Tek tds 3054b vs tds 744a

[tmbinc]

Though this stuff gets interesting on the low-end scopes, like the TDS2xx-series (and all the "modern" variants of it). They don't have high-speed ADCs, but a switched-capacitor array (which is whey they are limited to 2500 points; that odd non-power-of-two number) coupled with a slow ADC. That's why they exist with up to 2GS/s at moderate cost.

Now, the TDS224 for example also offers Peak-Detect. How would that work, given that the ADCs operate after "decimation"?

Point is, it doesn't. The manual states that "Peak Detect functions at SEC/DIV settings of 5us/div or slower. For SEC/DIV settings of 2.5us/div or faster, the acquisition mode switches to Sample automatically." 5us/div are 2500/50us = 50MS/s. That's the rate of the external ADC - at those slow sample rates, they don't use the SCA but sample the signal directly...

Upgrading from a TDS224 to a TDS2024 was a questionable decision (but I got the exchange for free). Upgrading from the TDS2024 to a DPO4034 (+software bandwidth hack) was a phenomenal (yet expensive) decision. I'm a happy DPO5034 (+hardware bandwidth hack) owner now, though the user interface sometimes drives me crazy.

[nctnico]

The TDS744A has a boat load of ways to acquire and display a signal. Modern scopes are dumbed down a lot.

The TDS744A's manual only lists these:

• Sample Mode - the normal sample mode
• Envelope Mode - collecting Extrema over multiple acquisitions
• Average Mode - collecting average over multiple aquisitions
• Peak Detect - essentially Envelope Mode over only two acquisitions
• Hi-Res Mode - oversampling to increase the vertical resolution

It's certainly a good selection of options for a scope from 1994, but by today's standards none of these modes are anything special. What modes do you miss on modern scopes?

On a TDS744A you can use 20 or so combinations of acquisition and display modes. Most with the ability to tweak some parameters.

[cncjerry]

Hotdog, Not sure if it is 20 because when you switch on InstaVue it doesn't allow Hi-res, for instance.  Also, InstaVue does it in full false color whereas the DPO is more like Z brightness.  From what I can tell, the DPO features are constant across all modes, but since it doesn't have hi-res which is disabled in InstaVue that isn't a fair statement.

So this morning I have the 3054b sitting on top of the 744a and I am looking at an old Instek GFG-8255A signal generator.  Should really be called a signal + noise + glitch generator.  What has always bothered me about this generator is that at the transition from up to down slope and the inverse, there is a little glitch. More obvious with a triangle, less so with a sine but still there.  I had to make some trade-offs in the two scope pictures but here they are below. 

Since this thread is getting to the point where people are putting on their mitres, I thought I would switch to that mode as well.

The first is from the 744a.  In this picture the step wouldn't be called out unless thou knew it was there.   I can see it a little as the slope is more acute at the top (maybe). The second is from the 3054b and though thou can see the glitch as that little down step, even after playing with the 3054b for 30 minutes and I am new to that scope, I can't get better resolution.  I left both in sample mode for the picture.  Thou will notice the step is more obvious on the 3054b.

No real conclusions here.  When I started this thread the intent was to garner opinions from thee (or is it thou?) on whether I should keep the 744a or sell it.  I don't need the $400 I would get for it (if lucky) but it does take up space.

If I had the cash, I would buy a 500Mhz or better Agilent or whatever they are called today.  Maybe Lecroy.  I think the value is in those brands more so than current Tektronix.  If I sold both of these scopes for lets say optimistically $2k total, I don't know if I could do better.


Thanks.

[nctnico]

You are using a shorter acquisition memory on the TDS744A which is why the distortion doesn't show when zoomed in. Just set the TDS744A to 25ksamples (or so) memory length and set 'fit to screen' to 'ON'.

[MarkL]

The first is from the 744a.  In this picture the step wouldn't be called out unless thou knew it was there.   I can see it a little as the slope is more acute at the top (maybe). The second is from the 3054b and though thou can see the glitch as that little down step, even after playing with the 3054b for 30 minutes and I am new to that scope, I can't get better resolution.  I left both in sample mode for the picture.  Thou will notice the step is more obvious on the 3054b.

One trick to getting "better resolution" is to use the scope's DC offset control to center the area of interest in the ADC's working range.

