LF Line Level Stage design to preceed class D amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi there,

I'm looking for a high performance line level buffer circuit that I can implement on a 2 layer board with surface mount components.

USE CASE: Preamplifier/buffer stage for TPA3116D2 class D amplifier.

- Will be fed with approx +/-10v regulated DC.
- Volume potentiometer required
- No balance control!

They will be built into the same chassis. With this in mind, I want to optimise the performance of the unit as a whole by matching the input impedance / gain settings of the preamp and the TPA3116D2 amplifier.

I understand that typically, lowering the gain of the TPA chip yields better sound quality (comments?), assuming the input is being driven by a capable device (the preamp). The lowest gain setting on the TPA chip is 20dB. Highest is 36dB. So to achieve the best sound quality, I thought the high quality preamp circuit should add at least 6dB of gain on its own.

From the TPA data sheet:
GAIN / INPUT IMPEDANCE / INPUT CAPACITANCE / HIGH-PASS FILTER
20 dB / 60 kΩ / 1.5 µF / 1.8 Hz
26 dB / 30 kΩ / 3.3 µF / 1.6 Hz
32 dB / 15 kΩ / 5.6 µF / 2.3 Hz
36 dB / 9 kΩ / 10 µF / 1.8 Hz

GOAL: Present an easy to drive load for any connected source, and drive the TPA in an optimum configuration for max performance.

Right now I'm thinking an opamp design would be better in this case than discrete, but I'm a newbie... Something like Rod Elliot's P88 minus the balance control.

Or... maybe something class A :p

1. General comments?
2. How does the preamp's (output?) impedance fit into the 'input impedance' figures quoted in the datasheet excerpt above?
3. Suggested designs/schematics?

Any help/suggestions would be appreciated!

Thanks,
Chris
 
Jfetboz? As long as you can get some 2sk170...although you need 16v but you could match voltage with the TPA...jfetboz likes 16-19v. I think diyaudio store has replacement to the 2sk170 if you can't find them.

Gain might be adjusted to be lower than "stock" circuit

Another nice one is the b1 buffer but it has zero gain, only buffer. 20db for your whole system might be light without super sensitive speakers.

I have built both and still use the jfetboz regularly. Circuit is super simple if you are already supplying clean voltage.
 
Thanks for the suggestion DUG, that is exactly the type of clever thinking I'm looking for! I'm excited to receive your PBTL TPA boards. I need the payment details though.

Could you give any suggestions on single-ended -> differential circuits I could use for the preamp?

I want to design a PCB for the circuit utilising surface mount components. High performance is important. I understand this will take lots of time and many design tweaks/revisions...

cjkpkg, thanks for the suggestion too! I have 24v single supply DC coming into the case. I was planning on using a LM317 + LM337 regulator combo to clean the power up a bit and get a split +/-10vdc rail, as most preamp circuits seem to want a split supply.

So, accounting for regulator losses, I can either do 20v single rail, or +/-10v split rail.

Do you know of any suitable circuits given the power supply constraints?

Thanks again,
Chris
 
...
I'm excited to receive your PBTL TPA boards. I need the payment details though.

you are on the list. :)

...

Could you give any suggestions on single-ended -> differential circuits I could use for the preamp?

...

Do you know of any suitable circuits given the power supply constraints?

Thanks again,
Chris

There are many ways/choices. All could be made to work.

Discrete transistors/FET's
Audio operational amplifiers
"Industrial" operational amplifiers
With/without transformers

You can either pick a method and experiment with it or build one of each and compare.

Or research each and as you gain knowledge you can decide which one you want to try.

If you want my opinion of what to try then I would chose the road less traveled:

http://datasheets.maximintegrated.com/en/ds/MAX4447-MAX4449.pdf

:)
 
Jfetboz will work at 20 on a single rail. Previous tests showed 16 had lowest distortion though.

86899d1181083159-jfet-boz-n-pass_2sk170-boz_1.png


That's all you need...a couple jfets a couple resistors and a couple 10uF decent film caps.
 
There are many ways/choices. All could be made to work.

Discrete transistors/FET's
Audio operational amplifiers
"Industrial" operational amplifiers
With/without transformers

If you want my opinion of what to try then I would chose the road less traveled:

http://datasheets.maximintegrated.com/en/ds/MAX4447-MAX4449.pdf
:)

Thanks again DUG.

As this is meant to be a fun and educational exercise, I'm going to think aloud (via my keyboard).

I'm not really interested in pursuing the transformer approach, as it seems comparatively high cost, and I get no signal gain (which I need).

The MAX444x series looking interesting. Would the DRV134 be an alternative in function?

Which chip would you select? I think the 4447 has a fixed gain of 2, while the others have selectable gain via a resistor to ground. The 4449 operates at a higher bandwidth, so that would be my super-newbie choice given that I need a gain higher than 2.

I note what seems to be a very high slew rate of 6500V/uS (compared to other datasheets). I have no idea how the slew rate affects the circuit and its performance though.

I also note that the MAX4449 chip runs on +/-5v. I was hoping to take the single 24v DC supply coming into the chassis for the TPA3116 boards, and regulate it into a +/-5v split supply in parallel with the amp boards. Is this okay?

The MAX4449 chip has a typical PSRR of 75dB, which doesn't seem great compared to some other chips. This means the circuit that splits the supply should have a very clean output.

One other question: How would the volume control be implemented?

