Distortion Tests - HifimeDIY T2 / LJM L20D / SURE TDA7498

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I have run harmonic-distortion vs frequency and distortion vs output-power tests on the three switching amplifier boards I have bought in the last couple of months.

For the HifimeDIY T2, I connected a Meanwell (clone) 27V 5A supply, and a dummy load which is configurable to both 4 and 8 ohms.
I measured Harmonic Distortion vs Frequency at about 4 watts output power into 8 ohms. Excellent results -- with distortion over most of the spectrum ranging between 0.1% (-60dB) and 0.3% (-50dB). I repeated this at 35 watts power level, and a respectable half a percent (approx) of distortion was maintained nicely over the usable audio range. Using a 1KHz tone, I then ran Distortion vs Output Power with an 8 ohm load (attachment 3) and also with a 4 ohm load. Very good results, even with the 4 ohm load (remember that -40dB = 1% distortion). The T2 is a nice, clean, little amplifier, especially at 8 ohms.

I then tested ljm_ljm's new L20D amplifier, which uses the IRS2092 delta-sigma modulation chip for PWM conversion. For a power supply I used a toroidal transfomer rated at 28VAC 100watts, with 88,000 uF of filter capacitors. The results with the L20D were nothing short of superb. Many linear amplifiers would be struggling to achieve such a high linearity, especially at the high power levels the L20D can handle. Attachment 5 is distortion vs frequency into 8 ohms, at 35 watts output level. There is no point in showing lower output powers, the L20D is exceptional throughout its whole power range. I repeated that sweep for a 4 ohms load. Attachment 7 shows distortion vs output power into an 8 ohm load. Even putting out over 100W into a 4 ohm load (attachment 8), the L20D generated virtually no harmonic distortion. Everything was right at the lower limit of my test setup, less than 0.1% over every reasonable operating condition. I can see why ljm_ljm was so excited when he finished his L20D board :)

So now to the new SURE 100W+100W board using the new(ish) TDA7498 chip. This board was unstable when it arrived, and I had to add Zobel damping to the output network (described in another thread). Attachment 9 shows that the harmonic distortion, even at 4W into 8 ohms, is unacceptably high, hovering just below 1%. Similarly, the distortion vs power curve was not even close to that of the T2 or the L20D.

I haven't tested InterModulation distortion for these amplifiers yet, I should find time to do that sometime this week. I also will try and figure out exactly how SURE managed to mess up their TD7498 design so completely...
 

Attachments

  • DistF_T2_8ohm_5r4Vrms.gif
    DistF_T2_8ohm_5r4Vrms.gif
    14.8 KB · Views: 7,479
  • DistF_T2_8ohm_17Vrms.gif
    DistF_T2_8ohm_17Vrms.gif
    15.1 KB · Views: 7,295
  • Pdist_T2_8ohm_55w.gif
    Pdist_T2_8ohm_55w.gif
    14.2 KB · Views: 7,227
  • DistF_L20D_4ohm_17Vrms.gif
    DistF_L20D_4ohm_17Vrms.gif
    15.8 KB · Views: 1,237
  • DistF_L20D_8ohm_17Vrms.gif
    DistF_L20D_8ohm_17Vrms.gif
    15.9 KB · Views: 7,201
  • Pdist_T2_4ohm_100w.gif
    Pdist_T2_4ohm_100w.gif
    15.7 KB · Views: 7,225
  • Pdist_L20D_8ohm_200W.gif
    Pdist_L20D_8ohm_200W.gif
    16.4 KB · Views: 1,208
  • Pdist_L20D_4ohm_200W.gif
    Pdist_L20D_4ohm_200W.gif
    16.8 KB · Views: 1,164
  • DistF_SURE_8ohm_4r4Vrms.gif
    DistF_SURE_8ohm_4r4Vrms.gif
    14.6 KB · Views: 1,079
  • Pdist_SURE_8ohm_55W.gif
    Pdist_SURE_8ohm_55W.gif
    14 KB · Views: 1,019
Last edited:
What you see is very typical. Tripath based amps typically have higher average THD than many other comparable amps (though much lower than many other designs) but because of it's sigma-delta comparator technique it's completely flat unlike what you see from the I-R based which shows a rising tendency with higher frequencies.

