1970s Design Indulgence

Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic

   Initial notes on the stereo pair

Both channels being driven has some effect which I was prepared for, and it is obvious that with two 8 Ohms loaded power amplifiers demanding current from a single power supply, that the repeatability of the single channel measurements would not materialise.

Notably however, the high frequency distortion remained pretty much the same. It was the 1kHz THD+N which grew. Up from the 0.04% of the earlier single channel, it went to just over 0.05%, and I may as well just call it 0.06%.

S/N also worsened down to 70dB from the single channel's 75dB.

It still achieved 60 watts per channel but at just under 0.7% THD+N. You will recall the better measurements were all made at 50 watts.

The way the power amplifier is designed is very similar to the Proprius in that it has its own "reservoir" capacitor (4,700uF) rather than the usual 220uF decoupling capacitor which only helps high frequencies.

However, the speaker returns here are taken from the power supply ground, where on the Proprius they are taken from the local "reservoir" capacitor (being powered from a remote power supply it couldn't be from anywhere else).

It might help if I experiment by taking the speaker grounds from each "reservoir" capacitor ground.

I also note the amount of "hash" (spurious random HF noises) when filtered by the AP's low pass filters (set at 300kHz), helped lower the distortion reading, which might suggest the dominant pole is set too high.

An audio analyser isn't capable of reading at the frequencies I would like to read, the 0dB crossover frequency being around 13MHz according to the simulator. With as many parasitics as I can envisage being modelled in, I have to assume that the simulation has it right.

The calculated slew rate (2mA/0.022nF) 91V/uS isn't really required, and by using a 47pF dominant pole capacitor it will reduce to 42V/uS, which is still more than adequate. The higher frequency "hash" should then be 6dB lower, and could lead to better 1kHz distortion and noise readings.

47pF has the 0dB crossover at around 7MHz, and the input slew rate model 0.3mFt suggests a slew rate of 20V/uS. One day I shall have to do the long mathematics regarding 0.3mFt as I think the value of 0.3 must apply to a differential input and if 0.6 with a single ended input, the slew rate would correlate with the VAS slew rate (42V/uS) much better.

Anyway, 20V/uS or 42V/uS, both are good, and if setting the dominant pole such improves distortion it will be a good move.

To get at the dominant pole capacitor on each power amplifier will require a considerable strip down and rebuild, such are the joys of working in power amplifier development. I can't say I enjoy this part of design.

Another effect I noted was that of fuse distortion. I'd initially left in the current limiting resistor across the HT fuse used to test the second power amp, and that amplifier fared better on distortion, sending the other amplifier's distortion up.

Having removed the resistor both distortions equalled out, but were, as you have read above, worse. The resistance has the obvious effect of helping the fuse, but as the job of the fuse is to isolate fully a faulty channel, it cannot be left in situ.

A larger fuse would help by reducing its impedance/resistance under load conditions, but we run the risk of taking away the protection it offers. Maybe an incremental enlargement to 2.5 amps would help? I would not want to risk blowing transistors for the sake of a non-audible improvement in measured distortion.

And obviously as the HT fuse contributes to distortion, so must the speaker output fuse. Here the negative feedback take off point is after the fuse, so as to include it and use some of the loop gain to minimise its effect. I am sure if this fuse were to be over-rated it would also lead to better measured distortion as it is using up loop gain to correct for its heating.

This might be something which will be explored simply because I need to test at 4 Ohms load, and the 2A output fuse blows at just under 70 watts into 4 Ohms. This indicates the fuse rating is a little too low, and maybe it warrants the use of a 2.5 or 3 amp fuse.

As usual, there are still things to do before I can be happy with the design so far.

