Class T-amps

[stevekr]

I'm using a Virtue Audio One class T in power amp mode with a 130v supply, It's the best amp amp I've ever owned. There's a review on TNT and they have a forum on audio circle. Would highly recommend them. The Virtue amps can drive 'normal sensitivity speakers too. I love the Tripath sound, the warmth of valves coupled with the power of SS. The smaller T amps are great, sonic impact, trends ect. but you do need to have high sensitvity speakers to get the best from them I've found, Lowther owners take note!

[Krisbee]

I've long been intrigued by t-amps. I can't do DIY and came close to buying a trends audio ta 10.1 about 18 months ago for a desktop system (PC + Stan's DAC etc.). It was both the sound per pound and green factors that attracted me, together with its compact size.

But UK prices doubled overnight just before I took the plunge and I ended up press ganging my Rotel RA-01 into use instead. I always had doubts about a T2024 based amp driving my 86db speakers.

Saw a kingrex T20U and kingrex psu for sale s/hand for £225 a few weeks ago but someone beat me to it. Like Ali's 90w t-amp, the kingrex is based on the T2020 chip which kingrex rate at about 25W so it stood a better chance of driving my speakers.

Nearly all this stuff seems to originate in the far east, and if the t-amp can be a giant killer as described in reviews at TnT, 6moons and elsewhere, I wonder why they are not more widely available in the UK.

Templeaudio's Bantam amps seem to be about the UK produced T-amp on the market. Both their bantam and bantam gold are described as “hand built in Engand”. I've seen inside a bantam amp and it looks to be based on a cheap Sure T2024 board. But they could have devised their own tweaks and from the comments left on their website you'd think it was a minor miracle.

In the end I bought a s/hand Quad 306 two weeks ago. Compact yes, green no. I like the Quad sound, it's not as harsh as the Rotel. I might do the DADA electronics thing on it if I ever feel flush.

But if anyone wants to loan me/bring round a t-amp for comparison, let me know.

[Ali Tait]

Hi Chris,
The Hlly T amp 90 uses the TA2022 chip,not the 2020.More powerful in that it'l output 90w into 4 ohms,and about 40 or so into 8.Plenty for all speakers except those with silly loads.

[Ali Tait]

Hi Chris,
The Hlly T amp 90 uses the TA2022 chip,not the 2020.More powerful in that it'l output 90w into 4 ohms,and about 40 or so into 8.Plenty for all speakers except those with silly loads.

[dave2010]

Vincent

I'm not very good in electronics but what I understand is that T uses pulse width modulation and Class D amplitude modulation.
Class D is not digital, it uses a oscillator to amplify, in essence it is AM, (amplitude modulation) well known from AM radio. http://www.hypex.nl/docs/allamps%20hypex%20layout.pdf
Both are pretty frugal, Class D > 90%
This is probably the reason why D is implemented in all kind of portable devices.

Yes, maybe it is not the best possible Class D implementation but as a 'proof of concept' a very interesting experiment.
I drive my Apogee's with 2 Class D mono blocks, UCD 400 with HxR regulators: http://www.hexateq.com/

I'd say that broadly I'm in agreement with you, though I wouldn't call Class D amplitude modulation, though maybe I'm not looking at it the "right" way. The article you highlight is interesting, and it does point out different ways of looking at data and waveforms.

My understanding of Class D is that essentially they are switching amplifiers which work on a train of on-off pulses, and the way they operate means that effectively they should get close to 100% efficiency. When the switch is off, there is no current flow, so no power consumed. When the switch is on, the MOSFET design should have minimum impedance - close to zero - so again the power consumption is low. The switch can be triggered by 0-1 or 1-0 transitions in the input pulse stream.

You are right that PWM is analogue, because it's not (necessarily) directly linked to a clock, so the representation of the magnitude of the signal is effectively a representation in continous time. It could be implemented by a ramping circuit. A sawtooth or triangle wave oscillator could be used to generate the pulses to feed to the power output switch circuit. The frequency of the ramping oscillator is typically at least 10 times the highest frequency of amplification required. As the voltage level of the ramp rises, it is compared with the sampled input, and if the input is higher than the ramp there is a pulse - 1 output level. Once the ramp level becomes higher, the output is set to zero. For good results the ramp has to be well done too.