But it's also a quite unfair comparison. The IRS2092 should be compared to the TA3020, not the TC2000. The TDA7498 isn't even in the same league so I'm not sure why that's being tested against the 2 others.
 
Last edited:
A rising tendency? That's not really what I see.
The Tripath is indeed flat with frequency, but at a relatively high level of -45dB, or about 0.3%, and while the I-R based L20D is indeed rising, the I-R is just much, much lower distortion. never going above -70dB (0.03%) (both systems measured into 8 ohms at 17Vrms). It is important to look at the magnitudes, not just the shape :) The I-R leaves the Tripath behind in every distortion comparison, it's just that the Tripath is arguably adequate in its performance, remaining below half a percent at all (audible) frequencies, especially when you don't drive it near its power ratings.

With newer loudspeakers, some would argue whether 0.3% is indeed enough. I have measured my Peerless 830983 at about -60dB THD throughout the midrange, for example, and they may be capable of outperforming a 0.3% amplifier in overall distortion if you don't drive them too hard. Who knows? For quality, I would choose the I-R if I could. But the Tripath is more convenient for smaller systems, working with a single-rail power supply (or power brick)...
.
 
Last edited:
Hmm.. Lots of data in that 41Hz report, probably too much for anybody to absorb at one sitting, Showing spectra is a little misleading, as THD is indeed TOTAL harmonic distortion, when one looks at spectral lines one usually fails to sum up the individual contributions. Noise floors are of little importance these days, noise in any decent power amp is inaudible, especially when compared with system noise -- and the noise in the listening environment. The 41Hz report would seem less overwhelming if they would stick to the normal performance measures and presentation.

But the 41Hz distortion data is interesting, as it begs the question, how could 41Hz get that data from the same chip as I just tested in the HifimeDIY? There just are no passive components in either circuit which could account for distortion differences of that magnitude. It must come down to 'switching dead time' and other difficult design issues. I am always intrigued when I see good data for which there is no viable explanation -- there has to be an explanation somewhere, and it rarely hides in the brand of bypass capacitors one or the other system is using. Perhaps 41Hz might explain to me what they are doing which HifimeDIY are not? The T2 looks a good clean board, well thought out, with quality components in all the correct places.

I am not really tempted to spend the money on a 41Hz board just to test it directly against the HifimeDIY using the same test procedures, but I suspect that is what would be required to perform a true comparison...

Oh - with respect to the TDA7498, the data sheet shows distortion in the -60dB region (see below), why do you say it's not even in the same league?
.
 

Attachments

  • TDA7498.gif
    TDA7498.gif
    38.4 KB · Views: 1,154
Last edited:
Oh - with respect to the TDA7498, the data sheet shows distortion in the -60dB region (see below), why do you say it's not even in the same league?
.

The data shown in that 41hz info is for the TA3020 driver chip, the one that is similar in design and application to IRS2092, not the TC2001 used in HifiDIY D2 (or amp11).

Have you heard the TDA7468? It doesn't compare to any of the chips mentioned above but is a low cost alternative.
 
Saturnus,
The TA3020 chip is nothing whatsoever "similar in design and application to IRS2092." The I-R technology uses low-order Delta-Sigma (fuzzy frequency) technology, a PWM encoding method I helped develop in the 1970's (when I published several papers showing methods of implementing delta-sigma). The TA3020 chip, on the other hand, uses the Tripath clocked-sample encoding technology. You can see from the test measurements in my first post that the I-R Delta-Sigma technology produces higher fidelity, distortion numbers which are much better than any of the Tripath chips. The sound is superb, rivaling the linear class AB amplifiers which I used to design.

