EEVblog #1228 - Do Digital Scopes Have REAL Verniers?

[Circlotron]

When you use the vernier volts/div I'd like if there was the option instead of all these fractional volts/div figures, you could shrink the waveform back inside the bounds of the screen and also shrink the spacing between the vertical amplitude lines (horizontally drawn lines). That way the volts/div figure could remain constant but the size of the divisions would change instead.

[Rerouter]

At least for the siglent, I can provide some evidence to the fact they are using the VGA for all vernier steps down to 500uV (cant see video because youtube is processing), (technically the front end can measure to about 313uV / div if you manually set it)

for a given channel you can use the CX:VGAC? command to read the current setting of the VGA, equally you can manaually set it with CX:VGAC = 255

So to read the VGA for channel 1, you would send C1:VGAC?

If you have the 4 channel model you can do this through the Web UI.

Attached is the calibration file of gain and offset it runs for "Most" VGA steps for each attenuator range, I would recommend opening it in notepad++ so it can present it in a nice formatted way.

To best explain it, you have the channel, the attenuation range, and then the VGA code, next up is the channel offset DAC code it needed to use to make the ADC read 0 during calibration, the other 2 codes I am less sure on at this time,

[EEVblog]

At least for the siglent, I can provide some evidence to the fact they are using the VGAC for all vernier steps down to 500uV (cant see video because youtube is processing), (technically the front end can measure to about 313uV / div if you manually set it)

Not on the scope you can't it stops at 500uV/div in vernier mode.

[rf-loop]

Here is small help to uderstand this case.

There is VGA before ADC.  But its gain steps resolution is not enough. (only one 8bit control byte)
Programmer can even select what is best combination how to use ADC digital cain combined with VGA gain. (optimize for noise etc... )

For get full control over whole voltage range including "vernier" aka fine steps in this case.
Most wide voltage band is band I  from 500uV to 118mV/div. It include over 350pcs  V/div steps. Where some steps are these major steps 500uV, 1mV, 2, 5,...etc up to 100mV/div and fine steps between these and then still some fine steps from 100mV/div  to 118mV/div. After then, (120mV/div) it change to voltage band II.

ADC output is 8 bit.  Simple solution is magnify this digitally. Just drop resolution. Like example in some Keysight and also some others. Also Siglent do this kind of magnification in SDS2000 series from 2mV/div to 1mV/div.

But in series SDS1000X-E there is Analog Devices HMCAD1511  ADC.  Also this ADC output is 8bit.
Secret is inside this ADC. Internally it is 13 bit.  5 bits more than output resolution 8 bit.
Analog Devices tell that it can use even 32X magnification without missing codes. Just this 5^2=32.
And Siglent use this, together co working with front end Variable Gain Amplifier for get full resolution including all major and fine steps starting from 500uV/div, then 510uV/div and so on... until 118V/div. Then change front circuit and go to 120mV/div to 1.18V/div with all fine and major steps and then change again front circuit and start from 1.2V/div to 10V/div with all fine and major steps.

So we do not need digitally magnify ADC 8 bit output at all. We must not mess and mix this system with systems what use 8bit ADC output magnification method at all. It leads to missing codes, (reduced resolution)
In Siglent vernier and major steps mode (they are not different except relay selected three Voltage bands) We can use VGA + ADC internal 13bit resolution for do all this without missing codes  full 8bit ADC output.

 

Note "digital gain", not "digital magnification" done for ADC output as some Siglent and other brands some models.

[Rerouter]

The command I detailed lets a user manually over-ride the VGA, if you max out the VGA with code 255, that is what I measured its equivilence to be, its noisy as all heck down at that level, but it does measure it (the scope in this situation does not correct its VDIV values, becuase you have essentially over-ridden the UI,

With this method, you could set 1 channel to say 5mV, another to 500uV, and manually override the first channel to 500uV's VGA code to compare if they are doing any other trickery. as this bypasses the entire scope UI

[The Soulman]

The VGA makes 1 dB steps?
That doesn't sound like fine adjustment to me.

And also from the ADC datasheet:

The full-scale voltage range of HMCAD1511 can be adjusted using an internal 6-bit DAC controlled by the fs_cntrl
register. Changing the value in the register by one step, adjusts the full-scale range by approximately 0.3%. This
leads to a maximum range of ±10% adjustment. Table 9 shows how the register settings correspond to the full-scale
range. Note that the values for full-scale range adjustment are approximate. The DAC is, however, guaranteed to be
monotonous.
The full-scale control and the programmable gain features differ in two major ways:
1. The full-scale control feature controls the full-scale voltage range in an analog fashion, whereas the
programmable gain is a digital feature.
2. The programmable gain feature has much coarser gain steps and larger range than the full-scale
control.

