Any textbook recommendations to get up to speed with logic circuits?

[benb]

... digital control of analog circuits.

What you may especially want to look at are "analog switches." In the original CMOS 4000 series there's the 4016, 4066, 4053 and a few others - you should learn about these mainly for when you see then in older equipment. There's the newer DG4xx series made by several makers that's much better in every way - separate digital interface, analog switching up to +/-15V, lower on resistance, faster switching, all at a "moderately" higher cost.

There are classic analog circuits such as full-wave rectifiers that use signal diodes around op-amps to do the switching, but if you really need high precision with such a circuit, forget the diodes and use a comparator driving analog switches.

The Art of Electronics mentions this DG series.

 

[Gene Pink]

...and use a comparator driving analog switches.

That just brought back a memory.

The Roland SDE 3000 from the '80's used a 4066 as part of the D-A. They advertised 16 bit performance, and used a 12 bit DAC both ways with successive approximation, along with every trick in the book to get to that 16 bit level.

A 12 bit DAC, along with a 4066 driven with bits 13 and 14 with 0.1% resistors for switching attenuation, and then this whole thing surrounded with a 4:1 analog compander.

Talk about jumping through hoops to get there. Damn.

Gene

 

[JohnRoberts]

That just brought back a memory.

The Roland SDE 3000 from the '80's used a 4066 as part of the D-A. They advertised 16 bit performance, and used a 12 bit DAC both ways with successive approximation, along with every trick in the book to get to that 16 bit level.

A 12 bit DAC, along with a 4066 driven with bits 13 and 14 with 0.1% resistors for switching attenuation, and then this whole thing surrounded with a 4:1 analog compander.

Talk about jumping through hoops to get there. Damn.

Gene

A bad memory?

For A/D I don't know if 0.1% Rs are accurate enough, while linearity is not important for D/A. The on resistance of the 4066 while better than 4016 could still be an issue for accuracy (at 16B?) .

I'm not sure how to make a 4:1 compander cheaply. Perhaps two 2:1 companders stacked in series.  Since  a 1:2 expander doubles any path frequency response errors, a 1:4 would double them again.

A 2:1 compander wrapped around a 12B codec would double that dynamic range so in theory  140dB, in practice the noise floor and dynamic range of the gain element would be the limiting factor...

I killed a lot of brain cells trying to think about ways to extend the dynamic range of early digital systems but it always came back to linearity being the limitation.  Any front end scaling needs to have the final resolution or an A/D conversion might zig when it should zag. For D/A who cares about tiny errors.

One of the more promising early digital systems was DBX who wrapped a high quality 2:1 compander (their pro NR design)  around a decent delta-mod encoder. The main downside was that the digital data was not in a format that supports DSP manipulations. (They recorded to a variant VCR format.)

JR

 

[Gene Pink]

A bad memory?

No, the RAM is fine.

You ain't kidding there.  Cut me a break, it has been ~20 years since I have been all up inside of one of these things. I dug out the SDE 1000/3000 service manual (my daughter's head would explode), the 4066 is in the 1000, and appears to be a mute, with all control inputs tied together. The 3000 has a different switch with a house number, two inputs, each controlling separate switches, one N.O. and one N.C. per input wired like a single pole, double throw. It may be a low speed scaler, as it is driven by a block called "Date(sic) Companding Logic". It switches taps on a resistor string, but at what speed, is it part of the binary word? It comes from the ram, but through a different path. It is also within the successive approximation loop, so technically it is part of the word. I'll scan it later tonight, and would appreciate any help.

For A/D I don't know if 0.1% Rs are accurate enough, while linearity is not important for D/A. The on resistance of the 4066 while better than 4016 could still be an issue for accuracy (at 16B?) .

Actually John, I don't believe resistor accuracy means much in this case. It is only a delay, there is no DSP or math going on. The word derived from analog comes back the same as it went into memory, just later. Being successive approximation with the same DAC and scaler used both ways, should cancel out any errors.

I'm not sure how to make a 4:1 compander cheaply. Perhaps two 2:1 companders stacked in series.  Since  a 1:2 expander doubles any path frequency response errors, a 1:4 would double them again.

The 1000 has a NE570N compander, I'll scan and post it later, so we can figure out  what they are really doing, I would appreciate the help. The 4:1 that I was thinking was probably the scaler of the 3000.

A 2:1 compander wrapped around a 12B codec would double that dynamic range so in theory  140dB,...

140Db?  I'm sure you meant double the voltage.  ;D

Gene

 

[user 37518]

Embedded control obviously makes the life vastly easier but there are times it can be a bit like cracking a walnut with a sledge hammer.

I suggest Digital Electronics Fundamentals and Applications by Roger Tokheim.  Also look for the Experiments Manual. Older prints go very cheap on e-bay. Great, practical, entry level book without getting too bogged down in theory.