As for your question "have you heard the TDA7468," I actually have a TDA7498 board in my possession, I published the test data above, and yes, I have listened to it extensively. There are very few DIY boards using this chip, primarily just the new SURE range, and so I suspect there are very few other members here who have actually had a good look at (or listen to) this new chip. I think the TDA7498 has potential to fill a gap at the low-end, for a lightweight, tiny, board.

IMO the IRS2092 I-R technology is disruptive technology. It takes the performance of switching amplifiers to a new level, both in reproduction fidelity and increased power level. I can understand how people who have a vested interest in the previous technologies would like to deny the existence of change, but IMO it is inevitable, as, unlike Tripath, International Rectifier is still a flourishing company, continuing to develop its products.

I will take another look at the SURE TDA7498 board later today, and report my finding here.

Meanwhile, I did measure the distortion curves for one of the low-cost YuanJing TDA8920B boards that I bought so cheaply on Ebay, and you can find the data in the attachments below. The performance is not bad, which is probably why it didn't sound so bad, either :) Distortion vs frequency is shown at a 5 watt power level, followed by distortion vs output power at 1KHz into 8 ohms, on a 25-0-25 volt DC power supply.
.
 

Attachments

  • Yuanjing_closeup.jpg
    Yuanjing_closeup.jpg
    109.3 KB · Views: 1,674
  • DistF_TDA8920_5V9rms_8ohm.gif
    DistF_TDA8920_5V9rms_8ohm.gif
    15.5 KB · Views: 1,513
  • Pdist_TDA8920B_25-25V_8ohm.gif
    Pdist_TDA8920B_25-25V_8ohm.gif
    14.9 KB · Views: 808
Last edited:
I have no vested interest in any of the above. My requirements for amp chips are fairly simple.

- Must use single rail supply only.
- Max supply voltage preferably optimzed for under 40V.
- Must be monolothic in design, ie. must not be a external FET driver.
- Ok, if it's a DDX output chip driver.
- Must have sound quality equal to or better than available Tripath chips

What I'm trying to explain to you is that the chips you are comparing are completely different in purpose and use.

The IRS2092 is +-100V dual rail voltage FET driver. From Tripath that would compare to TA3020 as that is a +-70V dual rail voltage FET driver.

The TDA7468 is a dual rail voltage monolithic design. From Tripath that would compare to TA2022 which is also a dual rail voltage monolithic design.

The TC2001 is a DDX driver (and therefore naturally single rail voltage). Distortion would primarily depend on the DDX output chip. In T2 that can be both TP2050 or STA505. You don't specify which version. The T4 with STA517B would be far better though. I has much lower THD than both the types used in the 2 versions of the T2.
 
Last edited:
I have retested the amplifier I received from SURE Electronics which was labelled as: "2 X 25Watt Class D Audio Amplifier Board - TDA7492" and which came with a chip whose top was laser-embossed as 'TDA7498'.

This amplifier seems to be using a chip which does not conform to the distortion specifications in the data sheet for either TDA7492 or the newer TDA7498. My best guess is that is a "seconds" chip, which, eg, 'fell off the back of a truck'. I have tested the TDA7920B chips from ST microelectronics, which largely perform to their data sheet. But the SURE amplifier has harmonic distortion which is unacceptable, and way above the ST data sheet values of either chip. I cannot believe that SURE could mess up their implementation so completely, although that was my first reaction, and have come to conclude it must be a second-rate chip.

Today I removed the tiny 33uH output inductors used on the SURE board and replaced them with the larger 22uH inductors which work well on the TDA7920B board. The distortion was still high, although a little better, certainly not as good as the same inductors produced on the TDA8920B Yuanjing board.