Author	Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic
Graham Slee Members Profile Send Private Message Find Members Posts Add to Buddy List Admin Group Retired Joined: 11 Jan 2008 Location: South Yorkshire Status: Offline Points: 16298	Post Options Post Reply Quote Graham Slee Report Post Thanks(0) Quote Reply Posted: 17 Sep 2018 at 3:46am
	Initial notes on the stereo pair Both channels being driven has some effect which I was prepared for, and it is obvious that with two 8 Ohms loaded power amplifiers demanding current from a single power supply, that the repeatability of the single channel measurements would not materialise. Notably however, the high frequency distortion remained pretty much the same. It was the 1kHz THD+N which grew. Up from the 0.04% of the earlier single channel, it went to just over 0.05%, and I may as well just call it 0.06%. S/N also worsened down to 70dB from the single channel's 75dB. It still achieved 60 watts per channel but at just under 0.7% THD+N. You will recall the better measurements were all made at 50 watts. The way the power amplifier is designed is very similar to the Proprius in that it has its own "reservoir" capacitor (4,700uF) rather than the usual 220uF decoupling capacitor which only helps high frequencies. However, the speaker returns here are taken from the power supply ground, where on the Proprius they are taken from the local "reservoir" capacitor (being powered from a remote power supply it couldn't be from anywhere else). It might help if I experiment by taking the speaker grounds from each "reservoir" capacitor ground. I also note the amount of "hash" (spurious random HF noises) when filtered by the AP's low pass filters (set at 300kHz), helped lower the distortion reading, which might suggest the dominant pole is set too high. An audio analyser isn't capable of reading at the frequencies I would like to read, the 0dB crossover frequency being around 13MHz according to the simulator. With as many parasitics as I can envisage being modelled in, I have to assume that the simulation has it right. The calculated slew rate (2mA/0.022nF) 91V/uS isn't really required, and by using a 47pF dominant pole capacitor it will reduce to 42V/uS, which is still more than adequate. The higher frequency "hash" should then be 6dB lower, and could lead to better 1kHz distortion and noise readings. 47pF has the 0dB crossover at around 7MHz, and the input slew rate model 0.3mFt suggests a slew rate of 20V/uS. One day I shall have to do the long mathematics regarding 0.3mFt as I think the value of 0.3 must apply to a differential input and if 0.6 with a single ended input, the slew rate would correlate with the VAS slew rate (42V/uS) much better. Anyway, 20V/uS or 42V/uS, both are good, and if setting the dominant pole such improves distortion it will be a good move. To get at the dominant pole capacitor on each power amplifier will require a considerable strip down and rebuild, such are the joys of working in power amplifier development. I can't say I enjoy this part of design. Another effect I noted was that of fuse distortion. I'd initially left in the current limiting resistor across the HT fuse used to test the second power amp, and that amplifier fared better on distortion, sending the other amplifier's distortion up. Having removed the resistor both distortions equalled out, but were, as you have read above, worse. The resistance has the obvious effect of helping the fuse, but as the job of the fuse is to isolate fully a faulty channel, it cannot be left in situ. A larger fuse would help by reducing its impedance/resistance under load conditions, but we run the risk of taking away the protection it offers. Maybe an incremental enlargement to 2.5 amps would help? I would not want to risk blowing transistors for the sake of a non-audible improvement in measured distortion. And obviously as the HT fuse contributes to distortion, so must the speaker output fuse. Here the negative feedback take off point is after the fuse, so as to include it and use some of the loop gain to minimise its effect. I am sure if this fuse were to be over-rated it would also lead to better measured distortion as it is using up loop gain to correct for its heating. This might be something which will be explored simply because I need to test at 4 Ohms load, and the 2A output fuse blows at just under 70 watts into 4 Ohms. This indicates the fuse rating is a little too low, and maybe it warrants the use of a 2.5 or 3 amp fuse. As usual, there are still things to do before I can be happy with the design so far.
	That none should be able to buy or sell without a smartphone and the knowledge in how to use apps

