5.1 preamp with pt2323/2322 processors

[Nadis]

built pt2323/2322 processors as a preamp for my 350w amplifiers
( Ampcontrol
first schematic used second pt2322 insted of tda Electro help: 12/01/13 )

. As they(350w amps) need quite high input voltage, processors by them self are to weak to drive them so i aded tl071 per channel with 2x or so amplification.
i`m having this problem of white noise coming from pt processors which is amplified quite to much thrue tl071...
Have anyone played with theese processors and can asist me with info on are they supposed to have such white noise even with no input or volume set to -79db. its like pt2322 has op amp in its back end with high amplification.
Maby these processors are not meant to be in "hifi" systems. or maby i should put op amps in front of pt because they can take close to 4v input.
maby i`m missing something?

btw i`m 100% sure that this noise originates in pt processors because connecting amp directly to usb dac noise is like 10- 20 times smaller even at max volume i can barely hear it with head close to speakers. noise is not digital pt and atmega has different power supply's. osciloscope doesnt show anything while using atmegas functions.


as i understand white noise is noise that you hear when amp is set to max volume and nothing is playing. soft sssssssssss sound

 

[sgrossklass]

These PT2322s don't even have proper specs in the datasheet. No max input level, max output level, output noise, no gain (if any). You're pretty much out on your own, having to find out all of these things by yourself.

Your hunch may well be correct - suggested applications certainly don't imply much beyond personal audio, and there is only so much dynamic range you can expect from a PGA running on +9V. There's a reason why serious Hi-Fi PGAs (e.g. PGA2320) operate on up to +/-15 V.

At least I was able to turn up a block diagram here; they also state the chip will operate from +5 to +15 V.

IMO, this chip features less than ideal gain staging. Looks like they start off with a digital volume pot, only to follow it with a tone control section (which have a tendency of being a bit on the noisy side, but at least this one can be bypassed), 3D effect and finally an "output volume trim" which may include some gain. I would want to determine experimentally what sort of gain that would be.

How much output level do you need? I would imagine that you should pretty much always get 1 Vrms out of a chip like this. Mind you, that's usually fine for driving 100 Wpc amps (@Gv = 29.5 dB), but a 350 Wpc job may be pushing it indeed. Do you really need that much power anyway?

What sort of source levels are you going to run into this thing? Your idea of including some gain ahead of the PT2322 sounds solid enough (I might use something with better common-mode distortion properties than a TL07x though, but if you insist on that one, go no lower than 10k/10k for feedback). I would also aim to get the chip's supply closer to +15 V since it can apparently take that. Maybe with some extra passive RC filtering just in case (see whether you can get away with 220 ohms - 220 µF).

You might also consider adding different input gain per input depending on which source is supposed to be connected there. You typical Android phone is likely to be far quieter than a typical CD or DVD player.

The sad part is that quality wise, you may be better off running a decent DAC directly into the power amp, assuming it's sufficiently free from power-on/off pop noise (and other potentially speaker-frying gremlins, as well as low-level idle tones). Something that'll deliver upwards of (real) 110 dB of dynamic range @ 2 Vrms isn't too exotic these days. Not really a solution for multiple sources though.

 

[Mark Tillotson]

I found a datasheet for PT2323 saying max input level 3.75Vrms (despite saying power supply of 9V, but I guess that's a suggestion), and 4µVrms output noise (ie 28nV/√Hz, so pretty bad).


If you use an opamp in front of the processor, maybe use an NE5532 with low impedance feedback network (1k), not a TL072 with 10k - you want low noise after all, injecting 18nV/√Hz into the thing's input isn't going to help.

 

[sgrossklass]

If you use an opamp in front of the processor, maybe use an NE5532 with low impedance feedback network (1k), not a TL072 with 10k - you want low noise after all, injecting 18nV/√Hz into the thing's input isn't going to help.

Nah. Last time I checked, volume controls tend to be turned down quite a bit (as in >20 dB) most of the time, and sources tend to have more than their share of noise of their own, easily swamping a TL072's contribution (which is just ~2.8 µV after all even with 10k/10k). Plus whatever the PT2323 in front puts out.

 

[Mark Tillotson]

The OP states the noise is in the PT2323, not the source, and the gain is maxed out on the PT2323. Clearly the gain structure needs attention, and the figures found, if reliable, mean 10k in an opamp stage will increase noise measurably - 350W amp needs very quiet inputs to avoid speaker hiss (in a PA application you don't care, for home use you may). The 28nV figure on the input to the power amp will be boosted by 40-fold or so to give maybe 0.15mV rms at the speaker terminals (20kHz bandwidth). A good power amp might have an input noise of 5nV/√Hz, so 28nV represents a 15dB increase.


A good 50W power amp with 5nV/√Hz input noise might have 15µVrms at the speaker terminals, 20dB less than this setup.

 

[sgrossklass]

A good 50W power amp with 5nV/√Hz input noise might have 15µVrms at the speaker terminals, 20dB less than this setup.

You rarely see less than about 30 µV in practical amps though - it'll generally be swamped by preamp noise anyway, unless you have a very fancy preamp with a two-stage volume control. Good integrated amps tend to be at around the 100 µV mark. Plenty good enough for normal-sensitivity speakers up to 90 dB / 2.83 V / m.

I didn't think that the OP implied that he was running with volume maxed all the time, he's just not getting enough maximum level for full power output, which might be ear-splittingly loud for all we know.