Do they use a combination of VGA and ADC analog domain adjustment for the vernier control?

I also don't see how they would keep real 8 bit resolution at 500uV range if 1mV is also 8 bit,
the datasheet doesn't mention adjusting the v-ref. (although the analog 10% adjustment may be doing just that).

Could they run the ADC at a higher sample rates at lower analog level input (lower settling times) and oversample??

(they=siglent)

edit: rf loop beat me.

[rf-loop]

The VGA makes 1 dB steps?
That doesn't sound like fine adjustment to me.

And also from the ADC datasheet:

The full-scale voltage range of HMCAD1511 can be adjusted using an internal 6-bit DAC controlled by the fs_cntrl
register. Changing the value in the register by one step, adjusts the full-scale range by approximately 0.3%. This
leads to a maximum range of ±10% adjustment. Table 9 shows how the register settings correspond to the full-scale
range. Note that the values for full-scale range adjustment are approximate. The DAC is, however, guaranteed to be
monotonous.
The full-scale control and the programmable gain features differ in two major ways:
1. The full-scale control feature controls the full-scale voltage range in an analog fashion, whereas the
programmable gain is a digital feature.
2. The programmable gain feature has much coarser gain steps and larger range than the full-scale
control.

Do they use a combination of VGA and ADC analog domain adjustment for the vernier control?

I also don't see how they would keep real 8 bit resolution at 500uV range if 1mV is also 8 bit,
the datasheet doesn't mention adjusting the v-ref. (although the analog 10% adjustment may be doing just that).

Could they run the ADC at a higher sample rates at lower analog level input (lower settling times) and oversample??

(they=siglent)

Read my previous msg.

[EEVblog]

The full-scale voltage range of HMCAD1511 can be adjusted using an internal 6-bit DAC controlled by the fs_cntrl
register. Changing the value in the register by one step, adjusts the full-scale range by approximately 0.3%. This
leads to a maximum range of ±10% adjustment. Table 9 shows how the register settings correspond to the full-scale
range. Note that the values for full-scale range adjustment are approximate. The DAC is, however, guaranteed to be
monotonous.
The full-scale control and the programmable gain features differ in two major ways:
1. The full-scale control feature controls the full-scale voltage range in an analog fashion, whereas the
programmable gain is a digital feature.
2. The programmable gain feature has much coarser gain steps and larger range than the full-scale
control.

Do they use a combination of VGA and ADC analog domain adjustment for the vernier control?

Don't know, interesting question.

[rf-loop]

The full-scale voltage range of HMCAD1511 can be adjusted using an internal 6-bit DAC controlled by the fs_cntrl
register. Changing the value in the register by one step, adjusts the full-scale range by approximately 0.3%. This
leads to a maximum range of ±10% adjustment. Table 9 shows how the register settings correspond to the full-scale
range. Note that the values for full-scale range adjustment are approximate. The DAC is, however, guaranteed to be
monotonous.
The full-scale control and the programmable gain features differ in two major ways:
1. The full-scale control feature controls the full-scale voltage range in an analog fashion, whereas the
programmable gain is a digital feature.
2. The programmable gain feature has much coarser gain steps and larger range than the full-scale
control.

Do they use a combination of VGA and ADC analog domain adjustment for the vernier control?

Don't know, interesting question.

Of course.

[EEVblog]

The full-scale voltage range of HMCAD1511 can be adjusted using an internal 6-bit DAC controlled by the fs_cntrl
register. Changing the value in the register by one step, adjusts the full-scale range by approximately 0.3%. This
leads to a maximum range of ±10% adjustment. Table 9 shows how the register settings correspond to the full-scale
range. Note that the values for full-scale range adjustment are approximate. The DAC is, however, guaranteed to be
monotonous.
The full-scale control and the programmable gain features differ in two major ways:
1. The full-scale control feature controls the full-scale voltage range in an analog fashion, whereas the
programmable gain is a digital feature.
2. The programmable gain feature has much coarser gain steps and larger range than the full-scale
control.

Do they use a combination of VGA and ADC analog domain adjustment for the vernier control?

Don't know, interesting question.

Of course.

And once again, what are you insinuating?

[The Soulman]

Read my previous msg.

Yes, thanks.
Could you post a link to the datasheet that shows that graph and mentions 13 bit?

