X,
Fast and furious prototyping; I doffs me cap Sire........
You know, I believe you could drive your cans straight off this simple BF862 circuit, nothing more. You don't really need more than 7Vpp any way, and a couple of 9V batteries are fine. For 280R cans, I think I would use three jfets total, rather than four, and this would reduce the current from 38mA down to 25mA to conserve energy with batteries.
This is all you need for mobile use; for a serious listening system all you need is a source, a stepped attenuator, then four BF862 jfets, and your 280R cans. With a smps at 18V and a C mult reg it would be dead quiet.
If you polarise good quality electros - that is, put some DC voltage across them - the sound quality is stunning, and they automatically protect the cans from DC. I have never seen a direct coupled system which can compare with this circuit for low phase shift; you could not improve on 0.25 degrees at 100KHz with a power amplifier. This is the result of an open loop circuit, with no global feedback.
HD
Fast and furious prototyping; I doffs me cap Sire........
You know, I believe you could drive your cans straight off this simple BF862 circuit, nothing more. You don't really need more than 7Vpp any way, and a couple of 9V batteries are fine. For 280R cans, I think I would use three jfets total, rather than four, and this would reduce the current from 38mA down to 25mA to conserve energy with batteries.
This is all you need for mobile use; for a serious listening system all you need is a source, a stepped attenuator, then four BF862 jfets, and your 280R cans. With a smps at 18V and a C mult reg it would be dead quiet.
If you polarise good quality electros - that is, put some DC voltage across them - the sound quality is stunning, and they automatically protect the cans from DC. I have never seen a direct coupled system which can compare with this circuit for low phase shift; you could not improve on 0.25 degrees at 100KHz with a power amplifier. This is the result of an open loop circuit, with no global feedback.
HD
Last edited:
Hmm... that's a bold claim! Ok I am game to order 100 units. I have 50 at present. 50 more is about $16. I will need to etch a small test rig to have pads and a small spring loaded clamp to hold these SOT23's in place.
So need to measure Vgs, Idss, Vp. Then
An externally hosted image should be here but it was not working when we last tested it.
Gives mutual transconductance?
Is there a better way to measure? And it should be done at approximate operating conditions?
Electros. There are many; Nichicon,
Black Gate, Cerafine, Rubycon, Kendeil etc.
Matching.
Use same 18V
Set up a jig with 1k drain resistor, 10R source resistor to ground and a 220k direct from ground to gate.
Install BF862 and quickly measure voltage to ground.
You simply need three devices with identical drain to ground - around 8.5V.
You do not need a particular voltage; you want three identical voltages only. This process gives identical Vgs for 10ma drain current and identical dc gm. AC gm is slightly different but it is close enough.
These devices warm up under test. Ensure all start at same ambient temperature and read off the drain voltage within 4 seconds with a fast DMM like a Fluke.
HD
Black Gate, Cerafine, Rubycon, Kendeil etc.
Matching.
Use same 18V
Set up a jig with 1k drain resistor, 10R source resistor to ground and a 220k direct from ground to gate.
Install BF862 and quickly measure voltage to ground.
You simply need three devices with identical drain to ground - around 8.5V.
You do not need a particular voltage; you want three identical voltages only. This process gives identical Vgs for 10ma drain current and identical dc gm. AC gm is slightly different but it is close enough.
These devices warm up under test. Ensure all start at same ambient temperature and read off the drain voltage within 4 seconds with a fast DMM like a Fluke.
HD
Here's an EBay test rig that will get you in the right ballpark for matching. M328 LCD 12864 Transistor Tester DIY Kit Diode Triode Capacitance LCR ESR Meter You just hold the SMD version in place with tweezers.
Ground to gate of 220R or 220k? The parallel power JFETs use 220R whereas input uses 220k. Are we trying to match all to input JFETs.
220k from gate to ground.
Source through 10R to ground.
Drain through 1k to +18V supply.
