Well, my config was a Raspberry PI -> I2S DAC -> modded 3 6N3 russian tubes preamp -> LJM L15D 2092
Maybe need an improvement to avoid fatigue? I never had this issue with TPA3116
I have no experience of tubed preamps but from what I understand of tubes they don't like low impedance loads. Is there a trafo between the tubes and the output?
OK thank you. I was taught to use TX as the shorthand for transformer although with Bluetooth part of the audio armoury, I suppose it could be confused with Transmission..
@camelator
If you look at the Iraudamp7 reference design which most (if not all) of these boards are a flavor of, the input resistor is only 3k, which is too low for a tube preamp. Reference design can be seen over here:
http://www.irf.com/technical-info/refdesigns/iraudamp7d.pdf page 30.
You need a beefy cathode follower to dive this amp properly. But it gets worse! Take a look at AN-1138 over here:
http://www.irf.com/technical-info/appnotes/an-1138.pdf page 3 figure 2.
Here you can see how the feedback loop is implemented. The output square wave is fed back to the summing node of the inverting input OTA through Rfb, and Rin is also connected to the summing node. This means that a square wave current of about 0.5 to 1 mA P-P is injected into this node (!), which will partially (about 10%) flow through Rin. This scheme only works well when the source that drives the amp has a low output impedance, up to and including the switching frequency and some harmonics. In practice, this amp needs a stiff driving source that keeps its output resistance low at frequencies up to a couple of MHz. So even if you use an output transformer from your tube buffer, the leakage inductance of the output transformer will still cause the driving impedance to be too high at the switching frequency.
IMHO that means this amp requires a good (fast) opamp buffer or one of these fancy diamond buffer IC's, preferably on board. As the IRAUDAMP7 is meant as an evaluation platform for the IC, and never as a finished product, all the boards based on it will suffer from this, and will only realize their full potential when driven from a good buffer that is situated close to the amp board. Also, do not connect a volume pot to the input of the board!
If you look at the Iraudamp7 reference design which most (if not all) of these boards are a flavor of, the input resistor is only 3k, which is too low for a tube preamp. Reference design can be seen over here:
http://www.irf.com/technical-info/refdesigns/iraudamp7d.pdf page 30.
You need a beefy cathode follower to dive this amp properly. But it gets worse! Take a look at AN-1138 over here:
http://www.irf.com/technical-info/appnotes/an-1138.pdf page 3 figure 2.
Here you can see how the feedback loop is implemented. The output square wave is fed back to the summing node of the inverting input OTA through Rfb, and Rin is also connected to the summing node. This means that a square wave current of about 0.5 to 1 mA P-P is injected into this node (!), which will partially (about 10%) flow through Rin. This scheme only works well when the source that drives the amp has a low output impedance, up to and including the switching frequency and some harmonics. In practice, this amp needs a stiff driving source that keeps its output resistance low at frequencies up to a couple of MHz. So even if you use an output transformer from your tube buffer, the leakage inductance of the output transformer will still cause the driving impedance to be too high at the switching frequency.
IMHO that means this amp requires a good (fast) opamp buffer or one of these fancy diamond buffer IC's, preferably on board. As the IRAUDAMP7 is meant as an evaluation platform for the IC, and never as a finished product, all the boards based on it will suffer from this, and will only realize their full potential when driven from a good buffer that is situated close to the amp board. Also, do not connect a volume pot to the input of the board!
I have been fortunate to have been driving this from a mixer preamp. When I used a miniDSP to drive it, it seemed to work OK too. That is interesting that the feedback contains the pre LC filtered square wave. Are you sure it is not feedback after the LC filter?
The schematic shows the feedback comes direct from the MOSFETs - i.e. prior to the output filtering. On the IRAUDAMP5 there's some filtering (an RC low pass network) prior to the feedback take-off point but that's absent on the AMP7.
If you want post filter feedback and stable lowww 2 ohm and buffer gainboard and don't mind soldering, Liteamp pcb's will be re-run, ~+/-40V psu required if you use identical parts as in thread here in classD. Gain for main ampboard is around 5, so buffer is also for gain, noise is 1/5th of other 2092's, distortion lower too and speakerindependant frequencyrespons.
