digitally controlled stepped attenuators..??

[Rochey]

PRR -

Thanks for the supportive post... Just to clear up a few things -- this is personal time for me - TI doesn't even know I venture into the newsgroups & forum domain. In other words, to cover myself, the opinions expressed in my posts are my own, not TI's.

First things on monday, I'll zap off part of your post to the app engineer for the PGA2310. I don't promise answers, I don't promise a timely response - in all fairness, these guys are usually running at 110% capacity as it is - answering DIY'ers questions will most probably take low priority.

I've read through your post, and there's one thing that strikes me - that people see the PGA2310 in many different parts of the circuit (some at the mic in, some before an ADC etc.)
The main idea behind the PGA23xx family was to use it after the DAC in the system as a volume control - so that volume control does not have to be done digitally with signal processing. The idea being that you're feeding it a line level signal that will most probably be attenuated. Some gain is possible on the device, although it's not it's main strength. I guess you could think of it like using a ferrari to pull a trailer... you could do it... but it's not really what it's designed for. :grin:

If your looking for a part that'll work within a gain circuit (such as a mic input on a desk etc.) then I strongly recommend you look at the PGA2500 digitally controlled Mic Pre. This little honey does 0db then 10dB through to 65dB in 1dB steps. This is the only device of it's kind on the market at the moment (unless you build it discretely... lots of relays etc).

Again, my apologies for seeming a little pushy-- I feel strangely 'proud' when we have a really great device that people want.

If you want to know what TI has in it's 'pro audio' offering (generally anything over 110dB SNR) then take a look at: http://focus.ti.com/lit/ml/szzb002a/szzb002a.pdf (Audio Solutions Selection Guide). There's a Pro Audio section in it that has details of what's available - no vapourware!

If you have any questions, then as usual, start a thread, and I'll answer in there.

Cheers

R

 

[PRR]

> The main idea behind the PGA23xx family was to use it after the DAC in the system as a volume control

Of course. And it is an excellent part for that. In consumer audio, I might design-in this part without testing. And I have no doubt it could replace the faders and sends on a mix-desk. Natch in that market there would be a lot of critical listening before committing, but also the budget allows buffers and even work-arounds.

> Some gain is possible on the device, although it's not it's main strength.

Don't undersell the part. The usual volume pot is loss-only, which means in a system you usually follow it with gain. And darn-near-always about 20dB of gain. Sometimes more would be nice to let the previous stage work easier, but you can't accept a high fixed gain because of noise. Here you have 30dB gain available, but you don't have to use it, and (I think) if you don't use it you have little or no noise penalty. And in most (not all) volume-pot / line-amp applications, a low source impedance is available so the THD rise is not an issue.

> ...mic input on a desk etc.... look at the PGA2500 digitally controlled Mic Pre. This little honey does 0db then 10dB through to 65dB in 1dB steps.

VERY INTERESTING part. Cute to have the four extra output bits too. Don't tell secrets, but the wish-list for this part didn't originate with semiconductor dudes, did it?

It does live in our Brave New World of +/-5V supplies.... I'm an old +/-18V man and cut my teeth (and burned my fingers) on +250V designs. But then I'm used to sending audio across the room or across town: these days you just move it an inch into a ADC and after that there's no worries. That certainly is the MainStream Wave; but some of the DIY crowd is very retro.

> ... lots of relays etc

Relays good. I like clack. Rd(on) is milli-ohms, not dozen/hundred-ohms. And you can debug them with a toothpick instead of a logic-scope.

But it do take a lot of relays to implement a log-law gain switch. And the queen of relays, the Bell Telephone System, rushed away from relays as soon as they could make transistors that didn't suck: relay reliability is that bad.

> post to the app engineer for the PGA2310. I don't promise answers, I don't promise a timely response

Feel free, but no need for response. My total orders would not cover his reading time, and even if a bunch of folks here bought these chips the sales-profit would not justify his response writing time.

The hi-end audio market is too small for chip makers: it used to be that every design was different, each maker wanted "his own" design, and fab-setup costs were greater than hiring a few low-pay workers to stuff boards.

There are exceptions: the long-running 5534/5532 seems to be an audio design that probably got fabbed as much for lo-noise non-audio markets as for the few audio systems that used it at first. Over the decades its audio sales must have exceeded its fab-setup, except you can't justify 1982 up-front investment on 1999 sales because of time and interest costs (and there was no way to know that so many 5532s would ever be sold to audio). A whole lot of TL072 have been sold into audio, but probably many-many more to non-audio. (As much as audio-philes knock 072 designs, it is still a brilliant piece of work, both for performance and for die-area/cost.) Not counting such low-end products as 5-Watt speaker amps, car-sound EQ, and CD-type ADCs, there aren't many "audio" chips, certainly not like other-market parts (logic, power conversion, multiplexers and ADC/DACs).