Match identical voltage from drain to ground, somewhere from 7V to 9V.
We really match the two, three or four jfets comprising the output stage. The input stage is locally constrained by local feedback to a gain of 9.1dB so pretty much any BFI62 in the input stage will works as predicted by the impedances at source and drain.
This will ensure three matched BF862s will all match identical Id and Vgs for the operating point we have selected for this circuit.
Judging by Mark Johnson's data there will be quite a large variance!
HD
Source through 10R to ground.
Drain through 1k to +18V supply.
Match identical voltage from drain to ground, somewhere from 7V to 9V.
We really match the two, three or four jfets comprising the output stage. The input stage is locally constrained by local feedback to a gain of 9.1dB so pretty much any BFI62 in the input stage will works as predicted by the impedances at source and drain.
This will ensure three matched BF862s will all match identical Id and Vgs for the operating point we have selected for this circuit.
Judging by Mark Johnson's data there will be quite a large variance!
HD
Last edited:
It's inherent in FET manufacture.
It's why we see the wide range of parameters for a single device. It does not ssem to matter whether it's a jFET, or a lateral mosFET, or a Vertical mosFET.
All have very wide parameter ranges.
Take the BF862 as being typical and it shows Idss as varying between 10mA and 25mA (250%), Vgs off from -0.3V to -1.2V (400%)
If one uses a single device, then this variance can often be ignored, if the duty suits that device.
Wherever devices are used in parallel, then selection is virtually mandatory.
Even BJT devices used in parallel arrangements benefit from selection.
Just to be clear. I am using Fluke DMM to measure volts below 1k at the Drain pin of JFET relative to GND right? Looking to match in the circa 8v to 9v range or wherever it may be. This is measuring Vgs at operating condition basically.
I have to ask if this circuit will work as well as the sims predict, why are typical amps so much more complicated than 3 or 4 JFETs? There are 35 active part preamps out there. In fact I am planning on building one. It has a huge activation energy to gather that many transistors. 7 or 8 varieties.
I have to ask if this circuit will work as well as the sims predict, why are typical amps so much more complicated than 3 or 4 JFETs? There are 35 active part preamps out there. In fact I am planning on building one. It has a huge activation energy to gather that many transistors. 7 or 8 varieties.
Last edited:
X,
Really complex amps are designed to get very low THD and direct coupling, so that need high gain stages, buffers, global feedback, offset control and protection. Nothing destroys cans than 10V across the voice coil.
But you have found that even though the THD is high, it is more than 99% H2, typically at -64dB down, and they sound wonderful. But you could not easily sell them to technocrats because their figures are dreadful....... you have to make a decision about what you want, and what you believe.
If you put the DMM across the 1k drain resistor, you will measure the mA through the jfet. But if you put the DMM across drain to ground, you are measuring very close to the Vds, the operating point at the circuit, and something less affected by the voltage supply.
So, yes, DRAIN to GROUND. Your first might be 7.72 V, the second might be 8.43V, the third might be 9.34V. Draw up a matrix, a table, with 1cm x 1cm squares. Top left, start with 6.50, move to the right with 6.55, next 6.60, etc. Then you will have a progressive increase of Vds left to right and through three or four rows to 10.50V. As you pass through your 100 jfets, you will find a couple of 6.50V (might be 6.48V and 6.52V), three in 6.55V, and so on.
Eventually you will have maybe ten or twelve in the same square, say 8.65V. You can then take all these, and ensuring they are all at ambient again, sort them down to the last 10mV. This takes time but you will become quick at it. I have done this with bipolars for years and it's surprising how quick it can be once you are organised.
Mas Didiet:
For 32R cans you might need five jfets in the output stage; a total of six in the entire circuit. I would use 120R for the passive load; this will give you around 13.9mA in each device. BUT, finding five matched jfets is going to be difficult! HD
Hugh
Really complex amps are designed to get very low THD and direct coupling, so that need high gain stages, buffers, global feedback, offset control and protection. Nothing destroys cans than 10V across the voice coil.
But you have found that even though the THD is high, it is more than 99% H2, typically at -64dB down, and they sound wonderful. But you could not easily sell them to technocrats because their figures are dreadful....... you have to make a decision about what you want, and what you believe.
If you put the DMM across the 1k drain resistor, you will measure the mA through the jfet. But if you put the DMM across drain to ground, you are measuring very close to the Vds, the operating point at the circuit, and something less affected by the voltage supply.
So, yes, DRAIN to GROUND. Your first might be 7.72 V, the second might be 8.43V, the third might be 9.34V. Draw up a matrix, a table, with 1cm x 1cm squares. Top left, start with 6.50, move to the right with 6.55, next 6.60, etc. Then you will have a progressive increase of Vds left to right and through three or four rows to 10.50V. As you pass through your 100 jfets, you will find a couple of 6.50V (might be 6.48V and 6.52V), three in 6.55V, and so on.
Eventually you will have maybe ten or twelve in the same square, say 8.65V. You can then take all these, and ensuring they are all at ambient again, sort them down to the last 10mV. This takes time but you will become quick at it. I have done this with bipolars for years and it's surprising how quick it can be once you are organised.
Mas Didiet:
For 32R cans you might need five jfets in the output stage; a total of six in the entire circuit. I would use 120R for the passive load; this will give you around 13.9mA in each device. BUT, finding five matched jfets is going to be difficult! HD
Hugh
Last edited:
Hi Hugh,
Is it absolutely necessary to match the Vgs for the circuit to work? or it is only for improvement in SQ.
regards
Prasi
Prasi,
This is a reasonable question because the matching is tiresome and time wasting.
The problem is that you have two, three, four or five devices with identical gate potential.
This means that because of the huge variance of BF862s - and indeed all jfets and quite a few mosfets too - you finish up with one identical Vgs from this one gate potential. The result is that one jfet will pass 6mA, another 10mA, and another 12mA, and the y(fs) - transconductance - will vary markedly from one jfet to the next.
This will introduce distortion, which does not sit well in my mind.
Another option is to use a small mosfet which passes large current and, if cooled well, will avoid these matching issues. The one I have in mind is the ZVN2110A which is typically ten times y(fs) (250mS) and rdson of 5R and up to 700mW in TO92. The Ciss is 75pf, which is OK in this role, and it will run from 20V, with max at 100V. With one device in the output stage, we do not need matching, and the N enhancement means the source is about 1.5V below the voltage of the gate.
Hmmm....... thimks..........
Hugh
This is a reasonable question because the matching is tiresome and time wasting.
The problem is that you have two, three, four or five devices with identical gate potential.
This means that because of the huge variance of BF862s - and indeed all jfets and quite a few mosfets too - you finish up with one identical Vgs from this one gate potential. The result is that one jfet will pass 6mA, another 10mA, and another 12mA, and the y(fs) - transconductance - will vary markedly from one jfet to the next.
This will introduce distortion, which does not sit well in my mind.
Another option is to use a small mosfet which passes large current and, if cooled well, will avoid these matching issues. The one I have in mind is the ZVN2110A which is typically ten times y(fs) (250mS) and rdson of 5R and up to 700mW in TO92. The Ciss is 75pf, which is OK in this role, and it will run from 20V, with max at 100V. With one device in the output stage, we do not need matching, and the N enhancement means the source is about 1.5V below the voltage of the gate.
Hmmm....... thimks..........
Hugh
There is a lot of tapping in the dark here so guys please do yourselves a favor and find out how JFETs really work.
It's a short and interesting read - inside these two articles by late Erno Borbely you'll find the answers to all the questions that bug you right now: all the basic stuff and a lot of practical circuits.
It's a short and interesting read - inside these two articles by late Erno Borbely you'll find the answers to all the questions that bug you right now: all the basic stuff and a lot of practical circuits.