Below please find attachments in support of my conclusion. First I have a photo of my board modified with the two larger output inductors on the left channel of the board, and with the original small 330 inductors on the right. Second is distortion vs power into 8 ohms with a 24V supply and the original 330 inductors, third is with the larger 220 inductors. Then the distortion vs frequency into 8 ohms with a 24V supply at an output power of 6 watts is shown with the unmodified board. Finally, to show that the larger inductors can handle the power levels without distortion, I attach the distortion vs power curve of the TDA8920B YuanJing board which I 'borrowed' them from.
.
 

Attachments

  • Pdist_SURE_24VDC_8ohm_22-22inductor_crop.gif
    Pdist_SURE_24VDC_8ohm_22-22inductor_crop.gif
    15.1 KB · Views: 526
  • Pdist_SURE_24VDC_8ohm_crop.gif
    Pdist_SURE_24VDC_8ohm_crop.gif
    15.2 KB · Views: 1,058
  • SURE_TDA7498_ww_Inductors.jpg
    SURE_TDA7498_ww_Inductors.jpg
    187.5 KB · Views: 1,297
  • DistF_SURE_24VDC_5Vrms_8ohm.gif
    DistF_SURE_24VDC_5Vrms_8ohm.gif
    15.9 KB · Views: 459
  • Pdist_TDA8920B_25-25V_8ohm.gif
    Pdist_TDA8920B_25-25V_8ohm.gif
    14.9 KB · Views: 472
Last edited:
Member
Joined 2004
Paid Member
Thanks for doing this work. I've made a couple of TDA8920 boards and always wondered how well they performed in a commercial design. This is good information. And it's also good to see a quantitative comparison of the output inductors.

Looks like the IR technology really is a huge step forward from the older designs. And when people are selling high quality 250W amps for $37/pr, that really qualifies as "disruptive technology".

I wish Analog Devices would come out with a scaled up version of their AD1994 (RIP) chip, as it would probably provide some good competition to the IR design. It would be nice to see some tests of the Zetex circuits as well.
 
The TA3020 chip, on the other hand, uses the Tripath clocked-sample encoding technology. You can see from the test measurements in my first post that the I-R Delta-Sigma technology produces higher fidelity, distortion numbers which are much better than any of the Tripath chips.

No. The TA3020 like all Tripath chip use Sigma-Delta converters. No input frequency is around 700MHz. With input it's spread spectrum from 200KHz to 1.5MHz.

The TA3020 as you can see in the link above has a THD+N vs. Power that over the whole range until clipping is beneath your testing equipments noise floor at -80dB (0.01%), so it's a far cry better than the IRS2092 in that respect.

How it sounds, I don't really know but I would assume it would perform similar to the TA3020. Though probably not as good.

It's cheap though... and still in production so for a new design the choice is obvious.
 
Show me the results, please. Where is your test data showing that a real circuit board can give a distortion vs power and distortion vs frequency with total distortion levels remaining below -80dB? I don't want to show disrespect, but please point me to a third party test where such a level of performance was achieved :)
.
 
Last edited:
I trust the measurements from Jan at 41hz completely. They have always been accurate.

That's also why the noise floor of the testing equipment was shown so that you can test for yourself... if you got good enough testing equipment.

An externally hosted image should be here but it was not working when we last tested it.


That's for the assembled finished version of the Amp15, ie. a complete amp. Not just the chip.
 
Last edited:
Saturnus,
I found the curve attached below on Page 8 of the Tripath application note for the TA3020, titled "Class T Digital Audio Amplifier Evaluation Board," part number Bridged RB-TA3020-1-3 – MC/1.0/06-01, EAD003.

It seems remarkably similar to the curve I measured on the L20D (also attached below), not to the curve the manufacturer gave you for its AMP15 performance. Here is an explanation of why such differences are frequently seen in power electronics, and why they are unimportant.

My test equipment actually goes down to the noise floor of the 16bit input A-D converter. There are problems involved in measuring a distortion vs power curve such as you have shown, the input level at 1W (vs 100W) is already 40dB below the peak threshold, requiring a 120dB total dynamic range in the measuring circuitry. That sweep you are showing for the AMP15 must have been very carefully measured indeed.