-------

Then there's the JFET BOZ... I think I'm eager to try this as well, at the very least to have a point of comparison with the MAX444x. Being discrete, does the JFET BOZ run in class A? The distortion profile looks impressive: http://www.diyaudio.com/forums/pass-labs/103050-jfet-boz-130.html#post3984462

I suppose this is a single ended solution and wouldn't offer the advantage of breaking a ground loop between the amp boards?

Is this a circuit that would work well using SMD components?

How would the volume control be implemented with the JFET BOZ?

-------

Any other options?

What about LME49724 | Audio Operational Amplifier | Audio | Description & parametrics ?

The specs look amazing, but it looks like its differential to differential... Does that mean it can't accept a single ended source?

-------

Thanks again guys!

Chris :)
 
Transistor amplifiers are not for the faint of heart. (to design from scratch)

The MAX444x series is for video so all devices' BW exceeds your requirements.
This volume control implementation may work but the device may limit the pot resistance value to something lower than you wanted.

The JFET BOZ looks to have a great amount of gain. Just looked at beginning of thread...gain of 50.
Do you need that much?
You could tone that down some...read the rest of the thread for the whole story.
Volume control easy.

The LME49724 does look like an amazing device.
You can drive it with a single ended signal.
However.
It is recommended to have the power pad soldered down to something:
"The LME49724 should always be soldered down to a copper pad on the PCB for both optimum thermal performance as well as mechanical stability.
The exposed pad is for heat transfer and the thermal plane should either be electrically isolated or connected to the same potential as the VEE pin."
To benefit from the specs you will also end up with low input impedance (~1k). This sort of defeats the purpose of a buffer to drive the 9K+ input of the 3116.

Happy learning.

:)
 
Transistor amplifiers are not for the faint of heart. (to design from scratch)

The MAX444x series is for video so all devices' BW exceeds your requirements.
This volume control implementation may work but the device may limit the pot resistance value to something lower than you wanted.

The JFET BOZ looks to have a great amount of gain. Just looked at beginning of thread...gain of 50.
Do you need that much?
You could tone that down some...read the rest of the thread for the whole story.
Volume control easy.

The LME49724 does look like an amazing device.
You can drive it with a single ended signal.
However.
It is recommended to have the power pad soldered down to something:
"The LME49724 should always be soldered down to a copper pad on the PCB for both optimum thermal performance as well as mechanical stability.
The exposed pad is for heat transfer and the thermal plane should either be electrically isolated or connected to the same potential as the VEE pin."
To benefit from the specs you will also end up with low input impedance (~1k). This sort of defeats the purpose of a buffer to drive the 9K+ input of the 3116.

Happy learning.

:)

Hmm, perhaps I should first attempt a proven circuit before trying something like the MAX444x.

50 gain is way too much. I'm looking for <= 10.

1. Do you think the JFET BOZ will be worthwhile trying with SMD parts at 10 gain (if possible)?

2. For the LME49724, could I not just have a top layer pad under the IC (avoiding the pins), and extending out of either side of the chip with some open surface area to dissipate heat, connected to VEE potential?

3. So, the ~1k input impedance of the LME49724 is significantly more difficult for the source to drive than the ~9k input impedance of the 3116?

That does sort of defeat the purpose! The buffer needs to be exceptionally easy for the source to drive, but also drive the TPA to clipping with minimal distortion.

The requirements list for this circuit is beginning form:

- Powered by max single rail 20vdc, or max split rail +/-10vdc.
- Accept single ended input
- Differential output
- Easy to drive (high input impedance?)
- High current output ability (class A operation bonus)
- Needs to be, for the most part, SMD compatible
- Needs to be a stable circuit, well documented

At this point, I'm thinking discrete circuits and audio op-amp based circuits are the answer.

4. What about using the ever popular LM4562? I have no idea if/how that's possible...

Thanks for all the help thus far.

Chris
 
I'm in the process of building this little circuit to use with my TPA3116. Only change I did was to use 715 ohm source resistors.

I hooked up one channel to my EDU01 kit oscilloscope, and here are the results so far.

I ran a 1kHz test tone through my Rasberry Pi/HifiBerry DAC, and it outputs about 2.1V:
attachment.php


After running that same 1kHz tone through the little pre, it's amplified to about 2.9V:
attachment.php


I also ran a spectrum analysis, and from what I can see there is some 2nd order harmonics. (The signal looks really distorted, but it looks about the same directly from the DAC.):
attachment.php


Finally a picture of my little buddy:
attachment.php


I won't have time to listen to it yet, too many other things to do atm.
 

Attachments

  • IMG_20140916_151012.jpg
    IMG_20140916_151012.jpg
    222.7 KB · Views: 264
  • IMG_20140916_151451.jpg
    IMG_20140916_151451.jpg
    240.6 KB · Views: 188
  • IMG_20140916_151412.jpg
    IMG_20140916_151412.jpg
    230.6 KB · Views: 192
  • IMG_20140916_144618.jpg
    IMG_20140916_144618.jpg
    228.2 KB · Views: 182
Very cool skrodahl! Thanks for sharing. Interesting that this is also a JFET pre. I wonder how it compares to the JFET BOZ.

I would imagine that the Jfet BOZ would perform leaps and bounds over this one. In simulations (using Multisim with the J309 transistor) it has a THD of about 1%. This MPF102 circuit has way less gain, which is important for me.

I still haven't figured out how to fit the pre in my home made enclosure yet, as a RIAA stage is more important at the moment. It will still be some time before I can properly test it. (And I have not built the jBOZ, so I wouldn't be able to compare anyway.)
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.