Well, I just mentioned it because I think that OS needs a product lineup.
One size doesn't fit all.
So maybe:
AX-Xtreme (big power) (assembled board, assembled amp)
AX (about 150w) (long kit, assembled board, assembled amp)
FA (about 90w**) (Simplified so it can be "short" kit)
and an optional computer-centric preamp to wake up dull computer sound
and a miniature AX with hybrid current drive for the full range people
and some power supply options, especially the cap multi
* Support costs: Long kit has seriously big advantages for Availability because he can (and should) charge a reasonable markup on all parts, and, with a long kit, problematic support labor (hassle) is seriously decreased with a more complete kit. This makes for much better quality control, a speedy application and a generally trouble-free experience for the customer.
OS, please don't fail to charge enough--that affects availability.
So, please charge enough that you Want to do it. Instead of feeling "guilty" about price, just have different products so you can accommodate different customers.
** Wallet size: The least priced amplifier in the lineup, and that should exist in the near-certainty that somebody wants to buy it, may be arranged on operating voltage so that 100v rated caps take the beating at the rectifier and elsewhere 50v caps may be used for lower cost. And, that's just an example. Usually the bottom of the line is the most difficult to design because of price point. So, the simplified DC servo, magic diode, and other interesting things might help. But, you're going to need different price points--the wallet size of potential customers varies considerably.
One size doesn't fit all.
So maybe:
AX-Xtreme (big power) (assembled board, assembled amp)
AX (about 150w) (long kit, assembled board, assembled amp)
FA (about 90w**) (Simplified so it can be "short" kit)
and an optional computer-centric preamp to wake up dull computer sound
and a miniature AX with hybrid current drive for the full range people
and some power supply options, especially the cap multi
* Support costs: Long kit has seriously big advantages for Availability because he can (and should) charge a reasonable markup on all parts, and, with a long kit, problematic support labor (hassle) is seriously decreased with a more complete kit. This makes for much better quality control, a speedy application and a generally trouble-free experience for the customer.
OS, please don't fail to charge enough--that affects availability.
So, please charge enough that you Want to do it. Instead of feeling "guilty" about price, just have different products so you can accommodate different customers.
** Wallet size: The least priced amplifier in the lineup, and that should exist in the near-certainty that somebody wants to buy it, may be arranged on operating voltage so that 100v rated caps take the beating at the rectifier and elsewhere 50v caps may be used for lower cost. And, that's just an example. Usually the bottom of the line is the most difficult to design because of price point. So, the simplified DC servo, magic diode, and other interesting things might help. But, you're going to need different price points--the wallet size of potential customers varies considerably.
[QUOTE So, the simplified DC servo, magic diode, and other interesting things might help.][/QUOTE]
There is NO magic diode. It is simply a means towards getting the collector voltages of the CM much closer together, which helps to ensure that if you started off with very closely matched LTP devices, that they would stay closely balanced due to allowance being made for the bias current of the VAS/ Emitter follower. You can then vary slightly the emitter resistor value of the VAS to finetune , so as to get the collector voltages of the LTP identical.
Have a look at the Digi 125 thread's later posted schematic, and visualise an added resistor of approximately 1.6Kohms inserted in the LTP collector that presently goes direct to the supply rail of +37V. IF the LTP devices were closely matched, there will be a similar SQ improvement as both devices will then track each other thermally, especially if they are thermally coupled.
The same applies even when using closely matched LTP devices where both halves of the LTP are loaded by another LTP.
The 2nd LTP must also have a means of keeping the dissipation through both halves similar, otherwise the different VBEs will unbalance the matched input LTP. I achieved this in a much older thread about the ETI5000, where i inserted an 18V Zener diode in one of the collectors of the 2nd differential pair.
Suzy J actually got to test (rescued from a friend's garage) the ETI 5000 where i had done this.
SandyK
There is NO magic diode. It is simply a means towards getting the collector voltages of the CM much closer together, which helps to ensure that if you started off with very closely matched LTP devices, that they would stay closely balanced due to allowance being made for the bias current of the VAS/ Emitter follower. You can then vary slightly the emitter resistor value of the VAS to finetune , so as to get the collector voltages of the LTP identical.