Graham Slee Members Profile Send Private Message Find Members Posts Add to Buddy List Admin Group Retired Joined: 11 Jan 2008 Location: South Yorkshire Status: Offline Points: 16298	Post Options Post Reply Quote Graham Slee Report Post Thanks(0) Quote Reply Posted: 17 Sep 2018 at 7:57am
	Further notes on the stereo pair After the above modifications (47pF dominant pole capacitor and speaker returns to local "reservoir" capacitor grounds), the following measurements were noted. I will add that I also increased quiescent current to 100mA. The output stages consuming 7.5W each. Is it class-A? No, not really, although some might call it "pure class-A", but I call it class-AB. HT: 74.8V (within the Low Voltage Directive) Iq: 100mA Max power: 58.77W into 8 Ohms at 0.1% THD at 1kHz, both channels driven for some considerable time. Distortion at 52.38W: THD+N at 1kHz: 0.035% THD+N at 10kHz: 0.11% THD+N at 20kHz: 0.2% Distortion at 1W: THD+N at 1kHz: 0.07% THD+N at 10kHz: 0.13% THD+N at 20kHz: 0.24% Frequency response: 12Hz - 50kHz -0.5dB Gain: 28.37dB (as before) S/N: 73dB A-wtd., 70dB CCIR 20Hz - 20kHz, 68dB unweighted 20Hz - 20kHz IMD: 0.1% at 52.38W You will note that the 1kHz 52.38W distortion is down from the 0.5-0.6% in my last post, but is up at 10kHz and 20kHz, which is only to be expected because of the lowering of the dominant pole turnover frequency by changing from a 22pF capacitor to a 47pF capacitor, which reduces the amount of negative feedback at high frequencies. Whether or not the speaker return modification is of any use, I did note that the channels were more matched on distortion readings. Signal to noise is improved by 3dB (A-wtd.) because of the reduction in "hash" at high frequencies due to the dominant pole shift, which also allowed for the increase in quiescent current, whereas before the current had to be reduced to obtain a good noise reading. All of which demonstrates that the dominant pole was too high. All of the above will be considered laughable by modern standards, but those modern standards see ultralow distortion paid for by the smallest of open-loop frequency response - often less than 100Hz! This design has its upper open-loop at 2.7kHz leading to a more valve-like "solidity" - 27 times better than 100Hz. But its distortion is not 27 times worse! And it does it with just 6 transistors.
	That none should be able to buy or sell without a smartphone and the knowledge in how to use apps

Fatmangolf Members Profile Send Private Message Find Members Posts Add to Buddy List Moderator Group Joined: 23 Dec 2009 Location: Middlesbrough Status: Offline Points: 8998	Post Options Post Reply Quote Fatmangolf Report Post Thanks(0) Quote Reply Posted: 17 Sep 2018 at 8:27pm
	I am learning more about hifi every time I read a post on this blog and can't thank you enough for sharing this Graham. The fine tuning of the pole I just read is a great example.
	Jon Open mind and ears whilst owning GSP Genera, Accession M, Accession MC, Elevator EXP, Solo ULDE, Proprius amps, Cusat50 cables, Lautus digital cable, Spatia cables and links, and a Majestic DAC.