Silly little ADC that is, 13 bit resolution on a 8 bit output..

Glad this isn't the metrology forum, also from datasheet:

ENOBexcl Effective number of Bits, excluding interleaving spurs
Single Ch Mode, FS = 1000 MSPS 7.9 bits
Single Ch Mode, FS = 1000 MSPS, FIN = 170 MHz 7.4 bits
Single Ch Mode, FS = 1000 MSPS, Gain = 10X 7.6 bits
Single Ch Mode, FS = 500 MSPS 7.9 bits
Single Ch Mode, FS = 500 MSPS, Gain = 10X 7.6 bits
Dual Ch Mode, FS = 500 MSPS 7.8 bits
Quad Ch Mode, FS = 250 MSPS 7.9 bits

And linearity is actually not to bad for a 8 bit high speed ADC,
but if you translate that to 12 or 13 bit, it is pretty horrible.

DNL Differential non linearity ±0.2 LSB
INL Integral non linearity ±0.5 LSB

[The Soulman]

The full-scale voltage range of HMCAD1511 can be adjusted using an internal 6-bit DAC controlled by the fs_cntrl
register. Changing the value in the register by one step, adjusts the full-scale range by approximately 0.3%. This
leads to a maximum range of ±10% adjustment. Table 9 shows how the register settings correspond to the full-scale
range. Note that the values for full-scale range adjustment are approximate. The DAC is, however, guaranteed to be
monotonous.
The full-scale control and the programmable gain features differ in two major ways:
1. The full-scale control feature controls the full-scale voltage range in an analog fashion, whereas the
programmable gain is a digital feature.
2. The programmable gain feature has much coarser gain steps and larger range than the full-scale
control.

Do they use a combination of VGA and ADC analog domain adjustment for the vernier control?

Don't know, interesting question.

Of course.

And once again, what are you insinuating?

I can't speak for rf loop, but I think he is implying that 1 dB aren't really fine adjustment and so they must use a other way on top of that and only other way is the analog adjustment inside the ADC.

[rf-loop]

The full-scale voltage range of HMCAD1511 can be adjusted using an internal 6-bit DAC controlled by the fs_cntrl
register. Changing the value in the register by one step, adjusts the full-scale range by approximately 0.3%. This
leads to a maximum range of ±10% adjustment. Table 9 shows how the register settings correspond to the full-scale
range. Note that the values for full-scale range adjustment are approximate. The DAC is, however, guaranteed to be
monotonous.
The full-scale control and the programmable gain features differ in two major ways:
1. The full-scale control feature controls the full-scale voltage range in an analog fashion, whereas the
programmable gain is a digital feature.
2. The programmable gain feature has much coarser gain steps and larger range than the full-scale
control.

Do they use a combination of VGA and ADC analog domain adjustment for the vernier control?

Don't know, interesting question.

Of course.

And once again, what are you insinuating?

I can't speak for rf loop, but I think he is implying that 1 dB aren't really fine adjustment and so they must use a other way on top of that and only other way is the analog adjustment inside the ADC.

No. It is not analog adjustment inside ADC. It is digital adjustment. From internal 13 bit we scale out 8bit depending amount of digital gain. No missing codes up to 32x. If go over 32x then of course missing codes because there is not anymore enough resolution.  It can also see in this ADC bloc diagram where it happen. After ADC. So agter signal is digitized. But ADC resolution is more than 8bit (of course 13 bit resolution is what it is... if look it with metrologist eyes... terrible.). But also it is not digital magnification if we talk about this kind of digital magnification what magnify displayed waveform on the screen like example what Keysight do. 4x magnification there mean that they take half of ADC data and magnify it so that it is now full screen so they can write 1mV/div  (display scale) and resolution is 64 bits, or 2mV/div with 7 bit resolution.

But there is 1511 ADC we can select and scale 13bits to 8bit (up to 32x magnification) (also there can use FS fine adjustment but I believe this is used only for cal) and output is 8 bit resolution (but with reduced some guality things) I do not "believe" that Siglent use this max 32x digital gain at all. But some amount  what need together in VGA what give acceptable compromise. (result we can see also in BodePlot II and FFT what can not do what these can if there is just simple digital magnification after ADC 8 bit output)

[EEVblog]

No. It is not analog adjustment inside ADC. It is digital adjustment.

How do you know it doesn't include the ADC reference DAC adjustment as well?