Have a look at the Digi 125 thread's later posted schematic, and visualise an added resistor of approximately 1.6Kohms inserted in the LTP collector that presently goes direct to the supply rail of +37V. IF the LTP devices were closely matched, there will be a similar SQ improvement as both devices will then track each other thermally, especially if they are thermally coupled.
The same applies even when using closely matched LTP devices where both halves of the LTP are loaded by another LTP.
The 2nd LTP must also have a means of keeping the dissipation through both halves similar, otherwise the different VBEs will unbalance the matched input LTP. I achieved this in a much older thread about the ETI5000, where i inserted an 18V Zener diode in one of the collectors of the 2nd differential pair.
Suzy J actually got to test (rescued from a friend's garage) the ETI 5000 where i had done this.
SandyK
Last edited:
Thank you SandyK!
I hadn't hoped to understand, but reading through your post about six times just about got me there.
It didn't answer the question of whether those technologies could be used to make an "extra simplified" design with rather fantastic sound quality.
And I can't answer that question either--I'm really very bad with discrete solid state design, although I can assemble almost anything.
Daniel, you imagine that using the special Diode technique we may achieve a simple and well-performing amplifier. We have imagination too! And we are constantly trying to simplify and improve performance at the same time. The idea is simple. You try and make the operating conditions equal for a matched LTP pair, so that they match well during operation. This is only one bit in the entire topology of the circuit. If the rest could easily be simplified, it would have been. The ironic thing about your question is that if we knew how to make the circuit simpler using the magic diode, we probably would have already done it. And we are as clueless as you as to how we might go about trying to figure out how we might. So really you are asking us to do what we're already trying to do. By all means, help out!
I am not saying your idea does not hold merit. Say that in your mind is the spark of a truly brilliant idea about making a simplified and well-performing amplifier with a magic diode. Somehow that inspiration needs to reach the escape velocity of your intuition, and in order for that to happen, you either need to communicate your intuition clearly to those who have the technical skills, or learn the technical skills yourself. And before that you must make sense of your intuition.
Anyways I hope this makes sense.
- keantoken
I am not saying your idea does not hold merit. Say that in your mind is the spark of a truly brilliant idea about making a simplified and well-performing amplifier with a magic diode. Somehow that inspiration needs to reach the escape velocity of your intuition, and in order for that to happen, you either need to communicate your intuition clearly to those who have the technical skills, or learn the technical skills yourself. And before that you must make sense of your intuition.
Anyways I hope this makes sense.
- keantoken
Why mess with the diode ? Go all the way and do the improved wilson mirror.
I have tried it , it works. Subjectively , I have listened to both the diode and the
wilson above... and can not tell the difference on the AX. Electrically , increased impedance and 10X greater current "mirroring" can be seen (measured). Loop gain increases slightly , but this is good for TMC or TPC compensations .
OS
An externally hosted image should be here but it was not working when we last tested it.
I have tried it , it works. Subjectively , I have listened to both the diode and the
wilson above... and can not tell the difference on the AX. Electrically , increased impedance and 10X greater current "mirroring" can be seen (measured). Loop gain increases slightly , but this is good for TMC or TPC compensations .
OS
Why mess with the diode ? Go all the way and do the improved wilson mirror.
I have tried it , it works. Subjectively , I have listened to both the diode and the
wilson above... and can not tell the difference on the AX. Electrically , increased impedance and 10X greater current "mirroring" can be seen (measured). Loop gain increases slightly , but this is good for TMC or TPC compensations .
OS
Os
But have you then taken the next step with the diode, and adjusted the VAS emitter resistor value slightly to obtain a virtually complete balance of the LTP when doing this ccomparison with the improved Wilson Mirror, like you did originally way back in November 2008 ? I will say it again, that the balancing doesn't work for just circuits that use a current mirror.I used a similar technique many years ago with an RIAA phono preamp where the closely matched BC549C input devices were loaded by a very high HFE (1,300!) LM394H dual transistor.