Below is an example on a TDS3054 using approx the same signal you used (12V Pk-Pk triangle at about 3.6kHz).  The DC offset is set to +5.5V and this allows the vertical sensitivity to be set to 100mV/div.  The delay has also been enabled and tweaked to bring the peak to the horizontal center.  Other settings of interest are "Avg 4", and trigger "HF Reject" to help stabilize the image.

100mv/div is as far down as you can go on the TDS3054 for this particular signal.  Any less and the offset range max changes to +/-1.000V, which isn't enough to bring the peak into range.

As you can see, this old FG504 function gen also has some transition issues at the peak, but you'll see that on any triangle generator to some degree.

[David Hess]

Though this stuff gets interesting on the low-end scopes, like the TDS2xx-series (and all the "modern" variants of it). They don't have high-speed ADCs, but a switched-capacitor array (which is whey they are limited to 2500 points; that odd non-power-of-two number) coupled with a slow ADC. That's why they exist with up to 2GS/s at moderate cost.

Most or all of their non-CCD oscilloscopes have non-power of 2 record and display lengths as well.  Their very early DSOs provide some interesting exceptions to this.  Some of their earlier CCD based DSOs had much higher record lengths but of course they were not expandable.

The TDS2xx series only samples at up to 1 GS/s per channel according to the documentation.

Now, the TDS224 for example also offers Peak-Detect. How would that work, given that the ADCs operate after "decimation"?

Point is, it doesn't. The manual states that "Peak Detect functions at SEC/DIV settings of 5us/div or slower. For SEC/DIV settings of 2.5us/div or faster, the acquisition mode switches to Sample automatically." 5us/div are 2500/50us = 50MS/s. That's the rate of the external ADC - at those slow sample rates, they don't use the SCA but sample the signal directly...

Take a look at a 2440 series service manual to get an idea about how this works in a CCD (charge coupled device) based FISO (fast in slow out) digitizer.  Peak detection is done in the analog domain before the analog shift register as a separate operation.  The cost reduced version of the 2432A, the 2431L, and the TDS600 series up to the B models dropped this capability.  I assume it was not until the B models of the TDS600 series that they were able to reimplement it to operate at sample rates higher than 500 MS/s.

I did not know that Tektronix continued to use FISO based sampling after the TDS600 series but it makes sense in low cost real time DSO designs if you already have the technology.  Tektronix had already divested itself of semiconductor manufacturing by the time that TDS220 series came out but it looks like they had National fabricate the ICs for them.  I wonder if Maxim made the ICs in the much faster TDS600 series.

This explains the rather odd TDS220 series specifications which indicate a maximum input bandwidth of 50 or 75 MHz when peak detection is used although the earlier TDS600 series lack such a limitation.

Upgrading from a TDS224 to a TDS2024 was a questionable decision (but I got the exchange for free). Upgrading from the TDS2024 to a DPO4034 (+software bandwidth hack) was a phenomenal (yet expensive) decision. I'm a happy DPO5034 (+hardware bandwidth hack) owner now, though the user interface sometimes drives me crazy.

I was not happy with the MSO/DPO5000 series interface either and a lot of other things about them as well.  They struck me as a good oscilloscope for ATE (automatic test equipment) environments.

[cncjerry]

nctnico,
I was using 5k points in 100 divs on the 744a.  Even moving it up to 50k in 1000, I get roughly the same picture.  I don't understand the difference in record length between the two scopes.  My 744a has options (depending on horizontal base) from 5k points in 100 to 50k in 1000. I had the 'fit to screen' turned on as well.  Since we are quoting chapter and verse :-), on page 3-12, Normal mode on my 3054b is 10k points but it doesn't say for instance, 10k in 10 divisions.   Since it displays 10 divisions, would it be 10k points in 10 divisions or 1k per?  Comparing to the 744a then, 50k in 1000 would be only 50, so that can't be correct, or maybe it is as the scope is so much older as everyone points out.  By the way, I'm getting a little tired of everyone saying how old my stuff is, it is all I can afford.  I don't have some monster company paying for my toys, it is all me!!!  Just kidding, I know the stuff is old but not as old as some of the things I have.

I was messing around the 3054b again and clearly I need to learn how to operate it better.  I wonder how many people on here make that statement every day? 

Anyway, success is in the pictures below.  The tds744 still has much higher contrast to me, again it could be just the vision defects I have playing into it.  The hardcopy mono prints look pretty much the same though there is variance in the sample time.  One thing that is obvious between the two scopes is the processing engine on the 3054b is much, much faster than the 744a at the same record length.

