The GWinstek just falls a bit short in a bit too many areas versus the 1054Z, enough to push me towards the Rigol. It's nice to see competition in this area though.

[Lukas]

Seems like far-east scope manufactureres really like the HMCAD1511 ADC these days: Rigol DS1054Z, Siglent 1000X and the Instek 1000B. I'm wondering why they didn't start using this one earlier. It's by far the cheapest 1GSa/s ADC. Maybe because it requires an FPGA that does 8×1Gbit/s SERDES?

I really like the simplistic design of this DSO, it won't get much simpler than frontend + ADC + FGPA/SOC.

[tautech]

Would have been even nicer if they had conformed to the IEC spec on the earth wire connector, blue is not the right colour for a PE conductor, it should be a green/yellow striped one.

The blue wire is line/mains and not an earth wire.

Neutral here in NZ.

With the orientation they used for the C14 inlet, the pins are X, G, and W. X is hot/service, G is ground, and W is neutral. The blue wire is connected to X and looks like it might be jumpering X to the fuse.

If that is carrying mains directly then the product fails the Australian standards for certain just on that. Looks like there might be a product investigation in the near future for the importers of these.

Sounds like Dave needs to recheck this, no it would not meet standards here either.
4 screws and 2 tabs and we'll know for sure.

[Hydrawerk]

According to the user manual, there is no Trigger Output available at GDS-1000B.

[free_electron]

what would really make this a competition killer:
Toss out the little lcd and the me u buttons around it. Make that whole grey area around the lcd, the new lcd area. And make it a touch screen.
The rubber buttons areound the lcd are gone. If you pull up a menu they slide in, touc to activate.

Touch screens cost next to nothing. Larger lcds are also a dime a dozen.

Leave the roatry encoders and buttons in that area as they are. No need for pop up windows or pinch to zoom. Just a simple zone touch to reclaim the wasted area for the menu buttons around the lcd.

You would not need the pcb area for the buttons, would need the rubber button molding. And it would have an ipad mini sized screen. Something in the 1280x600 resolution...
It would add 10$ tops to the bom...

[mikeselectricstuff]

Just stumbled on this Xilinx  article on GW using the Zynq - rather buzzword heavy
 
http://www.xilinx.com/publications/prod_mktg/zynq7000/goodwill-casestudy.pdf

[mikeselectricstuff]

The ZYNQ is $60

$54 1-off at Digikey.
I doubt they're paying more than $15-20. real-world FPGA pricing in any qty bears little resemblance to advertised prices once you get into the wierd world of "supported pricing".
In the past I've had quotes for less than half DK prices for just 100x
 

[mikeselectricstuff]

I suspect a reason for the seperate processor PCB is that it can be used in multiple products - remember it's not limited to scopes -  could easily be the basis of a powerful AWG, or even a spectrum analyser with different analogue board.

This architecture, maybe with a more integrated front-end, could be the basis for a pretty small portable scope.
 

[nixfu]

Now if only someone could do something similar with a frontend+adc+fpga design to make a low cost ($500) khz-1Ghz Spectrum Analyzer for the hobbyist market.

This model probably works ok for the price, but what I came away with is that it really does show how much Rigol really seems to "do things right" as best they can for the price and I am that much more appreciative of what sort of engineering and build quality I got for the money I paid for my DS1074z.

[tom66]

I suspect the reason it's split in two PCBs is cost driven. They may even be able to get away with a 2-layer PCB for the analog front end and ADC section. If you have few traces on the bottom layer then you can make it all ground plane and hit the required trace impedance.  The controller is probably a 6/8 layer PCB so quite expensive.

IIRC, the DS1054Z series is implemented on just a 4-layer PCB, given the high performance it's pretty surprising they can manage that.

[MBY]

The Jamicon caps on the mainboard are from Taiwan. I don't think they are necessarily bad - I have used many Jamicons through the years, with no reliability problems.

Yes! Maybe Jamicon isn't a brand to compare with Nippon Chemi-con, but I was surprised by Dave not having heard of Jamicon or call them "not a name brand". They are certainly okay and decent all-purpose capacitors and the brand is well established for decades. I've also used Jamicon for years with no problem whatsoever. Jamicon is *not* (well, to my knowledge) a "Wun Hung-Lo" company with made up optimistic parameters in their datasheets. If a datasheet says "100 mOhm @ 100 kHz ESR", it probably is 100 mOhm at 100 kHz or better... I've seen Jamicon caps (and fans) in gears and computers from the eighties and onwards. Price is often mid-range.

Edit: As for the scope. I wasn't impressed and I share Daves view in that the scope is ugly and bulky. I really liked the layout of the knobs and buttons, but there is no reason to space everything out. Surely they could have the same basic layout but in a form factor more like Rigols 1104z...

[rx8pilot]

The race to the bottom continues. How many ultra-low cost entry level scopes does the industry need? GW is only making a few dollars off each after distribution. Why bother?

Sent from mobile device.... Keeping it short and mis-spelled

[wraper]

Now if only someone could do something similar with a frontend+adc+fpga design to make a low cost ($500) khz-1Ghz Spectrum Analyzer for the hobbyist market.

This model probably works ok for the price, but what I came away with is that it really does show how much Rigol really seems to "do things right" as best they can for the price and I am that much more appreciative of what sort of engineering and build quality I got for the money I paid for my DS1074z.

You got Crapxons in your shielded PSU, so It's questionable which is of better quality. Really don't see anything seriously wrong with instek, there is no reason to make PSU with 2 layers PCB.