The improvement in sound stage and apparent improvement in S/N when adjusted, were just as obvious there.
Alex
I think the TMC'ed AX is just so good ... that my 46 YO ears can not hear the subtle differences anymore. 
I am keeping the modular system , so I can further refine the wilson. I did notice that a certain emitter R on the VAS (680r) DOES have the best sound. This must be because it is resulting in a better LTP balance.
OS
I am keeping the modular system , so I can further refine the wilson. I did notice that a certain emitter R on the VAS (680r) DOES have the best sound. This must be because it is resulting in a better LTP balance.
OS
I have never run into a problem in that way. Voltage stage wise, I have gone from 0C to 50+C(with the hairdryer) and observed no changes if the heating was uniform. In normal operation the only thermally active area is the VAS (with it's small HS). It stays about 5C above ambient winter or summer. The current source will hold .01mA or less variation regardless of temperature.
The drawback to your suggestion is possible FB from the huge voltage swing of the VAS to the small signal area. Best to keep them segregated.
OS
The improved wilson mirror doesn't look either complex or expensive, and it seems that the transistors may be finer resolution for audio signal.
Is the Wilson mirror the three transistor version? Q1, 2 & 3?
Is the improved Wilson all 4 transistors?
Is there another name it goes under?
What is the Widlar mirror? Which transistors are retained?
In the standard mirror the "error" is the abstraction of two base currents from one side.
This can be compensated in a number of ways
Ost has referred to using the VAS base current as the compensation.
Or one adds resistors to the mirror bases, or one adopts one or both of the extra transistors.
Or one adds the assisting transistor that reduces the error but does not eliminate it.
The important part in all of this is to get the two halves of the LTP operating as closely as possible as identical amplifiers.
This needs the LTP pair and Q1,2 pair and Q3,4 pair all be be hFE matched and Vbe matched, while all are tested/measured/matched at operating Ic and thermally coupled.
Is the improved Wilson all 4 transistors?
Is there another name it goes under?
What is the Widlar mirror? Which transistors are retained?
In the standard mirror the "error" is the abstraction of two base currents from one side.
This can be compensated in a number of ways
Ost has referred to using the VAS base current as the compensation.
Or one adds resistors to the mirror bases, or one adopts one or both of the extra transistors.
Or one adds the assisting transistor that reduces the error but does not eliminate it.
The important part in all of this is to get the two halves of the LTP operating as closely as possible as identical amplifiers.
This needs the LTP pair and Q1,2 pair and Q3,4 pair all be be hFE matched and Vbe matched, while all are tested/measured/matched at operating Ic and thermally coupled.
Last edited:
My first contact with this 4 transistor version was in New Electronics March 1981 by Dr B Wilson who called it a Precision Current Source. Is that why it's a Wilson mirror?
I've used it in my Symasym and FC-100 versions.
Brian.
Matching Q3 and Q4 won't do much good, unless Q4 is cascoded, because Early effect will vary the base width region and Vbe, not to mention Q4 will have higher Vce, higher dissipation, and lower Vbe because of this. This goes for the plain current mirror as well. This is why the 3-transistor Wilson makes bounds in linearity above the 2-Q mirror, but subsequent modifications present little if any improvement. Cascoding the output side of the 2-Q mirror solves thermal and Early effect issues.
- keantoken
- keantoken
The 3/4 Q wilson is nice... , but to solve the problem at hand one must "fiddle".
I used perfectly matched (2Q single package) LTP and CM pairs. The AX had .5mV offset at this point ( offset trimmer removed). This was the load of the VAS on the non-inverting side of the LTP. I removed the 2Q device and added a SLIGHTLY mismatched pair (221/219 Hfe) with the slightly higher Hfe on the non- inverting side of the LTP. This resulted in .1mV offset. So even with a gain of (410/ksa992 X 155/ksa1381) 63,000, the VAS still loads the IPS.