I do have to figure out how to change the battery in the 744 as one of the previous posters has me concerned now.  When I bought it, it was last serviced and calibrated in like 2011 so I wonder if they replaced the battery.  Also if anyone knows, is there a way to dump the calibration either to diskette or over GPIB in case I lose it the battery?

By the way, thanks for everyone's help.  It's been a very productive thread from my perspective.

Jerry

[cncjerry]

just noticed the difference in horizontal.  Here's the same 3054b at 400ns:

[David Hess]

By the way, I'm getting a little tired of everyone saying how old my stuff is, it is all I can afford.  I don't have some monster company paying for my toys, it is all me!!!  Just kidding, I know the stuff is old but not as old as some of the things I have.

Your TDS744A makes my newest oscilloscope, a 2440, look like an antique.  The the oldest oscilloscopes I regularly use were first produced in 1973 and one of those is still faster than anything Rigol makes for less than $4000.

The tds744 still has much higher contrast to me, again it could be just the vision defects I have playing into it.

The TDS displays were pretty nice and CRTs naturally have high contrast.  Tektronix used that weird LCD color shutter system because color CRTs would have sacrificed display resolution.

I would prefer your TDS744A to the alternatives simply because it supports magnitude/phase FFTs but that is a pretty esoteric feature.

[nctnico]

About the battery in the TDS744A: In the one I had I desoldered the NVRAM and backup it up using a $25 eprom programmer. I soldered the NVRAM in a socket and soldered a socket in the board. I never needed to replace the NVRAM. I still have a TS510A in which the NVRAM seems to be just fine (at least the last time I used it).

[Wuerstchenhund]

Acquisition interval (Tektronix still uses this term.) is not the same as waveform acquisition. 

You're right and I was wrong. My bad, I had a look at some other Tektronix documentation from that aera, which confirms what you said.

Are you thinking of peak detection as Rigol defines it on their older DSOs where they confound peak detection with envelope detection?  They work the way you describe when in peak detection mode which on a Tektronix would be envelope detection mode.

Possibly. The last time I have used a TDS700A scope was probably around 1997, and Peak Detect maybe a couple of times over a decade ago. For the type of work I did it didn't add any benefit. Most shops I worked were predominantly HP (and later Agilent).

But one point still stands, which is that the TDS700A Series was great in 1994 but even though some features like InstaVu work a bit differently than I remembered I can't see anything that can't be done in modern scopes (although I'm not saying that this is necessarily true for Rigols, though). After reading a it more about how the functions in the TDS700A work I can see why Tek had implemented it that way, considering that this is based on 1994 technology with very little processing power. And for a scope from 1994 it's a very good scope.

But I also see the limitations of the technology of that era, which in case of the TDS744A lies in the fixed combination of acquisition modes, the simple Hi-Res mode, the simple and slow math and FFT modes (although the TDS744A at least uses 10kpts for FFT which is much more than the the 700pts or less Rigol uses on their scopes), and the lack of any advanced analysis capabilities.

On a TDS744A you can use 20 or so combinations of acquisition and display modes. Most with the ability to tweak some parameters.

That sounds a lot (and for a scope from the 90's it is) but in comparison with newer scopes it's not much, especially when considering the limitations of most of these modes (i.e. Peak Detect works only on slow acquisitions). For example, if I'd count the possible combination of acquisition and display modes of my oldest scope (LT264M) then even in basic config I'd certainly get to at least 3x that number (and with the options I'd say its probably over 100 possible combinations), and that is a scope from 2003 which is positioned in the same spot as a TDS700/7000. A 9354 (which is from around 1995) should allow a similar number of combinations.

[cncjerry]

I would prefer your TDS744A to the alternatives simply because it supports magnitude/phase FFTs but that is a pretty esoteric feature.

Wait, magnitude, yes, phase?  I have to go back and dig thru my manual as I don't remember seeing phase.  Would that be a two channel math operation?  You have me thinking here tonight as for phase you would need I/Q, current, or a known impedance sense resistor, no?  Don't tell me I bought all that VNA, spectrum analyzer, etc, etc. stuff for nothing...  I wish it wasn't so late to go read-up on it.

[David Hess]

I would prefer your TDS744A to the alternatives simply because it supports magnitude/phase FFTs but that is a pretty esoteric feature.

