A/D converter for software defined radio, oscilloscope applications, etc.

[AB2EZ]

Hi!

We (the readers of this board) have been using a number of different types of hardware/firmware sampler + A/D converter subsystems for: SDR, audio processing, and for special and general purpose test equipment applications. E.g. using a plain vanilla sound card in a computer; and products like: SDR-I/Q, SDR-1000, QS1R, USRP1, PERSEUS, CGR-101, etc.

Here is a look at the high end. [I don't plan to purchase one of these for my shack]

http://www.guzik.com/product_detail_WDM5000.asp

Best regards
Stu

[w3jn]

Not sure how high end that is, its sampling rate is impressive but only 8 bits resolution is definitely bottom-barrel.

Here's something quite a bit more useful:

http://cp.literature.agilent.com/litweb/pdf/5989-1505EN.pdf

[AB2EZ]

In terms of the hardware, there is the usual tradeoff:

The Guzic WDM 5121 product (based on what is published on their web site) has a bleeding edge sampling rate of 40GSPS, but only 8 of bits of accuracy in the A/D converter. [I remember when a 15 MHz sampling rate and 8 bits of accuracy was the state of the art for commercially available A/D converter products, when I was a much younger engineer, circa 1979.]

The Agilent 3238S product (based on what is published on their web site) heterodynes the incoming VHF (or higher frequency) signal to an i.f with a 36 MHz bandwidth, and then samples (presumably at around 95 MSPS, as shown in the block diagram) with a larger number of bits of accuracy (not explicitly specified) in the associated N6830A A/D converter subsystem. [My guess would be 16 bits of accuracy]

Of note: a factor of 2 in the (over) sampling rate (combined with amplitude dithering) corresponds to the equivalent of another 1 bit of accuracy... if one over-samples a lower bandwidth signal. I.e. a factor of 2 in the (over) sampling rate reduces the quantization noise by 6dB (if done perfectly).

The Guzic product (based on what is published on their web site) has a sampling rate of around 400x that of the Agilent product. Using 256 = 2**8,  one could obtain the equivalent of up to 16 bits of accuracy, using the oversampling + amplitude dithering method.

Having said the above, I would much rather have the Agilent product in my lab at NJIT... given the research projects we are conducting, and the powerful software that comes with it. [Actually, I would much rather have the money to spend on less expensive items]

Stu

[Steve - WB3HUZ]

Tektonix has off-the-shelf scopes that sample at 50GS/s. 40GS/s is not bleeding edge.

[WA1GFZ]

I remember making video with a DAC 08
Ah to have another 8 bits of resolution with a saturated input of 5 volts.....

[w3jn]

Indeed, the A/D in the E3238S is nothing special these days.  It uses the tried-and-true method of a synthesized RF converter whose IF is dumped into the A/D.

What's impressive about the E3238S is the analysis software behind it.   An A/D converter, no matter how fast, is pretty useless without a good backend.

Unfortunately, I suspect we'll not see many of these in hobbyists' hands in working order.  I saw an earlier one on eBay a few months ago, with the microwave tuner.  No computer though.  It went for about a thousand dollars, as without the software it was pretty much a boatanchor.

[WA1GFZ]

John,
There are some sharp software guys out there who could make it talk to a computer. I'm really impressed with some of these guys. Give them a data stream and it will get decoded. A boat anchor to us hardware pukes is gold to them.

[AB2EZ]

Steve

Perhaps I am misinterpreting the materials on line but:

The highest speed scopes on the Tektronix web site appear to be "sampling scopes" ... where the sampling gate width is very narrow (<10 picoseconds), but the number of samples per second is much less than 1/the sampling gate width. These types of oscilloscopes are described (in their specifications) as having a "sampling rate" that is 1/the sampling gate width, even though they don't produce that many samples per second. Therefore, these scopes are useful for analyzing very high bandwidth periodic waveforms (where you can sample the same repeating waveform many times in succession).

I think the Guzik boxes are straight A/D converters that actually sample at the rates quoted. Therefore I think they can be used to sample non-periodic waveforms having bandwidths up to 0.5/the quoted sampling rate.

Separately... the highest speed pieces of test equipment that are sold by commercial firms (like Tektronix, Hamamatsu, Anritsu and Agilent) are bleeding edge. That's why they cost so much. They are typically hand made and hand tweaked under the supervision of development engineers (and members of the research staff).

Stu

[Steve - WB3HUZ]

Tek claims up to 50 GS/s real time sample rate. I guess it comes down to what exactly they mean by real time. The analog bandwidth on the scope is only (only  ) 16 GHz, so it would appear they are playing some games.

[AB2EZ]

Steve

Yes... this model Tektronix scope looks like it has a real time sampling rate of 50 GSa/S. I would bet that these get built and sold with continuous consultation with the Tektronix development engineering and research staff to deal with problems that come up in unit testing and system testing, as well as problems that come up with systems in the "field".

Stu

http://www2.tek.com/cmswpt/psdetails.lotr?ct=PS&cs=psu&ci=14764&lc=EN

http://www.trs-rentelco.com/Model/TEK_DPO71604.aspx [As low as $93,230, including power cord and accessory pouch (for a used system)]

[Steve - WB3HUZ]

Could be. I know people using other scopes in the same line (slightly lower sampling rate) that have not had any problems. Same for their RSA analyzers. They are a ton of money, so they should come with a full time engineering too!   

Steve

Yes... this model Tektronix scope looks like it has a real time sampling rate of 50 GSa/S. I would bet that these get built and sold with continuous consultation with the Tektronix development engineering and research staff to deal with problems that come up in unit testing and system testing, as well as problems that come up with systems in the "field".

Stu

http://www2.tek.com/cmswpt/psdetails.lotr?ct=PS&cs=psu&ci=14764&lc=EN

http://www.trs-rentelco.com/Model/TEK_DPO71604.aspx [As low as $93,230, including power cord and accessory pouch (for a used system)]

[KD6VXI]

Steve

Yes... this model Tektronix scope looks like it has a real time sampling rate of 50 GSa/S. I would bet that these get built and sold with continuous consultation with the Tektronix development engineering and research staff

Stu

Stu,

I was gonna chime in with, I'd bet the contracts they sign come with engineers cell phones with them, on sales of quantity one or two.

When I purchased a MAJOR (1000+ seat) license for McAfee's products, immediately after they had been purchased, I noticed that I also got a LOT of phone numbers.

Purchasing a quarter of a million dollars in Dell hardware also got me a team lead in Austin to personally bring me a hard drive replacement on a (at the time SUPER HIGH TECH) 18K rpm scsi Cheetah drive.  He was there for literally, 10 hours, trying to get the raid back.  In the end, it became a rewipe and restore from DAT.

People who can afford a 100K dollar scope often DON'T need tech support, but when they do, believe you me, they don't have to wait in ANY call center line.  They call someone who designs the software for it, or the chips inside it, on their cellphone.  And it's answered while they are at lunch, with the wife, at little league games, etc.

At least, that's my review of "corporate America, Houston Style!"

--Shane

[Steve - WB3HUZ]

I've purchases tons of stuff from Tek over the last 25 years. It's all done through what they've called field engineers and/or account managers. These guys serve a region or a set of specific customers. They have good call back into the company, if you need it, but you won't generally be talking to any designers unless you have a very unique application, or you want to modify equipment. They may open the curtain, but you will usually sign an NDA if company IP is involved.