I tried all 5 of my 2Q matched pairs in the initial test and this resulted in .4-.5mV offset. This eliminated the devices as a factor. This was done with just a 2Q standard CM !!! With a 4Q wilson/2Q matched pair the readings were 5X better (.1mv). With a small change in the beta enhancement device's (KSA992-VAS) emitter resistor (560-680R) , I could get nearly 0mV !! The adjustment of the BE emitter resister may be a better method than having an offset adjust across the LTP (thanks sandy). This will require slightly better input stage matching.
The IDEAL configuration is a closely matched IP pair , 4Q improved wilson, and that adjustable BE emitter resistance. My end results were hard to measure (<.05mV total offset , <1mV collector voltage difference)and the dual trace scoped waveforms were perfectly in sync at 1K.
I think we are at the climax for the frontend of this topology , any improvement would be in the OPS (triple's / LATFET's) , my next amp will be LFET's.
OS
I used perfectly matched (2Q single package) LTP and CM pairs. The AX had .5mV offset at this point ( offset trimmer removed). This was the load of the VAS on the non-inverting side of the LTP. I removed the 2Q device and added a SLIGHTLY mismatched pair (221/219 Hfe) with the slightly higher Hfe on the non- inverting side of the LTP. This resulted in .1mV offset. So even with a gain of (410/ksa992 X 155/ksa1381) 63,000, the VAS still loads the IPS.
I tried all 5 of my 2Q matched pairs in the initial test and this resulted in .4-.5mV offset. This eliminated the devices as a factor. This was done with just a 2Q standard CM !!! With a 4Q wilson/2Q matched pair the readings were 5X better (.1mv). With a small change in the beta enhancement device's (KSA992-VAS) emitter resistor (560-680R) , I could get nearly 0mV !! The adjustment of the BE emitter resister may be a better method than having an offset adjust across the LTP (thanks sandy). This will require slightly better input stage matching.
The IDEAL configuration is a closely matched IP pair , 4Q improved wilson, and that adjustable BE emitter resistance. My end results were hard to measure (<.05mV total offset , <1mV collector voltage difference)and the dual trace scoped waveforms were perfectly in sync at 1K.
I think we are at the climax for the frontend of this topology , any improvement would be in the OPS (triple's / LATFET's) , my next amp will be LFET's.
OS
I still may include the 1K trimmer between Q1-2 ... for the newbiees.
R18 becomes a 220R + a 500R trimmer. That's all , folks (besides a 4-device L-FET OPS )
PS - here's a thought ... if you tweak a circuit and you can measure an improvement electrically (offset /balance/current) but hear little or no sonic improvement, have you really improved it ??? Another example - the last change with the CCS's eliminated the "thump/click" , but did not change the sonics.
OS
R18 becomes a 220R + a 500R trimmer. That's all , folks (besides a 4-device L-FET OPS
)PS - here's a thought ... if you tweak a circuit and you can measure an improvement electrically (offset /balance/current) but hear little or no sonic improvement, have you really improved it ??? Another example - the last change with the CCS's eliminated the "thump/click" , but did not change the sonics.
OS
Attachments
Last edited:
Even if you can't hear it there may be others that are able to. Welllllllll - at least they will claim to be able to hear the difference
the mirror accuracy is affected by the base currents that are abstracted from the side that they are connected to.
the 2Q version takes all it's base current from one side only. That leaves the "error" to be equaled by some compensation system/method.
Adopt your compensation method.
Now change the mirror devices. If they are not matched then the compensation can no longer match the error.
Similarly with the 4Q version.
Matching of the devices allows the compensation to be set up to remove mirror "error".
replace either of the pairs with mismatched devices and the compensation no longer matches the error.
If the mirror is not accurately set up then the LTP cannot be balanced. It is this balance that OST's range of module has shown is critical to getting the best results from the input stage.
I have been preaching this for years. Few if any wanted to believe.
I was a bit like John the Baptist, I believed but I did not have the evidence. OST is a bit like Christ who came along a bit later with the real message and the proof that it works and how to do it.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.