Let's not rush forward too fast, or we might meet some 'dead ends' that are expected, a few years later. It is best to look a little more carefully as to what we found between 1973-1978 first, then go forward. To summarize, we started by TRYING 20KHz open loop response, low feedback 20-40dB, and getting far faster slew rates than typical IC's or even discrete designs at the time. Also, we were listening for differences between these designs and the traditional designs by Marantz, GAS, SAE, etc.
Now, 'class' let us turn to the JAES paper pp. 170-73. The first 2 pages describe the TIM test and some predictions based on IM distortion theory.
When I first arrived, to research the paper, we had not set the rise-time of the 'easier' TIM test. We experimented with TIM(10), TIM(20), TIM(30), as well as TIM(100) that we set as a worst case test. We settled with TIM(30) because Tom Holman had published a similar, but simpler test using a 30KHz filter, and it made about a 10uS rise-time, that is about what the best 1/2 inch diameter instrumentation microphones did, that were used by several high end audio recording companies, at the time.
10uS is a very good approximation of what can be recorded and reproduced, it turns out. More later.
When I first arrived, to research the paper, we had not set the rise-time of the 'easier' TIM test. We experimented with TIM(10), TIM(20), TIM(30), as well as TIM(100) that we set as a worst case test. We settled with TIM(30) because Tom Holman had published a similar, but simpler test using a 30KHz filter, and it made about a 10uS rise-time, that is about what the best 1/2 inch diameter instrumentation microphones did, that were used by several high end audio recording companies, at the time.
10uS is a very good approximation of what can be recorded and reproduced, it turns out. More later.
John,
You brought us a great preamp in the Blowtorch.
There is also a Blowtorch Phono Pre which has not drawn much attention.
But how about a no global feedback balanced power amp ?
Patrick
PS I hope you would forgive me for messing around with your circuit to turn it into an IV.
http://www.diyaudio.com/forums/pass-labs/173291-zen-i-v-converter.html#post2298531
You brought us a great preamp in the Blowtorch.
There is also a Blowtorch Phono Pre which has not drawn much attention.
But how about a no global feedback balanced power amp ?
Patrick
PS I hope you would forgive me for messing around with your circuit to turn it into an IV.
http://www.diyaudio.com/forums/pass-labs/173291-zen-i-v-converter.html#post2298531
Last edited:
I find that making a no global feedback power amp difficult, if not impractical for most commercial designs. In fact, in recent months I have been working on a couple of exotic power amps and modifying my Parasound line of power amps (4 separate models at different price points). We actually have to worry about REDUCING negative feedback in our power amps, because the Holman based THX spec is hard to meet without feedback, and we need the THX certification for home theatre installation. Even line preamps without global negative feedback are difficult, and the Parasound JC-2 preamp and the new Orion phono preamp (Constellation Audio) in development, has global negative feedback in the second gain block. It is almost impossible to meet 'specs' without it.
However, op amps of traditional type are ONLY used in the servos.
However, op amps of traditional type are ONLY used in the servos.
Hi Patrick,
See page 523, Figure 25.14 in my new book. 200 wpc balanced into 8 ohms, no NFB, ThermakTrak output transistors (8 pair). It is a fairly complete illustrative design, but was not built. Design considerations are discussed. John is right, no-NFB designs are difficult and challenging.
Cheers,
Bob
OK everyone, those of you who have done your homework can see something interesting. On page 172 in the JAES, or page 3 of our paper, there are 3 separate figures, done with an HP 3581 wave analyzer and its attendant HP graph recorder.
Figure 2 is the TEST WAVEFORM, for TIM (30). Figure 3 is the total TIM(30) distortion through a 741 IC op amp, with 20 dB gain, at 5V PEAK-TO-PEAK. Notice the 'grass' growing up between the test waveforms. Figure 4 is the SAME circuit, but only IM (30) generated by just flipping a switch on the square wave generator to triangle wave. Note the slight difference in the test waveform and the MAJOR difference in distortion products by comparing to the original test waveform in figure 2. This is essentially what the Crown IMA would measure at the same pk-pk output, and why we were sure that SMPTE IM had missed a major distortion mechanism. This is important, More later.