 

[Rochey]

TI's group doing the Pro Audio stuff really had their head screwed on. (I don't normally give friends that much credit... but in this case, I think it's worth it).

They must've spent 6 months talking to customers about this part to see what people wanted from the spec. Those 4 GPIO pins are for switching things like Phantom Power, Phase Switch, Pad etc etc etc.

-- Definately something that your run of the mill opamp designer wouldn't think of

You'll see the PGA2500 being used in a lot of high end systems being released very soon. I can't mention any names, but next time a digital desk or workstation is released, see if you can have a look inside. It's almost worth having a PGA2500inside type logo

I was told a story by one of our guys in the Audio products group about the time they met Mr Neve himself (he now lives in Texas apparently). They were telling him about the PGA2500, when he turned to them and said something like... "in my original desks we had over 144 dB Signal to Noise... how can you even compete with that?"

To which TI's reply was "What voltage rails were you working between?"

+ and - 18V.... :grin:

Every time remember that story, it brings a smile to my face. All I can say is that those poor opamps must have been cooking

As channel density grows, as well as the desire for completely silent audio workstations, I think lower voltage, highter performance integrated components are a must.
Many believe that this comes at the price of performance, but I would ask that those skeptics give such solutions a chance, a decent field trial.
I know of quite few people that have put such devices up against their own discrete solutions that they've had in the field for years, and are now making the switch over.

It's food for thought - I know most of you guys still believe that valves sound better, with a board full of relays etc. I don't want to deny you that, but I also think that there's so much potential with new semiconductors on the market, that it'd be a shame to discredit them because of a bad experience or two when integrated audio IC's were in their infancy.

Cheers

R

 

[mcs]

I just received this photo from Bauman:

I don't know how he made it work without installing resistors... :wink:

I changed the control board a bit (adding the encoder and EEPROM) as you can see above, but kept the size (almost).

Best regards,

Mikkel C. Simonsen

 

[Bauman]

Oh Mikkel!!



Shots after unpacking, very excited at that point... the camera works with or without the resistors... :razz:
Now gotta find that piece of paper with the position numbers to api312 application.

It really works great.


cheers!
Fabio

 

[rafafredd]

UAU!!!

It really looks fantastic! How much in parts did it csot, Fabio?

 

[Bauman]

Dunno how much, Mikkel sent it all set!

:thumb:

 

[kruz]

..cool..
but i don't like the control pcb.. IMHO it's not that great if you really have to use it in the studio..
a better "user friendly" design is needed i think

 

[PRR]

> Mr Neve himself ... said something like... "in my original desks we had over 144 dB Signal to Noise...

Sure. 0.2 microVolt input noise, 3V overload level.

> ...how can you even compete with that?"

A pair of any large-junction transistor, even 10/$1 2N4401s, working at several mA, with 10 or 15 volts supply.

Although, while you can get 0.2 uV noise at high gain, and 3V overload at low gain, with most transformerless designs you can't get both at the same time. If you make the feedback resistor small enough to not add noise (well under 100 ohms), then at low gain the feedback network is a heavy load. Transformerless mike amp noise tends to rise at low gain; how much is a design/cost issue. The real classics like early Neve and Jensen shift the issue with step-up transformer (now the feedback resistance does not have to be so insanely low) and huge output drive (the Jensen can drive 75 ohms).

> put such devices up against their own discrete solutions that they've had in the field for years, and are now making the switch over.

Sure. There's nothing "magic" in a mike-amp. Use clean silicon (now a given), make it big, run it rich. There have been several fairly-good chips around. But they were just a little short of the best, and often came from small shops that did not have TI's longevity. You don't re-design a system around a chip from a maker who may not be in business next month, or one using a rent-a-fab where the silicon may be "good enough for digital".

> "What voltage rails were you working between?" + and - 18V....

Early Neve tended to be a single +24V rail. Class A output with transformer step-up. The Jensen not only drives 75 ohms, but will run on +/-24V rails. +/-24V was really quite common before chip-makers stuck their oar in. One clue that the 5534 had an audio market is that it is one of the very few "normal 40V process" chips rated for more than +/-18V ab-max (5534 is rated +/-22V if I recall). Since the "40V process" really gives a spread from 30V to 60V, the 5534 doesn't get special processing (that would be expensive) but its spec gives lower yield than a +/-18V-rated part because you have to discard the occasional weakly-doped wafer.

