Very interesting. That's a very interesting amplifier. Could be used without feedback and it would be an acceptable amplifier (not a good one but OK)
One of the dangers with the Blomley is that it's so very high frequency. Peter says he was really interested in it for RF amplification and that he thought it too good for audio use...
One of the dangers with the Blomley is that it's so very high frequency. Peter says he was really interested in it for RF amplification and that he thought it too good for audio use...
Thank you Mr. Herron. Honestly, if I hadn’t read the great post you wrote (which isn’t free of emotional overtones either) I would never have built this amp. I built it and I didn't regret it. Such great sound quality is very difficult to achieve with other simple topology. Many people have sent me private message about the PCB, so I decided to design a new board that will be much more compact, easier to install and will preferably one-sided. I hope I have time for it. This amplifier is too good not to be built. Those who have listened to it all speak of its sound in superlatives. And they are right.
Member
Joined 2009
Paid Member
Mr. Blomley wrote in him paper: "The amplifier design is hopefully only a source of ideas which may encourage further research into the whole approach to design"
If my simulation is correct, we can state the following: the loop gain corner frequency is low, as expected from a single pole compensation, beyond that the gain drops relatively quickly. The UGF surprisingly low (164kHz), it would be much better at several 100kHz, say 800kHz or more. What if we use multipole compensation to achieve more or less constant loop gain in audio band? It will decrease the phase intermodulation distortion as well. Any opinion?
If my simulation is correct, we can state the following: the loop gain corner frequency is low, as expected from a single pole compensation, beyond that the gain drops relatively quickly. The UGF surprisingly low (164kHz), it would be much better at several 100kHz, say 800kHz or more. What if we use multipole compensation to achieve more or less constant loop gain in audio band? It will decrease the phase intermodulation distortion as well. Any opinion?
Last edited:
Member
Joined 2009
Paid Member
Yes, at least from my opinion you are on the right track. I added two pole compensation in the design of my TGM8 amplifier with good results. It works best with (requires) a compound VAS, so I’m not sure what you have shown here Is sufficient.
There is the risk that this is a slippery slope to ongoing improvements in the technical performance, likely ending when the amplifier sounds and measures great but has lost much of it’s original character and becomes similar to many others. Nevertheless, until you go down this rabbit hole you will never know what this topology is capable of and the exciting part is you are possibly the pioneer doing this. You may end up with a topology that is similar to my TGM8 (which is my best sounding amp) in some regards but the differences may elevate it further?
Last edited:
In the meantime i started designing the new PCB. I try to choose the most suitable transistors for the amplifier. The type of splitter transistors is crucial, as Dermot Herron mentioned. Why? Peter Blomley says:
"There is a problem with the use of transistors as signal splitters due to the emitter-base depletion capacitance. Under conditions of low injection this can add an additional phase lag during the crossover period. The problem can be overcome by using silicon planar devices with very high transition frequency (fT) or by selecting devices in which the fT is dominated by the diffusion capacitance as fT remains constant down to very low emitter currents."
Now I see that ZTX453/553 transistors are not the best choice for a splitter (these have only 100MHz ft), so I decided to use BC546/556 (fT 250-300MHz) in version 2.0. If anyone have better (faster) suggestion please let me know.
Last edited:
Dear sir, could you please send me the Sprint Layout files please, as I am using Abacom software for severla years. Thanks in advance. Regards, Willem
I tried hard to get the best sim out of blomley design using bc337/327/bd139 and mjl power trz.. but as far as i could do it i never saw a functional sim with 60mA idle current.Actually i can't get decent results below 1.2 Amp idle current .If the amp works as presented in real life, I find it impossible to sim with the original idle current using normal transistors so i concluded there must be something that either i don't get about it or the trz models i use can't replicate the original trz behavior.
File attached. However, it's worth waiting for the new version of PCB, it is 90% complete, I'll upload soon.
I have the same experience. Its sound, however, is impeccable. Similar to VHEX+ amplifier. The LT Spice simulations show a grossly poor result, but based on the measurements and listening tests, it is still a very good amplifier.
Oh, I certainly am interested in that pcb. I'll wait then. Thanks fort sending the file anyway.
Does it mean that we aren't really talking about a class B amp, but a full class A amp instead? 1.2 Amp idle current means bussiness...
Looking at the schematic, I see that R11 biases the OPS to whatever idle current you like while the class-B operation happens before the OPS in Q4+Q5. The output is not class-B but the pre-driver is. So I have to wonder, if you are going to run the OPS class AB, there is no efficiency benefit nor any thermal advantage, so the whole idea seems a bit pointless.
A slight modification to a typical 2EF achieves a similar result, but with less the cross-over distortion by running the drivers class-B with the OPs in AB. There is also "auto-bias" where the bias senses driver/op base current, which competes well with standard fixed VBE bias.
This Blomley does have a lot of driver stages, similar to 3EF, which may be the real key to its's sound.
A slight modification to a typical 2EF achieves a similar result, but with less the cross-over distortion by running the drivers class-B with the OPs in AB. There is also "auto-bias" where the bias senses driver/op base current, which competes well with standard fixed VBE bias.
This Blomley does have a lot of driver stages, similar to 3EF, which may be the real key to its's sound.
Changing all the resistor values to limit output stage current the min limit is arout 500mA...Unless your sim has a different schematic than the original one it is impossible to get below 500mA idle current measured across the 0.56 ohms resistor but even with 1.2 Amp idle current H2, H3...H7 can't go below -50db at 10v output on 16 ohms and -40 db on 8 ohm output.
I can't get anywhere near your results with the original schematic values.
I also tried the Hartsuiker variation which was only able to improve on the noise but nothing else.Even if you remove the bias resistors in series with the bias diodes entirely I can't get below 500mA...