Graham Slee Members Profile Send Private Message Find Members Posts Add to Buddy List Admin Group Retired Joined: 11 Jan 2008 Location: South Yorkshire Status: Offline Points: 16298	Post Options Post Reply Quote Graham Slee Report Post Thanks(0) Quote Reply Posted: 21 Sep 2018 at 8:58am
	A workable 50 watt power amplifier This update shows the power amplifier circuit which gave the measured results from my last post. I started this design by pushing for the highest power based on the power supply and then bargained it away for the best possible performance. In actual fact, and with 7% more mains voltage than specified (due to the UK adopting 230V standardization in name only), it does around 60 watts! But to specify it for universal use, and to reference distortion results at different frequencies, one has to quote the power they were measured at, and therefore that must be the declared power. It means that the same results should be capable of being measured anywhere rather than just in this location. There have also been modifications which improve calculated and subjective performance which may not jump out of the measurements as being so but give an incremental improvement. The power transistor base-emitter resistors are an example. These serve the purpose to discharge the dynamic capacitance of the base-emitter junction (and do not exist in a real Darlington/super-beta transistor to my knowledge). Should one power transistor delay its turn-off cycle whilst the other is turning-on it is obvious that two things will happen: 1. it will add distortion, and 2. the output stage will run hotter. The downside is they increase the load on the voltage amp by a small degree, but at large output excursions the base-emitter voltage will increase as the power transistors approach saturation, and so it can be seen that its current draw on the voltage amp is ever so slightly non-linear, which means distortion to some extent. So this is about trading distortions. The dominant pole capacitor having been increased to 47pF reduces open-loop gain at high frequencies meaning that high frequency distortion increases, but assures stability which reduces high frequency distortion. It is a contradictory see-saw, but I would rather have the stability. This all boils down to having to turn down the input signal a little here and a little there until the "best" set of distortion results can be gathered, and at that point the power measured should be what is indicated as its specified power output. So we have a 60 watt capable amplifier rated at 50 watts. Some might think this is good because it gives "dynamic headroom". Think again, it is only 0.8dB. I have starred what I call stability components. As I've said before, all emitter followers are Colpitt's oscillators just waiting to wreak havoc. In this circuit there are two: T4 with T6, and to a lesser degree T5 with T7. Due to 100% negative feedback T5 with T7 might be better behaved, so T4 with T6 might dominate, but there is still chance the output stage will want to do what comes naturally. The only way to prevent oscillation is by limiting its frequency extremes, but that would limit the audio bandwidth, so we have to tread carefully. C1 and C4 remove the possible open circuit input stimuli which is more possible than you might think. C3 stretches out the gain margin by dropping the gain just below the high frequency 0dB crossover point before 180 degrees is reached - this being easier to find using simulation but calculations show it tracks slightly higher than the dominant pole turnover. C11 and R26 form the zobel network which replaces the load at high frequencies where the speaker and associated crossover and wiring become inductive. C11 ensures R26 is high impedance at audio frequencies such that it is not stealing power. L1 at 5uH might be considered excessive but rescues the amplifier from too much capacitance on its output, making it more immune to speaker cable choice. Finally it should be noted that unconventionally the amplifier ground is not at the power supply but at the rather larger than usual decoupling capacitor (C13). This means that the user must be cautioned to ensure the speaker returns are not swapped between channels, as although this point is still 0V, it is a star point and the correct one as marked must be used for the amplifier to perform its functions properly.
	That none should be able to buy or sell without a smartphone and the knowledge in how to use apps

Farmer Members Profile Send Private Message Find Members Posts Add to Buddy List New Member Joined: 24 Dec 2010 Location: Soggy NW of UK Status: Offline Points: 12	Post Options Post Reply Quote Farmer Report Post Thanks(0) Quote Reply Posted: 21 Sep 2018 at 11:28am
	Hi Graham, (and other contributors) Thanks for the excellent thread... I've just spent the last 2 hours reliving my youth in reading this - and the time I spent (whilst supposed to be studying) building a John Linsley-Hood designed "integrated" amplifier - part by part, as my meagre student income (low grant) and work behind a bar (and often in front of it) allowed... That amp is still somewhere in my loft and this thread has enthused me to try and find it - and perhaps fire it up at the weekend.. I have just found part of the circuit diagram - it's been used as a bookmark for years in my RS catalogues!! I think this may have been an update to his original design - and hails from ETI or similar around 1990 at a guess.. I have two self build amps/bits of kit - both built when I was "hands-on" 1980-1995 working with everything from older DTL kit through to the then very new PLC's and software controlled stuff that used UVpROM's. Those were "interesting times" with one of my useful "tools" being a BBC Model B! It also gave me access to almost limitless supplies of components... (all actively encouraged by my employer - because it meant I was nearly always in the workshop and handy for call outs!) My other bit of kit that's still working and in use today (in my workshop driving a pair of old KEF speakers) is a self-build using hand assembled stepped attenuators on an adapted rotary switch in a passive "switching box". This connected to another Power Amp box - which housed a pair of (component provider bought) 120W MOSFET power amp modules that I simply built into a case and designed the two (each channel completely isolated) power boards for.. This comprised 2 x toroidals, big quality capacitor reserves plus a timing circuit to prevent the output "pop" on power up..! I think the modules were by a firm called JayTee or something like that - and I vaguely recall RS/STC/Farnell or someone even stocking them at one point (c.1986?) All I can say is I was wowed by the numbers and wanted something to drive a pair of Heybrook HB III's that I'd been given, and my Richer Sounds purchased cheap omni-box 20W amp which I had at the time couldn't move them!! I now know how little I knew back then (thanks Graham) - but actually this MOSFET kit still sounds way better than an old NAD 3020 I found lying about.. and certainly has the volume for an occasionally noisy mechanical workshop.. This thread has been a great re-education into analogue audio design and gain control, that if honest I don't fully understand yet - but will be re-reading the lot a few times... !! Again - thanks all. David (Part time Farmer - full time geek)