The Hypex UCD and nCore modules have post-filter feedback, but they use a different oscillator scheme. I have not yet looked if the IRS can be used in a post-filter scheme, but this is probably such a change of circuit that it is not worth bothering about with the cheap boards that are being discussed in this thread. Just drive them with a low Z source and you'll be fine.
@XRK: the minidsp probably uses the AD recommended passive output filter which has, as the last component, a 5.6 nF capacitor to ground. This is just fine for the IRS. The pre-LC feedback is vital to the functioning of its self oscillating circuit.
@XRK: the minidsp probably uses the AD recommended passive output filter which has, as the last component, a 5.6 nF capacitor to ground. This is just fine for the IRS. The pre-LC feedback is vital to the functioning of its self oscillating circuit.
Ahh, that is the topology - uses self-oscillation via feedback. Ok, thanks for explaining that. What is the approximate self osc frequency, and wouldn't that depend on all sorts of things like the input resistors, the output LC values, etc? Despite feeding the squarewave back, eventually it all gets filtered out and sounds amazingly clean to me. 🙂
The datasheet and application notes describe a frequency of around 400 kHz, but I thought that the Sure electronics board uses less than that. I haven't checked. I have one, but I have not yet fired it up. AN1138 page 3 describe the components that determine self oscillating frequency.
I suspect that IR were not thinking about Valve/Tube preamps when they designed the 2092.
Whatever your view on valves/tubes, they are obsolete to all intents and purposes.
Whatever your view on valves/tubes, they are obsolete to all intents and purposes.
This was moved over from the TDA7492 thread since it was off topic.
********
A little OT, but wanted to let you guys see what just came in the mail from Aliexpress...
These things are pretty compact - a little smaller than YJ blue/black 3116.
Very nicely made, appear to be quality components and very clean soldering work - IRS2092 200w monoblocks for $19ea shipping included. I am using qnty 4 x 19v laptop smps bricks in series for +/-38v supply (4.6amp limited) as the power supply. Testes it with some PA drivers and it is loud and clean. A great deal - easy to hook up and use - just need dual rail supply is the only downside. I may have to buy a Connexelectronic dual rail SMPS for $50 to get the real juice.
Recent price drop now $16 😀
IRS2092 200W Class D mono amplifier board-in Amplifier from Consumer Electronics on Aliexpress.com | Alibaba Group
Update (June 19, 2015): I testes this with the $20 Abletec dual rail 53v power supply and it works very well. Clean and loud.
TPA3116D2 Boards - diyAudio
Test setup:
Those RCA jacks look to be a very easy mount on a flat board. Can you tell me where to get them?
Thanks You
These are great binding posts - built like a tank and accept banana jacks or bare wire, internally can be soldered or crimped quick disconnects. Right price too. Gold plated ones also available for a little more.
Parts Express Banana 5-Way Speaker Wire Binding Post Terminal
Hi timpert,
Thanks for your comments,
Actually, my knowledge on audio electronic is really poor.
Can you please explain the impacts ?
I don't really understand what is the link from 3K input resistor to fatigue
Thank you very much!
Triodes work best when driving a high impedance load. In this case, distortion will be low and the sound will be pleasant. If the load impedance is lowered, gain drops and distortion increases. In order to achieve a given voltage across a low impedance load, the triode thus needs to be driven harder, further increasing the distortion. While this characteristic distortion works well in a guitar amp, it is best avoided in audio for the reason you experienced. In the "tubes" subforum, this topic is being discussed regularly.
I am not talking about distorsion but about Fatigue.
I have friends with pure tube amp, causing more distorsions than my IRS2092 and it is listenable without fatigue.
my IRS2092 has less distorsion that my TPA3116 (I measured it with REW)
but it causes me fatigue to listen
Fatigue is usually excessive 2k to 6k region. Did you ever do mic measurements of the audio frequency spectrum of the speakers? A 2-3dB bump at 3k or 5k can cause fatigue. Or if you have good ears like a dog and an amp has excessive output in ultrasonic range and you have tweeters that go beyond 20k, can cause fatigue. Some people think that a flat response causes fatigue and prefer a falling response of about 3 to 5dB from 100Hz to 20kHz. I wonder if your speakers just need a BSC like 0.8mH and 3ohms in parallel to take the bite off the mids and highs. If you hear it, it would be measurable in a spectrum.