To meet the traditional Pro Audio Line Level specs (16dB above +4dBm to 20dB above +8dBm) without an output transformer, you may need over 50 volts total supply. Hence +/-24V, or these days more often a bridge run from +/-15V. What the 2500 has done is shed that now-archaic level-standard. It isn't meant to drive long telephone lines, nor interface directly with studio patchbays descended from the old-old traditional line levels. These days we have a large need for preamps that butt right against an ADC. The digital side likes 5V. Maybe +/-5V or even push-pull +/5V, but we no longer have to over-power stray line noise, mixer loss, and other analog crap.

> those poor opamps must have been cooking

TL072 runs cool as a stone at +/-18V. (But don't ask a 072 for low noise voltage.) 5534 is, what? 4 or 5 mA idle current? 5mA at +/-20V is 200 milliWatts, well inside the 300mW rating of a DIP-8. I have used some hot-running chips. But mostly they run quite cool, because the op-amp industry has a historical obsession (and continuing market edge) with low supply power. This is where boutique audio tends to diverge from chip maker traditions: to hell with power consumption, we like to run things rich so they loaf under signal loads.

Speaking of cooking: I was startled to see the PGA2500 eats something like 30mA, far more than any general purpose op-amp and more than many discrete mike amps. The logic isn't eating much power. The input stage -should- run rich, but several-mA is enough for most good big devices to be very low noise from 150 ohms. Sounds like the output current is enough to keep it Class A for most signals, with enough drive to spare to use some very low resistance feedback resistances.

 

[NewYorkDave]

Ah, the 2N4401... my favorite, ballsy little jellybean TO-92 NPN transistor. I buy 'em by the hundred and use 'em all the time.

Opamp Labs still standardizes on +/-24V, but I realize that they're an anomaly.

 

[mcs]

[quote author="kruz"]but i don't like the control pcb.. IMHO it's not that great if you really have to use it in the studio..
a better "user friendly" design is needed i think[/quote]
That one was just made in a hurry so Fabio had something to test - it wasn't meant to be great...

The new one I made (you can see the drawing somewhere here) can be installed behind a panel. You just have to make holes for the displays, buttons and/or the encoder. What else do you need for studio use?

Best regards,

Mikkel C. Simonsen

 

[mcs]

[quote author="rafafredd"]How much in parts did it csot, Fabio?[/quote]
I don't generally worry about that. I just use the parts I can find in my drawers. And there are many drawers to choose from, as you can see here: http://stiftsbogtrykkeriet.dk/~mcs/Parts.jpg :grin:

But here is the approx. cost in Denmark - I have no idea about Brazil.

Relay board:
Parts DKK 11
Relays DKK 84
PCB (from Olimex) DKK 64

Control board:
Parts DKK 93
PCB (from Olimex again) DKK 30

The most expensive part is the encoder (DKK 28.50), so if you only use the up/down buttons you can save some money. The relays are also quite expensive, but you can often get them cheap from surplus dealers. You could also make a control board for more than one relay board at almost the same cost.

Best regards,

Mikkel C. Simonsen

 

[clintrubber]

from Rochey:

Again, I know I sound a bit too pushy on TI's parts, but all I can say in my defence is that it's really hard working with these parts day in day out, and not having great ideas of where they would fit in audio projects!

Don't feel bad - I try to push Philips' chips as much as possible as well :grin: :grin:

Just kidding - have a good day,

Peter

 

[Rochey]

Peter

- cheers mate... Don't all engineers in the Netherlands work for phillips at one time or another??? *kidding*

PRR-

I'm not going to even try to even answer your post -- it really is a work of art. I have no doubt that you have YEARS of analogue design behind you (and hopefully many more).

You've got the idea behind the PGA2500 in one sentence - right next to an ADC.

I've had a few of my own customers (they will remain nameless) ask about a +/-15V PGA2500. I think if these was really enough demand for such a device, the product group may consider it, but in my own opinion, as the market moves towards digital systems, the main volume of business is in the +/-5V market.

Oh, BTW - the comments you make regarding the PGA2500 is something I can't comment on... I simply don't have the knowledge. (I feel like such a disapointment :roll: ) -- Maybe another 20 years experience will do the trick. :green:

The PGA2500 is available in sample at the moment, straight through TI's website. (it only requires registration to my.ti) If any of you fancy doing some tests on the device with your own design, or a copy of our evaluation module, I would be really *really* interested to see your findings.

Maybe it's time to DIY with some more digital? (oh god... here come the flames!!!) :grin:

 

[mcs]

[quote author="Rochey"]Maybe it's time to DIY with some more digital? (oh god... here come the flames!!!) :grin:[/quote]
I'm building a harddisk recorder - is that digital enough?

I'm even using a TI ADC currently (PCM1800). The MCU I'm using is a Winbond though, sorry! :wink:

I'm also working on a clock generator for it using the PLL1708...