Note that PT2323 is an input selector and PT2322 is a PGA / tone control.

 

[Nadis]

First off all thanks for feedback. Been trying to solve this for few weeks. As sgrossklass say its basically poor choice of processor. i was afraid of that. but looking at datasheets of all ics that i can use shows that most of them( TDA7439 TDA7312 TDA7313/PT2313 TDA7314 TDA7315 TDA7318 PT2314 TDA7448 PT2323/PT2322 PGA2310/11) has output noise around 10uV
+-5 uV. pt2322 has 14 and with tone defeat it has 7 even less than pga2310 ( 9.5)
am i looking at the right place in a datasheets?

Btw trying to drive with this preamp 350w amp at 4 ohm but i only have 8ohm spec : 1.7v 200w 8 ohm so basycaly half power at 1v input .
my speakers are 180-220w rms driving them without tls at 0db on pts sound is not nice to listen to ( probably distorted a lot)

P.s. i`m runing pt2322 at lowest volume possible even without input that noise is always. pt2323 has way lower noise close to having them removed in from signal path (dac to amps).

 

[sgrossklass]

First off all thanks for feedback. Been trying to solve this for few weeks. As sgrossklass say its basically poor choice of processor. i was afraid of that. but looking at datasheets of all ics that i can use shows that most of them( TDA7439 TDA7312 TDA7313/PT2313 TDA7314 TDA7315 TDA7318 PT2314 TDA7448 PT2323/PT2322 PGA2310/11) has output noise around 10uV
+-5 uV. pt2322 has 14 and with tone defeat it has 7 even less than pga2310 ( 9.5)
am i looking at the right place in a datasheets?

Yes, but you've got to look at it in context, i.e. relative to signal levels. 7 µV isn't too bad - if the processor already includes the customary 14-16.5 dB of maximum preamp gain. If it does not and you've got to bring levels up, you've got a problem. There's no way you can accommodate even 14 dB (a gain of 5) ahead of the PT2322 while still being able to accept 2 Vrms input. Granted, you could include a gain of 2 stage ahead of the chip and add some more gain at a dedicated set of explicitly "high-sensitivity" inputs for mobile devices and such.

The PGA2310/20 for example includes a maximum gain of +31.5 dB and supports substantial output levels (5 Vrms easily, >9 Vrms max on +/-15 V), so you could easily bring down its level by 10 dB and cut down total output noise to the vicinity of 4 µV or on par with a good conventional preamp, with plenty of gain still available and maximum output level of >3 Vrms remaining. You could even push it to -15 dB max if you really wanted (e.g. with a resistor divider of 12k/2k7 and an NE5532 or similar buffer), even that would still give you ~1.7 Vrms while shifting output noise levels closer to a world-class 2 µV (or a power amp output noise contribution of <50 µV, in which case the power amp may actually become the limiting factor). It's all about shifting dynamic range around as needed.

Another option would be using two stages of PT2322s (easily done e.g. when you're only using two channels anyway and thus have some to spare), with a gain stage of ca. +16.5 dB in between. This is pretty much the equivalent of what high-class preamps used to do 30-40 years ago. It makes control a bit more difficult as you now have to juggle levels in a sensible manner, but that ultimately is a pure software issue. I would suggest keeping both stages' volume roughly in sync (but with coarser steps in the first so you don't lose resolution) until they are down to -30 dB or so, and then keep the first fixed there while further decreasing level in the second. Volume control range still is 109 dB like that... that should generally do.
For example, consider a -40 dB (total) setting: First PT2322 is at -20 dB, will accept everything up to 9 Vpp input, while a 2 Vrms signal gives a 200 mVrms output signal. The +16.5 dB stage then gives out 1.3 Vrms or so, which again is easily handled by PT2322 #2, which outputs 130 mVrms into the power amp. Total output noise, assuming both PTs are still at 7 µV, would be roughly 8.3 µV. A nominal 83.8 dB SNR at just a hair (0.6 dB) over 1 W / 8 ohm level sounds quite acceptable, it would be at least on par with an average-performing 100 Wpc integrated amp.
If you find PT2322 #2 noise increases too soon and needs to be kept down further, try varying relative attenuation ratios e.g. to 1:2. 15 + 30 dB for 45 dB puts some more strain on the intermediate gain stage but it should still be manageable, plus you can still reduce its gain to e.g. 10 dB - still more than you had before.

BTW, you guys are clearly better at finding comprehensive datasheets for these ICs than me, so pray tell, where did you find one with an output noise spec included?

Btw trying to drive with this preamp 350w amp at 4 ohm but i only have 8ohm spec : 1.7v 200w 8 ohm so basycaly half power at 1v input .

This comes down to an effective voltage gain of +27.4 dB. 350 W / 4 ohm output would require ~1.6 Vrms in.

my speakers are 180-220w rms driving them without tls at 0db on pts sound is not nice to listen to ( probably distorted a lot)

Not a major surprise right there. Speakers tend to be mechanically limited long before maximum power handling is reached.

Let's say we're shooting for a more realistic 100 W into 4 ohms, that would be 0.85 Vrms in.

 

[stratus46]

First off all thanks for feedback. Been trying to solve this for few weeks. As sgrossklass say its basically poor choice of processor. <snip>

That's why I used 3 NJW1119A tone control chips and 3 PGA2311 volume control chips. The PT2323
looks to be similar to the Mitsubishi M62446 which I used for a while until the new 'preamp' was done.

G²