Graham Slee Members Profile Send Private Message Find Members Posts Add to Buddy List Admin Group Retired Joined: 11 Jan 2008 Location: South Yorkshire Status: Offline Points: 16298	Post Options Post Reply Quote Graham Slee Report Post Thanks(0) Quote Reply Posted: 22 Sep 2018 at 4:25pm
	When I first saw the above circuit which was part of a JLH preamp design, I spotted one error: the original input load was 4.7k and the passive output of the disc stage could be fed straight to it, meaning that the high frequency RIAA attenuation would shoot up in frequency. Then I noticed the class A-A symmetrical config of the above circuit and went "oh dear". Bias current flows out of PNP bases and flows into NPN bases, so which way does it flow through Ro? and the gain pot to/from ground? It cannot flow in both directions at the same time. As it's DC bias it must be one way or the other. And how stable will that be? Within 10 years I found myself working freelance on some power amplifiers based on the same principle. The slightest instability made them fall-over - the output swung to one rail or the other - and pop went the fuse on that rail simultaneously with the domino effect of destroyed transistors. But they all had servos And looking back at the above design it is interesting to note the output balance is within the ..?.. At first look the ..?.. seems to be output stage protection, but the 100uF capacitors introduce a delay of 0.01 x 100 = 1 second... too long for that. So, it's a servo of sorts. Although I've heard it praised to heaven for its sound, I don't think it's one of JLH's better ideas.
	That none should be able to buy or sell without a smartphone and the knowledge in how to use apps

Farmer Members Profile Send Private Message Find Members Posts Add to Buddy List New Member Joined: 24 Dec 2010 Location: Soggy NW of UK Status: Offline Points: 12	Post Options Post Reply Quote Farmer Report Post Thanks(0) Quote Reply Posted: 22 Sep 2018 at 4:57pm
	Hi Graham, As I said in my previous post, how very little I knew back then! I did build the complete amplifier and it was okay, and certainly all that I could afford at the time - bit by bit! It also taught me a few things along the way. As an impressionable young engineer, I read all the magazines cover to cover (no Internet then) and so these designers were real gurus... It’s only later you realise the effects of opinion, monetary gain and vested interests - oh and the power of marketing... 😳 I think many designs were similar then, and I can’t remember why I felt that was the one to build. It may have been as simple as my access to the right components. I think a place in Shrewsbury sold kits of parts - but I couldn’t afford that, so just scrounged and had the pick from my beneficial and supportive employer! I must have been careful using it, since on power up last night, the amp is still working one channel - although the fuse on the other is blown, so I suspect that your thoughts may be correct! Replacing the fuse made no difference so suspect the transistors are dead. In any case, I’m much harder on the junk I retain these days - so it’s gone in the WEEE recycling bin now, and I’ve only retained the PSU’s and a few of the aluminium knobs as these always seem to come in handy... The encapsulated MOSFET power amps - I now remember were by ILP (MOS248) and my pair with my own PSUs are still going strong (and also served a purpose at the time) - even if it’s a little less elegant now - a bit battered and dust covered in my workshop! Do keep us updated on progress!!

1970s Design Indulgence

Cadman Enterprises Limited
7 Kirkstall Close
South Anston
Sheffield S25 5BA
England

+44 (0) 1909 568739
+44 (0) 7947 107188

john at gspaudio.co.uk

1970s Design Indulgence

Cadman Enterprises Limited7 Kirkstall CloseSouth AnstonSheffield S25 5BAEngland

+44 (0) 1909 568739+44 (0) 7947 107188

john at gspaudio.co.uk

Cadman Enterprises Limited
7 Kirkstall Close
South Anston
Sheffield S25 5BA
England

+44 (0) 1909 568739
+44 (0) 7947 107188