[nctnico]

I like to point out that a variable gain amplifier isn't the only way to do vernier control. Another way is to use a DAC to drive the ADC reference voltage. So even if an osciloscope doesn't have a VGA it may do vernier control in the analog domain.

[EEVblog]

I like to point out that a variable gain amplifier isn't the only way to do vernier control. Another way is to use a DAC to drive the ADC reference voltage. So even if an osciloscope doesn't have a VGA it may do vernier control in the analog domain.

All things considered though I think it's always the most desirable to do it in the front end amplification though.
But chips like the HAD1511 are purpose designed to do some or all of it more cost effectively inside the ADC.

[goaty]

Very interresting Video.

I read the HMCAD1520 datasheet (its from Hittite/AD) and it can do 8, 12 and even 14bits and mentions:

"For the high speed interleaved modes, there will be no missing codes when using digital fine gain, due to higher resolution internally (1 bit)."

So could this be a 1bit ADC ? Because it nowhere mentions internal '13bit' or alike. Only on some public announcement there is mentioned 13bits.

Thank you

[ogden]

I like to point out that a variable gain amplifier isn't the only way to do vernier control. Another way is to use a DAC to drive the ADC reference voltage. So even if an osciloscope doesn't have a VGA it may do vernier control in the analog domain.

All things considered though I think it's always the most desirable to do it in the front end amplification though.

Unless 40dB range of VGA is insufficient, "digital" gain of ADC does not help because it's gain granularity is 6dB 1dB compared to 1dB VGA steps. I think ADC reference adjustment is plausible. Isn't that (ADC VREF) good idea to check in follow-up video?

[edit]  After (more) careful reading of datasheet I did find out that ADC gain step is 1dB. Anyway it does not change my statement that ADC gain does not improve anything for vernier.

[Kleinstein]

The SNR vs "digital" gain curve does not look like just digital domain scaling. It looks more like scaling down the reference, maybe add analog gain before the actual ADC.  So I would consider the 13 Bit number more like marketing talk combining 8 Bit real ADC and a gain of up to 32 (2^5) gain adjust, even though SNR drops at a "gain" of more than about 6. So the actual useful gain adds more like 2-3 bits and not 5 bit to the resolution at the fine end.

This still may be enough to get the extra boost from 2 mV/div to 500 µV /div..

[ogden]

The SNR vs "digital" gain curve does not look like just digital domain scaling. It looks more like scaling down the reference, maybe add analog gain before the actual ADC.  So I would consider the 13 Bit number more like marketing talk combining 8 Bit real ADC and a gain of up to 32 (2^5) gain adjust

They would not lie about internal 13-bit architecture. Accusing (ex)Hittite of something like that is EE blasphemy.

[tautech]

Video link on YouTube:

https://youtu.be/7v-P75MOZ5o

Where Dave adds this note:
NOTE: The Rigol DS1054Z does have the ability to do vernier control. It uses the same HAD1511 ADC as the Siglent, and that ADC has two methods of vernier control. One is the digital gain after the ADC (available because the ADC is actually up to 13 bits internally). And second, the ADC reference voltage has an internal DAC that can be changed.
So either or both of those techniques can be used, technically no need for a Variable Gain Amplifier on the front end.

So why then is it used ? 

[ogden]

Where Dave adds this note:
NOTE: The Rigol DS1054Z does have the ability to do vernier control. It uses the same HAD1511 ADC as the Siglent, and that ADC has two methods of vernier control. One is the digital gain after the ADC (available because the ADC is actually up to 13 bits internally). And second, the ADC reference voltage has an internal DAC that can be changed.
So either or both of those techniques can be used, technically no need for a Variable Gain Amplifier on the front end.

So why then is it used ? 

Because VGA provides 40dB (100x) dynamic range. 13bits of ADC reach around 20dB (10x) with good SNR, ADC reference does not count because it is for small (<1dB) adjustments.

[Mr. Scram]

Looks like we don't want Siglents on the channel in the future. The hardware isn't terrible but the crowd is a nuisance.

[ogden]

Looks like we don't want Siglents on the channel in the future. The hardware isn't terrible but the fanboys are a nuisance.

What are you talking about? Mind to quote?

[StillTrying]

After reading the HMCAD1511 data sheet I'm quite sure it doesn't have 13 ADC bits internally, although it does have 8 ADCs.

If the fine gain could be adjusted by 2^9 the gain could only be adjusted to X 2 or X 0.5
The 2^13 means amplitude wise 1LSB can be adjusted in steps of +/- 1LSB/(13-8) = +/- 1LSB/32, but there's still only 8 bits.
I think.