Figure 2 is the TEST WAVEFORM, for TIM (30). Figure 3 is the total TIM(30) distortion through a 741 IC op amp, with 20 dB gain, at 5V PEAK-TO-PEAK. Notice the 'grass' growing up between the test waveforms. Figure 4 is the SAME circuit, but only IM (30) generated by just flipping a switch on the square wave generator to triangle wave. Note the slight difference in the test waveform and the MAJOR difference in distortion products by comparing to the original test waveform in figure 2. This is essentially what the Crown IMA would measure at the same pk-pk output, and why we were sure that SMPTE IM had missed a major distortion mechanism. This is important, More later.
Please note that the comparison of fig 3 and 4 shows what TIM measures, compared to IM measurement. Also note that we deliberately used a 4:1 ratio, peak to peak, of the square wave with the added sine wave, in order to most closely emulate the SMPTE IM measurement long held as a standard for testing audio distortion.
Finally, 5V pk-pk is about 2V rms, so we have headroom in the 741 of about 5 or 14 dB, before we would expect the IC to voltage clip.
You might think of this test as a typical line amp driving a power amp. The gain levels and the output voltage levels are similar.
However, although this test is VERY informative, we seldom use it these days, because it is so hard to interpret to give a simple number, as we have to average all the added distortion together. This is a great research approach, but not very practical. More later.
Finally, 5V pk-pk is about 2V rms, so we have headroom in the 741 of about 5 or 14 dB, before we would expect the IC to voltage clip.
You might think of this test as a typical line amp driving a power amp. The gain levels and the output voltage levels are similar.
However, although this test is VERY informative, we seldom use it these days, because it is so hard to interpret to give a simple number, as we have to average all the added distortion together. This is a great research approach, but not very practical. More later.
It should be mentioned that uA741 with its 0.5V/us, is only capable to transfer less than 8kHz at 10Vp before it starts distort by SR. This opamp has never been considered as convenient for audio, even in times when it was introduced. Only ignorants could use this circuit. A liittle bit better option was uA748, decompensated uA741, which allowed less compensation for higher gains than unity and thus higher slew rate.
I do not understand why that uA741 is still mentioned today, 35-40 years after its era.
As a rule of thumb, one should use 10x higher slew rate than that of the highest audio sine dv/dt. For this reason, uA741 had NEVER been good for audio, even in 1970, BEFORE the AES articles were published.
I do not understand why that uA741 is still mentioned today, 35-40 years after its era.
As a rule of thumb, one should use 10x higher slew rate than that of the highest audio sine dv/dt. For this reason, uA741 had NEVER been good for audio, even in 1970, BEFORE the AES articles were published.
PMA, we are DISCUSSING a paper that was written in 1976, that is 34 years ago. LATER, in the paper, other IC's will be discussed. AT THE TIME, the uA741 was one of the standard IC's commonly available. Even the 5534 was NOT out yet. The 741 was considered by people, you know, engineers, as adequate for the task. The Crown IMA didn't show any serious problems. Yes, whole studio boards, master recorders, and preamps were made with the uA741, or its rough equivalent. Even today, while the slew rate has moved up somewhat, the same circuits are being used in the VAST MAJORITY of audio equipment available to the public.
Now, back to the 1976 article. It should be noted that we are already into virtual clipping with the easier test signal, at approximately 1/5 the actual nominal measured clipping point with a Crown (or equivalent) IMA. Look at the distortion numbers of some of the bigger distortion products, by noting the vertical scale at the % distortion it denotes. Everybody see the -40 to-50 dB scale where the distortion components are?