Thank you! I found that the real culprits for the high idle currents were q7 and q8 .I used qbd139/140 ...any other small power transistor was better than that.
This schematic is so interesting and clever! It has this special synergy in the many functions of a single component that shows the touch of a very gifted designer. Much different from most newer circuits that are designed more like blocks with specific purpose...
Anyway, after reading the article I thought it's something I'd like to try out of curiosity... I have now a working channel on the bench and did some measurements. Looks good to me and people shouldn't be afraid to build it. The circuit is tolerant to supply voltages - I test with 35V and the adjustments work fine, but the plan is to use 45V. Peter Blomley obviously did it with 50V without issues.
I haven't heard the amp yet. Also, it's not tested at high power since the board is on a small aluminum L-profile. What I can say is that up to 1W of power I measured extremely wide bandwidth and no sign of visible distortion, thermal instability or oscillation. This thing is flat practically from DC up to 2MHz, which is the limit of my signal generator! With such a wide bandwidth squares are obviously reproduced very well even above 200kHz, if anyone cares about squares at all. To me they mostly tell the stability of the circuit... There is no overshoot or ringing in pure resistive load. I'll test soon with capacitive load to see how the circuit handles reactive power.
Here are the little changes that I've made so far (component numbers by the schematic in the original article):
1. The gain - 100x is huge by today standards. I dropped it to some 30+ by using R3=3k3, R5=100, C5=82pF. R4 was adjusted to 430 ohms to keep the impedance seen by the emitter of Tr1 same as Peter Blomley intended. C2 was changed to 330uF low leakage type.
2. Diode OA47 replaced by BAT46 low drop Schottky for obvious reasons of "unobtainium".
3. C3 changed to 330uF
4. C7 made 6.8mF because I'll drive 4 ohm speakers. This capacitor was also bypassed with another 330uF for better impedance control at high frequencies.
5. All transistors in the input and splitter section are BC546B/556B. In the triples I use MPSA06/56, ZTX453/553 and MJL3281/1302.
6. I found that the circuit may go into self oscillation if the input is left open (disconnected). This can be a disaster with your speakers connected. Be warned! It happens only if you touch and measure on the schematic, but I also saw it happening without any encouragement from me. So, I added 150k resistor from input to ground to give the circuit something to chew in case the input is disconnected. That affects the input impedance a little... I actually intend to limit the input bandwidth by putting a capacitor in parallel to this resistor. 2MHz is insane bandwidth for audio and all sorts of noise find their way in such a wide open circuit. Will test if capacitor of few hundreds pF is useful... Generally this part requires more experimenting.
BTW, the simulation that I downloaded somewhere from the early pages in this thread works fine and I can adjust the currents as desired by playing with the trimmer values... I never trust simulations though. At least not on the level of settings that I'm able to set. I use simulation mostly for troubleshooting of obvious functional errors in a circuit and never expect to see the noise, distortion etc that I'll see when I wire the whole thing and put it in a box.
Sorry for the long post! ... I thought it might be somewhat useful...
Cheers,
Wes
Anyway, after reading the article I thought it's something I'd like to try out of curiosity... I have now a working channel on the bench and did some measurements. Looks good to me and people shouldn't be afraid to build it. The circuit is tolerant to supply voltages - I test with 35V and the adjustments work fine, but the plan is to use 45V. Peter Blomley obviously did it with 50V without issues.
I haven't heard the amp yet. Also, it's not tested at high power since the board is on a small aluminum L-profile. What I can say is that up to 1W of power I measured extremely wide bandwidth and no sign of visible distortion, thermal instability or oscillation. This thing is flat practically from DC up to 2MHz, which is the limit of my signal generator! With such a wide bandwidth squares are obviously reproduced very well even above 200kHz, if anyone cares about squares at all. To me they mostly tell the stability of the circuit... There is no overshoot or ringing in pure resistive load. I'll test soon with capacitive load to see how the circuit handles reactive power.
Here are the little changes that I've made so far (component numbers by the schematic in the original article):
1. The gain - 100x is huge by today standards. I dropped it to some 30+ by using R3=3k3, R5=100, C5=82pF. R4 was adjusted to 430 ohms to keep the impedance seen by the emitter of Tr1 same as Peter Blomley intended. C2 was changed to 330uF low leakage type.
2. Diode OA47 replaced by BAT46 low drop Schottky for obvious reasons of "unobtainium".
3. C3 changed to 330uF
4. C7 made 6.8mF because I'll drive 4 ohm speakers. This capacitor was also bypassed with another 330uF for better impedance control at high frequencies.
5. All transistors in the input and splitter section are BC546B/556B. In the triples I use MPSA06/56, ZTX453/553 and MJL3281/1302.
6. I found that the circuit may go into self oscillation if the input is left open (disconnected). This can be a disaster with your speakers connected. Be warned! It happens only if you touch and measure on the schematic, but I also saw it happening without any encouragement from me. So, I added 150k resistor from input to ground to give the circuit something to chew in case the input is disconnected. That affects the input impedance a little... I actually intend to limit the input bandwidth by putting a capacitor in parallel to this resistor. 2MHz is insane bandwidth for audio and all sorts of noise find their way in such a wide open circuit. Will test if capacitor of few hundreds pF is useful... Generally this part requires more experimenting.
BTW, the simulation that I downloaded somewhere from the early pages in this thread works fine and I can adjust the currents as desired by playing with the trimmer values... I never trust simulations though. At least not on the level of settings that I'm able to set. I use simulation mostly for troubleshooting of obvious functional errors in a circuit and never expect to see the noise, distortion etc that I'll see when I wire the whole thing and put it in a box.
Sorry for the long post! ... I thought it might be somewhat useful...
Cheers,
Wes
Last edited: