Everything that looks like aside looks sexy, if don't understand how it works.
No free cheese in this world exist, except in a mouse trap. All decisions have to be carefully weighted. I use servos in dynamic load of tubes on gyrators, but I don't add more of gain in the servo loop than is needed to stabilize working point. If additional amplifiers are used in servos, you have to remember that this amplifiers have own properties, like noises and distortions, and this noises and distortions are added to the amp that you stabilize using servo. Each and every particular case have to be weighted, what it brings more: harm, or goodness.
And of course the amp with even matched JFETs like in John's favorite topology would benefit from a servo that uses good opamp, but it does not mean that any amp would benefit equally.
No free cheese in this world exist, except in a mouse trap. All decisions have to be carefully weighted. I use servos in dynamic load of tubes on gyrators, but I don't add more of gain in the servo loop than is needed to stabilize working point. If additional amplifiers are used in servos, you have to remember that this amplifiers have own properties, like noises and distortions, and this noises and distortions are added to the amp that you stabilize using servo. Each and every particular case have to be weighted, what it brings more: harm, or goodness.
And of course the amp with even matched JFETs like in John's favorite topology would benefit from a servo that uses good opamp, but it does not mean that any amp would benefit equally.
Last edited:
When a servo does replace even a high quality capacitor, the sound often improves becuase of the low freq extention --- but not because of freqs but because of lowered group delay..... the audible group delay effect builds with multiple roll offs from various connected equipment if they all used coupling caps..... so go all direct coupled from source to speaker.
The little (Marsh) headphone amp talked about here (DIYAudio) and published in Linear Audio does not use a DC servo and is rock solid dc stable on the output vs temp and drift and has thd below -100db into 30 Ohms. No IC's used. I dont see why any other low gain line level stage shouldnt be able to do as well without servo. High dc gain circuits is another matter. -RNM
The little (Marsh) headphone amp talked about here (DIYAudio) and published in Linear Audio does not use a DC servo and is rock solid dc stable on the output vs temp and drift and has thd below -100db into 30 Ohms. No IC's used. I dont see why any other low gain line level stage shouldnt be able to do as well without servo. High dc gain circuits is another matter. -RNM
Last edited:
Right !
Back to best capacitor / application
rated by :
DC voltage accross
AC voltage accross
AC current through
Same chip as in the long-gone 2SK147. Make sure you buy enough for everyone
Very low noise. 700pV/rt Hz at 10mA, typical.
From wiki Blauert and Laws on the audibility threshold for group delay
Frequency Threshold
500 Hz 3.2 ms
1 kHz 2 ms
2 kHz 1 ms
4 kHz 1.5 ms
8 kHz 2 ms
I am not a proponent of capacitive coupling where it's not necessary, but if you put the -3dB points at least 2-3 octaves above and below the audio band, will group delay really be a factor?
The LF -3dB point on my e-Amp is about 2Hz.
I don't think Toshiba tests (or tested) for noise other than getting data for the typical spec. Then they set some comfortable limit above this for the maximum spec, probably 5 or 6 sigma limit. And it sounds like from John's experience that sometimes those limits may still be exceeded
But at 10Hz you just have to measure, if it is that critical. See, iirc, Scott Wurcer's comments in the first of his Linear Audio articles about the BF862, most of which but not all have exemplary noise at 10Hz.
I'm sure the differences between the 147 and 363 data are in characterization, including using a bottom-of-the-range Idss for the transconductance in the 147, and a higher gm for the 363 along with wider limits.
I took several hundred 2SK364BL parts when Harman was shutting down a program (with their blessing --- they would have been trashed). I have yet to find a single one that was any much better or worse than any 2SK170BL. The 364 was to be used in the cancelled program's product as an analog switch, but looking at datasheets I realized that it was the same chip as the 170. Of course they were all of the same date code. So despite the favorable statistics within the batch, your mileage may vary. And I didn't measure each one at 10Hz.
The first approach to the Harman attenuator, for the subjective evaluation lab that has the speaker mover, was to use matched pairs of the 364s in the front ends of the line-level processor. We did a two-parameter sort on hundreds of 364s of pinchoff voltage and Idss, and selected nearby pairs (the plot was an interesting elliptical cloud; the guy doing the work got very tired of the effort). Each pair was mounted in an aluminum disc along with heaters and temp sensors, the idea being to run the devices just slightly above a cool ambient. I still have the 32 pairs in the disks.
Unfortunately the cost mounted, and despite the notion that the attenuator was a "cost-no-object" project, I scaled back to a less extravagant design. Then even that was deemed too expensive to finish, although some boards were made and performance results obtained, and the envisioned system described in an AES conference paper. The system, which included a bunch of Vishay S102 resistors and mercury-wetted reed relays, did nearly achieve a S/N below maximum output of ~144dB (the objective being to manage a true 24 bit system), although the paper wrongly states that the voltmeter used for some of the measurements was a true-rms type, making the results a bit worse than shown. Had the original design been done it would have easily met the objective.
Looking at it today, I'd do it a lot differently.
Last edited:
Yes, very good in fact. Somewhere in the BT (?) thread I posted a picture of the Quantech face (digital version) with the BF862 as DUT (2.5nV @ 10Hz iirc). Probably no one can afford a dirty fab these days. I have not found a noisy one yet though I only have gone through 50 or so.
Speaking of mods...
A customer brought his secondhand HCA1200 II to me for idle current adjustment because one of the channels was getting much hotter than the other one. He also agreed to get his amp modified after I told him what I've learned about the Parasound amps from that era from various posts by John Curl so far.
I know that you, John, put this particular amp into 'home theater' category, but I thought that it will be interesting exercise to hear what happens and whether if confirms your statements when I replace feedback and input resistors with Resistas (I have MN-3's here, same as MK-3, but additionally epoxy coated), remove all multiple Taiwanese bypass caps and replace them with one in the 0.1-0.22uF range (ROE MKP1841 probably, nothing exotic) and get rid of input level adjustment pots.
I'm not sure if I want to deal with internal wiring, though. Should I?
I promise to report back.
Best,
P.S. John, what's the recommended idle current per transistor in 1200 II? Thanks!
From wiki Blauert and Laws on the audibility threshold for group delay
Frequency Threshold
500 Hz 3.2 ms
1 kHz 2 ms
2 kHz 1 ms
4 kHz 1.5 ms
8 kHz 2 ms
I am not a proponent of capacitive coupling where it's not necessary, but if you put the -3dB points at least 2-3 octaves above and below the audio band, will group delay really be a factor?
The LF -3dB point on my e-Amp is about 2Hz.
I dont know. Group delay thresholds have been studied. Note- if a person 'likes' the sound of tube equipment, there are many caps and transformers which if measured as a total system would have worse total group delay from system input to system output. Direct coupled thru out is better in this regard and that is just another rreason why ss has best potential for audio. Not to mention, as JC points out, the size and space of the cap(s) and their added costs.
So, if your cap is the only one in the 'system' and the system group delay falls under the threshold, no problem-o. But, why have it if we can design circuits with out it and save space and costs?
Last edited:
John Curl:
"Generally, to set the Iq of any of my amps, you should measure across one Emitter resistor and get about .022V. The range of .015-028V is OK, but the REAL test is the final temperature. It should be warm, but you should not burn your hand touching the heatsink."
Thanks for desiging such an enjoyable amp John.
For my similar, but later, version HCA1500A, I added to the existing heat sink area, and then increased the bias current into the upper range for a very nice sound. You could probably trick me into blind-guessing it was a A21. (Yes, had two.)
The bias took several hours to optomize, as I always waited for the temperature to stablize after each adjustment. This was done with the top cover in place and with a shelf located 4 inches above the top cover. Measurements were done with a type k thermocouple via a Fluke multimeter.
The added HS consists of a full length piece of 1/2*1/2*1/2*1/16 inch alum "U" channel. The channel is held in place by epoxy and self-tapping screws using the existing pilot holes in the top of the HS. The Artic Silver heat transfer epoxy enhances the cooling efficiency. Additional 1/2" dia holes in the chassis below the HS also aid in the airflow cooling.
HTH, w/ Regards,
DavidT
"Generally, to set the Iq of any of my amps, you should measure across one Emitter resistor and get about .022V. The range of .015-028V is OK, but the REAL test is the final temperature. It should be warm, but you should not burn your hand touching the heatsink."
Thanks for desiging such an enjoyable amp John.
For my similar, but later, version HCA1500A, I added to the existing heat sink area, and then increased the bias current into the upper range for a very nice sound. You could probably trick me into blind-guessing it was a A21. (Yes, had two.)
The bias took several hours to optomize, as I always waited for the temperature to stablize after each adjustment. This was done with the top cover in place and with a shelf located 4 inches above the top cover. Measurements were done with a type k thermocouple via a Fluke multimeter.
The added HS consists of a full length piece of 1/2*1/2*1/2*1/16 inch alum "U" channel. The channel is held in place by epoxy and self-tapping screws using the existing pilot holes in the top of the HS. The Artic Silver heat transfer epoxy enhances the cooling efficiency. Additional 1/2" dia holes in the chassis below the HS also aid in the airflow cooling.
HTH, w/ Regards,
DavidT
I agree. One cap is not a problem and the group delay should not be an issue if selected correctly.
- Status
- Not open for further replies.