Wait, magnitude, yes, phase?  I have to go back and dig thru my manual as I don't remember seeing phase.  Would that be a two channel math operation?  You have me thinking here tonight as for phase you would need I/Q, current, or a known impedance sense resistor, no?  Don't tell me I bought all that VNA, spectrum analyzer, etc, etc. stuff for nothing...  I wish it wasn't so late to go read-up on it.

The FFT function in the TDS500 and later series can display the phase information as well as the magnitude so if you have an impulse response measurement in the time domain (or differentiate the step response), then you can use the oscilloscope for low performance vector network analysis.  Some LeCroy oscilloscopes can do this as well (and probably better) and both Tektronix and LeCroy have application notes which discuss the procedure.

http://www.tek.com/document/application-note/fft-applications-tds

[Ivan7enych]

These board ID resistors (R1061<->R1064), whit which the model can be set, are near these EEPROMs as well. Soldered is 1.

TDS784A: 7,   0 1 1 1
TDS754A: 8,   1 0 0 0
TDS744A: 6,   0 1 1 0

Is there any mistake?
I opened my TDS744A and found another picture,
1061 soldered, 1062 open, 1063 open, 1064 soldered.
Can you help me here, what should I made to change it to 784? Can I brick if I make invalid ID?

[blacknoise]

Are all four placeble resistors "0-Ohm jumpers"?

Do all four mounting places for the resistors / jumpers share a common signal or a common level (e.g. GND, +5V, +3.3V...)?

Then i could not resist my curiosity to connect a rotary hex switch to the common line and the four ID select lines to start "turning out" all 16 possibilities... ;-)

Kind regards

[cncjerry]

Are all four placeble resistors "0-Ohm jumpers"?

Do all four mounting places for the resistors / jumpers share a common signal or a common level (e.g. GND, +5V, +3.3V...)?

Then i could not resist my curiosity to connect a rotary hex switch to the common line and the four ID select lines to start "turning out" all 16 possibilities... ;-)

Kind regards

as long as one of the 16 options isn't "CLRNVRAM".

[Ivan7enych]

Finally my TDS744A thinks it is 784A.

TDS784A: 7,   0 1 1 1
TDS754A: 8,   1 0 0 0
TDS744A: 6,   0 1 1 0

In this table, soldered is actually 0, open is 1.

The solution was to remove 0-resistor from R1061 place. Now I'm testing the scope...

[tinhead]

Can I brick if I make invalid ID?

oh well, even if, you made for sure EEPROM and NVRAM backup, right?

NVRAM and these two EEPROMs are protected by defualt, so as long you not change the switch position nothing can be saved there (except the profiles/setups, but they can be purged).

[cncjerry]

Finally my TDS744A thinks it is 784A.

TDS784A: 7,   0 1 1 1
TDS754A: 8,   1 0 0 0
TDS744A: 6,   0 1 1 0

In this table, soldered is actually 0, open is 1.

The solution was to remove 0-resistor from R1061 place. Now I'm testing the scope...

So it reports as a 784a but what happens when it switches to the higher rate, does it work correctly?  As in basically function setting aside that it might need calibration or must it be calibrated first?

[cncjerry]

I got a lot out of this thread and appreciate all the input, very helpful.  A couple more questions and I will move on:

1) on the 3054b, I read tektronix released/unlocked all the module software in version FV4.1.  I have 4.1 on my scope and don't see the advanced analytical functions under the math menu.  Were there two versions of 4.1 or was that not included?

2) I looked around and don't see anything, but was wondering if the 3054b could be hacked up to a 3064B?  It's only a marginal increase but why not if I could?

3) one thing has always bothered me a little on the 744a is the calibration signal seems really noisy.  It could be all the overshoot and I see the same noise testing it with the 3054b.  Both calibration signals on the scopes are the same frequency yet the rise time on the 744a calibration signal is much faster than that on the 3054b: 1.320ns on the 744a;  206.5ns on the 3054b. 1/1.32E-9 is .75Ghz on the 744a and 1/206.5ns is 4.8Mhz on the 3054b.  I'm not using the correct terms, what would 1/(rise time) be called? I don't think this means anything, just trying to get a better understanding of the scopes.

I use my scopes for RF work approaching 1G today and higher in the future.  I've been looking at the low end Agilent scopes, MSOX2024a for example, and the displays look good to me with my vision problem.  I doubt they are special in any way, I think it is more how they present the trace than the display.   If someone made a high contrast OLED display I would jump on it if it was within the reach of a typical independent inventor, say less than 4k or so. 