I didn't worry about the rising noise floor in my L20D measurements (caused by my test input's 'limited' dynamic range) because the THD+noise values were already so low as to be negligible. No listening room on this planet can achieve a 120dB ratio between background noise and peak loudspeaker level. A typical noise room level rarely falls below 20dBA and most listeners don't want sound levels above about 120dBA, a total dynamic range of 100dBA. The situation is more complex than that, of course, but that is a good rule of thumb.

If you are going to directly compare my L20D data to that AMP15 curve, however, it looks like I need to make a better effort at identifying the actual L20D noise level below 10W, rather than have readers assume that the AMP15 is actually better than the L20D.

The IR amp did hit -80dB at higher signal levels, and may well have maintained it down to 1W, just as the AMP15 curve says it does. Who knows? I certainly paid little attention to any THD below -60dB, which is lower than any available speaker can reproduce :) Although it is always nice to publish great specs on an amplifier, they usually turn out to be meaningless.

I don't really want to split hairs, the quickest thing is for me to do is look around for a TA3020 chip and test it out more closely :)
.
 

Attachments

  • Tripath_TA3020_Pdist_curve.gif
    Tripath_TA3020_Pdist_curve.gif
    46.3 KB · Views: 1,893
  • Pdist_L20D_8ohm_200W.gif
    Pdist_L20D_8ohm_200W.gif
    16.4 KB · Views: 1,731
The FETs used by Tripath in their application note and the application note board are very inferior to those used by 41hz, and other manufacturers who actually use the TA3020 chip, so you can't read anything into that. The vast majority of distortion will be a in the output FETs and the board layout of the output section. This will be especially note worthy at lower power levels.
 
Last edited:
Are you suggesting that International Rectifier don't have the best range of output FETs in the business right now?

When I see a one on one comparison of the L20D with the AMP15 I can draw some meaningful conclusions. Until then I will continue to accept that the L20D is a superb amplifier.

But look, I have had enough discussion of phantoms. I was careful to use identical test conditions when I prepared my own data for this series of posts. Please post a third-party comparison data set to support your assertions that '41Hz rules the world' :) :) :)
.
 
Your assumptions are completely wrong. Anyways, until you get a TA3020 based amp (41hz or other manufacturer) for comparison I'd regard the whole subject here as irrelevant.

Let me also note that even though the amp15 is the amp for 41Hz with least distortion. It's not their best sounding amp. There's a lot more to sound quality than distortion.

And I'm sure that the L20D is a superb sounding amp. Never doubted that for a second. You just don't seem to graps that the whole point of this comparison was unfair as it involved chips and chip sets developed for completely different uses and as such the whole subject is moot.

If you're comparing something, at least compare fruits to other fruit types.
 
Wow. Well, excuse me for breathing! :( Am I indeed completely wrong, Sir?

Fundamentally, all amplifiers are designed to take an electrical signal and produce enough energy to drive loudspeakers in such a way that the brain appreciates the result as "pleasant."

I have designed many amplifiers, speakers, and other audio systems, some of which have been acclaimed as "pleasant" (or better). Would you rather I not share my perspectives? I have tried to be totally objective... :)
.
 
Lithoc,
I have a YuanJing board with the chip desoldered right now, so I could mount a TDA8950 on that one, I guess. In any case, I will post a complete set of graphs for the YuanJing TDA8920B boards, and I might also look and see if I can see any differences between the two fully assembled boards which I have working now. One has a chip I bought from Mouser, one has the original Chinese-sourced TDA8920B on it :) It will take me a day or two to run the tests, I will post them here when they are ready :)

I also have a TDA7498L winging its way from Mouser to me right now, so I can check whether the SURE board works better with a (presumed) full-spec chip :)
.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.