CMOS Cookbook by Don Lancaster is also a great practical book. (While you are at it grab a copy of his Active Filter Cookbook too.)

I respecfully disagree about the Active Filter Cookbook, it only features the most basic topologies and approximations. A good filter book is Electronic Filter Design Handbook, by Arthur B. Williams, its best if you can get an older edition (I have the 2th) because it includes tables for elliptic filters.

 

[JohnRoberts]

apologies for a very late response

No, the RAM is fine.

You ain't kidding there.  Cut me a break, it has been ~20 years since I have been all up inside of one of these things. I dug out the SDE 1000/3000 service manual (my daughter's head would explode), the 4066 is in the 1000, and appears to be a mute, with all control inputs tied together. The 3000 has a different switch with a house number, two inputs, each controlling separate switches, one N.O. and one N.C. per input wired like a single pole, double throw. It may be a low speed scaler, as it is driven by a block called "Date(sic) Companding Logic". It switches taps on a resistor string, but at what speed, is it part of the binary word? It comes from the ram, but through a different path. It is also within the successive approximation loop, so technically it is part of the word. I'll scan it later tonight, and would appreciate any help.

Actually John, I don't believe resistor accuracy means much in this case. It is only a delay, there is no DSP or math going on. The word derived from analog comes back the same as it went into memory, just later. Being successive approximation with the same DAC and scaler used both ways, should cancel out any errors.

If it is only a 12b DAC with 4 bits of gain in the front end the accuracy of those gain steps needs to be 16b or better  to avoid erroneous conversions. 2X needs to be 2.000000x.  Perhaps if you used the exact same gain resistors for input boost and output cut there should be a first order cancellation of errors, assuming output is a 1:1 replica of the input. Any crunching inside the digital domain and all bets are off.

The 1000 has a NE570N compander, I'll scan and post it later, so we can figure out  what they are really doing, I would appreciate the help. The 4:1 that I was thinking was probably the scaler of the 3000.

a single ne570 can do 2 channels of either 2:1 compression or 1:2 expansion, so one 570 can cover both ends of a 2:1 compander,

12B conversion is in theory 72dB dynamic range.... in theory a 2:1 compressor in front of a 12B ADC would extend that dynamic range to 144 dB.... of course in practice the noise floor of a cheap NE570 gain element is at best -80 to -90 dBu.  A premium VCA would still have a noise floor in the -100dBu range.

140Db?  I'm sure you meant double the voltage.  ;D

Gene

No companding doubles the dynamic range in dB....  this is why we used companding on noisy BBD delays to get usable audio from a delay path with maybe -50dBu noise floor  (or worse).  The real time path S/N isn't changed by companing but is like constantly optimizing the input and output levels...  when music is present you still only get the raw path S/N...

As i shared DBX made a very wide dynamic range recording system back in the day, but there were several compromises,,, Companded delta-mod digital data could not be easily crunched in the digital domain and dynamic behavior of even professional 2:1 companding is not perfect. Not as good as modern straight digital conversion using enough bits.

JR

 

[PRR]

> Lancaster ....Active Filter Cookbook, it only features the most basic topologies and approximations.

It was a revelation when it came out. Heavy-math reduced to cookbook form. "Basic" topologies were what we needed, and often sufficient. Not sure your "approximations" point: you can work "exact" values.

Elliptics were not audio-common before the digital thing came along. Now of course it would be chaos without them.

 

[user 37518]

> Lancaster ....Active Filter Cookbook, it only features the most basic topologies and approximations.

It was a revelation when it came out. Heavy-math reduced to cookbook form. "Basic" topologies were what we needed, and often sufficient. Not sure your "approximations" point: you can work "exact" values.

Elliptics were not audio-common before the digital thing came along. Now of course it would be chaos without them.

By approximations I meant filter response approximations such as Butterworth, Chebyshev, etc...

As for the book, I can see your point of having a heavy-math reduced cookbook, but "Electronic filter design handbook" by Arthur B. Williams is also very old (I believe the 1st edition is from 1981, but Lancaster's is even older)  and offers a cookbook approach, but with much more circuit and filter options, elliptic included.

Regarding the OP question, I would like to add my 2 cents by recommending "Digital Systems: Principles and Applications" by Tocci, It is a great book for a digital overview ranging from binary numbers, logic gates, flip flops, counters, registers, memories, ADC, DAC, etc... it also includes implementation with VHDL, and an introduction to PLDs, so taking the jump to FPGAs should be easier.

If you are interested on microcontrolers, then I would recommend microcontroller specific books, there are hundreds of them, some people like AVRs, some like PICs, etc, choose your poison and start coding. I wouldnt rule out using Arduinos as they are becoming more and more powerful, for example the Arduino DUE lets you do some DSP processing if thats a thing for you. And then of course there are the mini computers such as the Raspberry Pi, which use a Linux operating system and are meant to be programmed with Python.