I completely lost a little over half of my vision. The remaining has two aspects: The first can be described by us technical types as low dynamic range hence the need for high contrast.  If I make the scope bright enough then it washes out other details like the grid and text.  The second problem is that bright lights have blooming, not unlike cranking the trace up too high on an analog scope.

Anyway, thanks again.  If anyone ever needs high end CNC work done, parts made, drilling, etc, please send a PM.  I generally do work gratis for people that share my interests, especially if the need is one-off and interesting.  I don't make parts that can be bought elsewhere unless they Are ridiculously expensive and/or made of unobtanium.

Jerry

[EV]

..
1) on the 3054b, I read tektronix released/unlocked all the module software in version FV4.1.  I have 4.1 on my scope and don't see the advanced analytical functions under the math menu.  Were there two versions of 4.1 or was that not included?
..

I do not know about 3054b modell but in TDS3032 only Adv trigger and FFT modules are unlocked with the latest FW, nothing else.

[Ivan7enych]

Can I brick if I make invalid ID?

oh well, even if, you made for sure EEPROM and NVRAM backup, right?

No. I still have no idea where the battery is inside this scope. Is it possible to read both with GPIB commands? I looked at your source code, and I see some commands there, but I don't have any native GPIB adapter to run your code wihout modifications. I have prologix GPIB adapter, which works as if it is a com port.

[Ivan7enych]

So it reports as a 784a but what happens when it switches to the higher rate, does it work correctly?  As in basically function setting aside that it might need calibration or must it be calibrated first?

Main difference, there appears 4GS\s for single channel and 1GS\s for 4 channels. Analog bandwith, high frequiency rolloff (above 500-600MHz), rise time didn't change at all.

Also, second improvement - now I can see in FFT signals up to 1.6..1.8GHz. There is a significant rolloff, but many radio signals are visible there.

There appears some misaligment between ADCs when I switch single channel aquisition from 2GS\s to 4GS\s, but after warming and self calibration (from scope menu) everything in general looks fine.

[tinhead]

Can I brick if I make invalid ID?

oh well, even if, you made for sure EEPROM and NVRAM backup, right?

No. I still have no idea where the battery is inside this scope. Is it possible to read both with GPIB commands? I looked at your source code, and I see some commands there, but I don't have any native GPIB adapter to run your code wihout modifications. I have prologix GPIB adapter, which works as if it is a com port.

to backup nvram you can use tektool

http://www1.tek.com/forum/viewtopic.php?f=5&t=5527

or this backup tool (booth are working ok)

https://groups.yahoo.com/neo/groups/TekScopes/conversations/topics/79198

There are two nvram chips, one doing clock and what so ever, and the second to store settings.
They mapped to :

straddr = 0x04000000L; /* Start of TDS700A DS1486 - clock chip */
endaddr = 0x0407FFFFL; /* End of TDS700A DS1486 - clock chip */

straddr = 0x04080000L; /* Start of TDS700A DS1650/DS1250 NVRAM */
endaddr = 0x040FFFFFL; /* End of TDS700A DS1650/DS1250 NVRAM */

the cal data, as already said, is in EEPROM on TDS700A, so use that code which posted already (or part of it) with your GPIB adapter.

[David Hess]

3) one thing has always bothered me a little on the 744a is the calibration signal seems really noisy.  It could be all the overshoot and I see the same noise testing it with the 3054b.  Both calibration signals on the scopes are the same frequency yet the rise time on the 744a calibration signal is much faster than that on the 3054b: 1.320ns on the 744a;  206.5ns on the 3054b. 1/1.32E-9 is .75Ghz on the 744a and 1/206.5ns is 4.8Mhz on the 3054b.  I'm not using the correct terms, what would 1/(rise time) be called? I don't think this means anything, just trying to get a better understanding of the scopes.

The calibration signal is usually only intended for checking and adjusting passive probe compensation (and gain and offset on the 744A) so there is little reason for it to have a fast transition time.  I have noticed before though that Tektronix calibration outputs sometimes have fast transition times but there seems to be no pattern as to why.  Some of the LeCroy ones are fast enough to be suitable for TDR applications.

The noise seen on the calibration signal is usually just ambient RF noise picked up by the probe because of its long ground lead.  On oscilloscopes which use a coaxial connection for the calibration signal output, it will be noise free.