Why I am putting so much emphasis here on these charts is to eliminate 'glossing over' and not really understanding what is on the graph. We, the original testers did hundreds of tests, of everything we could think of, and virtually all levels, so WE know what we are looking at, but many of you will be not so able, until you do your homework.
Now, we should look at Fig. 5 on the next page.
Here is a SUMMATION of many tests done with many op amps and their SUMMED TIM at a given pk-to-pk output level. We can also see the 741 from fig. 4 and it SUMS to about 1% TIM at 5V pk-to-pk.
Also, there is the 301, the 709, and the 739, a popular audio device at the time.
There are some somewhat better devices, such as the 1456 and the 3140. Also, de-compensated versions of some op amps are presented that do much better than the unity gain compensated versions. This might seem like an easy fix, EXCEPT that common phono stages of the time or the inclusion of feedback tone controls might FORCE unity gain compensation. More later.
Why I am putting so much emphasis here on these charts is to eliminate 'glossing over' and not really understanding what is on the graph. We, the original testers did hundreds of tests, of everything we could think of, and virtually all levels, so WE know what we are looking at, but many of you will be not so able, until you do your homework.
Now, we should look at Fig. 5 on the next page.
Here is a SUMMATION of many tests done with many op amps and their SUMMED TIM at a given pk-to-pk output level. We can also see the 741 from fig. 4 and it SUMS to about 1% TIM at 5V pk-to-pk.
Also, there is the 301, the 709, and the 739, a popular audio device at the time.
There are some somewhat better devices, such as the 1456 and the 3140. Also, de-compensated versions of some op amps are presented that do much better than the unity gain compensated versions. This might seem like an easy fix, EXCEPT that common phono stages of the time or the inclusion of feedback tone controls might FORCE unity gain compensation. More later.
Pavel,
You are right almost no one uses the 741 anymore they use the 1558 and the lower noise version the 1559. These are used in great quantities in almost all the semi-pro audio gear. The TL072's are also used primarily in the tone control sections of mixing boards, so they really have to work at the upper frequency limits.
Yes there are still products in production that are the top sellers in their market using these products!
ES
Now back to early and present day slew rate, and TIM It is obvious that the 741 tests as pretty sorry with this test, but what about other IC op amps at the time? Some, tested very well, others tested at least significantly better. This is noted in the text.
Later, the 5534 came out, in fact I got the first samples for myself on the very day this paper was presented in NY, in 1976. The answer to all our needs in audio design? More later.
Later, the 5534 came out, in fact I got the first samples for myself on the very day this paper was presented in NY, in 1976. The answer to all our needs in audio design? More later.
Many of the first designs using the 5534 oscillated! This was due in part to the designers not understanding the importance of layout and parasitic capacitances as much as the variation in Gain-Bandwidth among the early production runs.
Now, in 1976, the 'missing link' to what the IM analyzer was missing, was put forth, and the 5534 was introduced. Since about 5V/us was considered about all we needed at minimum in a preamp, the 5534 should have been, and sometimes is, a darn fine answer to our audio needs. However, I made an A-B comparison in 1978, and subjectively found the 5534 wanting. I put this is print, in correspondence with Dr. Lipshitz in 1979, as an LTE in 'The Audio Amateur'. More later
Finally, I would like to show something I did, working at the Finnish lab, that did not get in the JAES paper. It is based on the same test equipment, but I tested a typical phono preamp stage for TIM, rather than a line stage. To do this, we have to make a typical phono stage, add a 'differentiator' or 6 dB/octave booster to more accurately represent the real signal coming off the phono cartridge, and add yet another low pass filter, to stop the 6 dB/octave differentiator from making too fast a signal. I call this test signal, (tim 30,30). To see the results of the test you have to go the the IEEE Audio paper that I presented in 1978, go to page 264, Graph 1. There you will see the SUMMED results of many tests and two phono test signals, tim(30,30) and tim(15,10). The slower test waveform was designed for moving magnet cartridges only. Some people use them exclusively, of course. More later.
- Status
- Not open for further replies.