Best regards,

Mikkel C. Simonsen

 

[daArry]

and here too plannin an ad/da box with some PCM4104 and PCM1804's inside... :thumb:

 

[PRR]

> ask about a +/-15V PGA2500.

Ah, but why?

If they already have a +/-15V supply, duh, sell them some 5V regulators.

While there are +/-15V ADCs, I assume you have some very-fine 5V and +/-5V ADCs to sell them. Changing ADCs is trivial compared to replacing a manual-gain discrete preamp with the 2500.

As a stand-alone standard analog output preamp, it is "incomplete". You need at least some kind of brain-chip (or a rotary switch, a ROM, and a counter) to program (and display) the gain setting. Adding a 5532 for +/-15V +24dBm output is trivial.

Maybe they would rather not use a pad for inputs over 2V. That's not an uncommon level with today's hot mikes and drumsets. But few general-purpose mike preamps will take more than 3V input. And at those high levels, pads are not a real problem.

FWIW: in my concert-hall work, I now run only "semi-pro" levels. "-10dB" (2V full scale) unbalanced. With everything in one rack and semi-permanently patched, there is no need to run +8dBm 600 ohm balanced. The 2500 will do 2V with one output disconnected, with enough current to feed 10 or 15 recorders. If I never did "fade to zero" (and I rarely do), this could be a one-chip rig: preamp, limiter (via computer control), and distribution amp. (OK, 2 chips for stereo, plus the control logic.)

Another point. The 2500 eats 30mA. Scale that up to 30V power at the same current, and you have 900 milliWatts, a lot for that package. It will live-long, but the heat will raise noise enough to measure. And if the goal is higher level in 600 ohm loads, the output stage current probably has to be increased. Say half the total current is output stage, 15mA, and going from 5V to 15V suggests tripling that or 45mA, plus 15mA for the rest of the chip, is 60mA. Almost 2 Watts! It won't live long in that package, and the noise will be significantly higher. The low-volt is part of the low-noise feature. (Discrete is different because the hot output parts don't have to cook the input parts. Altho, the 990 and some other modules do bind the whole amp in a thermal block.)

> our evaluation module

That is a slick item, especially with the PC interface. But how did the price get set to $99? TI is not going to get rich selling $99 eval boards.... on the other hand, they won't go broke selling those boards for say $49. That, to me, is more of an impulse-buy, while $99 is a Purchase. At $49, TI may lose several dollars on each board, but if some impulse-buyer stirs up interest in the right places, it could sell a thousand 2500s plus all the trimmings (ADC, PIC, glue-logic chips...). And while mighty TI may not want to fool with DIY sales, outfits like Velleman might be happy to rent the PCB mask and stuff boards/bags and sell to anybody. Give them a nice price on 2500s and they could probably profit at $49.

 

[Rochey]

PRR - on the eval boards...
let me give you a hint (and anyone else here reading this)... the full schematics and bom for the PGA2500EVM are available in the EVM manual...
If we could bring the size of the board down (i think it's doable), we could get olimex to make a few boards for us. (just a thought)

In terms of $99... I think it's a fair price - The device itself is $10, plus the PCB and the parts, and paying an external company to assemble them.
Also, the build quality on them is pretty good -- all measurements that go in the datasheet are actually taken from the eval board (unlike other cases, where manufacturs have the datasheet spec, and then the actual spec on the eval board...grrrr )


Speaking of taking others works and developing them further
let me feed some ideas here (which have just popped into my head)...

TI has a TUSB3410, which is essentially an 8052 with a USB connector...
Now let me whisper PGA2500, PCM4204 (118db, 4 channel, 600mW ADC).
PCM4104 (118dB 4 channel DAC, 200mW)

Essentially, there's the potential to develop a complete I/O solution, controlled from a nice USB interface (if that floats your boat).
Interfacing into a PC is the next challenge... although that's a completely different thread! :grin:

I know very little about firewire and usb2 - and I have no desire to even look at PCI :!:



BTW -- I'd just like to say thanks to you all for making me feel welcome here... I was expecting to get flamed a LOT more. It's been a nice surpised to see you guys a lot more open to new products and ideas.

Cheers :sam:

 

[mcs]

[quote author="Rochey"]PCM4204 (118db, 4 channel, 600mW ADC).[/quote]
Are you talking about the two channel PCM1804, or is it a new secret product the TI website hasn't been told about?

Best regards,

Mikkel C. Simonsen

 

[Bauman]

Rafael,

Mikkel solution was about R$130 (US$40). Still more than a rotary switches
BUT it is the cost for 1 piece...

And if you're intended to have it remote controled or want to store the presets, this is very useful


cheers!
Fabio