These pulses then drive the output power switch circuits, and the output is then low pass filtered. The effect of this is to give an efficient amplifier. In practical switching amplifiers I think the switching rate is pretty high compared with audio frequencies, and although this generates noise, it's mostly very high frequency, and filtered out relatively easily.

I really don't know what happens exactly with Class T, as there are several patented features of the chips. One obvious thing is that PWM is not going to work very well with low level signals, as most of the time the encoding is going to be 0, with only very short periods of 1 level outputs. This can be improved in several ways, which may be represented by the patented approaches in the chip. Some processing may take place in the analogue domain, such as modifying the gain before generating the 0-1 bit stream, with a corresponding modification to the output gain. That should be an option.

From the description of Class-T which can be gleaned, it seems that some of the processes may be similar to adaptive delta modulation. It is clear that methods such as first order sigma-delta operation also are capable of giving significant improvements - S-D is used in SACD players and in the DSD encoding method. S-D brings integrators into the circuitry, and the pulse train represents the mean average level of each signal sample. It's possible to use higher order S-D encoding, presumably to give better results.

It is fairly clear that these should give significantly better results than simple PWM, but having said that it's still hard to tell what exactly is going on inside the chips, and different ones may use slightly different algorithms to generate the high frequency pulse trains for switching.

It is possible to analyse these by considering z-transforms (see Richard Hamming's excellent book on Digital Filtering - http://www.amazon.com/Digital-Filter.../dp/048665088X

There are some good web pages which deal with bit stream approaches. I did find some, and will post details later.

One thing I've not seen, but again maybe they are buried somewhere or in patents, is the use of 1 0 -1 encoding, which when applied to sine wave power generation for DC - AC conversion, gives a significantly better approximation than a 1 -1 (logically equivalent to a 1 0) encoding, being accurate up to the 6th harmonic, which means that even simpler filters can be used on the output to clean up the required output sine wave. It is quite possible that the Class T amps use multi-level circuitry, and multi-level codes, which would explain why they are seemingly very good at low level resolution.

My suspicion is that surround sound amps use some chip based designs which are very similar to Class-T designs. I will try my Sony with my regular loudspeakers, and hopefully be able to let you know what it sounds like. I have been very surprised at how good it sounds with the fairly cheap looking speakers it was supplied with, so maybe it'll be really good with more regular speakers. I think mine's a Sony BDVE360 5.1 Blu-Ray Home Cinema - the latest model is a 370.

[hornucopia]

I've just stuck my Kingrex USB amp on to drive my (quite efficient) Ocellia Calliopes (Full rangers) as my Leben has lost a valve. Each time I do this, I start wondering if i really NEED this expensive valve amp! I was just testing out my newly acquired 'Hewitt" lightspeed control (you know; that rare handbuilt-in-Lancashire gem), but the Kingrex may stay a while. Normally it runs my Zigmahornets (see 6 Moons) as a desktop (side!) system for the iMac.
this set up was good enough to convince me to ditch horns and 'go fullrange' for my main system.
I also have a Sonic Impact 'Son of T' amp, the one mentioned as reviewed on 6 Moons, which is pretty good. It's a De luxe cased version of the basic 'T'. (You can have it for £50.....)
They work well with reasonably efficient speakers. Tannoys too I'd guess!

A/v amps? My Panasonic SA-HR70 receiver was raved about in hi-fi circles about 5 years ago. I've run Glastonburys, and latterly, SD1s from it. Impressed folk just running CD from a DVD player into it/them, so not bad at all?

[dave2010]

I said I'd come back with some links.

Here's one which seems OK, and seems quite readable - http://www.beis.de/Elektronik/DeltaS...eltaSigma.html

A problem which can occur in Class D amps is called ground bounce - it's a technical issue - and here is something about that - http://www.pericom.com/pdf/applications/AN005.pdf

There may be more.

[AndrewR]

Thanks guys, my PC mouse just 'slipped' and I bought a Sure Electronics 2*100W TK2050 board and 24V 14A Meanwell PSU.

It is going to be a fun little project and will be performing some of the diyaudio recommended mods (like the input cap replacement).

Still um-ing and ah-ing about what case to use. Thinking of going for link, which would end up giving a retro lab-equipment style. Sadly after changing jobs I no longer travel to Japan, where they sell very nice boxes from companies like Takachi.

Will let you know how it goes.

Andrew

[Ali Tait]

Maplin do some nice boxes too.

[Ali Tait]

Maplin do some nice boxes too.